摘要:

Ultra-violet and X-Ray Photoe€ectron Spectroscopy Studies of Oxygen Chemisorption on Copper, Silver and Gold BY S. EVANS,* E. L. EVANS, D. E. PARRY, M. J. TRICKER, M. J. WALTEM AND J. M. THOMAS Edward Davies Chemical Laboratories, University College of Wales, Aberystwyth, SY23 1NE Received 29 May, 1974 A photoelectron study (HeI, HeII and Mg Ka) of the oxidation at ordinary temperatures of poly- crystalline copper, silver and gold, by both ground-state molecular and excited oxygen has established which oxidation states of these metals are formed during and after the dissociative chemisorption of oxygen. Up to about a monolayer coverage, copper is converted to the Cu(1) state and thereafter, as oxygen uptake proceeds, to the Cu(I1) state. Silver also initially forms a f l oxidation state, but both the magnitude of the line widths of, and the significant absence of satellites associated with the core-level peaks suggest that a diamagnetic Ag(II1) oxide rather than Ag(I1) is formed upon further oxidation. Gold does not proceed beyond the Au(1) state.The densities of states (as reflected in the HeII spectra) of the three metals offer less guidance as to the conditions or ease of formation of the M(1) oxide overlayers, than the energy differences between the s-band edges and the vacuum level. The high sensitivity of XPS and UPS as techniques for monitoring composi- tional and electronic changes at solid surfaces has led us to embark upon a compara- tive study of a group of closely-related metals, the general bulk chemical behaviour of which is widely appreciated.The principal aim of the present work is to assess the importance of such factors as the density of states (DOS) close to the Fermi level, the position of the latter with respect to vacuum, and the relative positions of the d- and s-regions with respect to one another in governing the ease and extent of surface oxide formation when a simple oxidant is introduced to a clean dlOs type metal, A full account of our work on these metals is in preparation. It is a fortunate fact that recent independent XPS studies 9 of a range of transition metal compounds have convincingly demonstrated that shake-up satellite peaks are usually associated with paramagnetic metal ions, but are generally absent when the ionized species are diamagnetic. This fact, along with others 5* discussed below, enables the oxidation states of the parent metals at various stages of oxidation to be identified.Moreover, this information greatly facilitates the understanding of the UPS data which would otherwise be rather difficult to rationalise. EXPERIMENTAL All measurements were made on an AEI ES 200 electron spectrometer between 25 and 80°C. The u.v.(HeI and HeII) and X-ray (Mg Ka) sources were mounted at 180" to each other and at 90" to the analysed electron beam; the sample could be rotated to permit excitation by either source without loss of vacuum. The base pressure was ca. lo-* Torr. Samples of polycrystalline copper, silver and gold were cleaned by alternate argon-ion bombardment (2x Torr, 700 V, beam energy 4-8 PA) and exposure to microwave- excited oxygen (2450 MHz, Torr, 80 W) : this oxidant removes residual carbonaceous material (resistant to the Ar+ treatment alone) rapidly at room temperature, while the 97 5s-D98 STUDIES OF OXYGEN CHEMISORPTION resulting oxide can easily be removed by further Ar+ bombardment.The clean surfaces were characterised by both UPS and X P S ; the latter indicated that the surfaces were essentially free of surface contamination. Freshly cleaned surfaces of each metal were exposed to both ground-state molecular and microwave-excited oxygen at Torr, and the progress of each oxidation monitored 3 d -vv 15 I0 15 10 binding energy /eV FIG. 1.-HeI (left) and Hell (ri&ht) spectra of metallic copper (a) immediately after argon-ion etching : the dashed line shows the effect of submonolayer quantities of carbonaceous contamination, deposited during a period of approximately 10 min ; (b) after completion of the first stage of oxidation (approx- imately monolayer oxygen coverage) ; and (c) after completion of the second stage of oxidation.continually by UPS and XPS. Changes in the work function as chemisorption proceeded were estimated using secondary emission threshold methods similar to those previously described * : however, U.V. rather than X-ray excitation of the secondary emission was generally used as higher intensities were thereby produced.EVANS, EVANS, PARRY, TRICKER, WALTERS AND THOMAS 99 RESULTS COPPER He1 and He11 spectra of freshly-cleaned copper are reproduced in fig. la. Recontamination of this clean surface (unlike those of silver and gold) by carbonaceous inaterial was rapid, the characteristic UPS band (centre ca.8 eV from the Fermi level i' I I l l l l l l l l l l l l l l i l 1 I l i l 1 l I energyleV FIG. 2.-XPS Cu 2p3 signal: (a) clean copper, (b) after completion of first stage of oxidation, and (c) after completion of second stage of Oxidation, as determined by UPS ; further oxidation results in progressive loss of the underlying Cu metal XPS signal (arrowed) as incorporation of oxygen proceeds. 9 5 0 94 0 930 'TABLE BINDING ENERGY AND WORK FUNCTION DATA (eV) FOR COPPER, SILVER AND GOLD AND THEIR SURFACE OXIDES Wk fn increase work from clean metal 0 Is BE core-level metal function &0.1 eV O l s B E b rel. to vacuum BE increase cu 4.5 0.45 530 535.0 Nil (2p) 0.9 529.4 534.8 ca.1 eVC(2p) 15311 Ag 4.3 0.45 530.6 535.4 Nil (3p, 3d, 4p) 0.9 529.5 534.7 Au 5.3 0.7+ 0.2 529.6 535.6 ? 1.4d (4f) ref. (1 1 ) ; 0 relative to the FL : via Cu 2p3 = 932 for Cu, Au 4fi- = 84 eV for Ag and Au ; repro- ducible within 0.2 eV ; C band maximuin : exchange splitting not resolved ; dnew peak very weak- value obtained by estimated deconvolu tion. f.w.h.m. ca. 4 eV) often appearing within a minute at lo-* Torr. The effect of sub- monolayer quantities of this contamination on the UPS of copper is indicated in the figure. Torr the same products (as charac- terised by their UPS) could be obtained using molecular oxygen as with the microwave- excited oxygen. Build-up of the 0 1s XPS signal was rapid : initially a single sharp (f.w.h.ni.1.2 eV) line appeared, soon reaching a " plateau " intensity beyond which On oxidation of a clean surface at100 STUDIES OF OXYGEN CHEMISORPTION growth became progressively slower. Completion of the first stage of the adsorption was followed by the similar growth of a second, weaker, 0 I s peak at higher binding energy (BE) as the first peak moved to lower BE. During the first stage, the Cu 2p XPS signal was gradually reduced in intensity, but showed no detectable broadening, chemical shift, or associated " shakeup " structure. All three of these features, a -1 I t I I I I I I I I I I I I I I I I I I i 15 10 5 0 15 10 5 0 binding energy lev FIG. 3.-HeI ( l e f ) and Hell (right) spectra of metallic silver: (a) immediately after argon-ion etching ; (b) after exposure to ground-state molecular oxygen [note that the Ag 5s band (unlike the Cu 4s band in Fig.lb) is still clearly visible], and (c) after prolonged exposure to microwave-excited oxygen. however, developed concurrently with the second 0 1s signal, as shown in fig. 2. The UPS of the surface was frequently monitored during the oxidation; spectra after completion of each stage are reproduced for comparison with those of the clean metal in fig. 16 and Ic. BE and work function data are given in the table.EVANS, EVANS, PARRY, TRICKER, WALTERS A N D THOMAS 101 SILVER He1 and HeII spectra of freshly-cleaned silver are reproduced in fig. 3a. Exposure to molecular oxygen at 1 0-4 Torr resulted in the initially fairly rapid growth of a single, sharp 0 Is peak (Fwhm 1.4 eV), the intensity of which again tended towards a " plateau " value with continued exposure.No significant change occurred in the Ag 3d signal throughout the oxidation other than some loss of intensity. UPS spectra of the oxidised surface are given in fig. 3b. Exposure to microwave-excited oxygen initially produced the same product as molecular oxygen, but on continued exposure a new 0 1s signal appeared at lower BE than the original 0 1s peak, as previously reported.1° The first peak slowly declined in intensity as the new peak developed. \ 6s A f u 15 10 i l l l l l 1 l l l I 1 i l 1 l 1 l l l l l l l l l 5 0 15 10 5 t, binding energy/eV FIG. 4.-HeI (left) and HeII (right) spectra of metallic gold: (a) immediately after argon-ion etching, and (b) after prolonged exposure to microwave-excited oxygen.The band IabelIed B may correspond with either B or, more probably, C of fig. 1 and 3. The third band expected (see text) would then be largely masked by the intense 5d band structure. The growth of the new 0 Is signal was accompanied by a slight broadening of the Ag 3d peaks, but no significant chemical shift was detected. UPS spectra of the final product are given in fig. 3c, while BE and work function data are included in the table.102 STUDIES OF OXYGEN CHEMISORPTION GOLD He1 and HeII spectra of freshly-cleaned gold are given in fig. 4a. Exposure to molecular oxygen at Torr resulted in no detectable change in either the UPS or XPS of the surface. Attack by microwave-excited oxygen, however, was quite rapid, 5-min exposure producing a substantial 0 1s signal, rather broader (fwhm 2.5 eV) than with either copper or silver.The rate of increase of the 0 1s signal again gradually decreased with continued exposure ; UPS spectra after prolonged exposure are reproduced in fig. 4b. BE and work function data are given in the table. DISCUSSION Our U.V. photoelectron spectra of cleaned Ag and Au foils (fig. 3a, 4a) are in excellent agreement with those previously reported l2 for evaporated films, but the agreement for Cu is less good. The most intense component of the d-band in our He1 spectrum is much more pronounced than in those previously reported,lO* l 2 but for the spectra of ref. (10) at least the discrepancy is probably due to carbonaceous contamination of the evaporated film used in the earlier work, the characteristic UPS contamination band (cf.fig. la) being clearly evident. The ready detection and identification of such contamination is an important advantage of having UPS and XPS facilities available together in the same apparatus, especially when vacuum conditions are less than ideal. We are consequently confident that our spectra relate to surfaces essentially clean at least as far as XPS is concerned : but the possibility remains that differing surface preparation might result in variation in the observed surface band structure. In each case the ns-band is clearly visible on the low BE side of the much more intense (n-1) d-band structures. This latter band is much more complex in the He1 spectra as a result of direct interband transitions which become progressively less important as the exciting photon energy increases.12 The HeII spectra may therefore be expected to reflect the true surface DOS much more closely than the He1 spectra. A prerequisite for the adequate interpretation of the UPS of oxidised surfaces is a knowledge of the effective escape depth for the ejected photoelectrons. Previous work in these laboratories on the oxidation of lead indicated that neither He1 nor HeII radiations were able to eject significant number of primary photoelectrons from below the surface layer of lead atoms, a conclusion supported by Auger studies of the Au/Pb system by Biberian and Rhead l4 who found that the inelastic mean free path for 71 eV electrons in lead was 1.0+0.05 nionolayers.In addition, we have noted during the present work that changes in the electronic band structure resulting from the progress of the oxidation manifest themselves simultaneously in the He1 and HeII spectra. This implies, in confirmation of our earlier work on lead, that the escape depth varies only slightly with electron energy in the 10-50 eV range. The escape depth for gold is probably not very different from that for lead and so the fact that even after prolonged oxidation the 6s band of Au metal is still clearly visible may be taken as an indication that surface coverage is not complete. Similar arguments may be applied in the oxidation of silver by O2 : many previous workers have also concluded that the surface is only partly covered at relatively low oxygen pressures.However, with copper the 4s band has diminished below detectability by the time the first stage of oxidation is complete (fig. lb); here effective coverage of the surface is readily attained. Making allowance for these differing surface coverages, the He11 spectra of the The assignment of these spectra is not in doubt.EVANS, EVANS, PARRY, TRICKER, WALTERS A N D THOMAS 103 initial oxidation products of the three metals (fig. 16, 3b, 46) bear a close resemblance to each other. In all three cases, one new band (A) appears near the FL, together with bands at higher BE than the d-band. Two distinct bands (B and C) can be distinguished in this region for Ag and Cu : only one (labelled B) is readily detectable for Au but the other could well be too weak or broad to be distinguished as a separate band.These resemblances are also apparent in the He1 spectra, although band C in the case of Ag and band B in that of Au are here barely detectable above the rising secondary emission background. These similarities suggest that the metal is present in the same oxidation state in each of these three products. This conclusion is consistent with the work function changes accompanying the oxidation : a similar increase is observed (ca. 0.5 eV, table 1) for all three surface oxides. In the case of copper an unambiguous distinction between the two possible oxidation states Cu(1) and Cu(I1) may be made, since Cu(I1) compounds all show intense satellite structure on the 2p XPS peaks,s* 16* as well as the expected broaden- ing of the parent 2p peaks resulting from exchange interaction with the unpaired electron in the 3d shell.Both these effects are absent in the XPS of this first Cu oxidation product : their appearance as the second 0 1s signal develops confirms that the initial product is a Cu(1) surface oxide, which oxidises to a Cu(1I) oxide on con- tinued exposure to oxygen at room temperature. We consider (by analogy) that the initial surface oxides of Ag and Au also involve M(1) species. This conclusion is consistent with the observation that the profiles of the d-bands in the HeII spectra of these oxidised surfaces are in all three cases very similar to those of the corresponding d-bands in the He11 spectra of the clean metals. We infer from this that the metal- oxygen bonding predominantly involves mixing of the 0 2p orbitals with the s-type states of the metal rather than with the d-type states.In the He1 spectra of oxidised copper (especially), however, the fine structure clearly evident in the spectrum of the clean metal diminishes in intensity as surface coverage increases, and when the process is essentially complete (fig. lb) this structure can no longer be definitely identified. The He1 d-band profile then resembles the HeII profile much more closely. Similar effects are observable with silver and gold. However, because the surface coverage is not complete in these cases the fine structure is only diminished in intensity and not lost entirely. This interesting behaviour presumably arises as a result of drastic changes in the final states involved in the direct transitions responsible for the presence of such fine structure in UPS excited by relatively low-energy photons.Clearly, HeII spectra give a much more direct indication of changes in the DOS on surface treatment than do He1 spectra, which must therefore be interpreted with caution. The 0 1s XPS signal is quite symmetrical and unusually narrow (fwhm 1.2-1.4 eV) during the initial stages of the oxidation of both Cu and Ag, implying that the adsorp- tion is dissociative. We would expect adsorbed O2 (or other paramagnetic oxygen species such as 0;) to show an exchange splitting (1.1 eV in gaseous 02)18* l9 or at least a rather broad 0 I s signal. The fact that the initial products with microwave excited oxygen (which contains some atomic oxygen 20) have the same UPS as the products obtained using molecular oxygen also suggests that adsorption of the latter is dissociative.For gold, reaction occurred only with the excited oxygen. A reaction involving (Ozaas.) is thus again unlikely, even though the 0 1s signal here is rather broad (fwhm ca. 2.5 ev). The reason for this is not clear : broad 0 1s lines are not, however, uncommon in diamagnetic compounds. The 0 1s BE referred to the FL for these three M(1) oxides range irregularly over ca. 1.0 eV. However, if they are corrected to the vacuum level by the addition of the work function of the metal and the work function change on oxidation (see table), this spread is reduced to only 0.65 eV and a smooth trend emerges.104 STUDIES OF OXYGEN CHEMISORPTION With copper and silver, surface oxides of different electronic structure (fig.lc, 3 4 can also be produced on further oxidation. In these oxides, the " metal d-bands " have been radically altered in profile in both the He1 and He11 spectra. We interpret this as evidence of mixing of the 0 2p orbitals with the d-band of the metal, i.e. that the metal is here in a higher formal oxidation state. With copper, the oxidation state involved is almost certainly Cu(I1) because of the concomitant development of the core-level satellite structure mentioned above : but with silver we believe that Ag(II1) is produced. The A g 3d XPS peaks are only slightly broadened (0.2 eV) compared with those of Ag metal, and no shakeup structure could be detected for the 3p, 3d or 4p core levels.This behaviour would be expected only for a diamagnetic Ag species ; much broader 3d lines (ca. 2.5 eV wide), resulting from exchange interaction with the unpaired electron in the valence shell, have been reported for many Ag(I1) com- plexes. The 0 1s XPS signal which develops as the oxidation progresses is 1.1 eV to lower BE than that of the initial Ag(1) surface oxide. This suggests that greater charge transfer is involved in this new Ag-0 system and so argues against the alternative possibility that fig. 3c refers merely to a restructured Ag(1) system rather than to an Ag(II1) surface oxide. The fact that this second stage of oxidation is accompanied by a substantial further increase in work function for both Ag and Cu also argues for the formation of higher oxides.This interpretation of the spectra is in accord with the normal chemistry of these elements,21 copper existing predomi- nantly in oxidation states I and 11, Cu(1) frequently oxidising readily to Cu(II), while silver generally prefers the oxidation state I. Both Ag(I1) and Ag(II1) are also known, but the fact that bulk Ag(I1) oxide consists of Ag1AgrIrO2 21 is not without relevance in the present context. Thus the similar surface DOS profiles of these three metals lead to similar electronic structures for the surface M(1) oxides, once these have been formed. However, the shapes of these metallic DOS plots give us little indication of the conditions or ease of formation of M(1) oxide overlayers. These properties seem more closely linked with the work functions of the metals concerned, i.e., the energies of the s-bands of the metals relative to the vacuum level. The order of reactivity of the metals, Ag > Au, is roughly paralleled by the work function difference Ag - Au( - 1 .O ev), but the small work function difference-of opposite sign-between copper and silver, in contrast with the pronounced change in reactivity, shows that this cannot be the only factor involved. Again, the much more vigorous conditions required to oxidise silver beyond the + 1 oxidation state might be tentatively correlated with the increased stability (2-2i eV) of the d-band relative to copper; but our failure to oxidise gold beyond the + 1 state is not reflected in a similar increase in d-band BE, irrespective of whether the BE are referred to the FL or to the theoretically preferable vacuum level.Clearly, one must be wary of using thermodynamic arguments such as these, which consider only the reactants and products, to explain the predominantly kinetic problem of reactivity. The activation energies for the various oxidation processes may be related not only to the electronic band structure but also to other factors, such as the facility of surface rearrangement to accommodate new surface oxide phases : incorporation of adsorbed gases is more facile for copper than for gold. The activation energies for production of the higher oxides may also be related, inter alia, to the extent of surface coverage of the M(1) surface oxides, the formation of which precedes any further oxidation; complete coverage is readily attained only for copper.Such hypotheses must at present be speculative, but they would be consistent with both the marked difference in reactivity between copper and silver and the increasing difficulty of formation of the higher surface oxides on going from copper throughEVANS, EVANS, PARRY, TRICKER, WALTERS AND THOMAS 105 silver to gold despite the general chemical trend of increasing stability for the higher oxidation states. We thank the Science Research Council for support. J. M. Thomas, E L. Evans, M. Barber and P. Swift, Trans. Faraday SOC., 1971,67, 1875. C. R. Brundle and M. W. Roberts, Proc. Roy. SOC. A , 1972,331, 383. M. Barber, E. L. Evans and J. M. Thomas, Chem. Phys. Letters, 1973, 18,423.T. E. Madey, J. T. Yates Jr. and N.E. Erickson, Chem. Phys. Letters, 1973, 19,487 ; J. T. Yates Jr., T. E. Madey and N. E. Erickson, Surface Sci., 1974, 43, 257. D. C. Frost, A. Ishitani and C. A. McDowell, Mol. Phys., 1972, 24, 861. L. J. Matienzo, L. I. Yin, S. 0. Grim and W. E. Swartz, Inorganic Chemistry, 1973, 12, 2762. S. Evans, Chem. Phys. Letters, 1973, 23, 134. S. Evans and J. M. Thomas, J.C.S. Furuday 11, 1975, 71, 301. J. C. Rivikre, Work Function: Measurements and Results (AERE Report R5526, AERE, Harwell, Berks. U.K., 1967) : subsequently published in Solid State Surface Science, vol. 1, ed. M. Green (Dekker, New York, 1969). l2 D. E. Eastman, in Electron Spectroscopy, ed. D. A. Shirley (North Holland, Amsterdam, 1972), p. 487, and ref. therein. l3 S. Hiifner, G. K. Wertheim and D. N. E. Buchanan, G e m . Phys. Letters, 1974, 24, 527. l4 J. P. Biberian and G. E. Rhead, J. Phys. F., MetaZ Phys. 1973, 3, 675. G. Rovida, F. Pratesi, M. Maglietta and E. Ferroni, Surface Sci., 1974,43,230 and ref. therein. l6 A. Rosencwaig and G. K. Wertheim, J. Elect. Spectr., 1973,1,493 ; B. Wallbank, C. E. Johnson and I. G. Main, J. Phys. C., 1973, 6, L493. l 7 T. Robert, M. Bartel and G. Offergeld, Surface Sci., 1972, 33, 123. l8 K. Siegbahn et al., ESCA Applied to Free Molecules (North Holland, Amsterdam, 1969). 2o R. P. Wayne, in Adv. Photochem., 1969, 7, 330 and ref. therein. 21 F. A. Cotton and G. Wilkinson, Advanced Inorganic Chemisiry (Wiley Interscience, New York, ' D. P. Murtha and R. A. Walton, Inorgunic Chemistry, 1973, 12, 368. lo E. L. Evans, J. M. Thomas, M. Barber and R. J. M. GriBths, Surface Sci., 1973, 38, 245. S. Evans, unpublished work. 3rd. edn., 1972).

ISSN:0301-7249    

DOI:10.1039/DC9745800097

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

He1 Photoelectron Spectroscopy of Small Molecules Adsorbed on Metal Surfaces BY P. BILOEN* AND A. A. HOLSCHER Koninklijke/Shell-Laboratorium, Amsterdam (Shell Research B.V.) Received 24 May, 1974 Multilayer condensation of CBr4 on Pt causes the UPS bands to broaden by some 0.8 eV and to shift by 1 eV towards lower ionization energy. These observations are related to dielectric polarization in the adlayer. Larger shifts, observed in chemisorption, likewise can be explained from image charge screening of the support. For or-CO/Pt a two-band spectrum is observed, whereas for ,tI-CO/Pt only one band is observable. This is ascribed to either the effect of heavy-metal 5d orbital admixture in the CO 2po-derived band, or the formation of bridge-bonded CO. Demuth and Eastman’s results for C6Hs/Ni( 11 1) have been reproduced for polycrystalline nickel.Similar measurements on Pt, Cu and Cu/Ni, however, are severely hampered by the overlap in the spectrum of metal-d and carbon-2p derived bands. Annealing induces a change in the UPS spectrum of 0.33 Cu/0.67 Ni, probably due to enrichment of the surface in copper. Our understanding of the nature of the chemical bond between atoms for a large part stems from spectroscopic studies of the electronic structure of molecules. Likewise, unifying theories on chemisorption are expected to benefit from data on the electronic structure of the gas/solid interface. Exploratory studies indicate that ultra-violet photoelectron spectroscopy (UPS) might furnish such data. Quite a few basic phenomena interfering with the UPS study of chemisorbed species have as yet hardly been investigated, e.g., the effects underlying the broadening and shifting of ad-orbitals-derived bands.In view of these effects it is an open question how useful gas-phase data are for the identification of the molecular structure of adsorbed species. Exploratory studies, such as this one, therefore seem in order. EXPERIMENTAL Our home-built spectrometer uses a cylindrical mirror analyser which, contrary to common practice, has an aperture between the target and the annular entrance slit. This simple alternative for a double-pass analyser allows the rather stringent restrictions on source dimension to be relaxed to some extent, and opens the possibility to scan the spectrum either in the sample potential (pre-retardation) or in the analyser deflection voltage mode.Test runs on argon yielded line-widths between 40 and 50 meV (fwhm). The analyser, surrounded by a double-walled Mo-metal shield, has a bakeable stainless steel housing. With the help of a 90 l./s turbomolecular pump we obtained base pressures in the 10-9-10-10 Torr range. Residual gas analysis with the He-lamp in operation shows P(He) w 1 x Torr, POI2) w 1 x Torr with all other partial pressures below the 5 x 10-l1 Tom level. Research-grade gases were admitted from glass ampoules via a baked metal system. As metal specimens we used polycrystalline materials of 99.99 % purity (Dryfhout) either in the form of thin ribbons (14x 4x 0.025 mm) or of small massive cylinders (3 mm dim., 10 mm length).The sample holder allowed for resistance heating up to 1400°C and cooling to - 12OOC. Temperatures were measured with a thermocouple in direct contact with the Torr, P(CO+C2H4) w 2 x 106P . BILOEN AND A. A . HOLSCHER 107 specimen. Specimens were usually cleaned by a one-hour argon ion bombardment at 900 V/2 FA. Spectra were accumulated in a MCA by repetition of 12.5 s scans over periods of time ranging from 125 to 1250 s. Contributions from low-energy stray electrons obscured the measurement of the total width of the spectra, and thus impaired in-situ measurement of sample work functions. All spectra have been plotted with the energy measured relative to the Fermi level of the substrate (AI) on the abscissa. The spectrometer work function, as observed from the position of the Fermi edge, varies little, and lies close to 5.0 eV.RESULTS AND DISCUSSION PLATINUM Different cleaning procedures, viz., 17-h firing in Torr O2 at 800°C (ribbon- shaped sample) and 1-h argon ion sputtering, both procedures being followed by a short annealing, resulted in identical UPS spectra. We therefore assume the spectrum displayed in fig. 1 to be characteristic for clean platinum. 25 000 ! 8 t 8 8 - 10 5 0 AIleV FIG. 1 .-Platinum after 1 h, 900 V/1.2 p A argon-ion cleaning. Full curue: as measured in the deflection voltage scanning mode; dotted curue: reconstructed from the drawn curve by correcting for the variable analyser band-pass. Comparison with 21.2 eV-photo-electron spectra published for gold 3* shows the structure in the 5d-derived band of platinum to be virtually identical, indicating the rigid-band approximation to be relevant in comparing gold and platinum.PLATINUM-CARBON TETRABROMIDE Exposure of the annealed and pre-cooled (T = - 120 "C) platinum substrate to 10 s x Torr CBr, produces four bands on top of the spectrum of platinum (fig. 2), located at A1 = 12.8, 8.4, 4.8 and 3.2 eV, respectively. Adding a work function of 5.7 eV for polycrystalline annealed platin~m,~ the band positions corres- pond closely to the positions of the a,, 1 t2 and a convolute of the 1 t,, 2t2 and le108 He1 PHOTOELECTRON SPECTROSCOPY orbital of gaseous CBr4,6 apart from a shift of approx. 1.1 eV of all bands towards lower ionization energy (called hereafter, " upward shift "). Similar observations have been made for condensation of H20 on Au and C6H6 on Ni.2 Compared to the gas phase, the It, level is broadened from 0.55 eV to 1.0 eV fwhm.Assuming the broadening function to be Gaussian, it should have a width close to 0.8 eV (fwhm). Such a convolution of the gas-phase spectrum indeed reproduces the overall features of the observed band (fig. 2, inset). 3 lO0OC !3 8 I I I I 15 10 5 0 Torr CBr4 exposure on pre-cooled platinum (T = - 120°C). AI/eV FIG. 2.-10s x Measurement in the deflection voltage scanning mode. Inset (full line) : region of the Itl, 2t2, le bands after sub- traction of the Pt 5d intensity. (dashed line) : corresponding part of the gas phase spectrum, con- voluted with a 0.8 eV-wide Gaussian. Considering the causes underlying a band broadening of 0.8 eV and an upward shift of 1.1 eV, it should be noted that the prevailing exposure conditions as well as the intensity of the CBr, derived bands are indicative of multilayer condensation.Lateral variations in the work function of the substrate thus will average out to a certain extent, although they cannot be ruled out completely. Broadening due to overlap of orbitals on neighbouring CBr, molecules (band formation) seems possible for the peripheral Br 4p orbitals, but unlikely for the localized It, orbitals. As argued below, both the upward shift and the broadening can be explained as effects due to dielectric polarization around the final, electron-hole, state. Polarization of the direct vicinity of the electron hole that is produced by the photo-ionization, will depress the final-state energy and thus also the ionization energy.P.BILOEN AND A . A . HOLSCHER 109 Estimates for dielectric solids vary around 1 eV;' those for metals are considerably higher.8 The upward shift now observed thus has a magnitude compatible with dielectric polarization. For not too large a number of adlayers the polarization energy is likely to vary with the position of the electron hole, viz., with its distance from the adlayer/metal and adlayer/vacuum interface. This would explain a broaden- ing of the order of the mean polarizatiofl energy itself, viz., a broadening of 0.8 eV versus a shift of 1.1 eV. Torr, only the It, derived orbital at AZ = 8.4 eV is observed. From the relative intensities observed in the gas phase, we can understand the absence of the lai derived band.However, despite the masking effect of the Pt 5d band, the Br 4p derived bands should be observable. We therefore conclude that for the first adlayer the Br 4p derived bands are anomal- ously broadened. Decomposition of the CBr, accompanied by formation of Br, or of Br/Pt, if thermo- dynamically feasible, is probably an activated process and therefore rather improbable at the prevailing low substrate temperature. We therefore speculate that one or more of the following factors : life-time broadening, delocalization of the ad-orbitals over the substrate, or inhomogeneous image-charge screening, underlie the anomalous broadening of the Br 4p derived levels. Life-time broadening, due to Auger-type transitions from the substrate valence levels to the electron hole in the adsorbate, has been suggested to be important for adsorbates on metal s~rfaces.~ As this process requires overlap between filled metal orbitals and the electron-hole orbital, it may well be favoured for peripheral orbitals such as the Br 4p ones. With lower exposures, of the order of 1 s x R ADLAYER IMAGE CHARGE SUB STR AT E FIG.3.--Image-charge screening causes ionization energy depressions of the order *e2 < 1 /(R+A-') >, in which A-' is the screening length of the substrate. Condensation on top of a low-temperature substrate is likely to produce an adlayer with an ill-defined geometry. A variation in l/R (cf. fig. 3) in going from one adsorbed molecule to another then should produce a variation in image-charge screening, resulting in band broadening.For a fixed value of AR the variation will increase with decreasing value of R. Thus, the broadening in the first adlayer is likely to be larger for peripheral orbitals near the metal surface. PLATINUM-CARBON MONOXIDE Room-temperature exposure to 20 s x lo-' Torr CO produces one single broad band at A1 - 8.7 eV (solid curve, fig. 4). We refer to this spectrum as the " single band " spectrum. In view of the underlying exposure conditions we assign it to CO chemisorbed in a p-~tate.~. lo With CO present in the gas phase during the measurement at pressures of 1 x Torr or higher, a second band at A1 = 11.7 eV appears (dashed curve, fig. 4). It vanishes upon prolonged pumping, and we therefore assign it to CO chemisorbed in an a-~tate.~ The band at 1 1.7 eV is not due to gas phase CO, as was evident from measurements110 He1 PHOTOELECTRON SPECTROSCOPY with the substrate removed from the ionization region.Only at a CO pressure as high as 2 x Torr does the 2pn band of gaseous CO, including its well-resolved vibrational fine structure, become observable. From subtle changes in shape of the AZ - 8.7 eV band on pumping off CO we infer that a-CO contributes also to that band. We thus assign to a-CO, bands at AZ = 11.7 eV and at A I - 8.7 eV, i.e., a " two-band '' spectrum. The locations of the two bands of a-CO correspond, apart froin an upward shift of some 1 eV and a considerable uncertainty in the precise location of the 2pa band, with the locations of the 2pa and 2pn levels of gaseous CO.Such a correlation has been observed previously for CO/Nil and CO/Mo (low temperature). Single-band spectra, as reported here for P-COIPt, have been observed for CO/W l2 and CO/Mo (low temperature).l' The observed band has been associated with the carbon/metal band, and the apparent absence of the 2pn derived band has been ascribed to a perturbation of the C-0 band. In view of the similarity between Ni and Pt for CO chemisorption 13* l4 one may wonder whether the spectrum reported here is not due to an artifact, such as the formation of carbonaceous residues or a distortion due to the transmission characteristics of the analyser. 10 ooc FIG. 4.-Pt/CO Solid line: 20s x lo-' Torr CO exposure ; dashed line: 1 x Torr CO present during measurement.The sharp peak at 8.7 eV corresponds to gas phase CO. (Measurements made in the deflection voltage scanning mode.) A similar single-band spectrum could in fact be produced by preferential chemi- sorption of C,H, and/or the formation of carbonaceous residues (cf. fig. 4 and 5). Under the prevailing vacuum conditions however, such events are highly unlikely. Alternatively, the analyser transmission characteristics inherent to the deflection voltage scanning mode could hamper the observation of a weak band in the AI = 12 eV region. However, for a 2pn/2pa intensity ratio comparable to that of CO in the gas phase, both bands are observable, as evidenced by observation on the gas- phase alone.P. BILOEN AND A . A . HOLSCHER 111 If it is thus inferred that p-CO/Pt is truly represented by a UPS spectrum which is significantly different from that of cr-CO/Pt and CO/Ni, one should wonder whether this is compatible with the conventional view according to which CO is chemisorbed as a carbonyl-like species.In view of published spectra of several gaseous carbonyls, notably Cr(CO),, Mo(CO), and W(CO),,6 we conclude the answer to be affirmative. If one compares the relative intensities of the 2p0 and 2px derived bands in carbonyls with those in CO it appears that the 2pa orbital derived band has gained enormously in intensity. Because the ionization cross-section of the metal d-orbitals is high compared to those of C2s,p and Ozs,p, such an intensity increase in the 2pa derived band indicates that it has, to a considerable extent, a metal-d-orbital character. The same phenomenon for /I-CO, leading to chemisorption-induced enhancement of the intensity of the 2pa derived band, may explain the " single-band " nature of the UPS spectrum.The p-CO spectrum could, alternatively, be connected with bridged-bonded CO. A large perturbation of the axial symmetry of the CO molecule will cause the band derived from the 2px orbitals to split into two components, whereas electron donation from the substrate will cause it to shift upwards relative to the 2pa derived band. Both effects are observable when comparing the UPS spectra of CO and H,CO (fig. 6). Thus, whereas dominant bridge-type bonding of CO on Pt seems rather difficult, to reconcile with infra-red data,I4 it would be quite compatible with the observed UPS spectrum.PLATINUM-ACETYLENE Exposure of pre-cooled Pt to 20s x Torr C2H2 produced a split band, centred around A1 = 9 eV (fig. 5). For C2H2/Ni, Demuth and Eastman observe a three-band spectrum with, compared to the gas phase,2 an upward shift of some 1000 lo 5 0 AIleV FIG. 5.-Pt/C2H2 : Lowerfull curue: 20s x lo-' Torr CO on Pt (- 120°C) ; upper fulZ curve: specimen temperature increased to + 50°C after low temperature exposure ; dashed curve: C2H2/Ni as measured by Eastman.*112 He1 PHOTOELECTRON SPECTROSCOPY 3 eV in the cr bands (fig. 6, dashed curve). They attribute this large upward shift to, inter alia, image-charge screening.2 For C,H,/Pt we observe an upward shift of some 2 eV. Unfortunately this shift is already sufficient to cause masking of a 20- 3 6 /19.7 16.9\\ / 14.0 1 1 1 I 1 1 \ \ 1 \ \ / i ',/ ' 4 / 20 I H,CO 16.315.9 14.I 10.9 /I 9.7 - 22 1 6 ZleV designated in the text as 2pn and 2pu. FIG. 6.-Comparison of the level positions of CO and H2C0. The I n and 30 levels in CO are discrete lnu orbital derived band, if there is any, by the Pt 5d band. In view of the similarity in position of the C 2p derived uppermost orbitals in the various hydro- carbons, we conclude that detailed studies of chemisorption of such molecules on platinum, and on many other transition metals, will be severely hampered by band overlap. PLATINUM-HYDROGEN AND PLATINUM-OXYGEN Room-temperature exposures, up to 100 L, did not result in the appearance of new bands in the spectrum. BENZENE/Ni, Cu, Cu/Ni For the fundamental study of chemisorption on alloys the system Cu/Ni, in which the variation with composition of the 3d band filling has been thoroughly investigated, is particularly attractive.Recently, Demuth and Eastman obtained promising results for the chemisorption of CzH2, CzH4 and CGH6 on nickel (1 11). We there- fore attempted to extend these measurements to Cu and Cu/Ni substrates. C6H on polycrystalline nickel produced results almost identical to those observed for Ni(ll1) (fig. 7). However, on Cu (fig. 8) and on 33 % Cu/67 % Ni the upper n bands of C6H6 appear to be masked by the 3d band structure of the copper. As with platinum, excessive intensity of the copper 3d band in the region AZ > 3 eV, combined with the upward shift in adsorbate levels due to image-charge screening, impairs the observation of the highest adsorbate bands.One would expect, as pointed out by Van Santen,15 that image-charge screening decreases with increasing delocalization of the (final-state) electron hole. This follows classically, for instance, from the decrease in Coulomb interaction between two charged rings, on increasing their radius while keeping their relative position andP. BILOEN AND A. A . HOLSCHER 113 i I I I I 15 10 5 0 AZleV FIG. 7.-C6H6/Ni: 25s x Torr CsH6 on a polycrystalline nickel substrate at room temperature. Only curve a corresponds to the AZ-span indicated on the adscissa. Curve b corresponds to a separate measurement over the indicated region (measurements in the retarding voltage scanning mode). This measurement and that of Eastman2 have been replotted on a common AZ scale (curve c and d re- spec t ivel y ).total charge fixed. The same effect arises in molecular orbital calculations, because the reduction in the value of the Coulomb integral on a given atom, which represents the image-charge screening, is approximately proportional to the fractional electron- hole charge on that atom. Therefore the upward shift should be larger for o orbitals 15 10 5 0 AIfeV FIG. 8.-20 s x lo-’ Torr C6Hs on Cu.114 He1 PHOTOELECTRON SPECTROSCOPY than for delocalised n orbitals. Preliminary observations seem to be in conflict with the above expectations. We finally report that annealing the argon-ion sputtered Cu/Ni specimen causes the Cu-derived band in the EDC to gain in intensity relative to the Ni-derived one (fig.9). We ascribe this phenomenon to annealing-induced copper enrichment of the 5 0 AIleV FIG. 9.-EDC of freshly sputtered 33 % Cu/67 % Ni before (full curve) and after (dashed curve) annealing. surface. With XPS and AES such composition changes, notably enrichment with respect to the component with the lowest heat of sublimation, have been observed pre- viously.16 Study of these phenomena at different photon energies in the vacuum U.V. region might yield additional information about surface enrichment in alloys, because in this energy domain the electron mean free path varies considerably with the electron kinetic energy. The authors gratefully acknowledge the contributions of Mr. F. Lieder in designing and constructing the U.V. spectrometer. They are indebted to Dr. Eastman and Dr. Demuth for making available a preprint of their recent chemisorption work. D. E. Eastman and J. K. Cashion, Phys Rev. Letters, 1971, 27, 1520. 1974 C. R. Brundle and M. W. Roberts, Surface Sci., 1970, 38, 234. D. E. Eastman and J. K. Cashion, Phys. Rev. Letters, 1970, 24, 310. R. Bouwman, H. P. van Keulen and W. M. H. Sachtler, Ber. Bunsenges, 1970,74,198. Molecular Photoelectron Spectroscopy, ed. D. W. Turner (Wiley-Interscience, London, 1970). ’ P. H. Citrin, R. W. Shaw, A. Packer and T. D. Thomas, Proc. Int. Con$ Electron Specfro- scopy (Asilomar, 1971), Contr. V.4. * C. D. Wagner and P. Biloen, Surface Sci., 1973, 35, 82. H. D. Hagstrum and D. E. Eastman, personal communication. * J. E. Demuth and D. E. Eastman, (a) 2nd. Int. Conf. Solid Surfaces, Kyoto, March 25-29, (b) Phys. Rev. Letters, to be published. l o W. L. Winterbottom, Surface Sci., 1973, 37, 195.P. BILOEN AND A . A . HOLSCHER 11s l 1 T. J. Atkinson, C. R. Brundle and M. W. Roberts, Chem. Phys. Letters, 1974, 24, 175. l 2 J. M. Baker and D. E. Eastman, J. Vac. Sci. Technol., 1972, 10,223. l 3 G. Blijholder, J. Phys. Chem., 1964, 68,2772. l4 D. 0. Hayward in Chemisorption and Reactions on Metallic Films, ed. J . R. Anderson (Aca- demic Press., London, 1971). R. A. van Santen, personal communication. l 6 R. Bouwman and P. Biloen, Surface Sci., 1974, 41, 348.

ISSN:0301-7249    

DOI:10.1039/DC9745800106

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

Photoemission Spectra of Adsorbed Layers on Pd Surfaces BY H. CONRAD, G. ERTL, J. K~~PPERs* AND E. E. LATTA Physikalisch-Chemisches Institut der Universitat Munchen, 8 Munchen, Germany Received 2 1 May, 1974 Adsorption of CO, NO, Hz and O2 on (1 10) and polycrystalline Pd surfaces was studied by means of ultra-violet photoelectron spectroscopy (UPS) using He1 and HeII resonance radiation. Generally, a marked decrease of the emitted intensity from the energy range just below the Fermi level was observed after adsorption. Adsorption of CO causes the appearance of two peaks at 7.9 and 10.8 eV below Ef ; after NO adsorption two maxima at 9.3 and 15.5 eV were observed. Adsorption of oxygen leads to a peak at 5.8 eV. Attempts are made to correlate these additional structures with energy levels of the free molecules.The usefulness of the UPS technique in studying surface reactions is demonstrated with the thermal decomposition of NO as well as with the displacement of adsorbed NO by CO. Recent work on ultra-violet photoelectron spectroscopy (UPS) with solids has demonstrated the surface sensitivity of this method which is mainly due to the small mean free (i.e., elastic) paths of low-energy electrons in matter. U.V. radiation from noble gas discharge lamps yields maximum photon energies of 40.8 eV corres- ponding to emission from the HeII resonance. Thus, an information depth of Q 10 A is achieved for electrons originating from electronic states down to about 15 eV below the Ferrni level. (With proper experimental conditions for the He gas discharge lamp even higher photon energies may be attained whose intensities, however, are too low for most practical purposes).In a pioneering work, Eastman and Cashion showed clear evidence for the variation of u.v. photoelectron spectra from metal surfaces by the presence of adsorbed gases and, moreover, were able to correlate the observed " extra "-emission with some of the orbital energies of the free molecules. In the meantime, surface effects in UPS measurements and the usefulness of this technique for adsorption and surface reaction studies were demonstrated by several authors. An important task is to correlate U.V. photoelectron spectra from adsorbate systems with information attainable by other techniques as, for example, LEED. It is an open question to what extent the UPS data may reflect structural changes or variations of the binding energies of the adsorbed layers.Problems of this kind can only be tackled by combining UPS with other methods in the same u.h.v. system. For characterizing the state of surface cleanliness the use of Auger electron spectroscopy (AES) is a necessary prerequisite. LEED may yield informations on the structural properties of clean and adsorbate-covered single-crystal surfaces. The apparatus used in the work described below combines UPS with the LEED and AES techniques and contains furthermore facilities for recording thermal desorption spectra. High- precision work-function measurements are one of the extensions planned for the future. The adsorption of carbon m~noxide,~' oxygen and hydrogen on Pd single- crystal surfaces has been studied extensively in the past by several techniques, and data on adsorption energies, work function changes and structures of adsorbed 116H.CONRAD, G . ERTL, J . KUPPERS AND E. E. LATTA 117 layers are already available. Therefore we started UPS with the adsorption of CO, H, and O2 on (1 10) single crystal and on polycrystalline Pd surfaces. In addition, some data for adsorbed NO are presented ; the detailed results on this system will be described elsewhere. ' UPS studies with the system HJPd have already been performed by Eastman et al. who found some " extra " emission from a region centred at about 5.5 eV below E,. Due to the chosen experimental conditions the authors claimed that this effect was not caused by the presence of adsorbed hydrogen but by the formation of a bulk hydride.In the present investigation the emission mainly from an adsorbed hydrogen layer was observed although, however, some small contribution from the bulk cannot be excluded. did not discover any effect on the UPS data after CO adsorption although with nickel surfaces it was clearly present.l Since the adsorption properties of Pd and Ni with respect to CO are similar, corresponding U P S spectra are to be expected after adsorption on clean surfaces, this indeed has been found in the present work. EXPERIMENTAL The experiments were performed with a combined LEED-AES-UPS system in which the sample could be positioned in two experimental planes by means of an extended linear motion feedthrough.As can be seen from fig. 1 the upper plane contains the LEED grid optics, an additional glancing angle electron gun (for AES by using the LEED optics as a retarding field analyzer), an argon-ion sputter-gun for cleaning the sample surface, and a quadrupole mass spectrometer. A hemispherical electron-energy analyzer is located in the lower plane together with the U.V. discharge lamp. The former is rotated by 90" with respect to the axis of the LEED optics, thus restraining the experimental set-up for UPS to a fixed geometric arrangement of analyzer entrance slit/sample/u.v. photon source. Only The authors of this earlier paper FIG. 1.-Combined LEED/Auger/UPS-system. Left : side view ; right : front view.118 UPS OF ADSORBED LAYERS ON Pd the angle 6 between the sample surface and the photon beam can be varied within certain limits.It was found that the total current of photoemitted electrons increased with decreasing the angle 6. Therefore, all the spectra reported below were recorded at the smallest accessible angle of 30". The analyzer resolution was normally adjusted at 0.4eV which gives a reasonable compromise between intensity and measuring time. Electron emission was excited by radiation from a cold gas discharge in a differentially pumped windowless lamp system. The sample chamber is evacuated by means of a turbomolecular/diffusion pump combination to give base pressures below Tom routinely without the use of liquid-nitrogen cold traps. The whole apparatus was constructed by Vacuum Generators. During operation of the lamp, the pressure in the u.h.v.chamber rose to 5~ loe9 Tom (He I) and 8 x 10-lo Torr (He 11), respectively. Desorption effects caused by the continuous beam of He atoms striking the surface as well as due to excitation of the adsorbed particles by the incident photons were never observed. These findings were mainly confirmed by the fact that the LEED patterns from adsorbate structures were never degraded even after exposing the surface to the U.V. radiation for several hours. However, it cannot be excluded that a small desorption effect (if existing) would have been compensated by readsorption from the residual gas atmosphere. Experiments were performed with a (1 10) surface and with a " polycrystalline " surface. The latter consisted mainly of domains of (111) and (100) orientation giving rise to a super- position of the corresponding diffraction spots in the LEED pattern. Cylindrical disks of about 5 mm diam.and 1 mm thickness were used as samples which were mounted on the manipulator by spot welding between two thin tungsten wires which themselves were fixed to supporting leads of molybdenum. The samples could be resistively heated up to 1OOO"C; the temperature was controlled by a fine Pt/Rh thermocouple spot-welded to the back side. J J 1 - 6 = 0 2 4 6 8 10 12 FIG. 2.-Photoemission spectra of a (a) clean and (b) CO-covered Pd(l10) surface (pco = 5 x lo-' Ton) ; (c) difference spectrum. binding energyleVH . CONRAD, G . ERTL, J . KUPPERS AND E. E. LATTA 119 The surfaces were cleaned in situ by prolonged argon-ion bombardment /annealing cycles until no impurities could be further detected by AES.The dominating contaminants were carbon and sulphur, as commonly observed. Gases of high purity could be introduced into the sample chamber through bakeable leak valves. The composition of the gas atmosphere was continuously monitored by means of the quadrupole mass spectrometer. RESULTS No significant differences were found for UPS spectra taken from polycrystalline and (1 10) Pd surfaces (either clean or adsorbate covered) ; therefore, only data obtained with Pd(ll0) are reproduced in the following. Spectra were recorded with He I (= 21.2eV) and He I1 (= 40.8 eV) resonance radiation. The features caused by adsorption were essentially not affected by changing the photon energy.However, the use of He 11-radiation offers a greater accessible energy range with simultaneously lower background intensity caused by secondary electrons. €6 =O 2 4 6 8 10 12 i4 binding energy/eV FIG. 3.-Photoemission spectra of a (a) clean and (b) NO-covered Pd(ll0) surface (PNO = I x lo-' Torr) ; (c) difference spectrum. Fig. 2 shows the energy distribution curves (EDC) of a clean and of a CO-covered Pd(ll0) surface together with the " difference spectrum " (DS) which results from a subtraction of both spectra. The EDC from the clean Pd surface is characterized by strong d-band emission between 0 and 5 eV below the Fermi level Ef in agreement with earlier result^.^ Adsorption of CO leads to the appearance of two new peaks centred at 7.9 and 10.8 eV and broadened to 1.5 and 1 eV, respectively.In addition, the emission just€+ =O 2 4 6 8 I0 12 binding energylev FIG. 4 -Photoemission spectra of a (a) clean and (b) 02-covered Pd surface (po2 = 1 x lo-' Torr) ; (c) difference spectrum. a &=O 2 L 6 8 10 12 binding energy/eV FIG. 5.-Photoemission spectra-of a (a) clean and (b) Ha-covered Pd(l10) surface @ H ~ = 1 x Tom); (c) difference spcctnuaH . CONRAD, G . ERTL, J . KUPPERS AND E . E . LATTA 121 below the Fermi level is considerably suppressed as becomes more evident from an inspection of the difference spectrum. The results with the polycrystalline Pd surface as well as the effects caused by variation of the photon energy and of the CO coverage are treated elsewhere. O Adsorption of NO leads to a new structure in the EDC at 9.4 eV and again to a strong decrease of the intensity just below the Fermi level (fig.3). It is evident from the DS that, in addition, a broad maximum is formed around 15.5 eV. The variation of the EDC of clean Pd( 110) caused by oxygen adsorption is shown in fig. 4. As can be seen from the DS the prominent features are again some decrease of the intensity below Ef and the occurrence of a broad maximum at 5.8 eV. This latter structure is hardly visible in the (non-subtracted) EDC. Ef =O 2 L 6 8 10 12 binding energylev FIG. 6.-Photoemission sDectra of a Pd surface with adsorbed CO and NO. Experimental conditions (sequence upwards) : (a) dlean surface ; (b) PNO = 1 x lo-' Torr ; (C)PNO = 1- x lo-' Torr, PCO = 1 x lo-' Torr ; ( d ) p ~ ~ = 1 x lo-' Torr,pCo = 2 x lo-' Torr ; (e)gNo = 1 x lo-' Tom, pco = 5 x lo-' Torr ; (f)pco = 5 x lo-' Torr.Only a very slight change of the EDC was observed after adsorption of hydrogen (fig. 5). It becomes evident from the DS that a small decrease of the intensity just below Ef is the only substantial effect, which, however, is clearly caused by adsorbed hydrogen and not by any experimental artefacts like instability of the lamp etc. Some applications of the UPS technique to follow surface reactions are illustrated by the following examples. Curve (a) in fig. 6 shows the spectrum from a NO- covered Pd( I 10) surface. When this surface was exposed to increasing amounts of122 UPS OF ADSORBED LAYERS ON Pd CO a gradual change of the spectra was observed until finally the two peaks character- istic of adsorbed CO were present (curve (f)). Clearly, adsorbed NO was con- tinuously displaced by the CO molecules.binding energy/eV FIG. 7.-Photoemission spectra demonstrating the effect of heating the sample in NO (a) PNO = 3 x Tom, 300 K ; (b) PNO = 1 x lo-" Torr, 700 K. If a NO-covered surface is heated in vacuo to about 300°C the spectrum charac- teristic for NOad (fig. 7a) changes to that of the clean surface due to desorption. If, however, the surface is held at about 400°C in an atmosphere of Torr NO an EDC characteristic for adsorbed oxygen (fig. 7b) appears. It was confirmed by means of Auger electron spectroscopy that under these conditions an oxygen layer was built up at the surface. It is concluded that NO dissociates at elevated temperatures.The oxygen atoms then remain adsorbed on the surface whereas the nitrogen atoms recombine and desorb. If some hydrogen is added the adsorbed oxygen is readily removed from the surface in the form of H20 and the decomposition of NO can proceed further. DISCUSSION The main purpose of UPS studies with adsorbed layers is to get informations about the electronic structure of the chemisorption bond. These chemisorption levels are expected to be deduced from the atomic or molecular orbitals of the free adsorbate particles whose energies are in many cases fairly well known from photoelectron measurements with gaseous molecules. However, the measured EDC do not represent directly the densities of electronic states at the surface.In particular, relaxation effects cause some shifts on the energy scale whose extent is usually un- known. The " effective " work function of the solid offers a further problem if attempts are made to adjust the energy scale of the free molecule (which refers to theH. CONRAD, G. ERTL, J. KUPPERS A N D E. E. LATTA 123 vacuuiii level) with that of the adsorbate layer (which is determined with respect to the Fermi level). In the following discussion a value of 5 eV (corresponding approximately to the work function of clean Pd) is tentatively assumed as the difference between the origins of both energy scales. The UPS spectra from CO-covered Pd surfaces are quite similar to those reported for the CO/Ni system which is in agreement with the general finding that the adsorptive properties of both metals with respect to CO differ only slightly from each other.After a theoretical treatment of CO adsorption on transition metals l2 the binding should mainly take place by " back-donation " of metallic d-electrons into the antibonding 2n* orbital of free CO thus leading to the formation of a chemi- sorption level just below the Fermi level. If we assume that the probability for photo-excitation of electrons in this level is smaller than that for pure metallic d-electrons this effect would explain the observed decrease below Ef of the intensity of the EDC from CO-covered Pd. Another reason for this effect could be that the emission of photoelectrons is generally attenuated by the presence of an adsorbate layer but it would be difficult to understand why this phenomenon should be restricted to a small energy range below the Fermi level.UPS spectra from free CO molecules exhibit peaks centred at 14.0 and 16.9 eV below the vacuum level, which are due to excitation from 50 and In molecular 0rbita1s.I~ If, for the work function of Pd, an amount of 5 eV is subtracted, these structures should be positioned at 9 and 11.9 eV below Ef which corresponds with observed peaks from adsorbed CO at 7.9 and 10.8 eV, taking into consideration the above-mentioned uncertainties in determining quantitative energy data. Whereas UPS spectra from free CO molecules exhibit a pronounced vibrational fine structure, in the adsorbate state the corresponding levels are broadened to 1 eV or even more due to the coupling with the electronic band states of the metal. The same observation holds for the other adsorbates studied.With free NO, photoemission peaks at 9.3, 16.5, 18.3,13 and -21 eV l4 below the vacuum level are reported, the latter value being determined by XPS. With respect to the Fermi level of Pd the values would be 4.3, 11.5, 13.3 and - 16 eV, which are to be compared with the peaks found for adsorbed NO at 9.3 and 15.5 eV, in addition to the " negative " peak at Ef. The highest level of free NO at 4.3 eV contains only a single electron and is therefore expected to be partially filled by electron transfer from the metal. This effect should cause a shift of this level in the vicinity of the Fermi level due to Coulomb repulsion.15 The observed peak at 9.3 eV is interpreted as being caused by a superposition of the 50 and 171 states of free NO, located at 11.5 and 13.3 eV.The energy shifts and the broadening are remarkable, but the observed peak is clearly characteristic for the N--0 bond in the adsorbed state. The 40 level at 16 eV is obviously only slightly affected by the formation of the chemisorption bond. Oxygen adsorption causes, besides a decrease of the emission below E,, the formation of a peak at 5.8 eV. This result is again similar to the results with nickel, where adsorbed oxygen caused a peak at 5.5 eV.' Recently, a theoretical interpreta- tion of the latter data was given by Messmer et aZ.16 The peak is ascribed to emission from non-bonding 02p orbitals. indicates some electron transfer from the metal to the adsorbed H atoms.The only effect observed by UPS is a small negative peak at Ef for which no straight-forward explanation can be presented. No peak was observed in the region of -5.5 eV below Ef as reported in an earlier paper by Eastman et aZ.* on (bulk) palladium hydride. Hz adsorbs dissociatively on Pd and the observed increase of the work function124 UPS OF ADSORBED LAYERS ON Pd Although a unique identification of the origin of the chemisorption levels and a quantitative evaluation of their energies from UPS data offer a series of theoretical problems (which probably may not be solved in the near future), a second important aspect of this technique should be emphasized. UPS yields unique informations on the molecular state of surface species. By using the spectra just as a " fingerprint " from the different adsorbed particles, a detailed insight into the mechanism of surface processes may be obtained as demonstrated with the reported reactions of displace- ment and decomposition of NO.We are grateful to the Deutsche Forschungsgemeinschaft (DFG) for financial support of this work. D. E. Eastman and J. K. Cashion, Phys. Rev. Letters, 1971 27, 1520. D, E. Eastman, Solid State Comm., 1972,10, 333 ; C. R. Brindle and M. W. Roberts, Surjuce Sci., 1973, 38, 234. C. R. Helms and W. E. Spicer, Phys. Rev. Letters, 1972,28,565 ; K. A. Kress and G. J. Lapeyre, Phys. Rev. Letters, 1972, 28, 1639. R. L. Park and H. H. Madden, Surface Sci., 1968,11, 158 ; J. C. Tracy and P. W. Palmberg, J . Chem. Phys., 1969,51,4852 ; G. Ertl and J. Koch, 2. Naturforsch., 1970,25a, 1906 ; H. Conrad, G. Ertl, J. Koch and E. E. Latta, Surface Sci., 1974, 43,462. G. Ertl and P. Rau, Surface Sci., 1969,15,443 ; G. ErtI and J. Koch, in : Adsorption-Desorption Phenomena, ed. F. Ricca (Academic Press, 1972), p. 345. H. Conrad, G. Ertl and E. E. Latta, Surface Sci., 1974, 41, 435. H. Conrad, G. Ertl, J. Kiippers and E. E. Latta, in preparation. D. E. Eastman, in Electron Spectroscopy, ed. D. A. Shirley (North Holland, Amsterdam 1972), p. 487. J. Kiippers, H. Conrad, G. Ertl and E. E. Latta, Japan. J. Appl. Phys., in press. D. R. Penn, Phys. Rev. Letters, 1972, 28, 1041. l 2 G. Doyen and G. Ertl, Surface Sci., 1974,43, 197. l 3 D. W. Turner et al., Molecular Photoelectron Spectroscopy (John Wiley, New York, 1970). l4 K. Siegbahn et a)., ESCA Applied to Free Molecules (North Holland, Amsterdam, 1971). * D. E. Eastman, J. K. Cashion and A. C. Switendick, Phys. Rev. Lett., 1971,27, 35. T. B. Grimley, in Molecular Processes on Solid Surfaces, ed. E. Drauglis, R. D. Gretz and R. I. Jaffe, (McGraw-Hill, New York, 1969). I6 R. P. Messmer, C. W. Tucker and K. H. Johnson, Surface Sci., 1974,42, 341.

ISSN:0301-7249    

DOI:10.1039/DC9745800116

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

GENERAL DISCUSSION Papers 7, 8, 9 Dr. C. R. Bmdle (Bradford) said: Evans states that for up to a monolayer coverage of oxygen, copper is converted to the Cu' state. The evidence seems to be based entirely on the fact that the Cu 2p+ peak shows neither satellites nor broadening (both characteristic of Cu"). Since the peak does not alter at all in shape or position, would not an equally valid interpretation be that the copper has at this point still largely retained its metallic character rather than being Cu' (though the oxygen may be close to GG 0 2 - '3 as discussed in my earlier comment)? The lack of a signz5cant change in the metal peak on adsorption of less than a monolayer is well known for many metals (Ni, Fe, Pt, V, Mo, W, Zn) for which a similar explanation can hold.It seems particularly dangerous to argue simply by analogy (as stated in the paper) to Cu/02 that Ag passes through Ag' to Ag"' on oxygen adsorption when in fact no changes are observed in the metal core lines during the adsorption. Here again metallic character can be maintained during the early stages of oxygen take-up. Dr. S . Evans (Aberystwyth) said: I would make two points in reply to Brundle's question regarding the Cu/O, system. First, we do not consider that loss of metallic character is necessary for the production of species describable as M', MI', etc. : many bulk oxide systems where the oxidation state nomenclature would be applied without question possess metallic characteristics. In the present case the description Cu' is intended to indicate that we believe the metal-oxygen bonding to arise largely from the mixing of the 0 2p adsorbate orbitals with Cu 4s orbitals, the Cu 3d orbitals remaining essentially unchanged.* Whether or not the resulting (surface) energy level scheme possesses a part-filled band, and hence a relatively high DOS at the Fermi level (this seems the simplest verifiable test of" metallic character ") is incidental. In fact, however, comparison of the UPS of the copper surface before and after adsorption of oxygen [fig. l(a)] shows that the occupied DOS at the Fermi level, high for clean copper metal, decreases as the adsorption proceeds, and is negligibly I icu Y' ' -- 3 2 I F.L. U A 4 3 2 I F.L. binding energy /eV FIG. 1 .-UPS Valence band profiles near the Fermi level for copper and silver and their surface oxides : (a) copper ; (i) clean, (ii) effective monolayer : (b) silver ; ri) clean, (ii) final surface oxide.* We emphasize here that we are not suggesting that a " monolayer of cuprous oxide " is formed : bulk cuprous oxide may well have substantially different structural and electronic properties from this surface CuI oxide, although subsequent work has indicated that there are some similarities. ' S. Evans, J.C.S. Faraday 11, 1975,71,1044. 125126 GENERAL DISCUSSION small by the time the first stage of reaction (effective monolayer coverage) is completed. It would thus seem that the surface copper atoms do in fact lose '' metallic character " on adsorption. This loss may be associated with some rearrangement of the surface atoms, and in describing the process as chemisorption we do not imply that the surface metal structure necessarily retains its original geometry.The only direct evidence electron spectroscopy can offer on this matter would seem to be the work function change, which can hardly be conclusive taken in isolation. On the second point, we did not argue " simply by analogy " that " Ag passes through Ag' to Ag"' ". Only the identification of the first surface oxide as Ag' rests on such an analogy, and, as indicated above, the description " Agl " is in any case not incompatible with retention of metallic character. Since here monolayer coverage is never attained, we are unfortunately unable to examine the point experimentally. Moreover, the principal similarities to the Cul surface oxide on which we based our assignment are (i) the presence of an equal number (3) of similarly positioned new bands in the UP§ after adsorption, (ii) the fact that again the metal nd band remains unchanged in profile by the adsorption, (iii) an 0 (Is) binding energy close to the value found in the Cu-oxygen system and (iv) a similar work function change on adsorption.These four points taken together seem to us reasonably conclusive. That the Ag core level spectra remain unchanged merely tends to confirm that Ag" (paramagnetic) species are not involved, although it remains an open question whether or not XP§ is sufficiently sensitive to identify a shift or broadening involving 10% or less of the total core-level signal. Incidentally, although (as Brundle indicates) there is often very little change in the substrate core level spectra on adsorption at up to monolayer coverages, more substantial effects are sometimes observed (e.g., Pb/02,2 and also Au/O, as reported in table 1 of our paper in this Discussion) : XPS thus does possess the requisite surface sensitivity to pick up such changes where they exist at monolayer coverages (say, 20% of total signal).The argument for Ag"' in the final surface oxide relies on the lack of change in the 3d XP spectrum only to the extent that we use this fact to eliminate (paramagnetic) Ag" species. However, since the XPS valence band 3* now also shows changes similar to those in the UP§ we have presented here, this surface oxide, unlike the Ag' surface oxide, must be rather more than one layer thick.A much higher proportioil of the Ag 3d XPE signal must therefore be associated with this oxide layer, and we consider that any substantial broadening would in this case have been obvious. If we then also rule out the possibility that we have merely another, thicker, Ag' surface oxide, on the basis of the new, and dramatically shifted 0 (1s) peak, combined with the further increase in work function, and the substantial changes in the valence band profile, the only remaining possibility is that the oxide layer now involves Ag'". Analogy with the Cu/0, system is not required here. The UPS near the Fermi level for clean silver and this final oxide are shown together in fig. l(6) : it is clear that his oxide, like the surface Cu oxides, does not possess a significant occupied DOS at the Fermi level.Consequently we take the view that " metallic character " (in the sense defined earlier) at the surface is in this case also lost on oxidation. Prof. M. W. Roberts (Bradford) said: Some caution regarding the data preseiitcd by Evans for the Pb+0, system and the assignments suggested for the XPS peaks, as S. Evans and J. M. Thomas, J.C.S. Faradczy 11, 1975, 71, 320. S. Evans, Chem. Phys. Letters, 1973, 23, 134. E. L. Evans, J. M. Thomas. M. Barber and R. .I. M. Griffiths. Siuface Sci., 1973, 38, 245 S Evans, unpublished work.GENERAL DISCUSSION 127 offering support for the interpretation of his Cu+02 data seems necessary. First, work function and krypton adsorption data indicate that chemisorbed oxygen is not stable on a Pb film surface even at 77 K and penetration into the sub-surface occurs with subsequent substantial oxide growth at room temperature.I am therefore doubtful whether the break in the O(ls) oxygen exposure plot can be taken to represent the completion of the chemisorbed layer, i.e., a situation where no sub-surface oxygen exists. Furthermore, his data suggest a sticking probability for oxygen during what is referred to as chemisorption of only about This seems very low and is more likely to include “ oxidation ”, i.e. lattice penetration. Finally, regarding his sug- gestion that the presence of satellites may be used to provide evidence for the oxidation state, Schon has reported satellites for both Cu20 and CuO. Is this attributable to poor definition of the surface stoichiometry of Schon’s copper oxides? Dr.S. Evans (Aberystwyth) said: As our work concerning the Pb/02 system (some relevant aspects of which I briefly mentioned in introducing our paper) is cur- rently in press We were in fact aware of, and indeed were able to reproduce using our secondary emission technique, the work function decrease on exposure to oxygen reported in the earlier work on lead referred to by Roberts4 We found that the decrease was progressive up to the “ break ” in the 0 (1s) intensity against log exposure plot.2 We interpret these “ breaks ” in the first place as meaning that the initial fast uptake of oxygen impinging on a bare metal substrate is complete, and that further oxygen uptake must be dominated by a different mechanism.The identification of these points as roughly “ equivalent to a monolayer ” then comes first from a comparison of uptake plots for Pb, Cu and Ni : “ break ’’ points come at roughly similar (within a factor of 2, Pb being the highest) absolute 0 (1s) XPS intensities in each case, and for Ni, the indicated oxygen exposure required to reach this point fits in well with coverage data reported by Horgan and King.’ Secondly, estimates based on the relative XPS signal intensities from clean substrate, “ oxidised ” substrate, and adsorbate, combined with experimental cross-section data and mean free paths, and assuming an exponential variation of electron escape pro- bability with depth, suggest that the effective thickness of the oxygen overlayer is of the order of one monolayer ( N 2 - 6 A) in each case.The value for lead is indeed towards the higher end of this range, but since some of the parameters involved (especially the mean free paths and cross-sections) are not accurately known, we would not attach much significance to this variation. Thus when an adsorbate is still in or close to the surface layer, this method cannot at present reliably distinguish between chemisorption without reconstitution and chemisorption accompanied by rearrangement of the surface. Consequently, the terms we have used to identify the “ break ” point exposure (e.g. “ effective coverage ”) should not be taken (and indeed were not intended) to imply that surface reconstitution does not occur: in fact it almost certainly does occur even at submonolayer exposures, for Pb/02,4 as pointed out by Roberts, and it may well take place to some extent for Cu/02 ’ as well.All the XPS data indicate is that the amounts of oxygen taken up in the first rapid stage are comparable, and roughly equiuuknt to a monolayer. The reported satellite structure for Cu20 has been investigated by a number of a detailed repetition here is not desirable. J. M. Saleh, P. B. Wells and M. W. Roberts, Trans. Furuduy Soc., 1964, 60, 1865. S. Evans and J. M. Thomas, J.C.S. Faruday ZI, 1975, 71, 320. S. Evans, Chem. Phys. Letters, 1973, 23, 134. J. M. Saleh, B. R. Wells and M. W. Roberts, Trans. Furada-v Soc., 1964, 60, 1865. A. M . Horgan arid D. A. King, Swfuce Sci., 1970, 23, 259. S. Evans, J.C.S. Faraday 11, 1975, 71, 1044.128 GENERAL DISCUSSION other worker^.^-^ It is now clear that the surfaces of Cu20 examined by Schon ' were in fact contaminated by CuI1, although the exact nature of the surface species involved remains uncertain.Prof. M. W. Roberts (Bradford) (communicated) : With relation to the comparison of the Aberystwyth work and our earlier work function/adsorption data, the oxygen pressure was probably a factor of about lo3 higher in our case ( N Torr). Since oxidation is likely to be pressure dependent, the growth of oxide crystallites reflected by our increase in surface area by a factor of six may not be as important in their work. The log t plot may therefore have more meaning with relation to the formation of a chemisorbed layer than appears at first sight.Dr. D. Briggs (1.C.I. Corporate Lab., Runcorn) said: Concerning the paper by Evans et al., it must be noted that there is no direct evidence to support the inference that " monolayer " oxidation results in the formation of an oxide layer, as opposed to a strongly chemisorbed layer of oxygen in the dissociated state. Only in the case of further oxidation of Cu is a species directly identified (CuX1 by virtue of chemical shift and accompanying satellite peaks), while the growth of a shifted very weak peak is claimed in the case of Au. Two important sources of information available by means of this kind of experiment which can be used to resolve this problem have been neglected in this work ; some recent results which we have obtained from the oxidation of zinc serve to illustrate my points. The experimental procedure resembles that described by Evans et al.except that an evaporated Zn film was prepared under ultra-high vacuum (base pressure - lo-'' Torr) and gas adsorption was carried out at Reaction was followed up to the point where growth of the 0 (1s) peak had " plateaued " (after ca. 200 L ex- posure). He I and Ne I UPS spectra at this point were virtually identical to those from cleaved ZnO crystals. It has already been clearly demonstrated that for transition and other metals 7 * * shifts in Auger peaks between metal and oxide can be much larger than core-level shifts. In the LMM series of Zn, the separation between metal and oxide peaks is 4 eV. During the initial stages of oxidation observed in our experiments the characteristic LMM ZnO peaks do appear superimposed on attenuated LMM Zn peaks.The LMM peaks from Cu, Cu,O and CuO are diagnostic and could be used to resolve the question of whether CuzO is formed by monolayer oxidation of Cu as suggested by Evans et al. My second point is to emphasise the potential of angular variation experiments in XPS In this procedure the angle of electron emission from the sample (0) is varied by rotating the sample surface with respect to the entrance slit : Torr. My first point concerns the use of Auger peaks in the XPS spectrum. for investigating the structure of the first monolayer. D. C . Frost, A. Ishitani and C. A. McDowell, Mol. Phys., 1972, 24, 861. T. Roberts, M. Bartel and 6. Offergeld, Surface Sci., 1972, 33, 123. I<.S. Kim, J. Electron Spectr., 1974, 3, 217. T. Novakov and R. Prins, in Electron Spectroscopy, ed. D. A. Shirley (North Holland, Amster- dam, 1972), p. 821. G. Schon, Surface Sci., 1973, 35, 96. R. A. Powell, W. E. Spicer and J . C. McMenamin, Phys. Rev. B. 1972,6, 3056. C. D. Wagner and P. Biloen, Surface Sci., 1973,35, 82. W. A. Fraser, J. V. Florio, W. N. Delgass and W. D. Robertson, Surface Sci., 1973, 36, 661. ' J. E. Castle and D. Epler, Proc. Roy. SUC. A, 1974, 339, 49.GENERAL DISCUSSION 129 t e - At low 8 (10-1 5") the surface sensitivity (surface/volume signal ratio) can be increased by x 10.l Applying this technique to the surface produced by exposure of Zn to N 200 L of O2 gives very interesting results. First, in going from 8 = 75" to 8 = 15" the Zn 2p+ peak shifts to higher binding energy by 0.3 eV and broadens slightly, hence the low 8 spectrum corresponds mainly to the surface monolayer and reveals the presence of ZnO by a core-level shift. Secondly, two 0 (1s) peaks (at 530 and 532 eV binding energy) which are equal in intensity at 8 = 75" behave differently to angular variation and only the 530 eV peak is enhanced at 8 = 15" (by ca.x 3). A tentative suggestion is that this peak corresponds to oxygen adsorbed on ZnO. In the work of Evans et al. this technique would have been particularly useful for enhancing the shifted weak Au peak which they suggested appeared with the production of an oxidised surface. Dr. J. C. Fuggle and Prof. D. Menzel (Munich) said: Some remarks have been made about the absence of large shifts of core level peaks in XP spectra of adsorbates and substrates as a function of coverage.For some time now we have been studying these shifts during chemisorption on W and Ru to define their size more precisely and see if any information about the physics and chemistry of the adsorption processes can be extracted from them. We have found small but definite shifts and mention these now in case there is an impression that no shifts at all should be observable. Although shifts in adsorbate peaks measured with respect to EF are much smaller that the work function changes, shifts of the 0 (1s) level up to approximately 0.5 eV have been observed, as fractional oxygen coverage on W changes from 0 to 1 .2 Results for shifts in substrate levels are complicated by the summation of contributions from surface and bulk atoms.Work at the National Bureau of Standards and in our laboratory has shown the shifts of whole metal peaks during chemisorption LO be typically about 0.1 eV ; in the special case of oxygen chemisorption on W( 1 10) there is evidence for another set of N6, peaks of the clean metal. We have investigated the shifts in substrate peaks from Ru(OO1) as a function of electron take-off angle and chemisorbed oxygen coverage and shown that at grazing take-off angle the shift of the Ru N5 peak is nearly 3 times as large as at 25". This is excellent evidence for the localization of the effect at the surface and stresses the independence of work function changes. peaks -0.6 eV from the N6, Dr. C. R. Brmdle (Bradford) (communicated): I am not certain that anyone has ever claimed that the binding energy shift for the top layer of metal atoms during chemisorption is non-existent.The point is that they are very small compared t o the change on, say, bulk oxidation, and are difficult to observe unless shallow ejection W. A. Fraser, J. V. Florio, W. N. Delgass and W. D. Robertson, Surface Sci., 1973, 36, 661- A. M. Bradshaw, D. Menzel and M. Steinkilberg, this Discussion p. 46. J. T. Yates and N. E. Erickson, Surface Sci., 1974, 44,489. J. C. Fuggle and D. Menzel, Chern. Phys. Letters, in press. 5s-E130 GENERAL DISCUSSION angles are used to enhance the surface layer contribution to the XPS intensity.l The data that Fuggle has just presented illustrate this. I have previously emphasised that the shifts were small to correct the impression of some erroneous earlier work which gave the impression that large substrate shifts were observable on chemi- sorption.Dr. Frank R. Smith* (Cambridge) (communicated) : Brundle suggests that all oxide 02- ions have binding energies for 0 (1s) of 530 eV. Does he argue that the binding energy data of table 1 of Evans et al. indicate identical binding energies within experi- metal error for chemisorbed 0 on Cu, Ag and Au? Should not, however, the comparison be made of binding energy relative to vacuum, in which event Evans et aZ.’s 0-on-Pb result of 529.1 eV becomes even more discrepant, the work function of lead being 3.8 eV? Furthermore, what is one to say of Kim, Sell and Winograd’s binding energies for adsorbed O( 1s) relative to the Fermi level of 530.4 eV (Pb), 532.2 eV (Pt) and 533.5 eV (Ni)? have made XPS measurements with the Au + 0 system, generated electrochemically.Anodisation for 5 min at + 1.9 V against saturated calomel electrode in 0.5 mol dm-3 aqueous H2S04 generated a surface “ oxide ” with intense Au 4f+, 3 peaks shifted to 89 and 85.9 eV (also observed with pure Au203) relative to Au 4fz of 84.0 eV. The same period of anodisation at + 1.8 eV (at which potential the cyclic voltammogram peaked) produced Au and “ oxide ” peaks of equal intensity. Intensity ratio evidence and the effect of H20 on Au,03 suggested that Au(OH), was the anodic product. The same workers anodised Ir in 0.1 mol dm-3 KOH to IrO, and film thickness estimates by comparison of 4f3 electron intensities from Ir, IrO, and the anodised metal agreed excellently with coulometric estimates in the range 30-60 A.It may be of interest that Kim et al. Dr. S. Evans and Professor J. M. Thomas (Aberystwyth) (communicated) : The binding energy data of our table 1 could in fact be taken to indicate similar chemisorbed 0 (1s) binding energy, especially after correction by the addition of the work function after adsorption (see our earlier comment). Moreover, a higher value (ca. 932.5 eV) for the Cu metal reference in the Cu/O system may well be preferable, 5 * and the corrected binding energies then do agree within experimental error. In these systems, however, we consider (as indicated in our paper) that the bonding to the metal is very similar in character in the three cases, so that this is not unexpected.Our Pb/O result, ’ for a different situation in that there the oxygen penetrates well into the surface, is, in contrast, significantly different. The action of O2 on Pb reduces the work function by ca. 0.5 eV (see earlier comments by M. W. Roberts and ourselves) and so our correction makes the discrepancy even greater ! We consider, as indicated in our earlier comment that such differences arise largely from a difference in the Madelung (lattice) potentials ignored in the demonstrably inadequate arguments advanced by Brundle in this Discussion. We would not, however, place too * permanent address : Memorial University of Newfoundland, Canada. T. A. Clarke, R. Mason and M. Tescari, Proc.Roy. SOC. A, 1973,331, 321. C. R. Brundle, J. Electron Spectr. 1974, 5, 291. J. C. Riviere, in Solid State Science, ed. M. Green (Dekker, New York, 1969). K. S. Kim, C. D. Sell and N. Winograd, paper presented at The Electrochemical Society, May 1974 Meeting, San Francisco. G. Johansson, J. Hedman, A. Bendtsson, M. Klasson and R. Nilsson, J. Electron Specrr. 1973, 2,295. S . Evans, J.C.S. Faraday 11, 1975, 71, 1044. ’ S. Evans and J. M. Thomas, J.C.S. Faruduy IZ, 1975, 71, 301.GENERAL DISCUSSION 131 much reliance on comparison with Kim et al.’s figures, reported in several paper~.l-~ Some of the assignments are not indisputable, and discrepancies between their data and those of other groups suggest 4* 5 * 6* that their calibration techniques may also be unreliable.Dr. C. R. Brundle (Bradford) (communicated) : The data in table 1 of the paper by Evans et al. do show a constancy of 0 (1s) binding energy, referenced to Ef within k0.5 eV. In my previous comment I stated that there are many other chemisorption and/or bulk oxide systems where an 0 (Is) binding energy of 530 f 0.5 eV can be found Ni, Fe, Na, Ag, V, Mo, W, Zn, Ru, Ir, Au, Mn, Cr and Co are examples. I was careful not to say that all oxygen chemisorption and metal oxide systems show such a value, and to point out that during oxygen chemisorption, one often finds additional higher binding energies O(1s) XPS features (Ni is the best documented example,8* with at least two extra O(1s) values at 531.4 eV and 533.2 eV (in addition to the major 529.5 eV feature).My suggestion is that the higher binding energy features represent adsorption at the surface and the “ constant ” O(ls) represents oxygen atoms in an oxide-like environment (formally 02- or O-II, but not implying a total charge transfer of 2). The stronger argument in favour of this interpretation is the similarity of the chemisorption O(ls) binding energy, that I consider to be oxide-like, to the O(1s) value for the bulk oxide, rather than any “ constancy ” in this value across a range of substrate metals. I entirely agree with the comments expressed by Thomas and Evans that if one wishes to say something definitive about the actual charge residing on 02- in a series of metal oxides one must consider the question of a suitable reference level (about which there is no agreement at present) and the role of Madelung po- tentials.Two other reservations about the “ constancy ” of O(ls) at 530k0.5 eV can be expressed. First, different groups use different calibration methods which may not be strictly comparable or appropriate (the constancy of 530.0 & 0.2 eV quoted by Joyner and Roberts lo for five chemisorption systems does not in fact hold to this accuracy according to other authors, including myself), and therefore we do not experimentally know exactly how constant is “ constant ”. The second point is that even if there were no reference level of Madelung potential corrections it is not yet clear what spread in binding energy would be theoretically expected for the range of s-values feasible for different oxides-i.e.f0.5 eV might turn out to represent a significant variation in s. We can now examine the data introduced by F. R. Smith. The Winograd et al. value of 533.5 eV for oxygen adsorption on Ni is confirmed by my own measure- ments’ (mentioned above) and this is one of the higher binding energy features which I believe represents a true surface adsorbed species. In both Winograd’s work and my own, this species is of minor intensity, the major species being at 529.5 eV, identical to that of bulk NiO, even in the initial stages of adsorption. The value of 532.2 eV for O2 on Pt l1 may be of similar nature, a true surface adsorbed species, K. S. Kim and N. Winograd, Chern. Phys. Letters, 1973, 19,209. K. S. Kim, N. Winograd and R. E. Davis, J. Amer. Chem. SOC., 1971,93,6296.K. S. Kim and R. E. Davis, J. Electron Spectr., 1973, 1, 251. S. Evans, J.C.S. Faraday ZZ, 1975, 71,1044. S. Evans and J. M. Thomas, J.C.S. Faraday ZZ, 1975, 71, 301. J. M. Thomas and M. J. Tricker, J.C.S. Faraduy ZI, 1975,71, 317. K. S . Kim and N. Winograd, Surface Sci., 1974, 43,625. C. R. Brundle and A. F. Carley, Chern. Phys. Letters, to be published. lo R. W. Joyner and M. W. Roberts, Chem. Phys. Letters, to be published. K. S. Kim, C. D. Sell and N. Winograd, paper presented at the Electrochemical Society, May 1974 Meeting, San Francisco. ’ S. Evans, unpublished work.132 GENERAL DISCUSSION but there are several results, both published and unpublished which give values ranging from 530.2 to 532.2 eV for chemisorbed oxygen and oxides of Pt. The value of 529.1 eV for oxygen chemisorbed on Pb (confirmed by several unpublished sets of results, including my own) is intriguing because though it is outside the 530f0.5 eV range it is close to the values for bulk PbO and Pb02, i.e.it still represents chemisorbed oxygen in an electronically oxide-like environment, as Thomas and Evans admit in their reply to Smith’s questions when they say the oxygen atoms have penetrated into the surface. My argument is that the presence of oxygen in an oxide-like environment on initial adsorption is a more general phenomenon than realised. The sign of work-function changes can be misleading here. The work function decreases for Pb/Oz, which is accepted as indicating incorporation. It increases slightly (0.3 eV) in the initial Ni/02 interaction, but as I am suggesting both incorporation and surface adsorption in this case (two or more O(1s) binding energies, one at 530f0.5 eV), the change could be due to the sum of two changes of opposing sign.Dr. D. E. Parry (Aberystwyth) (communicated) : Referring to Brundle’s second point, the change in core level binding energy 6B due to a change in atomic partial charge 6q is 6q e2/r to a first approximation, with 6q in units of lei and r the average radius of the orbitals which are gaining or losing electrons. This has been discussed quantum-mechanically by, e.g., Davis and S h i r l e ~ . ~ We have found that 6B/6q = 30 V for o~ygen.~ A spread of f0.5 eV in B thus corresponds, if only this effect is considered, to a variation in charge of less than +0.021el.In the solid state the neglect of Madelung effects is not justified. The effect on B of a significant change in partial charge through the term 6qe2/r above is strongly offset by the concomitant change in Madelung effects, so that the net effect can be a small, even zero, chemical shift. Dr. A. M. Bradshaw (Munich) said: It should be noted that the features B and C in the oxygen/silver spectra of Evans et al. could perhaps be attributed to co-adsorbed CO. The peaks seem to be positioned at -9.0 to -9.5 eV and at - 13 eV, which although somewhat high in binding energy, could correspond to associatively adsorbed CO. We have so far found no evidence for these features in our spectra from the Ag(ll0) surface. In an earlier paper we pointed out that it is necessary to keep the CO content of the oxygen down well below 1 % to obtain reproducible adsorption behaviour, characterising what is held to be a pure oxygen adlayer. Higher CO concentrations appear to result in a co-operative adsorption of CO and oxygen, although CO does not adsorb on silver by itself at room temperature. In this particular case, however, we may have to look for another explanation for the discrepancy, because the Aberystwyth system is, I believe, diffusion pumped and should produce very little CO during the dosing of oxygen. The problem of assignment in the CO spectrum notwithstanding, it is perhaps appropriate to take up a point made by Williams in the presentation of his paper. He notes that the characteristic CO features appear at - 9.2 and - 11.6 eV on poly- G.C. Allen, P. M. Tucker, A. Capon and R. Parsons, Electroanal. Chem. Interfacial Electrochem., 1974,50, 335. M. W. Roberts and B. R. Wells, Trans. Faraday SOC., 1966, 62,1608, and papers cited therein. P. H. Citrin, R. W. Shaw, Jr., A. Parker and T. D. Thomas, in Electron Spectroscopy, ed. D. A. Shirley (North Holland, Amsterdam, 1972, p. 691). D. W. Davis and D. A. Shirley, J. Electron Spectr 1974, 3, 137. D. E. Parry, J.C.S. Faraday 11, 1975,71, 344. H. A. Engelhardt, A. M. Bradshaw and D. Menzel, Surface Sci., 1973,40,410.GENERAL DISCUSSION 133 crystalline platinum and at - 8.0 and - 11 .O eV on polycrystalline nickel, representing not only a net shift to higher binding energy but also a dramatic change in separation. The comparison that we then might like to make between the two systems necessarily takes us into a discussion of reference levels and the work function correction.TO provide futher material for such a discussion it is useful to look at the table, which summarises some data for a-CO on the (100) and (1 10) surfaces of tungsten. TABLE 1.-UPS PEAKS FOR a-CO ON TUNGSTEN AT 300 K Torr CO in equilibrium ; energies (f0.2 eV) relative to Fermi level) (He I1 ; peaks energy/eV surface 1 2 W(100)2 - 8.8 - 12.0 W(110)l - 7.9 - 11.0 We note here that the binding energies on W(100) are shifted higher by approxi- mately 1 eV relative to those from the W(110) system. The separation, however, remains constant. We would expect the type and strength of bonding to be the same in the two cases and there are probably comparable relaxation contributions.The constant separation of the peaks lends some support to these assumptions. The observed difference in the binding energies relative to the Fermi level could therefore be put down to the difference in energy between the Fermi level and the reference level for each system. As Menzel has pointed out already in this Discussion, for an atom or molecule dissolved in the bulk, this would be the average inner potential. For a molecule adsorbed on the surface however, this reference level no longer has any meaning : the molecule is sitting in the surface dipole layer. In addition we have to consider the effect of the adlayer itself and for the CO/tungsten system the full coverage situation is further complicated by the presence of both a- and p-CO. Nonetheless we have a clear difference of about 1 eV, which could be connected with the difference in work function of the two surfaces.Dr. B. Biloen (Shell, Amsterdam) said : Bradshaw's data, which disclose a difference in the position of CO-derived bands when comparing W(lO0) with W(l lo), provide interesting material bearing on the question whether or not " the " surface electrostatic potential has an independent influence on band positions. A formal description might clarify the terms " independent " and " surface po- tential ''- In the following description we explicitly account for the fact that any photoelectron is subject to kinetic energy changes when there is a difference in electro- static potential between " the " site of photoionization and the site of detection : AEkin = e($s - $*).Then, band positions, i.e., Ekin, characteristic of substrate bulk features, follow from : in which Ekin is the photoelectron kinetic energy as measured by the velocity analyser, e4s and e$A are the substrate and analyser work functions, respectively and IF is the Fermi-level referenced, ionization energy, which for bulk structure is a constant characteristic of the material. From eqn (1) it follows that positions, viz. Ekin, of bands derived from the substrate A. M. Bradshaw and D. Menzel, Ber. Bunsenges. phys. Chem., 1974, 78, 1140. A. M. Bradshaw, D. Menzel and M. F. Steinkilberg, Chem. Phys. Letters, 1974, 28, 516.134 GENERAL DISCUSSION bulk are independent of the substrate work function, e& i.e., from the substrate surface electrostatic potential ( - 4& Another limiting case is provided by an " adsorbate " molecule located at a distance from the surface as relevant for a definition of work functions, i.e., a distance small compared to patch sizes but large compared to the distances involved in local chemical binding, say, - 20 A.We then consider the adsorbate molecule as unperturbed, but experiencing the surface as a local electrostatic potential - $s. Photoelectrons from " adsorbate " molecules will then have a kinetic energy given by : in which I' is the, vacuum level referenced, ionization energy, which is characteristic of the free molecule. Eqn (2) shows that in this limiting case the surface electrostatic potential, - # s , makes an independent contribution to the adsorbate band position (Ekin).For molecules which really are part of the adlayer, the situation is complicated, and the cautious remarks of Menzel and Bradshaw are well taken. Still, I would be interested to learn to what extent the observed changes in adsorbate-derived band positions correlate with work function changes (preferably with those which have been measured in situ). Therefore I would like to ask Bradshaw how this works out for his data on CO on W(100) and W(110). Ekin = hv - I -k e(4s - 4 A ) (2) Dr. A. M. Bradshaw (Munich) said: As Biloen would readily admit, it is dangerous to make too close a correlation. However, we do note that the work function of the clean tungsten (100) surface is generally accepted to be about 4.5 eV,' and that there are several values given in the literature for the (110) surface, one of which is as high as 5.4 eV.l So we have here a possible 0.9 eV difference in the right direction before even going on to consider the CO-covered surface.Unfortunately, however, there is, as far as I know, little or no work function data for the latter. Dr. S. Evans (Aberystwyth) said: Bands B and C in our Ag/oxygen spectra cannot be attributed directly to associatively adsorbed CO. Both C(1s) and O(1s)XP spectra were monitored as the adsorption proceeded, and the C(ls) signal was still a negligible intensity after saturation with oxygen. The CO content of the oxidant was not, however, monitored, and it is possible, though unlikely, that the course of the oxidation may have been affected by CO, even though the final product contained only silver and oxygen.However, the difference spectra reported by Bradshaw et al. do show maxima at N -9 eV, similar to our peak B, although they are not always above the zero level. Also, peak C is extremely weak, and at the oxygen exposures used by Bradshaw et al. was virtually undetectable. (The spectra reproduced in fig. 3b of our paper were obtained after exposures of the order of lo5 L.) Other factors which may have contributed to the greater prominence of adsorbate features in our spectra are that our results were obtained with polycrystalline silver, cleaned by a different procedure, and were recorded with a different source-sample-analyser geometry. Taken together, these differences may well account for the observed discrepancies.Dr. E. W. Plummer (Pennsylvania) said: When CO is adsorbed on a metal surface as a molecule, the photoelectron spectra show two peaks below the metal band. J. C. Rivike, in Solid State Surface Science, ed. M. Green (Dekker, New York, 1969), vol. 1, p. 197.GENERAL DISCUSSION 135 These two peaks, which usually have binding energies of - 13.5 and. N 17.5 eV, have been identified as the 50 and 1n levels of the gas phase mo1ecule.l I believe this labelling of the levels is incorrect for the following reasons. (1) The SCF-Xu scattered-wave calculation for CO by Rosch et aL2 shows. that the lowest binding energy level [50] is primarily the lone-pair orbit on the C atoms The next level (In) is the n bond between the carbon and oxygen atoms, and the next level [40] is primarily the lone-pair orbital on the oxygen atom.SCF-Xa calculations for the carbonyl compounds show that bonding the carbon to a metal atom inverts the levels. The ln level is shifted slightly while the new molecular orbital formed from the 50 ends up below the In, probably mixed with the 40 level. (2) A tabulation of the measured relaxation shifts of orbitals of chemisorbed molecules on Ni, W, Mo and Fe shows that a valence orbital should be shifted up- ward in energy by 2 to 3 eV from that in the gas phase.5 of the classical polarization energy of an electron gas due to the interaction of the induced screening charge with the hole ' shows that this is a reasonable energy shift for an orbital the size of the CO In orbital.Therefore, if the lowest binding energy level of adsorbed CO is the 50 level there is little or no relaxation shift for either the 50 or In. On the other hand, if this level is derived from the In gas phase level there is a reasonable relaxation shift of 2 4 eV.5 We have looked at both electropositive and electronegative adsorbed CO on W, so the discrepancy is not a true chemical shift problem. (3) The photon energy dependence of the relative strengths of these two adsorbed levels is not what one would expect from the gas phase measurements, if the lowest binding energy level is the 50. Therefore it now seems likely that the lowest binding energy level of adsorbate CO is derived from the l n level, while the next level is a mixture of 40 and the 50 orbit interacting with the substrate.Recent calculations for a Ni-CO cluster by Batra and Robaux * using SCF-Xo! indicated that the 50 level is mixed in with the 40 level, while CNDO calculations by Blyholder indicate that the 50 level is mixed in with the 1n level. A simple calculation Dr. C. R. Brundle (Bradford) said: The He I and He I1 two band spectrum of undissociated CO adsorbed on metal surfaces has previously been assigned as 50 (lowest binding energy) and In, following the assignment for gas phase CO. The separation between the bands is also always similar to that of 50 and 1n in the gas phase. The absence of an adsorbed CO band corresponding to the gaseous 40 level (2.9 eV deeper than In) was not confirmed until He I1 spectra became routine (it could have been present just beyond the cut-off point of a He I spectrum).Plummer has suggested that the adsorbed CO assignment should be reversed compared to the gas phase and that also the 50 and 40 levels are overlapped together in the second band. His arguments, based on the variation of the relative intensities of the bands with D. E. Eastman and J. K. Cashion, Phys. Rev. Letters, 1971, 27, 1520. N. Rosch, W. G. Klemperer and K. H. Johnson, Chern. Phys. Letters, 1973,23,149. K. Johnson, personal communication. R. Messmer, Battelle Colloquium, The Physical Basis for Heterogeneous Catalysis, September, 1974, Gstaad, Switzerland. E. W. Plummer, Topics in Applied Physics, ed. R. Gomer (Springer-Verlag), to be published. E. W. Plummer, Battelle Colloquium, The Physical Basis for Heterogeneous Catalysis, Sep- tember, 1974, Gstaad, Switzerland.' L. Hedin and S. Lundqvist, Solid State Phys., 1969, 23, 1. * I. P. Batra and 0. Robauz, J. Vuc. Sci. Tech., 1974, (August), in press. G. Blyholder, J. Vuc. Sci. Tech., 1964, 11, 865.136 GENERAL DISCUSSION photon energy in the adsorbed and gaseous cases, do not seem appropriate. First, the relative intensity of the second band to the first seems too weak in both He I and He I1 for it to be assignable to 50+40. Second, the change of relative intensity ob- served in going from He I to He I1 in fact supports the original assignment ; 50, In. Third, the X-ray valence region spectrum of Fuggle for Ru/CO given earlier in this meeting shows the CO bands so weakly as to be practically unobservable in the raw data (as opposed to the difference spectrum) and their relative intensities can therefore hardly be used to justify a reversed assignment.A further point is that a reversed assignment implies an enormous change in energy of 50 relative to 1n on adsorption (about 6 eV). Such an effect is perhaps conceivable for a uery strongly bound situation, but the two bands are still observable for y CO on Mo, a very weak re- versibly adsorbed state where such an enormous change is inconceivable. The definitive experiment would be physically to adsorb CO at low temperature (i.e., the weakest adsorption possible) and examine that spectrum. Dr. J. C. Fuggle and Prof. D. Menzel (Munich) said : Plummer and Lloyd have chal- lengedthe previously accepted assignment of photoemission peaks from chemisorbed CO.They suggest that the peaks found at 10.5-12 eV, and 7.5-9 eV below EF be re-assigned to (40 + 1 x) and 50, or 40 and (1 + 50) CO-orbitals, respectively. We have studied clean and CO-covered Ru(OO1) by XPS and have found that CO adsorbate peaks are found at - 10.7 and - 8.1 eV below EF with total integrated intensity - 1 /80 that of the Ru valence band and intensity ratio approximately 2 :l. Only two adsorbate peaks were found in a search to 25 eV below EF. Theoretical, and experimental 4 9 estimates of the relative intensities of XPS peaks from gaseous CO give intensity ratios 40 :1x :50 of approximately 2.5 :0.5 :1.0. Our results do not support the old assignment of the peak at 10.5-12 eV below EF to the 1 x level for we cannot then explain the absence of the peak due to the 40 level, which should be the strongest peak.However, if the electronic orbitals in adsorbed CO retain their molecular character then our results support the new assignment of Lloyd in which the peak at 10.7 eV is associated with the 40 level and should be the strongest adsorbate valence band peak in the XP spectrum. The relative intensities of the same peaks in He I and He I1 spectra also seem to support this assignment. Dr. T. B. Grimley (Liverpool) said: Regarding the assignments of the peaks in the photoemission spectra of CO on metals to the 50 and 1n levels, a thorough study of the angular dependences of the intensities would be extremely useful in deciding this question, assuming that final state effects do not dominate the angular dependence so that the symmetry of the initial state has observable consequences.Dr. D. R. Lloyd (Birmingham) said: Some information relevant to the assignment of the two bands observed in p.e. spectra of CO adsorbed on metal surfaces may be gained by a comparison with the p.e. spectra of the molecular metal carbonyls. A variety of these have now been studied, including Ni(C0)4,6 M(C0)6 (M = Cr, Mo, J. C. Fuggle and D. Menzel, this Discussion, p. 129. S. J. Atkinson, C. R. Brundle and M. W. Roberts, this Discussion, p. 62. M. Steinkilberg, J. C. Fuggle, T. E. Madey and D. Menzel, Phys. Letters A, in press. 4K. Siegbahn, C. Nordling, G. Johansson, J. Hedman, P. F. Heden, K. Hamrin, U. Geiius, T. Bergmark, L. 0.Werme, R. Manne and Y. Baer, ESCA Applied to Free Molecules (North Holland, Amsterdam, 1969), p. 77. J. T. J. Hung, F. 0. Ellison and J. W. Rabalais, J. Electron Spectr. 1974, 3, 339. I. H. Hillier, M. F. Guest, B. R. Higginson and D. R. Lloyd, Mol. Phys., 1974, 27, 215.GENERAL DISCUSSION 137 W, V) ‘ 9 and M(CO),Y (M = Cr, W, Mn, Re, Y = various ligand~).~-~ In all of these there are only two principal bands which can be assigned as ionization from orbitals mainly localised on CO. In the He I and the Me I1 spectra these appear as an intense band in the 13-16 eV region and a weaker band in the 17-19 eV region, and in the XPS spectra there is a substantial increase in intensity of the weaker band. The only reasonable assignment of these bands, which was first proposed at an earlier Discu~sion,~ is that the lower i.p.band corresponds to ionization from groups of orbitals derived from 5 0 and ln, and that the band at higher i.p. is due to orbitals derived from 40. This is fully supported by ab initio molecular orbital calculations, and detailed analysis suggests that the spread of the bands derived from 5a is greater than that from In but that the average ionization potentials for the two groups are very similar. Thus the intensity changes on forming a carbonyl referred to by Biloen and Holscher are the result of overlapping ionisations from 1n and 50 levels, and not of different cross-secti0ns.l The relative intensities of the two bands in XPS spectra for carbonyls are certainly compatible with those in the spectrum of CO adsorbed on Ru reported by Fuggle and Menzel, and the energy separations of the two bands of adsorbed CO are very similar to the separations of the two bands in carbonyls.I propose that a reassignment of the low i.p. band of adsorbed CO as 1 n + 50, and the high i.p. band as 40, be given some consideration. Dr. C . R. Brundle (Bradford) said: This suggestion of Lloyd is more reasonable than Plummer’s since it not only accounts for the absence of a third band, but also does not require an enormous shift of 5 0 relative to In on adsorption (about 2 eV only). A first band composed of 5 0 + In would also account for its extra intensity. One can in fact see in fig. 6 of the paper by Page and Williams an incipient splitting of the first band for CO/Ni, suggesting a 5 0 and In overlap.DP. P. Biloen (Shell, Amsterdam) said: As far as the interpretation of the UPS spectra of gaseous carbonyls is concerned, I fully agree with Lloyd. However, his assignment makes it clear that a major cause of the difference between the spectrum of CO(g) and the carbonyls is the attachment of several CO molecules simultaneously to the same metal atom. For chemisorption this is not excluded, but it is not likely to be a representative situation. Contrary to my original impression therefore, I believe that the carbonyls are not really suitable as a guide for interpreting the chemisorption spectra. I thus am reluctant to accept the reassignment based on carbonyl spectra, as proposed by Lloyd. Dr. D. R. Lloyd (Birmingham) (communicated): There is no evidence in the work I have cited which suggests that the superposition of 50 and In ionizations in the car- bonyls is due to attachment of more than one CO molecule to a metal atom.B. R. Higginson, D. R. Lloyd, P. Burroughs, D. M. Gibson and A. F. Orchard, J.C.S. Faradw ZI, 1973, 69, 1659. S. Evans, 3’. C. Green, A. F. Orchard, T. Saito and D. W. Turner, Chem. Phys. Letters, 1969,4, 361. B. R. Higginson, D. R. Lloyd, J. A. Connor and I. H. Hillier, J.C.S. Farday ZZ, 1974,70,1418. S. Evans, J. C. Green, M. L. H. Green, A. F. Orchard and D. W. Turner, Disc. Furaday Soc., 1969, 47, 112. S. Cradock, E. A. V. Ebsworth and A. Robertson, J.C.S. Dalton, 1973, 22. B. R. Higginson, D. R. Lloyd, S. Evans and A. F. Orchard, unpublished work. ’ I. M. Hillier and V. R. Saunders, Mol.Phys., 1970, 22, 1025. * M. F. Guest, M. B. Hall and I. H. Hillier, Mol. Phys., 1973, 25, 629.138 GENERAL DISCUSSION Dr. R. W. Joyner (Bradford) said; In the adsorption of CO on polycrystalline platinum, Biloen and Holscher observe the peak at 11.7 eV in the helium I spectrum only at CO pressures 2 1 x Torr, (I x Pa). In the helium I1 spectrum, however, this band appears immediately at a CO pressure of 1 x lo-' T0rr.l The failure to observe this feature at low coverages in the helium I spectrum is thought to be due only to a low ionization cross-section, and similar differences between He I and He I1 spectra are noted for CO chemisorption on copper at 80 K, 1* where there is only one adsorption state. In agreement with Biloen and Holscher, oxygen exposures up to (and well above) 100 Langmuir do not generate new bands in the UPS spectra of platinum surfaces.However, even below 100 L an oxygen 1s peak is observed at a binding energy of 529.8+0.2 eV in the XPS spectrum.l The UPS spectra show an insensitivity to oxygen chemisorption which is difficult to understand. Dr. C. R. Brundle (Bradford) said : Biloen has shown the He I spectrum of acetylene on Pt. Demuth and Eastman have previously observed the spectrum of acetylene on Ni (111) and I have recorded the spectra for C2H2 on polycrystalline nickel. Unfortunately the (n) upper acetylene level is obscured by Pt levels in Biloen's work, but for Ni it is quite apparent (fig. 1). It is, however, very broad and flat in 3 ' . * * a ' . - - - 6 L, I 2 3 4 5 6 7 8 9 1 0 1 1 1213 binding energy/eV FIG.1.-He I photoelectron spectrum of (a) gaseous acetylene (schematic) ; (b) clean Ni film at 273 K ; (c) Ni plus acetylene at 273 K ; (d) the difference spectrum of (b) and (c). the chemisorbed state (at 273 K) as opposed to the condensed state (77 K) or the gas phase where a sharp intense peak is recorded (symptomatic of non-bonding character). In the chemisorbed state, assuming n bonding, the (n) level will not remain degenerate if one n orbital is bonded to the surface and the other is parallel to it. One would expect the non-bonding parallel level to be largely unaffected by the chemisorption R. W. Joyner and M. W. Roberts, to be published. R. W. Joyner and M. W. Roberts, J.C.S. Favaday I, 1974,70, 1819.GENERAL DISCUSSION I39 bonding, and the other to move deeper owing to the bonding interaction with the metal surface.The severe broadening and flattening of the (n)' level observed presumably represents this splitting and provides a measure of the interaction of the n orbital with the surface. This illustrates the type of information one might hope to obtain from valence region photoelectron spectroscopy where degeneracies are lifted on adsorption. I thank A. D. Baker and P. Liotta for originally pointing out to me the discrepancy in shape of the n: band in the chemisorbed state. Dr. J. Kiippers (Munich) (Communicated) : A reinvestigation of the system Pd/ hydrogen at low temperatures has revealed a result different from that demonstrated in fig. 5 of our paper.' As can be seen from fig.1, after hydrogen adsorption the -7- 2 i 6-8 I0 I 2 I4 binding energy/eV Pd(ll1) h = 40.8 ev T = 130 K FIG. 1.-Photoemission spectra of a clean and a hydrogen covered Pd(l11) surface. Difference spectrum on the right. photoemission spectrum of a Pd( 1 1 1) surface is altered : the difference spectrum shows two maxima at 2 and 6.7 eV below the Fermi level. Actually only the maximum at 2 eV below EF is pronounced above the zero line; the peak at 6.7 eV does not exceed the zero level in the difference spectrum. This results from the total decrease of emission over a wide energy range. In addition to the two extra peaks, a strong decrease of electron emission just below EF is visible, this effect being strongest at EF. H. Conrad, G. Ertl, J. Kiippers and E.E. Latta, this Discussion.140 GENERAL DISCUSSION To understand the occurrence of hydrogen-induced changes in the photoemission spectrum at low temperature one must take into account the small adsorption energy of H2 on Pd. In recent work this was determined to be 20.8 kcal mol-' on the crystal face.' Lowering the temperature would cause a marked increase of hydrogen coverage (in the present pressure range). As photoemission seems to be of low sensitivity with respect to hydrogen adsorption, in contrast to CO-adsorption,2 this coverage increase then leads to measureable changes. In an earlier measurement, Eastman et al. found, after absorption of hydrogen in a Pd-film, a broad peak centred at 5.4 eV below EF using He and Ne resonance radiation. With hydrogen adsorbed on Pd changes of the spectra were only visible with He-, but not with Ne-radiation.A detailed description of photoemission spectra of hydrogen adsorption on Pd as well as on Ni and Cu surfaces will be given el~ewhere.~ Dr. J. Kuppers (Munich) said: We would mention some results about photo- emission from oxygen and carbon monoxide co-adsorbed on a Ni( 11 1) surface. We think that these measurements are useful in relation to two contradictory topics discussed : (1) the role of the work function in the comparison of peaks in the photoemission spectra of adsorbed species with the respective free molecule spectra; (2) the assignment of the two peaks which rise in the photoemission spectra of adsorbed CO on transition metal surfaces to the CT and IT orbitals of the free CO molecule.In fig. 1 the photoemission spectra of clean Ni (bottom), oxygen- and CO-covered Ni surfaces (middle) are compared with a spectrum which is revealed after pre- adsorption of oxygen (6 L exposure; this causes the build-up of a (2 x 2) super- structure) before admitting CO to the surface. At the right-hand side of fig. 1 with increased energy resolution the CO-spectra with and without oxygen preadsorption are displayed. (1) The first CO-induced peak at 7.8 eV below EF remains in its original position on oxygen preadsorbed and is not affected by the total work function increase of about 0.8 eV due to the coadsorbed oxygen. that only the local work function is involved in the measuring process. The same argument holds for the position of the 2p level of oxygen (fig.1,2nd and 4th spectrum on the left) with or without coadsorbed CO. (2) The second CO-induced peak at 10.8 eV below EF is shifted by 0.3 eV when oxygen is preadsorbed, the level being shifted downwards to higher binding energy. Supposing that the original assignment of the CO levels in the adsorbed state as given by Eastman is true, one must conclude that the Coulomb repulsion within the ln- orbital of CO is lowered by the presence of coadsorbed oxygen. It would be expected that the C-0 stretching vibration of the CO molecule is affected by this change. Indeed, Eischens found a remarkable shift of this vibration using an i.r. technique, indicating an increased bond strength between C and 0 when oxygen is preadsorbed. The consistency of these data tends to favour the original assignment of the two levels.This fact supports the remarks of Menzel H. Conrad, G. Ertl and E. E. Latta, Surface Sci., 1974, 41, 435. J. Kuppers, H. Conrad, G. Ertl and E. E. Latta, Japan J. Appl. Phys., in press. D. E. Eastman, J. K. Cashion and A. C. Switendick, Phys. Rev. Letters, 1971, 27, 35. H. Conrad, G. Ertl, J. Kuppers and E. E. Latta, to be published. D. MenzeI, this Discussion ; R. W. Joyner, this Discussion. E. W. Plummer, this Discussion ; C. R. Brundle, this Discussion. ' D. E. Eastman and J. K. Cashion, Phys. Rev. Letters, 1971, 27, 1520. * R. P. Eischens, personal communication.GENERAL DISCUSSION 141 E+=O 2 4 6 8 10 12 binding energy/eV FIG. 1.-Photoemission spectra from a Ni(l11) surface (hv = 40.8 eV). Left : clean surface (bottom) ; oxygen covered surface, 6 L exposure ; CO-covered surface ; surface with preadsorbed oxygen (6 L exposure) in CO atmosphere (top).Right : surface covered with CO (top) ; with preadsorbed oxygen, 6 L exposure (bottom). Dr. H. Killesreiter (Marburg) said: Energy transfer depends on correct energy relations on both sides of the phase boundary. Charge carrier formation by this method is at least a two step process and, thus, has a reduced quantum efficiency 4 = &T x &, &T being the efficiency for energy transfer, and c,bCT that for the sub- sequent charge transfer, both quantities being not greater than unity. The system p-chloranilldye does not allow efficient energy transfer as the strong transitions in the spectra do not overlap (A,,, = 412 nm for the lowest n-n* singlet state of p - chloranil crystals and 498,526, and 552 nm for the dyes oxacarbocyanine, Rhodamine 6G and Rhodamine B, respectively). Other transitions, such as those involving triplets, cannot hold for the high quantum efficiencies of up to unity that have been reported. Tunnelling has been used very successfully to describe the main step in the theory of electrochemical reactions. Of course, this conclusion cannot be drawn from an experimentally determined barrier height of eV that has been evaluated as " a striking point '' for discussion sake : either, because of the high electron affinity (-4 eV) of chloranil there is no effective barrier height, or, by an experimental error, the chromophores are not separated from the crystal surface by the aliphatic chains. However, the latter point would be in disagreement with the observed wavelength shifts and the quenching of excitation energy due to distance variations.1 42 GENERAL DISCUSSION It has been pointed out (fig. 2) that the photocurrent due to monomers and dimers is not usually a linear function of absorbance. So, an investigation of the function photocurrent = f(absorbance), restricted to values of ;1,,,, and Adimer, was needed to distinguish the contributions due to adsorption and those due to formation of species at the surface, by comparison with a rigid structure, the absorption of which could be checked by independent absorption measurements. The result was not a linear dependency of the photocurrent on the absorbance and the minor injection rates by dimers compared to those by monomers (table 1) have been discussed in terms of lowered excited levels and of distance variations of the reacting sites (fig. 5 and 1). Deviations from the proportionality at constant wavelength are ascribed to structural differences from A to C in the measurements of fig. 3. As, up to now, no information is available about additional quenching reactions or changes in the density of adsorbed dye molecules on both substrates, conclusions have only been drawn from the relative heights of monomer and dimer peaks.

ISSN:0301-7249    

DOI:10.1039/DC9745800125

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

B. PHOTO-ADSORPTION, PHOTO-DESORPTION AND PHOTO- REACTIONS AT SURFACES Oxidation of CO and Desorption of Oxygen by Ultra-violet Irradiation of ZnO Single Crystals under Ultra-high Vacuum Conditions BY F. STEINBACH* AND R. HARBORTH Institut fur Physikalische Chemie der Universitat Hamburg, D-2000 Hamburg 13, Laufgraben 24 Received 17th May, 1974 When the clean surface of a ZnO single crystal, mounted in an UHV flow system, is heated to temperatures of 400 or 450°C and, in the presence of Torr O2 and CO, is illuminated periodically by ultra-violet light from a high pressure mercury arc, rapid changes of the partial pressures due to photo catalytic reaction to COz or photo desorption of oxygen are monitored by a quadrupole mass spectrometer. In the presence of O2 at pressures of lo-* Tom or below, atomic oxygen is produced in a quantum process and, in the absence of a partner apt to reaction, emitted into the vacuum.The photo generation of 0 is accompanied by thermo-desorption of Oz. In the presence of CO at pressures of lo-* Ton: or below, no atomic oxygen may be detected, instead, C02 is produced by photo reaction. Atomic 0 is generated by photo dissociation of a ZnO bond, 4 eV being the upper limit of the energy necessary for this process. The investigation of a surface reaction by means of photo excitation of the reacting bonds together with direct mass spectrometric observation of the species being desorbed from the surface provides several major advantages. The direct mass spectrometric observation of the species being desorbed from the surface by means of a quadrupole mass spectrometer mounted in ultrahigh vacuum in the close vicinity of the surface allows the identification of unstable intermediates formed due to photo- reaction.The energy necessary for the reaction studied may be added in a controlled way by variation of the photon energy and of the photon density. Thus, threshold energies for the various reaction steps may be measured. By the use of chopped light together with lock-in amplification, the velocity of each single reaction step leading to desorption of a particle flux sufficiently high to be detected by the mass spectrometer may be determined. In the present investigation the photo induced formation of C02 at clean surfaces of ZnO single crystals and the interaction of the substrates CO and O2 with the ZnO surface have been studied under steady state conditions using unmodulated light of constant intensity of a full mercury arc.'. On powdered ZnO catalysts in the pressure range of several Torr to atmospheric pressure, the reaction has been studied in some detail under the aspects of the iduences of doping, of metal support and of the intensity and wavelength of the exciting light on the activation energy of the reaction.3-8 At illuminated surfaces two processes may occur, either reactions due to phonon excitation, i.e., production of thermal energy in the surface region, or else true photo reactions due to electronic excitation. In the present study the photo generated intermediate has been directly observed. In addition, the different patterns of response of the various reaction steps to interruptions of illumination, as well as 143144 OXIDATION OF co BY U.V.IRRADIATION OF ZnO several experiments with modulated illumination, furnish further arguments for a clear distinction of photo processes and thermal reactions. EXPERIMENTAL A stainless steel vacuum chamber (fig. 1) of 501. volume (a modified chamber of the Torr by a 70 I./s turbomolecular pump Varian 120 Leed system) was evacuated to some (Pfeiffer-Balzers) after bakeout at 300 or 400°C. FIG. 1 .-Schematic diagram of the UHV apparatus ; plan view : 1, mercury arc ; 2, U-V grade sapphire window ; 3, ZnO crystal ; 4, quadrupole mass spectrometer with rectangular mounted secondary electron amplifier ; 5, power unit ; 6, oscillograph or recorder ; 7, turbomolecular pump below the vacuum chamber ; 8, electrical feedthroughs ; 9, liquid helium cryopump.By the use of a 2000 l./s liquid helium cryopump (Balzers) having a pumping speed of 6OOO l./s with respect to hydrogen, a final pressure of 2 to 4x lo-'' Torr was reached. A single crystal of pure ZnO of 6 mm diameter and 20 mm length was situated in the cavity of a holder made from Pyrex glass. The holder could be heated to 550"C, it was mounted in close vicinity to an U-V grade sapphire window (Varian) and to the ionisation chamber of the quadrupole mass spectrometer (QMA 143, Balzers) outfitted with a rectangular mounted secondary electron amplifier. Oxygen (Linde) and carbon monoxide (l'air liquide) of high purity were introduced by two variable leak valves (Varian). Illumination was effected by a lo00 W or a 500 W d.c.high pressure mercury arc (Osram) mounted in a Schoeffel lamp housing (spectrum fig. 2). The light was concentrated on the surface of the ZnO crystal by a polished A1 mirror and quartz lenses. Previous to insertion into the UHV chamber, the crystal was cleaned by etching in syrupy phosphoric acid and I 5 25 I# 66 photon energy/eV FIG. 2.-Spectrum of the high pressure mercury arc and relative intensity of the lines in arbitrary units.F. STEINBACH AND R . HARBORTH 145 subsequent rinsing with distilled water, KOH solution and distilled water, again. Before photo catalytic experiments were carried out, the crystal was degassed for several hours at 550°C under 2 to 4x 10-lo Torr.The experiments were carried out in a continuous flow of oxygen, carbon monoxide or mixtures of both at pressures of lo-* Torr or below. The crystal temperature usually was kept at 450°C. No masking was-applied to the crystal surfaces ; th_us, the ambient gases were in contact with prismatic (llOO), polar (OOO1) and growing (1 120) surfaces. The effects caused by illumination of the polar surface only did not exhibit any fundamental difference from those caused by focusing the main intensity on the prismatic surfaces. RESULTS When CO and 02, both were present at pressures of about lo-* Torr, CO, was formed during the periods of illumination (fig. 3); the increase of the C02 pressure above the dark value was about 5 x Torr. This is due to photo catalytic formation of C 0 2 on the ZnO surface only, as was shown by blank experiments at the same conditions (fig.4). I . -. . 4 8 12 I 6 a- tlmin FIG. 3.--Pressure of C 0 2 during alternating periods of illumination and darkness of the ZnO crystal. Crystal temperature 450°C ; 1OOO W mercury arc ; 5 x Tom 01, CO, respectively. An increase of only about 5 x 10-14 Torr C02 was observed due to the reaction at the illuminated heated glass surface. Photo catalytic oxidation of CO was possible only in the temperature range above 400°C. At temperatures below 400°C neither a catalytic formation of CO, nor a photo catalytic effect was observed. Since the temperature of the ZnO surface rises and falls during the light and dark periods, respectively, the response of the reaction to changes of the crystal temperature was measured in the dark (fig.5). The response was much slower ; the increase or decrease of the C 0 2 formation due to temperature differences of about 150°C was several orders of magnitude smaller (5 x 10-14 Torr) than the effects due to illumina- tion. Simultaneously with the light induced formation of C 0 2 , desorption of O2 during the periods of illumination was observed (fig. 6); the response was compara- tively slow. A similar behaviour of the 0, pressure was observed in the absence of co.146 OXIDATION OF CO BY U.V. IRRADIATION OF ZnO When CO was present alone, the effects of periodic illumination on C 0 2 formation, though similar to the pattern shown in fig. 3, were, however, of a smaller magnitude, decreasing very rapidly during a small number of cycles.Off 1 4 8 12 16 tlmin FIG. 4.-Pressure of COz during alternative periods of illumination and darkness of the empty crystal holder. Temperature of the holder 450 "C ; lo00 W mercury arc ; lo-' Torr Oz, CO, respectively. The direct mass spectrometric observation of atomic oxygen as an intermediate of the photo reaction (fig. 7) exhibited a pattern completely different from the response of all other ambient pressures. Over a slowly rising background due to the rising pressure of 0, from the inlet of 0, by the leak valve, fast and very sharp rises and falls of the pressure of 0 were observed. They were more than ten times greater 10 20 30 tlmin FIG. 5.-Pressure of COz at different temperatures of the ZnO crystal in the dark.The various temperatures (lower scale of the additional abscissa) are reached after sudden changes of the heating current of the crystal holder (top scale of the additional abscissa). than they would have been, had they originated solely in the 10 % of 0 generated from O2 in the ion source of the quadrupole mass spectrometer. No photoproduced 0 whatsoever could be detected when equal amounts of CO were present together withF. STEINBACH AND R. HARBORTH 147 the 02. Furthermore, no 0 was detected during thermodesorption of oxygen in the dark up to temperatutes of 550°C, the ratio of 02/0 in the mass spectrum staying strictly at 10/1 in all experiments. L10-4 0 5 lo 15 t/min FIG. 6.-Thermodesorption of O2 during alternating periods of illumination and darkness of the ZnO crystal.Crystal temperature 440°C; 500 W mercury arc; 2 x lo-* TOK 01, CO, respectively. O f f 2 4 6 tlmin FIG. 7.-Photodesorption of 0 during alternating periods of illuminating and darkness of the ZnO crystal. Crystal temperature 450°C ; lo00 W mercury arc ; Torr 02. DISCUSSION THE DIFFERENT SURFACE PROCESSES Regarding the pressure of the species liberated from the crystal surface, three markedly different patterns of response to the change from light to dark and from148 dark to light are established by the results shown in fig. 3 to 7. The velocity of the rise and decay of the ambient pressure is a critical parameter with respect to the distinction between photon and phonon processes. For a proper appraisal of the pressure curves-especially of the decay periods-the high pumping speed together with the large dimensions of the vacuum chamber have to be borne in mind.The simplest pattern is exhibited by the pressure curve of atomic oxygen (fig. 7). Instan- taneous rise and fall of the pressure occurs when illumination begins and ends. Since atomic oxygen is generated during illumination only and is never observed in the dark, the generation of 0 is attributed to a true photon process. Accordingly, no slow rise and decay of the pressure as in a thermal process is observed. The opposite behaviour is exhibited by the desorption of molecular oxygen (fig. 6). A comparatively slow increase of the O2 pressure, caused by thermodesorption due to phonon processes during the light period, is followed by a slow decay according to slow cooling down of the crystal during the dark period.Comparison of fig. 6 and 7 shows that no relationship exists between the two patterns ; obviously, both processes occur simultaneously, but independently of each other. A hybrid pattern is seen in the pressure of CO,. Sharp rises and falls of repro- ducible amplitude are followed by periods of slow increases and decays, the difference depending on illumination time. The fast response is attributed to photo reaction, the " subsequent " slow response being caused by the accompanying temperature rise due to phonon processes. The amount of temperature rise naturally depends on illumination time. A comparison with fig. 5 confirms the pattern of the thermal increase and decay. From the magnitude of the CO, production in comparison to the dark experiment it may be seen that very high temperatures are generated on the surface during illumination.To summarize the above : atomic oxygen is produced in a quantum process and, in the absence of a partner apt to reaction, emitted into the vacuum. When CO is present, no atomic oxygen may be detected, instead, COP is produced by photo reaction. Both photo processes are accompanied by thermal desorption of O,, thermal formation and desorption of CO,, respectively. 0 is the reacting inter- mediate at the ZnO surface. The distinction between photoprocess and thermodesorption made on the ground of different patterns of response to alternating light and dark periods is well supported by lock-in experiments with light chopped at frequencies of 5 and 20 Hz.Only the pressures of 0 and CO, respond to these high frequencies. The pressures of CO, 0, as well as of H20 and H, do not respond, i.e., these are thermal processes. The phase shifts of the photo desorption of 0 and of the photo generation of CO, are different from each other, i.e., the rates of both processes are different. Similar to the experiments with continuous illumination, with modulated light the photo- processes were detected only at temperatures above 400°C ; however, as is character- istic of true photoprocesses, no temperature dependence of the rate was observed in the range from 400 to 500°C. OXIDATION OF CO BY U.V. IRRADIATION OF ZnO MECHANISMS FOR THE GENERATION OF ATOMIC OXYGEN From the spectrum of the mercury arc (fig.2) it is seen that the highest energy available for the photo process is about 4 eV. Therefore, a direct photo dissociation of molecular oxygen at the surface, followed by a desorption of 0 into the vacuum, is excluded for energetic reasons (eqn (1)). However, sufficient energy is available for the following processes (eqn (2) to (4)).F . STEINBACH AND R . HARBORTH 149 k V 0 2 -+ o+o ZnO -+ Zn+O hv hv Zn+02 + ZnO+O ZnO + Zn + +02 5.1-4 = 1.1 eV 4 -+5.1 = 1.45 eV. hv (3) (4) All three processes ((2)-(4)) are endothermic ; the energies given result from taking the highest available photon energy of 4 eV for the most endothermic process (2). Of course, for the processes (3) and (4) the activation energy is expected to be higher than the given endothermic energy; however, even for process (3) the activation energy should not be higher than 4 eV.Since no photo desorption of O2 is observed apart from the slow thermodesorption, it must be concluded that 0 desorbs faster from the surface than a recombination to molecular oxygen is possible (eqn (4)). Reaction (3) implies an oxidation of a reduced ZnO surface; the generation of 0 according to (3) should come to an end after the surface is re-oxidized. This has not been observed ; however, a decrease of the pressure jumps of 0 over long experimental runs has been observed in lock-in experiments. Thus, 0 is generated by a photo- dissection of a ZnO bond (eqn (2)), whereas in the presence of O2 the equilibrium of oxygen at the surface is restored by the reverse of process (4) as well as by process (3), which at the same time generates additional 0.Measurements of molecular reaction times and threshold energies will enable us to determine the contribution of each reaction to the whole process. The photodissociation of a ZnO bond according to (2) may be described by eqn ( 5 ) to (7). hv + p+e (5) p+ZnO -+ O+Zn+ e+Zn+ -+ Zn p+o, + 0 2 (8) The presence of 0; at the surface 9-13 at temperatures above 400°C may be excluded, since a reaction of the type (8) has not been detected. CONCLUSIONS In a direct photo process, atomic oxygen is generated at the ZnO surface by photo dissociation of a ZnO bond. 0 does not recombine with another 0 to form 02, however, it does recombine with CO forming CO,. 0 desorption and CO, formation have different molecular reaction times.In further experiments absolute molecular reaction times together with threshold energies of both processes have to be determined. We thank the Deutsche Forschungsgemeinschaft and the Fonds der Chemischen Industrie for financial support. We thank Prof. Dr. E. Mollwo and Dr. R. Helbig (Institut fur Angewandte Physik an der Universitat Erlangen-Nurnberg) for the supply of pure ZnO single crystals and Prof. Dr. A. Knappwost for his interest in our work. F. Steinbach, Proc. Vrh Int. Congr. Catalysis, ed. J. W. Hightower (North Holland, Amsterdam, 1973), p. 1008. F. Steinbach and R. Harborth, Ber. Bunsenges. phys. Chem., 1973, 77, 1024. G.-M. Schwab and J. Block, Z. phys. Chem. (Frankfurt), 1954, 1, 42.150 OXIDATION OF CO BY U.V. IRRADIATION OF ZnO F. Romero-Rossi and F. S . Stone, Actes ZZ-ikrne Congr. Znt. Catalyse (Edition Technip, Paris, 1961), p. 1481. G.-M. Schwab, F. Steinbach, H. Noller and M. Venugopalan, Z. Naturforsch., 1964. 1%, 45. W. Doaer and K. Hauf€e, J. Catalysis, 1964, 3, 171. ’ F. Steinbach, 2. phys. Chern. (Frmkfurt), 1968,60, 126. * F. Steinbach and R. Barth, Ber. Bunsenges. phys. Chem., 1969, 73, 884. G.-M. Schwab, F. Steinbach, H. Noller and M. Venugopalan, 2. Naturforsch., 1964,19a, 445. lo H. Chon and J. Pajares, J. Catalysis, 1969, 14, 257. K. Hauffe and R. Schmidt, Photogr. Korrespondenz, 1971, 107, 132. l2 H. G. Fitzky, Photogr. Korrespondenz, 1967, 103, 173. l3 S. Fukuzawa, K. M. Sancier and T. Kwan, J. Catalysis, 1968, 11, 364.

ISSN:0301-7249    

DOI:10.1039/DC9745800143

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

Interaction of Ultra-violet Radiation with Surface Energy Levels of the 2110-Li20-02 System as Revealed by Electron Spin Resonance Spectroscopy BY J. HABER,* K. KOSINSKI AND M. RUSIECKA Research Laboratories of Catalysis and Surface Chemistry, Polish Academy of Sciences, Krakbw, Poland Received 17th May, 1974 E.s.r. examination of the influence of u.v.-irradiation on surface processes at Li-doped ZnO shows that several paramagnetic centres are involved in the interaction with the gas phase. Annea- ling in oxygen at 773 K results in formation of zinc vacancies, with signal at g = 2.01 30. On oxygen adsorption at 293 K, zinc vacancies are formed at the surface with g = 2.0145. Both signals are visible only in Li-doped samples of low Fermi-level, their intensity depending on irradiation.The radical oxygen signals at g1 = 2.054, 9 2 = 2.0077 and g3 = 2.0033 accompany the formation of vacancies and increase on irradiation due to photo-adsorption. Signals at 9 = 1.96 and 1.975 are also present. On grinding, F-centres are formed with g = 1.997. They reversibly adsorb oxygen and interact with radiation. X-ray photoelectron spectra reveal two kinds of oxygen, the lattice oxygen 1s level depending on defect structure. The mechanism of the interaction of oxygen with the zinc oxide surface has aroused considerable interest in connection with the electrical and optical properties of this oxide. In order to obtain information on the correlation between surface properties and the electronic structure, the effect of irradiation on the processes at zinc oxide surface has been studied by many authors directly by following the kinetics of photosorption, as well as indirectly by measuring the photoconductivity, photo- e.m.f., photoelectric effect, e.s.r.spectra, etc. Additionally, valuable information on the type of adsorbed oxygen species and their role in oxidation reactions was obtained from these studies. Several recent review articles summarise the res~lts."~ we have shown that the sign of the photoeffect in the temperature range, wherein the adsorption equilibrium can be established, depends on the concentration of interstitial zinc atoms, changing from photodesorp- tion to photoadsorption when this concentration is increased. In terms of the electronic theory of chemisorption this may be explained by a shift of Fermi level.On the other hand, it has recently been pointed out that surface states may play an important role in chemisorption O It seemed, therefore, of interest to investigate the photosorption of oxygen at zinc oxide of different electronic structure, modified by Li-doping, and to identify some of the centres by e.s.r. spectroscopy. UPS and XPS techniques were used to get information on the influence of surface processes on energy levels of oxygen and zinc atoms as well as the band structure of the solid. In one of our previous studies EXPERIMENTAL Doped zinc oxide samples were obtained by calcination of zinc carbonate impregnated For the reference After drying in air at 393 K the samples were annealed with water suspensions of lithium carbonate of different concentration. sample, pure water was used instead.151152 SURFACE PROCESSES I N ZnO-Li,O-0, SYSTEM in air at 1273 K for 10 h, quenched and ground. In order to remove lithium ions which were not incorporated into the ZnO lattice, the samples were then repeatedly washed with distilled water until no lithium could be detected in the filtrate by flame photometric analysis. The quantity of lithium incorporated into the ZnO lattice was determined by flame photo- metry after dissolving the preparations in HCI. The following samples were obtained : A-pure ZnO, B 4 . 0 6 at %, (24.12 at % Li, D 4 . 9 5 at % Li, E--1.35 at % Li. Before being used for e.s.r. measurements the samples were standardized in the following way : sample outgassed at 773 K for 15 min, then : la-cooled and contacted at 293 K with oxygen at different pressures ; b-outgassed at 293 K ; 2a-contacted at 773 K with oxygen at different pressures, then cooled in oxygen; b-outgassed at 293 K ; 34ontacted with water vapour at 293 K.E.s.r. measurements were made with a Jeol JES-ME-X spectrometer. The sample cavity employed 100 kHz field modulation at 2 G amplitude. Microwave power was held constant at about 10 mW. The spectra were recorded in the temperature range between 77 K-293 K during cycles of dark periods and irradiation with u.-v.-light in situ in the spectrometer. An HBO-200 lamp was used for the irradiation, the light being filtered through water to cut out the infra-red. The g-values were obtained using a manganese standard. The UPS and X P S measurements were made with Vacuum Generators’ ESCA-3 spectrometer, equipped with aluminium X-ray anode and helium I u.v.-source.The samples were standardized in the preparation chamber of the spectrometer in the same way as for e.s.r. measurements. RESULTS As an example, fig. 1 illustrates the results of the e.s.r. measurements of sample D after outgassing at 773 K. In the region of g between 2.035 and 1.997 the initial FIG. 1.-Sample D heated at 773 K in vacuum and cooled to 77 K : 1. Before u.v.-irradiation. 2. During u.v.-irradiation. 3. After switching off irradiation. 4. During the second irradiation. spectrum (curve 1) shows only the presence of the line at g = 2.0035. After irradia- tion (curve 2), the strong signal at g1 = 2.0263, gll = 2.0026 appears, and is attributed to a positive hole localised at an 02- ion next to lithium substitutionally incorporated into the ZnO lattice, with characteristic hyperfine splitting from interaction with LifJ .HABER, K . KOSINSKI A N D M. RUSIECKA 153 nucleus of I = 3/2. It was observed in the spectra of all Li-doped samples registered at 77 K after irradiation with u.v.-light. At this temperature, the signal was stable even after switching off the irradiation, whereas at 123 K the signal disappeared after a few minutes of irradiation. Simultaneously a considerable change of the cavity Q was observed due to increased conductivity of sample; this effect disappeared after switching off the irradiation and thus corresponds to the thermoluminescence peaks and thermally stimulated current observed by Zwingel and interpreted as releasing of the electrons from traps to the conduction band and their recombination with holes localized at lithium acceptor centres.The spectrum of a sample irradiated at 77 K when registered at 293 K, showed no lithium signal. When however the sample was again cooled to 77 K the signal reappeared, although the sample had not been irradiated any more. This shows that heating to room temperature results only in the partial recombination of the holes localized at Li-centres. Simultaneously there appears a small signal with g = 2.0145 which has not been described in literature. Its intensity increases strongly after switching off the irradia- tion (curve 3) and decreases when the irradiation is again switched on (curve 4).The signal is slightly asymmetric and is observed only at 77 K. The line width is 1.2-2.4 G. It appears also in the spectra of non-irradiated samples, when oxygen is adsorbed at room temperature. This is illustrated in fig. 2, where the spectrum of 2.?023 1,9750 I I I 2.0265 2.0 I3 1 FIG. 2.-1. Sample D heated at 773 K in air and cooled to 77 K. 2. Sample D heated at 773 K in vacuum and then exposed to 8 x lo-' of oxygen at room temperature. 3. Same sample during u.v.- irradiation. Both spectra recorded at 77 K. sample D-la is shown before (curve 2) and during u.v.-irradiation (curve 3). A strong signal at g = 2.0145 is present even in the initial sample and its intensity remains unchanged by irradiation. The influence of irradiation depends on oxygen pressure : (a) when the oxygen pressure is lower than Torr, no signal is present in the sample outgassed at 773 K.It appears on irradiation. (b) On contacting the sample with 10-1 Torr a small signal appears. Irradiation increases its intensity. (c) When the oxygen pressure is higher, a large signal is observed even before irradia-154 SURFACE PROCESSES I N ZnO-Li20-Q2 SYSTEM tion, which either does not change the intensity of the signal or results in its decrease. In all cases the intensity of the signal increases after switching off the irradiation and decreases when it is again switched on. On outgassing at room temperature the signal disappears. When Li-doped samples are contracted for a longer time with oxygen at 773 K a new signal with g = 2.0130 is observed in the spectra, registered at 77 K and 293 K. As an example curve 1 in fig.2 shows the spectrum of sample D-2a. The signal can be removed by outgassing the sample again at 773 K. It is asymmetric, but no splitting was observed even by modulation of 1 G and k 10 G sweep field. The spectra represented in fig. 2 also show a weak line at g = 1.975. It appears in the spectra of all samples which were exposed to oxygen and is observed in the temperature range between 77 K and 293 K. It is not influenced by irradiation. 2.0050 2.0077 2.0023 1 I I FIG. 3.-Temperature dependence of 0; signal. Sample D heated at 773 K in vacuum and then exposed to 2 x lo-* Torr of oxygen at room temperature. Spectra recorded at : 1-293 K, 2-203 K, 3-123 K and 4-77 K.The spectra of all samples on which oxygen was adsorbed at pressures lower than 1 Torr, recorded in the temperature range between 77 K and 293 K revealed the presence of a signal at g1 = 2.054, g2 = 2.0077, g3 = 2.0033 as described in the literature 12-15 and assigned to adsorbed oxygen radicals 0;. It is illustrated in fig. 3, where sections of the spectra of sample D-la are recorded at different tempera- tures. At low temperature a symmetric signal is superimposed at g = 2.003 screening the triplet signal. Its origin is not clear and it will not be discussed in detail. ItJ . HABER, K . KOSINSKI AND M. RUSIECKA 155 disappears at 203 K and only the triplet signal remains at higher temperatures. It may be noticed that its intensity increases on raising the temperature to 293 K, it increases also on irradiation at 77 K.All spectra recorded at 293 K showed also the presence of a well known line at 1.96 assigned to conduction electrons. In agreement with the observations of Sancier l6 its intensity decreased on lowering the temperature. It also decreased with increasing Li content as noticed by Kasai so that it practically disappeared from the spectra of samples C, D and E recorded at 77 K. This is illustrated in fig. 4, where spectra 1 and 2 refer to samples A and E respectively, both recorded at 77 K. 2.0184 2.0131 2 0 0 2 3 1.9971 I I i I FIG. 4.-E.s.r. spectra of: 1-sample A, 2-sample E, 3-sample E after grinding, 4-same sample after exposing to oxygen, 5-same sample after u.v.-irradiation. All spectra were recorded at 77 K after evacuating at room temperature.An interesting phenomenon was observed when Li-doped samples were carefully ground. Grinding provoked namely the appearance of a new line at g = 1.997, as seen from comparison of spectrum 2 in fig. 4, registered for fresh sample E, with spectrum 3 taken after grinding the sample. Both samples were outgassed at 293 K before the measurements. When the ground sample was then exposed to oxygen at room temperature, the signal at g = 1.997 disappeared (spectrum 4). It could be induced again by irradiation (spectrum 5). Fig. 5 shows the section of the X-ray photoelectron spectrum in the region of 01s obtained from sample D after different pretreatment. The binding energy of oxygen 1s electrons in the sample outgassed at 773 K amounts to 530.4 eV.It apparently does not change after exposure to oxygen at room temperature. When, however, the sample was heated in oxygen at 773 K, outgassed and cooled the binding energy increased. It increased still further to 531.4 eV after heating in oxygen at 773 K and then cooling in oxygen. All oxygen lines have a hump on the high energy side. Their deconvolution showed that they are composed of two lines, the higher energy level being 532.4 eV. The intensity of this line was high for outgassed samples and decreased to a low value after exposure to oxygen at high temperature, its position changing to a small extent only. The pretreatment in vacuum, oxygen or hydrogen had little effect on the u.v.-photoelectron spectrum of zinc oxide (fig. 6).156 SURFACE PROCESSES I N ZnO-Li,O-0, SYSTEM FIG.5.-The 1s photoelectron line of oxygen from sample E : 1-outgassed at 773 K, 2-oxygen ad- sorbed at 293 K, 3-annealed at 773 K in oxygen, evacuated and cooled, 4-annealed at 773 K in oxygen and cooled. Upper insert : curve 1 deconvoluted. I I I , 0 2.2 4.6 7.0 9.4 11.8 14.2 16.6 19 FIG. 6.-u.v.-photoelectron spectra of sample E : l-outgassed at 773 K, 2-annealed at 773 K in oxygen, 3-heated in hydrogen at 723 K.J. HABER, K . KOSINSKI AND M. RUSIECKA 157 DISCUSSION E.s.r. examination of the influence of u.v.-radiation on surface processes shows that several paramagnetic centres are involved in the interaction of the zinc oxide surface with the gas phase. They are characterised by the following signals : g = 1.96 ; 2.0145 and g1 = 2.0263,gIl = 2.0026.The spectra of samples containing more than 0.06 at % of Li, which were annealed at 773 K in oxygen at higher pressures show the presence of a signal at g = 2.0130. It is not present in samples outgassed at 773 K and is not generated by adsorption of oxygen at 293 K or lower temperature. This indicates that it may be related to a defect created by the interaction of the solid with oxygen at higher temperatures. Such conditions also existed during the preparation of samples ; therefore, fresh samples always revealed the presence of this signal. This temperature range coincides with the onset of bulk diffusion, permitting the incorporation of oxygen. We may thus postulate that annealing in oxygen at higher temperatures results in the incorpora- tion of oxygen into the lattice, which shifts the defect equilibria in the surface layer towards the formation of zinc vacancies : g = 1.975 ; g = 1.997 ; g, = 2.054, g2 = 2.0077, g3 = 2.0033 ; g = 2.0130 ; g = +02 gas+Zn, + Zn,,+O, Zn,, +- Zn, + V,,.In pure zinc oxide the Fermi level is high enough for the vacancies to be doubly ionised. When however, the Fermi level is shifted down by incorporation of Li- acceptor centres, the concentration of singly ionised zinc vacancies increases due to localisation of a positive hole on one of the surrounding non-axial oxygen ions. We assign the signal at g = 2.0130 to such a paramagnetic centre. The signal is asym- metric because of the asymmetry of the oxygen tetrahedron around the cationic vacancy.On irradiation the signal slightly increases, further increase being observed on cutting off the irradiation. This may be explained by the shift of ionisation equilibria between neutral, singly and doubly ionised vacancies, only singly ionised vacancies being paramagnetic centres. When the Li-doped sample, annealed in vacuum at 773 K, is exposed to oxygen at room temperature, a new signal appears at g = 2.0145. It seems that the appear- ance of this signal may be associated with the first stage of oxygen incorporation, i.e., the formation of a zinc vacancy at the surface of the oxide. The influence of u.v.- irradiation on this signal is consistent with such a model. The signal can be generated in samples outgassed at 773 K even at very low oxygen pressure by u.v.-radiation.As shown in our previous paper,6 in such conditions photoadsorption of oxygen is observed. The surface zinc vacancies may be thus created by oxygen adsorption, they are however almost totally ionised under the influence of irradiation. Only when irradiation is cut off, does the localisation of a hole take place, the vacancy becoming paramagnetic. This is confirmed by the observation that at high oxygen pressures, when a strong line at g = 2.0145 appears after adsorption, its intensity decreases considerably under irradiation and resumes the initial or even higher value when irradiation is switched off. For the same reason the signal is observed only in Li-doped samples at 77 K when Li-acceptor centres effectiveIy trap the electrons. The line disappears at temperatures higher than 77 K because of the short relaxation time.This is confirmed by the fact that the signal reappears after cooling back to 77 K, which indicates that the centres are not destroyed by heating to 123 K. The signal is removed by outgassing at room temperature, which yields further support to the model. The very small asymmetry of the signal may be explained by the158 SURFACE PROCESSES IN Zn0-Liz0-02 SYSTEM almost axial symmetry perpendicular to the surface plane. The number of spins was of the order of 10l6 spins/g sample, corresponding to a surface concentration of vacancies less that 1 % of a monolayer. Formation of the surface zinc vacancies may proceed in several consecutive steps, one of them being formation of the adsorbed radical oxygen by localisation of the electron in a surface acceptor level created by a physisorbed oxygen molecule.The spectra of all our samples with adsorbed oxygen, which showed the line at g = 2.0145 also revealed the presence of the characteristic triplet signal due to 0; indicating, in fact, that it may be the intermediate in the formation of a surface vacancy. The triplet signal was resolved only at oxygen pressures lower than 1 Torr. At higher pressure, the line broadening occurred as described by Cope and Campbell. Con- trary to the behaviour of the line at g = 2.0145, which was observed only at 77 K and decreased on irradiation, the intensity of the radical oxygen signal increased on irradiation at 77 K and the signal was stable at 293 K. This may be easily understood when it is remembered that in this case formation of the paramagnetic centre consists in trapping an electron instead of a hole as in the preceding case.The irradiation at low temperature increases the concentration of surface donor states, increasing the ionisation of radical oxygen states. All samples investigated after adsorption of oxygen at room temperature also showed a symmetric signal at g = 1.975. Its intensity is independent of the con- centration of lithium and is not influenced by irradiation. A similar signal was observed by Larach and Turkevich l8 on zinc oxide impregnated with rose bengal dye. They related the signal to 0; species, such assignment is, however, inconsistent with the symmetry of the line. The signal at g = 1.96 was present in the spectra of all samples.Analysis of the behaviour of the two components of this signal led Sancier l6 to the conclusion that the g = I .957 component is due to conduction band electrons whereas the g = 1.963 component is due to electrons in regions of precipitated zinc. Our samples after outgassing at 773 K showed one non-split asymmetric signal. Adsorption of oxygen at 293 K resulted in splitting of this signal into two components, due to a considerable decrease of the intensity of the g = 1.957 component. This indicates that localisation of electrons on surface oxygen acceptor levels involves conduction band electrons and not those of excess zinc. This result supports Sancier’s conclusion. Interesting conclusions may be drawn from the behaviour of samples which were ground prior to experiments.After outgassing at 293 K, a line at g = 1.997 appeared in addition to those described above. A signal of similar characteristics was observed on irradiation of a ZnO monocrystal with high energy electrons 19* 2o and was in- terpreted as due to the formation of an F-centre. We suggest thus that such V, centres are formed on grinding. When such a sample is contacted with oxygen at 293 K, the signal disappears, but reappears on outgassing. Apparently oxygen is reversibly adsorbed on these centres, withdrawing the electrons. It may thus be concluded that the F-centres formed on grinding are localised mainly at the surface of grains. When the concentration of conduction electrons is increased by irradiation, electrons are again localised on oxygen vacancies and the signal reappears.This indicates that at 293 K adsorbed oxygen does not recombine with vacancies. Annihilation of the F-centres takes place only on annealing in oxygen at 773 K. Surface processes resulting in the formation of various defects also have a pro- nounced influence on the energy states of lattice oxygen. This is clearly manifested in the photoelectron spectra. For samples which were outgassed at 773 K the energy of the oxygen 1s photoelectron line amounts to 530.4eV. The concentration of donor centres in such samples is very high, and we may expect that lattice oxygen ionsJ . HABER, K . KOSINSKI AND M. RUSIECKA 159 assume the most negative charge attainable in the zinc oxide crystal and thus the binding energy of electrons is lowest.On heating the sample in oxygen, defect equilibria shift towards an increased concentration of zinc vacancies, resulting in a decrease of the average negative charge on oxygen ions. This is manifested in the shift of the oxygen 1s photoelectron line to higher binding energies, which for samples annealed in oxygen at 773 K attains the value of 531.4 eV. The spectrum also reveals that two kinds of oxygen are present in the samples. It is interesting that the binding energy of the oxygen species amounting to 532.4 eV is similar to that observed in oxygen adsorbed on metallic zinc. We suggest thus that these species represent oxygen ions adsorbed on clusters of interstitial zinc precipitated in the surface layer of the grains. This is consistent with the fact that the intensity of the higher energy line decreases on annealing the sample in oxygen at higher tem- peratures.In this connection we recall that precipitation of zinc in the surface layer was also postulated to explain the splitting of the e.s.r. signal at g = 1.96. It is noteworthy that the structure of the energy bands in zinc oxide remains practically unaffected by the surface processes and no formation of a donor band is observed as suggested by some authors.21 The authors thank Mr. W. Marczewski, for recording the ESCA spectra. One of the authors (M. R.) gratefully acknowledges a grant from the Polish Academy of Sciences. J. H. Lunsford, Catal. Rev., 1973,8, 135. F. Steinbach, Heterogeneous Photocatalysis; in Topics in Current Chemistry, vol. 25, Catalysis (Springer Verlag, Berlin 1972,) p. 117. E. Molinari, Electronic Phenomena in Adsorption and Catalysis, ed. K. Hasiffe (Proc. Symp. IV Int. Congress on Catalysis, Moscow, 1969). G. Heiland, E. Mollwo and F. Stockmann, Solid State Phys., 1959, 8, 191. F. F. Vol’kenshtein, Adu. Catalysis, 1973, 23, 157. J. Haber and A. Kowalska, Bull. Acad. Polon. Sci., ser. chim., 1965, 13,463. ’ A. A. Lisatchenko and F. I. Vilesov, Dokl. Akad. Nauk U.R.S.S., 1970, 192, 365. * J. Lagowski, S. Sproles and H. C. Gatos, Surface Sci., 1972,30, 653. E. Arijs and F. Cardon, J. Solid State Chem., 1973, 6, 310. D. Zwingel, J. Luminescence, 1972, 5, 385. l o E. Arijs, F. Cardon and W. Maenhout-van der Vorst, J. Solid State Chem., 1973, 6, 319. l2 J. 0. Cope and I. D. Campbell, J.C.S. Fwaday I, 1973, 69, 1. l3 M. Setaka, S. Fukuzawa, Y. Kirino and T. Kwan, Chem. Farm. Bull. (Tokyo), 1968,16,1240. l4 M. Setaka, K. M. Sancier and T. Kwan, J. Catalysis, 1970, 16,44. l5 J. H. Lunsford, J. Chem. Phys., 1968,44,1487. l 6 K. M. Sancier, J. Phys. Chem., 1972, 76,2527. la S. Larach and J. Turkevich, Appl. Optics (Suppl. on Electrophotography), 1969, p. 45. l9 J. M. Smith and W. E. Vehse, Phys. Letters, 1970,31A, 147. ‘O D. R. Locker and J. M. Meese, IEEE Trans. Nucl. Sci., 1972, 19, 237. P. H. Kasai, Phys. Rev., 1963, 130,989. A. Hausmann, 2. Phys., 1970,237, 86.

ISSN:0301-7249    

DOI:10.1039/DC9745800151

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

Photo-assisted Surface Reactions Studied by Dynamic Mass Spectrometry BY JOSEPH CUNNINGHAM,” EOIN FINN AND NICOLAS SAMMAN Chemistry Department, University College, Cork, Ireland Received 30th May, 1974 A dynamic mass spectrometer system is described which utilises fast detection techniques to follow pressure surges of gaseous product initiated by incidence of 30 ps flashes of U.V. photons onto ZnO or Ti02 surfaces pretreated in vacuo at N m-2. A difficulty inherent to the system is appearance of “system” transients corresponding to flash-initiated growth in ion currents at m/e = 44, 28, 16 and 12, corresponding to photo-desorption of carbon oxides from walls of the system. Addition of ZnO samples to the system caused appearance of a new transient at m/e = 32, consistent with photo- assisted release of O2 from the ZnO surface.Rutile samples similarly treated did not give rise to a comparable O2 transient. Presence of electron-attaching methyl halides at pressures - N m-2 over the ZnO or TiO, surfaces during flashes resulted in only a small transient at m/e = 19, corresponding to growth of CDJH from CD31/Zn0 with 4(CD3H)w Presence of adsorbed, hole-trapping methanol and ethanol resulted in photo-assisted formation of formaldehyde or acetal- dehyde and observed yields correspond approximately to quantum efficiencies of for these photoassisted dehydrogenation processes. Use of the deuteratedethanol showed that photodesorption occurred accompanied by H/D exchange involving one deuterium atom in the originally adsorbed alcohol. For flash-illuminated ethanol/ZnO systems the mass distribution of transients could best be accounted for as a combination of flash desorption of alcohol, photoassisted formation of acetal- dehyde and photoassisted formation of ethylene.The importance of photodegradation of strongly adsorbed ethanol in these photoassisted surface processes was demonstrated by observations that a ratio of 1 : 4 : 9 observed between them when ethanol was present over the ZnO surface during the flash changed to 1 : 10 : 80 when flashes were delivered to the surface after all gaseous alcohol was removed. For ethanol/Ti02 systems, flash desorption of alcohol, plus photoassisted formation of acetaldehyde and ethylene were the main flash-initiated processes. Results are briefly compared with various models of the flash illuminated gas/oxide interface The work described in this paper is mainly directed towards improved under- standing of photoinduced processes at the surface of zinc oxide and titanium dioxide and it is appropriate first to develop realistic models of the nature of these surfaces prior to incidence of U.V.photons. Such models should ideally take into account both specifically surface properties of the solid/gas interface and, also, bulk-related properties of the metal oxide surface. Several recent publications have emphasised that surfaces of zinc oxide and titanium dioxide samples exhibit a greater degree of metal-excess, nonstoichiometry than the underlying bulk lattice. 1-4 Heating in vacuo to temperatures 300-400°C, which is the process utilised in the present study to degas the surfaces of the powdered samples, facilitates formation of such surface nonstoichiometry through loss of oxygen from surface layers.For TiOz samples the resultant oxygen deficiency is accommodated in surface layers as a combination of oxygen vacancies and various metal-excess centres, e g. interstitial T i r or Ti3+ adjacent to an oxygen vacancy. The existence of centres involving oxygen vacancies on thermally outgassed TiOz surfaces has been supported by several lines of evidence, but particularly by electron paramagnetic studies.l* By contrast, there is a lack of convincing evidence for existence of oxygen vacancies in ZnO samples, except where samples have been 160J . CUNNINGHAM, E . FINN AND N . SAIUMAN 161 exposed to bombardment by high energy neutrons, protons or electrons.Oxygen vacancies are not thought to exist in significant concentration at the surface of vacuum- outgassed Zn8 powders. Different reactivities of ZnO and TiOz surfaces are there- fore to be expected in surface processes involving oxygen vacancies. On the other hand, similar reactivities may arise in surface processes involving electron transfer from metal-excess surface centres, which exist as Ti;+ on TiO, and Z n i or Zng on ZnO. A third surface property of the ZnO and TiO, samples, which must be taken into account in the present study is existence of residual hydroxyl groups 7 9 after thermal outgassing at temperatures of 250-400°C. Fig. 1 illustrates application of a bulk-related model and attempts to relate electronic energy levels at the surfaces of ZnO and Ti02 samples to the energy band description of the bulk lattice.Part (a) in the centre of fig. 1 schematically represents energy levels of valence and conduction bands in Ti0, or ZnO, together with energy levels of shallow donors originating from metal-excess centres distributed throughout the bulk. Both ZnO and TiO, are intrinsic n-type semiconductors at room tempera- ture with conduction-band electrons originating through thermal promotion of electrons from shallow donor levels (0.03 and 0.05 eV trap depths respectively). * - 7;\ c fib (c) (4 (b) FIG. 1.-Electron energy levels in bulk and surface regions of ZnO and TiOz. (a) " Flat band " situation within bulk, with ca. 3.0 eV separating bottom of conduction band I&.from top of valence band &.b. In the dark electrons originate from donor levels, Ed', ca. 0.05 eV below &.b. resulting in Fermi energy, E f . as drawn. (b) Downward band-bending at the surface, as could originate in dark from surface excess of ionizable donors, or from cumulative chemisorption. (c) Upward band- bending at the surface, as could originate from electron localization by surface states or from depletive chemisorption. Arrows denote direction of initial movement of photogenerated electrons e, or holes, h, in surface potential gradients. As already noted, the surface of vacuum-activated samples exhibit a greater degree of metal-excess nonstoichiometry than the bulk. Part (b) of fig. 1 depicts downward band-bending, which should result from a greater surface density of ionizable donors (whose numbers were established during prior activation at a high temperature in uacuo).In the dark at room temperature such band-bending is a self-limiting process at the " clean '' surface, since downward bending of the bands reverses the process of electron loss and results in localisation of conduction band electrons on the surface donors. The situation depicted in fig. l(b) then results. An alternative view of band bending at " clean " thermally-outgassed ZnO or TiOz surfaces results [cf. fig. l(c)] when " surface states " capable of localising electrons predominate over surface nonstoichiometry. In such cases an upward bending of bands should arise, 58-F162 SURFACE REACTIONS BY DYNAMlC MASS SPECTROMETRY and again this is self-limiting. The extent to which such band-bending can be affected for ZnO and TiO, by strong adsorption and by illumination with U.V.photons must be considered in the present study. Molecules which adsorb onto the surface of ZnO or Ti02 with localization of an electron from the conduction band at the site of adsorption are said to undergo depletive chemisorption. lo* This results in upward band bending similar to that depicted in fig. l(c). Calculations based on the boundary layer theory indicate that upward band-bending and hence depletive adsorption is strictly limited, e.g. for ZnO with initially flat bands, saturation would correspond to ca. 0.1 % monolayer coverage l 2 by chemisorbed species such as 0; from oxygen adsorption or X- from dissociative depletive chemisorption of methyl halides.In the present study methyl halides are used as adsorbates onto ZnO and TiO, to examine the possible role of depletive chemisorption and for comparison with adsorbates undergoing cumulative chemisorption. For ZnO or TiOz cumulative chemisorption involves electron donation into the largely empty conduction band by the adsorbate. This in turn results in downward band-bending at the surface [cf. fig. l(b)] but is not limited by boundary layer considerations to the small saturation coverages noted above for depletive chemisorption. The surface nonstoichiometry and band- bending just described can exert strong influence on recombination rates of holes (11) in the valence band and additional electrons (Ae) in the conduction band of ZnO and Ti02.Photons of energies greater than the forbidden band gaps (3.2 and 3.0 eV respectively) can produce h and Ae, but high extinction coefficient for U.V. absorption means that such production mainly occurs within ca. Penetration depths coincide approximately with thickness, D, of the diffuse electrical double layer existing adjacent to the surfaces of ZnO and TiO, particles exhibiting the band-bending depicted in fig. l(b) or l(c). Photogenerated electrons and holes are, therefore, largely " born " in an electric field gradient which tends to separate them and so decrease geminate electron-hole recombination. For the situation depicted in fig. l(c) (which is appropriate to " clean " surfaces dominated by electron trapping surface states or to surfaces carrying saturation coverage of depletively chemisorbed species) photoholes should be drawn towards the surface and photoelectrons towards the interior of u.v.- illuminated metal-oxide particles.The converse should obtain for the band-bending depicted in fig. l(b) and, in either case, u.v.-illumination should decrease band- bending. It follows that, where band-bending originates exclusively froin strong chemisorption, extent of such chemisorption should decrease on illumination, i.e. photodesorption should be observed in the present study. If the decrease in band- bending expected from this process is exceeded by photoinduced changes in specifically surface properties (e.g. changes in concentration of surface statcs, surface hydroxyl groups or surface defects) then photoadsorption may result. The probable location at which photoholes undergo recombination with conduction band electrons differs sharply, however, for the situations depicted in fig. 1 (b) and 1 (c).For 1 (b), recombination within the particles is more probable since photoholes diffuse to the interior. Recombination close to the illuminated surface appears more probable for the situations in l(c). By analogy with results for alkali-metal halides,17 energy released in hole-electron recombination events may be released via radiationless processes as energy of ionic displacement at the location of recombination. It will be important to examine in the present study whether the greater probability of recombination close to the surface for the situation depicted in fig.l(c) results in greater displacement and disruption of the surface. m of the surface.J . CUNNINGHAM, E . FINN AND N . SAMMAN 163 EXPERIMENTAL APPARATUS It followed from considerations presented in the Introduction that experimental study of the influence of " specific surface " or " bulk-related " properties upon photosorption and other photoinduced surface processes on ZnO and TiO, should be carried out as far as possible with " clean " surfaces and with low pressures of adsorbates. Such conditions should avoid excessive contribution to observed photoeffects by physically or weakly cherni-. sorbed molecules and should accentuate photoeffects in strongly chemisorbed species Consequently most of the results here reported were obtained using a stainless-steel ion- pumped high vacuum system, as schematically represented in fig.2. This system routinely attained background pressures of N m-* after thermal outgassing. Samples of ZnO or TiOz were introduced into this system as thin layers previously deposited onto cylindrical quartz substrates of geometric surface area 0.01 m2. These metal oxidelquartz samples were located inside a short glass section of the vacuum system. The walls of this cylindrical glass " window " in the vacuum envelope were Kodial glass for most experiments and served to prevent light of wavelengths < 300 nm from entering the vacuum system. A quadrupole mass spectrometer (q.m.s.) operating at 4 MHz for the m/e range 0-50 or at 2 MHz for the m/e range 2-200, was an integral part of the vacuum system.The electron multiplier and 15 cm quadrupole mass filter of the q.m.s. were located very close to the ion pump to achieve minimum system pressure at their location. Results described in this paper were all obtained in conditions such that this minimum system pressure did not exceed 10-4Nm-2, as indicated by the metre of the ion pump. Reactant gases were introduced to the high vacuum system via metal variable leak valves from an external gas handling system and their steady state pressures at various other locations in the vacuum system, as monitored with Bayert Alpert ionization gauges (cf. fig. 2), did not exceed N m-2. Output of the q.m.s. at appropriate m/e values was linearly related to these pressures over the range 10-5-10-2 N mU2.A marked disadvantage of the q.m.s. used was that it did not yield " standard " ion fragmenta- tion patterns, but exhibited greater sensitivity to low mass numbers. This necessitated extensive calibration to obtain spectra for comparison. T FA -C I 0 Y l , L QMS E 0 IP FIG. Z.--Dynamic mass spectrometer system for the study of flash-initiated surface reaction and made up of : (i) high vacuum system comprising : inlet leak valve, I ; pressure measuring gauges, B ; glass walled photo reactor, C ; metal oxide layer, MO ; high conductance ss tubing, E ; quadrupole mass spectrometer, QMS; 14-stage electron multiplier, EM ; ion pump, IP ; and liquid nitrogen cooled baffle, LNB. (ii) Fast detection circuitry comprising : trigger unit, T ; variable delay line, D ; quartz flash tube, FT ; oscilloscope, 0 ; and fast amplifier, FA.(iii) Appropriate electronic supplies : S(EM), S(QMS1 and S(FT). MATERIALS The ZnO and TiO, materials used in the present study were high-purity powdered samples obtained through the courtesy of the New Jersey Zinc Co. and coded respectively as164 ZnO- SP500-78115 and Rutile-MR-128. Impurity contents of these oxides were low (e.g. <0.001 % Fe, Cu or Mn in ZnO and <0.07 % C12 in Ti02). Materials were also alike in surface areas (4 and 5.4 m2 g-'), particle sizes (0.2-2 pm diameter) and reflectance spectra (onset of absorbance at ca. 390 nm rising to a maximum at 370 nm). Either material was taken into an aqueous slurry with triply distilled H20, or occasionally D20, and coated onto a quartz substrate as a layer of thickness ca.m which was dried in a vacuum oven at 80°C before introduction into the high vacuum system. After bake-out of the entire vacuum system at 250°C until system pressure fell to N m--2r a small heater was placed around the glass section of the vacuum system to bake out the metal oxide at 250-350°C for 16 h. A sequence of experimental observations on such samples was usually commenced within 1 h of cooling to room temperature. Residual gas analysis indicated N m-2 partial pressure of CO plus N2 as the major constituents of residual gases. Reactant gases N20 and O2 and ethylene were spectroscopically pure (BOC Grade X) from Pyrex break-seal vessels used as received. Alcohols and acetaldehyde were AnalaR samples dried by distillation from molecular sieve 3A and purified by freeze-pump-thaw cycles.Anhydrous deuterated alcohols or methyl iodide were obtained from Prochem and used as received. Reference mass spectra of each reactant gas entering the q.1n.s. via a by- pass which did not expose it to the metal oxide were determined prior to each experiment. SURFACE REACTIONS BY DYNAMIC MASS SPECTROMETRY SAMPLE ILLUMINATION Metal oxidelquartz samples were exposed to 30ps duration light pulses emitted by an oxygen-quenched xenon flash tube dissipating 200 J electrical energy per flash. An elliptical reflector housing, enclosing the flash lamp and the cylindrical glass window of the vacuum system, was used to deliver emitted light to the sample. Substitution of potassium ferri- oxalate actinometer and appropriate filters at the position normally occupied by the metal oxide/quartz samples indicated that 2x lo'* photons in the wavelength range 300-360 nm were delivered to the sample per flash.These high photon fluxes were thought appropriate in view of quantum efficiencies 10-3-10-5 associated, in the literaturey8P l5 with photo- assisted processes at zinc oxide surfaces. Where possible in the course of this study parallel experiments were made with continuous illumination at very much lower photon fluxes than achieved with the flash tube. Such illuminations employed a 150 W mercury arc lamp, a 250 W compact mercury-xenon arc or a 450 W xenon arc lamp. FAST DETECTION CIRCUITRY Rapid response in the electron multiplier detector of the q.m.s.was desired in order to follow any sudden changes in gas composition within the vacuum system, as occasioned by incidence of the high-intensity 30ps light pulses onto the metal oxide. For this purpose, fast detection circuitry very similar to that normally employed in flash-photolysis apparatus was utilised (see fig. 2) except that the normal photo-multiplier was replaced by an electron multiplier and the normal monochromator by a quadrupole mass analyser in our equipment. Response time of the detector system, with output of the electron multiplier fed into an oscilloscope via a low-pass filter (to eliminate 4 or 2 MHz) was < 100 p and was not the slow step in response of the system to flash-initiated changes in gas composition. Time-of- flight of gas molecules from the flash-illuminated metal oxide/quartz section of the high vacuum system to the ion-source of the q.m.s.was the rate-limiting step, as could be demon- strated by generating gases of different molecular weights with light flashes and showing that rise-time of the signals with the m/e values of the corresponding molecular ions was proportional to m-3. RESULTS AND DISCUSSION A. FLASH-INITIATED PROCESSES OVER ZnO OR Ti02 SUBSTRATES AT N m-2 In view of literature reports that U.V. photons incident on ZnO can photolyse it to yield O2 and zinc, it was of particular interest to monitor peaks at m/e = 32 and 65J . CUNNINGHAM, E . FINN A N D N. SAMMAN 165 before, during and after arrival of a high-intensity light pulse onto a zinc oxide surface under the lowest residual pressure N m-2) attainable with the high vacuum system.The q.m.s. was therefore set to continuously monitor ions with m/e = 32 and time profiles were measured for change in ion current at m/e = 32 caused by arrival of the first pulse delivered to a “ fresh ” ZnO surface. The trace shown in fig. 3(a) was obtained by photographing a slow, appropriately-triggered oscilloscope sweep before, during and after this flash. It demonstrates a large rise in ion current at m/e = 32 initiated by the single flash and observed on a time scale of 1 s/div. No such transient was observed at m/e = 65 or any other m/e values except those for “ system ” transients (see below). Fig. 3(b) demonstrates the rise in ion current at m/e = 32 on a much faster time scan (20 ms/div) and compares it with output of the flash tube, as monitored by a photodiode and displayed on the upper trace of the storage oscilloscope.The delay evident in fig. 3(b) between the two traces originated from time-of-flight of product gas between flash reactor and q.m.s. The slow decay of signal intensity at m/e = 32 after the pulse, evident in both fig. 3(a) and 3(b) originates from pump-down rate of product gas by the system ion-pump. Comparison of the latter with rate of flash-initiated gas evolution indicated that the maximum reached in traces such as 3(a) is a good approximation (within 10 %) to the true maximum expected in the absence of continuous pumping by the ion pump. The latter condition was not normally used in the present study to avoid possible evolution of contaminants from the ion pump when switched off.TABLE 1 .-MAGNITUDE OF FLASH INJTIATED OXYGEN TRANSIENTS FROM ZnO AND Ti0 metal oxide Zn0-SF500 TiOz 2110-SP500 TiOe (a) flash incident though quartz (A > 200 nm) Po2/ APO2/ U O , / steady state initial rapid change a slow secondary process N m - 2 N m - 2 4 Nm-2 4 Q 3 x + 1 x 10-4 5~ 10-5 - < 3 x +2x 10-7 7~ 10-8 1 x 10-5 + 6 x - 4 ~ 2~ loe6 8~ 10-5 + 5 x - 3 x lo-’ 1 x (b) flash filtered by WR 38A (1>360 nm) 4 3 x < 1 x 10-5 + 1 x 10-5 4~ 10-5 + 1 x 10-5 3 x 10-5 9~ 10-5 + 1 x 10-5 4 3 x < 10- 2~ 10-5 +3x 7~ - 2 ~ lo-’ 4~ 1W5 4x lo-’ + 3 ~ 1 0 - ~ 9 ~ 1 0 - ~ - l ~ l O - ~ 3xW5 6x +3x 7x -2x 4x a corresponds to photodesorption with T+ ca. 150 ms ; b corresponds to photooxidation of the TiOz surface occurring at times 2-20 s.If traces such as 3(a) originated solely from photolysis of the ZnO surface with release of oxygen, similar yields of oxygen could be expected from successive pulses. Photograph C of fig. 3 demonstrates, however, that the magnitude of flash-initiated transients monitored at m/e = 32 decreased progressively to a limiting value when successive flashes were delivered to the same ZnO surface at 1 min intervals. How- ever, if the flash-illuminated ZnO surface was kept in the dark for an hour or longer166 SURFACE REACTIONS BY DY NAMlC MASS SPECTROMETRY between sequences of flashes, behaviour similar to that shown in fig. 3(c) could be repeated several times. A possible interpretation of this behaviour is that the ZnO surface slowly acquired a saturation coverage of depletively chemisorbed oxygen in the high vacuum system in the dark by interaction with residual oxygen pressure (which was not measurable but < Flash-initiated desorption of much of this cheniisorbed oxygen would account for the observed high transient at m/e = 32 by the first flash.The lower yield shown in fig. 3(c) for subsequent flashes delivered to ZnO surfaces at 1 min intervals might also be understood on this basis, since this short delay would not suffice to restore saturation coverage by 0 2 at oxygen pressures < An alternative to this interpretation of the large initial transient as photodesorption of chemisorbed oxygen from ZnO was that photolytic surface zinc progressively inhibited ZnO photolysis.No transient comparable in intensity or behaviour to that illustrated in fig. 3 was, however, detected when a TiO,/quartz sample was prepared, thermally treated and flash illuminated in conditions identical to those used for ZnO/quartz. The very small transient at m/e = 32 detectable from fresh TiO,/quartz samples was of constant size for successive flashes and was more similar to the low residual signal achieved with ZnO/quartz samples after several successive flashes. In the absence of Ti02 photolysis, transients corresponding to oxygen photosorption were readily measured and are summarized in table 1. Photo- sorption in 02/Zn0 systems was measurable using Wratten filter 38A to eliminate photolysis (cf. table 1). Although no detectable transient was observed at m/e = 32 in blank experiments involving flash illumination through the Kodial envelope in the absence of ZnO, TiO, or other metal oxide, sizeable " system " transients were then observed at m/e = 44, 28, 16 and 12.These are thought to originate by flash-initiated desorption of carbon oxides from internal surfaces of the vacuum system. Their occurrence constitutes a significant disadvantage and limitation of the system, since these " system " transients can seriously interfere at low system pressures with observations on other transients at mle = 44,28, 16 and 12. N m-2). N m-2. B . FLASH-INITIATED PROCESSES WITH PREADSORBED MOLECULES PRESENT ON ZnO OR TiO, CD,I/ZnO Recent publications 8* l6 from our laboratory have shown that, under continuous low-intensity U.V.illumination, the electron-attaching gases N20, CD31 or CH,CI undergo photoassisted change over ZnO, but not Ti02 surfaces. Products of the reported photodissociation of N20 to N2(g) or of CH31 and CH3C1 to methane were not suited to study with the flash-initiated dynamic mass spectrometer system, because their parent ion component (N; with m/e = 28 or CHZ with m/e = 16) coincided with system transients. Previous data had, however, indicated that CD3H was the major photoassisted product from CD,T over ZnO surfaces which bore large numbers of surface OH groups after pretreatment at 250-300°C. Consequently a flow of CDJI at pressure 10-3-10-s N m-2 was established at room temperature in the high vacuum system over ZnO surfaces pretreated at 250 or 300°C.These pressures were maintained constant for at least 1 h prior to flash-illuminating the surface, in order to establish an adsorbed layer of CD31 on ZnO. The q.m.s. was then set to detect any flash-initiated transients with m/e = 19, corresponding to fast photoassisted formation of CD,H. Fig. 4 presents three oscilloscope traces illus- trating the reproducible time-profile of small flash-initiated transients with m/e = 19 from the CD,I/ZnO system. For this experiment unreacted CD3T, together with0 50 > E - 25 0 20 GO 60 80 * 1 - 2 . 3 1 ; tlms tls (c) FIG. 3.-Oscilloscope trace recordings of time profiles of variations in ions with m/e = 32, as detected by the q.m.s. upon incidence of a high intensity U.V. flash upon ZnO in z’acuo. (a) Slow scan of transient at m/e = 32 initiated by first U.V.flash incident on fresh ZnO in z‘acuo of N m-2 (through a quartz envelope). (b) Fast scan of transient (lower trace) compared with time profile of flash intensity (upper trace). (c) Slow scans of transients at m/e = 32 generated by successive flashes of equal intensity on ZnO at 1 min intervals. Largest vertical displacement corresponds to first flash and deereases progressively with flash number. 5- 25 t 2 ’ 4 * . I O T 2 ’ 4 t l s t l S (a> (b) FIG. 4.-Oscilloscope traces recordings of slow scan time profiles of flash-initiated transients from CD31/Zn0 system exposed through Kodial to high intensity U.V. flashes. (Condensible vapours intercepted by liquid N2 baffle). (a) Transients at m/e = 19 observed with three successive flashes incident on CD,I/ZnO.(6) Transients at m/e = 32 observed with four successive flashes incident on [To face page 166 CD31/Zn0.0 5 0 > I00 4 5 3 0 15 mle FIG. 5.-Mass spectral distribution of ion fragments in steady-state spectra of methanol at 4 x N m-2 compared with that in mass histograms of ion fragments from flash-initiated changes in gas composition. ( a and a’) Oscilloscope trace recording for gaseous CH30H present above Ti02 in the dark (lower plot) compared with mass histograms from gases present 100 ms after flash (upper plot). (b and b’) Oscilloscope trace recording for gaseous CD30D present above TiOz in the dark (upper plot) compared with mass histogram of gases present 100 ms after flash (lower plot). dark FIG. 6.-Oscilloscope trace recordings of q.m.s.mass scan illustrating selective enhancement of product peaks by a flash delivered during scanning. (a) Mass scans of the region m/e = 22-15 with CDJOD flowing over Ti02. No flash was delivered during the “dark” scan (upper trace) but arrival of a flash at the indicated point of the lower trace demonstrates selective enhancement at m/e = 20 and m/e = 19, corresponding to flash-initiated production of CD3H4 and CD3H. (6) Mass scan of region m/e = 48-38 with C2H50H flowing over Ti02. Flash delivered at ca. m/e = 45 on the upper trace resulted in selective enhancement of peaks at m/e = 44 and 42 corresponding to flash-initiated enhancement of acetaldehyde relative to the spectrum of C2H50H in the dark (lower trace).Lo 3'2 25 l b 12 mle FIG.7.- Comparison of flash histograms from C2H50H/Zn0 and C2H50H/TiOL with each other and with C2H50H spectrum. (a) Mass distribution of flash-initiated transients from C2H50H/Ti02 at 4 x 10-4N m-2 partial pressure of C2H50H. (b) Oscilloscope trace recording of q.m.s mass spectrum of C2H50H at 4 X 10-4N m-' in similar conditions. (c) Mass distribution of flash- initiated transients from C2H50H/Zn0 at 4 x lo-" N m-2 of C2H,0H.J . CUNNINGHAM, E. FINN AND N. SAMMAN 167 any other gases condensible at 77 K, were collected on the metal surfaces of the liquid- nitrogen cooled baffle after leaving the reaction chamber and before arrival at the q.m.s. (cf. fig. 2). This precaution was necessary to avoid degradation of the electron multiplier response by adsorption of CD31.No measurable flash-initiated peaks were detected with q.m.s. set for m/e 17, 18, 20, 21 or any other mass number (except those corresponding to the “ oxygen ’’ transient or “ system ” transients) so that traces in fig. 4 demonstrate the ability of the q.m.s. system to detect selectively the fast photoassisted formation of the expected CD3H photoproduct. The time profile of its growth and decay in the high vacuum system is similar to that of the oxygen transient (cf. fig. 3 and 4). Such fast time profiles represent a valuable feature of the system since it thereby discriminates fast photoassisted product formation from slow surface assisted conversion, which is known to occur simultaneously, as per expression (I), in CDJZnO systems. photo /+CD,H(g) etc.CD31(g) f ICD,I(ads) . __ + ZnB-OH! (1) \ - - - + C D , H ( g ) , CD,(g) etc. dark Fig. 4(b) illustrates that presence of CD31 on the ZnO surface during the flash did not inhibit the oxygen transient, which is seen to be similar in magnitude and be- haviour to that observed at the fresh ZnO surface. This observation would be consistent with the upward band bending at the CD,I/ZnO interface illustrated in fig. l(c). As argued in the introduction this should enhance rather than depress surface disruption through hole-electron recombination in the surface layers. These results on the CD31/Zn0 system thus lend support to the importance of specific surface properties (H of surface hydroxyl groups was needed to produce Cl3,H) and of the bulk-related energy band description.METHANOL/TiO2 Methyl halides were not suitable single reactants for study over illuminated Ti0, surfaces, in view of reports that they did not undergo photoassisted conversion to methane or other products even after prolonged low intensity u.v.-illumination. Instead CH30H/Ti02 or CD,0D/Ti02 were exposed to the high-intensity light flashes in the dynamic mass spectrometer system. Flash-initiated transients with time-profiles very similar to those already illustrated in fig. 3 and 4 were observed at several m/e values. These are summarised in fig. 5(a) in the form of a ‘‘ flash histo- gram ” showing the maximum reached by flash-initiated transient (if any) at each m/e value from 1 to 50. Each histogram value in fig. 5(a) was obtained in a separate flash but, since the monitoring photodiode showed that output per flash was repro- ducible, the flash histogram in fig.5(a) corresponds to the mass spectrum of gases “ seen ” by the q.m.s. at times ca. 100 ms after each flash. The metal liquid-N, baffle between flash-reactor and q.m.s. was not cooled in these measurements on CH30H/Ti02. Consequently, alcohol reached the q.m.s. continuously after flowing over the TiO,/quartz sample and fig. 5(a’) displays a low resolution mass spectrum with steady state pressure of CH30H entering the ion source of the q.m.s. in the absence of any flash. Comparison of fig. 5(a) with 5(a’) shows that flash desorption of CH30H unchanged from the TiOz surface can account for observed flash enhance- ment of peaks with m/e = 32, 31 and 29, but not for the large flash-initiated peak at m/e = 30.Unlike the flash initiated peak at m/e = 28, which can be accounted for, in part, by a system transient, the large transients at m/e = 30 or 29 do not coincide with168 SURFACE REACTIONS BY DYNAMIC MASS SPECTROMETRY a system transient and instead indicate flash-initiated production of a new gas-phase product from the illuminated CH30H/Ti02 samples. Formaldehyde appeared the most probable cause of these transients. A system transient obscured the origin of flash-initiated transient observed at m/e = 16 but the flash-initiated transients at m/e = 15 and mle = 14 for the CH30H/Ti02 system appeared consistent with formation of methane as a major flash-initiated photoeffect in the CH30H/Ti02 system.Similar experiments were carried out with CD30D flowing over TiOJquartz samples in the hope that flash-assisted photoproduct peaks would not then be obscured by system transients. Results are summarised in fig. 5(b). The composite " stick " spectrum in fig. 5(b) summarises the extent to which magnitude of flash- initiated transients at various m/e values with 6 x N m-2 CD30D flowing over the TiO, surface exceeded flash-initiated transients measured in identical conditions but without CD30D present. From this plot it can be seen that major additional flash-initiated transients related to CD30D existed at m/e = 36, 34 and 30, consistent with flash-assisted desorption of methanol (CD30D) as also deduced above for the CH30H/Ti02 system. The large transients observed at mje = 35 and 33 indicate that the flash desorbed methanol exhibited extensive H/D exchange at two deuterium positions of the CD,OD originally adsorbed onto the hydroxylated rutile surface.Enhancement of flash-initiated transients at m/e = 20, 18, 16, 14 and 12 by presence of CD30D is also shown in fig. 5(b) and is consistent with fast photo- assisted formation of CD, from CD30D/Ti02. A partial mass analysis of the " burst " of methane gas arriving at the q.m.s. ca. 100 ms after a high intensity flash was achieved by simultaneously triggering the scope and a mass scan and then arranging that a flash was delivered to the CD30D/Ti02 sample 100 ms before the the q.m.s. would scan through m/e = 20. The lower trace of fig. 6(a) shows a mass scan takefi this way.Comparison with a steady state spectrum [upper oscilloscope trace in fig. 6(a)] shows that peaks at m/e = 20 and 19 are strongly enhanced by the flash, thus showing that CD4 and CD3H were enhanced in the flash-initiated burst of methane gas. Subsequent peaks with m/e = 17 to J 2 are not significantly enhanced, partially because the rate of pump-down of the system was rapid relative to the rate of scanning used for the lower spectra of fig. 6. Limitations on the rate of scanning and response time of the q.m.s. meant that this scanning mode of operation did not yield accurate ratios of ion current at various m/e values. However, the lower trace in fig. 6(a) not only provides excellent evidence for flash-initiated formation of CD4 and CD3H but also provides reassurance that the q.m.s.system retained normal behaviour and resolving power for methane arriving at the ion source as a flash- initiated pressure surge. The flash-histogram presented in fig. 5(b) for the CD3OD/TiO2 system allows accurate values of the relative abundance of parent and fragment ions (CD,H,,,)+ from flash-initiated methane to be determined. Occurrence of significant peaks at odd-numbered m/e values in the flash histogram for m/e values 20-12 is consistent with some H/D exchange during the flash-initiated processes leading to methane production from CD30D/Ti02. Since odd-numbered peaks in this region are still significantly lower than even-numbered peaks, it would appear that H/D exchange was slow during methane production. More efficient H/D exchange during fast photo- desorption of methanol is indicated by the large odd-numbered peaks at m/e = 35 and 33 in fig.5(b). The flash-initiated transient at m/e = 32 was much larger than flash-initiated O2 production from the fresh Ti02 sample and too large to be accounted for as flash-desorbed alcohol. Consequently, ethane, ethylene or formaldehyde photoproducts were considered as possible origins of this large transient but pre-J . CUNNINGHAM, E . FINN AND N. SAMMAN 169 liminary analysis indicates that the observed transients having m/e values 36-26 can best be accounted for by a combination of flash-desorption of methanol (accompanied by extensive H/D exchange) and photoassisted formation of formaldehyde. Indica- tion that these photoeffects were originated principally by U.V.photons came from the observation that insertion of a filter which excluded all U.V. photons from the CD,OD/ Ti02 sample eliminated all the transients related to methanol. The filter used, (Wratten a), transmitted visible photons 450-620 nm and i.r. photons to the sample, although at reduced intensity. Results to date on methanol/TiO, and their possible mechanistic implications may be summarised as follows : (a) Marked photoassisted enhancement of H/D exchange at room temperature occurred, involving two deuterium atoms of CD30D originally adsorbed, which suggests molecular rather than atomic exchange ; (b) results were consistent with the fast photoassisted production of formaldehyde which suggests that flash illumination conferred additional dehydrogenation capacity upon the rutile surface ; (c) fast photoassisted formation of methane CD4 occurred, which would be consistent with ability of the illuminated TiO, surface to abstract oxygen from methanol CD,OD, but formation of CD3H indicates that surface hydroxyl groups also contribute.ETHANOL/Ti02 OR ZllO Similar experiments were carried out with freshly-prepared ZnO/quartz or TiO, / quartz samples. Flash-illumination with ethanol flowing over them, revealed larger flash-initiated transients than for the methanol/TiO, systems described. Further- more, as illustrated in fig. 7, the principal transients in flash-histograms of these ethanol systems were grouped in three m/e regions : 46-40 ; 31-24 and 19-12, rather than in the 36-26,20-12 regions predominating for CD30D/Ti02 (cf.fig. 7 with fig. 5). Flash-initiated transients in the ethanol systems greatly exceeded " system transients " measured by flashing the ZnO or TiOz surface prior to admission of C2H50H. Thus these ethanol systems possessed the great advantage that " system transients " could be ignored or subtracted out from the observed " flash histograms " to yield corrected flash histograms. The histograms shown in fig. 7 have been corrected in this manner and should be representative of the fast photoassisted gas products deriving from presence of ethanol over the flash-illuminated metal oxide sample. Their comparison with a steady-state mass spectrum [cf. fig. 7(b)] of the gaseous alcohol present above the samples reveals significant differences: thus, in the m/e region 46-41, ions with m/e == 43 are most abundant in the flash histogram from C2H5OH/TiO2, whereas such ions are not present in the steady-state spectrum which features ions with m/e = 45 as most abundant in that region; a similar downward shift by two mass units is also apparent for most abundant ions (m/e = 29) of the flash histogram in the 31-26 region from C2H50H/Ti02 relative to most abundant ions (m/e = 21) in the steady-state alcohol spectrum.These and other differences between the flash histograms and steady-state alcohol mass spectrum make it clear that the flash-initiated transients in these systems cannot be accounted for simply by photodesorption of alcohol. Fig. 7 also reveals marked differences between the flash histograms of transients from C,H,OH/ZnQ and C2H50H/Ti02. Thus, for example, in the region m/e = 46-42, ions with mle = 42 are most abundant in the flash histogram from C2H50H/ ZnO rather than ions with m/e = 43 as for C,H,OH/TiO,.Likewise in the region m/e = 32-26, the two most abundant ions in the flash histogram from C,H,OH/ZnO occur at m/e = 27 and 25, rather than at m/e = 29 and 28 as from CzH50H/Ti02.170 SURFACE REACTIONS BY DYNAMIC MASS SPECTROMETRY On the basis of these and other differences evident in fig. 7 between corrected flash histograms involving the same reactant (ethanol) over different metal oxides, a convincing case can be made that the identity of the metal oxide substrate exerted an important influence on the flash-initiated transients. Our further attempts to deduce from corrected flash histograms the molecular identity of transients delivered to the gas-phase as a flash-initiated pressure surge over ethanol/metal oxide samples involved the following steps.STEP 1 .-Histograms further corrected for the contributions made by flash desorbed alcohol were obtained by assuming, arbitrarily, that the peak height in the flash histogram at m/e = 31 (the most abundant ion fragment from CzHSOH in the steady-state spectrum) could be used to calculate proportionately smaller contributions by ion fragments of ethanol for observed peak heights at other m/e values. Calculated contributions by flash desorbed alcohol were then subtracted from the histograms similar to those of fig. 7 to yield " ethanol-corrected " histograms as depicted in fig.8. mle FIG. 8.-Ethanol-corrected flash histograms and extent of fit achieved for indicated acetaldehyde/ ethylene ratios. (a) Flash-initiated transients with CzH50H flowing over a ZnO surface during flash through Kodial glass. Open peak heights have been corrected by subtracting flash-desorbed ethanol. Closed height is the fit, achieved for CH3CHO/CzH4 = 4/9. (b) Flash-initiated transients with C2H50H flowing over TiOzsurface during flash through Kodial. Open peak heights corrected for flash desorbed ethanol. Closed height is fit achieved for CH3CHO/C2H4 = 6/5. (c) Flash- initiated transients with no C2H50H in gas phase above a flash-illuminated ZnO surface which pre- viously had adsorbed C2H50H. (Kodial envelope). STEP 2.-Histograms as in fig. 8 were empirically analysed, by comparison with standard spectra, into probable contributory molecular species.Thus histograms from C2HSOH/TiO2 appeared consistent with acetaldehyde, methane and ethylene as major components of the flash-initiated pressure surge, whereas histograms from C2H,0H/Zn0 were consistent with just acetaldehyde and ethylene as the major components.J . CUNNINCHAM, E. FINN AND N. SAMMAN 171 STEP 3.-Reference steady-state spectra of probable molecular components were run on the q.m.s. system at pressures and instrumental settings closely similar to those obtaining during the flash histograms. Attempts were made to account quantitatively for ethanol-corrected histograms as a combination of these reference spectra in appropriate proportions.Fig. 8 illustrates the extent of agreement that has been achieved to date for C2H50H/Zn0 between calculated histogram peak heights (shaded sections) and total histogram peak heights on the basis of 4:9 acetaldehyde-to-ethylene ratio C2H50H pressure of 4 x N m2. Agreement is adequate at most m/e values (except where system transients are known to contribute and at m/e = 26,15 and 29) to support this preliminary identification. For C2H50H/ TiO, preliminary analysis in terms of just acetaldehyde and ethylene, as in fig. 9(a), gives a less satisfactory fit and unexplained peaks at m/e = 30, 15 and 13 suggest that ethane also contributes to photoproduct in this system. Ions with m/e = 44 and 43 are most abundant in mass spectra of acetaldehyde, so that confirmation of flash-initiated acetaldehyde formation could be sought by scanning through those m/e values ca.100 ms after a high intensity light pulse was delivered to a C2H50H/Ti02 sampIe. The oscilloscope traces reproduced in fig. 6(b) illustrates the enhancement of m/e = 44 and 43 attributable to presence of acetalde- hyde 100 ms after a flash incident on Ti02 with 4 x N m-2 of C2H50H present above it. The validity of other aspects of preliminary identification just suggested, viz. identity of the hydrocarbon products and their abundance relative to acetaldehyde, is being checked by experiments with C2D50D over both oxides. Mass histograms of gases released by flash-illuminating a CH,CHO/ZnO sample demonstrated efficient photodesorption of acetaldehyde. This raises the possibility that selective photodesorption allowed acetaldehyde to make disproportionately large contributions to gas-release from flash-illuminated ethanol/metal oxide interfaces.The full mechanistic implications of present results towards defining the relative importance of various processes at these illuminated interfaces cannot, therefore, be formulated until distortions introduced by selective desorption are evaluated. Points emerging from preliminary analysis include : (i) if dehydrogenation and dehydration of ethanol/ZnO are assumed to proceed by routes depicted by Wolken- shtein, then comparable losses of H2 or H20 are consistent with availability of electron-hole pairs at the interface; (ii) excess photodehydration is consistent with added availability of holes ; (iii) hydrogen-containing surface sites contribute H/D exchange for flash desorbed C2D50H but not for production of C2D4 or CD,CDO.ROLE OF ADSORBED ALCOHOLS Attempts were made to distinguish the role of adsorbed alcohol from possible fast interactions of gaseous alcohol with the flash-activated surface. In the experimental investigation of this point ethanol or methanol was first adsorbed on to a Ti02 or ZnO sample at pressures ca. N m-2 for 0.5 to 16 h, after which times gas-phase alcohol (and any other gases) was mass-analysed and pumped out of the system. Mass analysis of alcohol thus stored over TiO, or ZnO in the dark provided evidence for slight " dark reactions " but any gaseous product of such reactions was removed with alcohol by prolonged pumping until system pressure fell to ca.N m-2. Exposure of these " alcohol pre-exposed " surfaces to flash illumination then produced transients giving flash histograms similar in some respects to those observed from the same surface when alcohol was previously flowing over it at pressures 10-5-10-3 N m-2 during flashing. Comparison of fig. 8(a) with 8(c) shows similarities in the m/e regions 12-18 and 24-30 between flash histograms from C2H,0H/Zn0 in flow and172 SURFACE REACTIONS BY DYNAMIC MASS SPECTROMETRY no-flow conditions. Our preliminary analysis of the flash histograms in fig. 8(a) and 8(c) indicate that ethylene product (from photodehydration) over ff ash-illuminated ZnO was largely unaffected by pumping away gas-phase C2H50H. Removal of gas phase CZH,OH did, however, sharply reduce acetaldehyde product [from photo- dehydrogenation e.g. cf.small peaks for CHO+ in fig. 8(c)] so that the acetaldehyde/ ethylene ratio fell to 1/8 from the value of 4/9 observed with C2H50H present over the same ZnO surface during flashes. One other difference between 8(c) and 8(a), which was also noted in experiments with methanol/TiO, is the enhancement of COz product in flash histograms obtained with preadsorbed alcohol. Parallel studies of steady state adsorption of C2H50H onto ZnO indicated that nett coverage of the ZnO surface did not exceed occupancy > 1 site per 500 surface sites in the conditions of the flash experiments. Results such as those in fig. 8(c) indicate, however, that even this sparse coverage should be further subdivided into: set I, which undergo photodesorption ; set 11 which undergo photoassisted conversion to ethylene and are most numerous ; and set I11 which are converted to acetaldehyde by U.V.flashes. PHOTOASSISTED VERSUS THERMALLY-ASSISTED PROCESSES It was essential to check by whatever means possible whether observed flash- initiated effects originated from surface heating or from true photoassisted processes made possible only while the metal-oxide or the adsorbed species existed in an electronically excited state. One, admittedly crude, method for examining this point was based on the argument that i.r. radiation should promote thermal heating whereas only U.V. could promote the metal oxide or adsorbed molecules into mle FIG. 9.-Flash histograms obtained through quartz from ethanollZn0 systems.(a) Histogram of flash- initiated photo-products (corrected for photodesorbed C2H50H) observed with 4 x N rnd2 C2H50H in gas phase above ZnO (total peak heights). (6) Observed without CzH50H in gas phase but with preadsorbed C2H50H. Blacked in peak heights correspond to fraction of total peak heights explicable by indicated combination of steady state spectra of CzH4 and CH3CH0 (see 'stick spectra' at top as read by q.m.s,).J . CUNNINGHAM, E. FINN AND N. SAMMAN 173 electronically excited states. Filters which cut off the U.V. component of the flash tube while admitting the visible and i.r. (Wratten No. 4 or No. 40) were therefore inserted to observe their effect and in all cases they reduced to zero the oxygen- related or alcohol-related transients. Additional precision and reliability was sought for such experiments by comparing results observed with Wratten No.40 and Wratten No. 38A, since the major difference between radiation transmitted by these filters was that only the latter transmitted any near-u.v. 380-400 nm. This near-u.v. was apparently essential for producing transients in the pre-adsorbed ethanol/TiO, system, since acetaldehyde transients at mle = 43 were observed with filter No. 38A but not with filter No. 40. Further indications of the importance of U.V. rather than i.r. photons for the observed flash-initiated transients were obtained when a quartz envelope (transmitting to 200nm) was substituted for the Kodial envelope (transmission limit 300 nm) usually employed.This resulted in an approximate doubling of the flash-initiated transients, as may be seen by comparing fig. 3(a) with 8(a). Such comparison also reveals greater complexity in the m/e range 12-18 as would be consistent with more extensive photolysis for light transmitted through the quartz envelope. Thus an acetaldehyde/ethylene ratio of unity [similar to those used in fig. 8(a)] is clearly inadequate to account for the transients in the m/e region 12-18 in fig. 9(a). Fig. 9(b) illustrates that complexity of these flash-initiated transients was present also for flash histograms obtained with only pre-adsorbed alcohol present on the ZnO surface. Data on the nature and rapidity of thermally assisted surface reactions in the alcohol-metal oxide systems were also sought by using the high vacuum system as a flowing catalytic reactor at pressures ca.4 x N m-2. In these conditions maximum contact times of alcohol with the metal surface were ca. 0.1 s and so comparable to the time-scale of flash-initiated processes. No appreciable de- hydrogenation or dehydration occurred even when a Ti02 substrate was heated to 258"C, making it improbable that fiash-initiated production of aldehyde or ethylene would result from high surface temperatures in the flash. Methane was produced from CD30D/Ti02 at temperatures > 150°C and H/D exchange occurred in two alcohol positions so that thermally-assisted processes might contribute to an unknown extent to methane formation and H/D exchange in flash-illuminated CD30D/Ti0, systems.There is, however, no evidence that they contribute to formation of acetaldehyde or ethylene. CONCLUSION Even in its present form the dynamic mass technique described has considerable utility for the examination of fast photoassisted surface reactions. Results for the CD,I/ZnO system illustrated its ability for discriminating fast photoassisted surface reaction from slow thermal reactions. Results with alcohol/metal oxide systems are consistent with the need to take into account both the bulk related and specifically surface properties of the metal oxide surfaces. There are significant differences between flash-initiated processes over ZnO and TiO, surfaces and together with wavelength dependence of the results these differences make it clear that the majority of results observed here originate from true photoassisted surface processes and not from thermal heating during the flash. Funds and equipment for this research have been supplied in part by the University and in part by a research grant from the National Science Council of Ireland. The authors are grateful for this support and also for the assistance of Dr. A. L. Penny174 SURFACE REACTIONS BY DYNAMIC MASS SPECTROMETRY in planning and setting up the equipment. We are also grateful to staff of the Department of Electronics and Electrical Engineering at Liverpool University for their assistance with the q.m.s. and to Mr. C. Bell for his help with the electronics. ' R. D. Iyengar and M. Codell, Ado. Colloid Interface Sci., 1972,3, 365. (a) J. Schreider and A. Rauber, Z. Naturforsch., 1961, A16, 712 ; (b) P. H. Kasai, Phys. Rev., 1963,130,989; (c) M. Codell, H. Gisser and R. D. Iyengar, Canad. J. Chem., 1968, 46, 2239. M. J. Duck and R. L. Nelson, J. C. S. Faraday I, 1974,70,436. T. J. Gray, C. C. McCain and N. G. Masse, J. Phys. Chem., 1959, 63, 472. J. B. Goodenough, Progr. Solid State Chem., 1971, 5, 344. J. M. Smith and W. E. Vehse, Phys. Letters, 1970, A31, 147. ' J. Hockey, Disc. Faraday SOC., 1971, 52. ' J. Cunningham and A. L. Penny, J. Phys. Chem., 1972,76,2353. (a) G. Heiland and W. Bauer, J. Phys. Chem. Solids, 1971, 32, 2605 ; (b) G. Heiland and H. Luth, Phys. Stat. Solid., 1972,14,573 ; (c) G. Heiland, W. Bauer and M. Newhaus, Photochem. Photobiol., 1972, 16, 315. P. B. Weisz, J. Phys. Chem., 1953, 21, 1531. l1 (a) F. F. Volkenshtein, The Electronic Theory of Catalysis on Semiconductors (Pergamon, Oxford, 1963) ; (b) F. F. Volkenshtein, Adv. Catulysis, 1973,23, 157. l2 (a) J. Cunningham, J . J. Kelly and A. L. Penny, J. Phys. Chem., 1970,74,1992 ; (b) 1971,75,617. (a) T. Kwan, Symp. Electronic Phenomena in Chemisorption and Catalysis oii Semiconductors, ed. X . Hauffe and X. Welkenstein, (Walter de Gruyter, Berlin, 1969), p. 185 ; (6) L. L. Basov, Y. P. Solonitsyn, A. N. Terenin, Doklady Akad. Nauk S.S.S.R., 1965, 164, 122. l4 G. Heiland, E. Mollwo and F. Stockmann in Solid State Phys., 1959, 8, 193. l 5 R. J. Collins and D. G. Thames, Phys. Reu., 1958, 112, 388. l6 J. Cunningham and A. L. Penny, J. Phys. Chem., 1974,78, 870. l7 (a) D. Pooley, J. Phys. (Proc. Phys. SOC. London), 1968,1,323 ; (6) F. J. Keller and F. W. Patten, Solid State Comm., 1969, 7, 1603.

ISSN:0301-7249    

DOI:10.1039/DC9745800160

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

GENERAL DISC USSlO N Papers 10, 11, 12 Dr. R. I. Bickley (Bradford) said: Although Steinbach’s results show that the rate of photo-oxidation of carbon monoxide is invariant during the interval of temperature from 400 to 5OO0C, it is not made clear why the reaction does not take place below 400°C. Since the two observations are apparently anomalous would he clarify the situation ? and I have carried out with oxidising films of evaporated zinc show that no photocatalysis occurred when an equimolar mixture of carbon monoxide and oxygen was exposed to the surface which was being illu- minated with ultra-violet light (A 2 300nm) within the temperature range 78-420 K, although oxygen photo-adsorption occurs rapidly to approximately 0.25 monolayer. Slow thermal catalysis is just detectable at 420 K ; carbon dioxide being produced at a rate of not greater than 1.7 x mol m-2 s-l.Such experiments suggest that the oxygen species produced on zinc oxide at lower temperatures under illumination do not participate in the photo-oxidation of carbon monoxide and that a dissociated species of oxygen (Oads or O&) rather than O;ads is required before the catalytic reaction can proceed. Experiments which Metcalfe Prof. G. Heiland (Aaclzeiz) said: I would comment on the photodesorption of atomic oxygen from ZnO in the absence of carbon monoxide. (1) I have studied photodesorption from ZnO together with W. Heinzel using an omegatron and ultra-high vacuum system. Ultra-violet light at room temperature caused desorption of atomic oxygen but not of molecular oxygen2 This observation is in agreement with the present paper.(2) (a) Optical energy.-Photodesorption of oxygen from ZnO results in an in- creased density of electrons in the conduction band near the surface. Therefore photodesorption has often been studied by observing the slow, steady and irreversible rise of surface conductivity during irradiation in a vacuum. The spectral distribution shows a threshold at 3 eV and a continuous increase towards higher photon energies, clearly distinct from photoconductivity. (b) Thermal activation energy.-From studies of surface luminescence an activation energy of 240meV has been estimated by H e ~ h t . ~ This is in agreement with our observation, that photodesorption comes to a complete stop at low temperature and is considerably accelerated above room temperature.Therefore thermal processes are connected here with the so-called photoprocesses. (3) Photodesorption or photolysis causing a rise of surface conductivity can be decreased by small amounts of impurity. I would mention copper,’ lithium and iron cyanide.’ The effect can amount to several orders of magnitude. L. P. Metcalfe, Ph.D. Thesis (University of Bradford, 1970). quoted in : G. Heiland, J. Phys. Chem. Solids, 1961, 22, 227. G. Heiland, E. Mollwo and F. Stockmann, Solid State Phys., 1959, 8, 281, fig. 75. H. Hecht, Thesis (Erlangen, 1971). H. Weiss, 2. Phys., 1952, 132, 338, table I . R. J. Collins and D. G . Thomas, P/iys. Ref,., 1958, 112, 358, fig. 6 and 9. S. R. Morrison, J. Vac. Sci. Tech., 1970, 7, 84.175176 GENERAL DISCUSSION Finally I would ask a question: the production of atomic oxygen might be an important step in the catalytic process studied. Could the parameters influencing photodesorption of oxygen also be important for the photocatalytic oxidation of carbon monoxide ? Dr. W. Hirschwald and Dr. E. Thull (Berlin) said: In addition to Harborth and Steinbach's results 1 give some experimental details of our own work, concerned with the photoactivation of oxygen and photolysis on zinc oxide surfaces. In an experimental setup, almost identical to that used by Steinbach, but without the liquid helium-pump, thin layers of ZnO deposited on a platinum support were irradiated with band-gap light of 3.2 eV photon energy at temperatures between 25 and 900°C following a 30h bakeout at 200°C.Thermal- and photo-desorption products were monitored by a micromass 6 mass spectrometer. Torr) for more than 2h at 500"C, no photolysis was observed with a 200 W high-pressure mercury lamp; thermolysis only became measureable above 700°C. (2) On oxygen-treated layers (15 h, 720°C, 100 Torr 0,) thermolysis and photo- lysis according to starts at about 350°C as shown by the appearance of a zinc peak corresponding to an increasing l60 peak. (3) Atomic oxygen is more abundant than the molecular O2 by a factor of about 10 in the initial stage of photo- and thermal-decomposition at 5OO0C, in contrast to Steinbach's findings, where atomic oxygen only in the photoprocess was detected. (1) On ZnO-layers treated in u.h.v. (< ZnO(s)+Zn(g)+O(g) (1) 50 1 initial state 40 20 10 steady state 10 50 too 770 1050 1100 1150 tlmin TlK (a> (b) FIG. 1 .-Plot of E.Izn [I = f ( t ) ] against (a) time (at 770 K), (b) absolute temperature. This is easily explained by the different pretreatments used, as the non-stationary behaviour created by our oxygen treatment vanishes during Steinbach's outgassing process at 550°C for several hours. In fig. 1 (a) the intensity of the zinc peak is plotted against time at a constant temperature of 500°C. The dependence shown in fig. 1 reveals decreasing thermolysis combined withGENERAL DISCUSSION 177 decreasing photolysis. Going to higher temperatures (> 700°C) " normal " steady- state dissociative sublimation is observed yielding an activation energy of about 78 kcal mol-l as observed earlier with vacuum-thermo-gravimetric methods [Fig I(b) and 21.By repeating the oxygen treatment low temperature thermolysis and photolysis could be reproduced several times for the same sample. From our results we conclude the following. (1) Thermal- and photo-desorption of oxygen from oxygen-treated zinc oxide surfaces at temperatures higher than 300°C is mainly a consequence of decomposing the solid. (2) Oxygen treatment extracts lo2 KIT FIG. 2.-Plot of In Z against reciprocal temperature. interstitial zinc or oxygen vacancies from the bulk (or even creates zinc vacancies), yielding a nearly stoichiometric surface layer. This layer formation was directly observed by scanning electron microscopy on the (0001) and (OOOi) basal planes.Extraction of interstitial zinc or oxygen vacancies was followed by simultaneous conductivity measurements. (3) The nearly stoichiometric (oxygen-rich) layer is much more reactive than the steady-state (zinc rich) underlying solid, as demonstrated by our low temperature (< 600°C) thermal- and photo-decomposition and first pointed out qualitatively by Faivre.2 As soon as the layer is completely decomposed, zinc and oxygen intensities go down to the background level below 700°C. As the activation energy of high temperature steady-state decomposition is about equal to the band-gap energy of 78 kcal mol-l (3.4 eV), a possible explanation for the observed behaviour is given by the following reaction scheme regarding bond excitation by electron transfer as the rate limiting step : Iiv-,o+e ki kz k3 O2-W + @-,OW (pl~otolysis) 1 O-@) + 0 +O(g) 0-(s) + @+02-(s) (recombination) from which the rate of photodecomposition can be deduced as d[O(g)l/dt = d[Zn(g)l/dt = k , k2 I 0 l 2 / k [03+k, [@I.W. Hirschwald and F. Stoh, 2. phys. Chem. (Frankfurt), 1972, 77, 21. Faivre, Ann. Chim., 1944, 19, 58.178 GENERAL DISCUSSION From eqn (2) it follows that at low electron concentrations as existing in the nearly stoichiometric layers the rate of photogeneration of oxygen should be proportional to the hole concentration (- light intensity). At high electron concentrations in zinc rich layers some orders of magnitude larger than the stationary concentration of photoproduced holes, photodecomposition decreases to very low rates due to en- hanced Further * support to the 350 400 450 500 A/nm FIG. 3.-Plot of log(Zo/f) against wavelength of irradiation at various temperatures.' statement, that oxygen treated samples behave differently from vacuum treated ones with respect to photoactivation, zinc abundance and lattice parameters is given by the work of Schelfaut and Maenhout van der Vorst,' showing that optical absorption in the band-gap region is increased by oxygen treatment (fig.3) and by the investi- gations of Allsopp and Roberts,2 concerned with zinc abundance and by Ciinino's work on the lattice parameters of thermally treated ZnO. The last two sets of data combined (fig. 4) show a decreasing zinc abundance with increasing lattice constant, probably due to the oxidation of interstitial zinc.Some further kinetic investigations on reduction of (0001)-, (0001)- and (1010)- faces of zinc oxide with CO give evidence that true chemisorption of oxygen (uptake of surplus oxygen) is favoured on the (000T)-oxygen face.4 Activation energy and pressure dependence (reaction order) for CO-oxidation by surface oxygen is found to be different for the three faces, as would be expected, if oxygen activation is the limiting step. The fact that no photodecomposition is observed below about 350°C is explained F. Schelfaut and W. Maenhout van der Vorst, Pliotogr. Korresp. 8. Sonderheft, 1966, p. 19. J. Allsopp and J. P. Roberts, Trans. Faraday Soc., 1959, 55, 1386. A. Ciniino, G. Mazzone and P. Porta, Z. Phys. C h e ~ . (Frankfurt), 1964, 41, 154. M. Grunze and W.Hirschwald, J . Vac. Sri. Tech., 1974, 11, 424 ; Proc. 12th Conf. Vacuum Microbalance Techniques, Lyon (France), Sept. 1974, to be published.GENERAL DISCUSSION 179 by the high activation energy of 32 kcal mol-1* necessary for zinc desorption. Hence a very small loss of oxygen from the surface causes zinc enrichment at low temperatures and the recombination mechanism becomes prevalent. Again, support for this concept is given by the fact that CO-oxidation by surface oxygen exhibits a low-temperature mechanism, governed by zinc desorption. 0 200 400 600 800 1000 1200 TIT FIG. 4.-Plot of zinc abundance and lattice parameter a against ternperat~re.~, The fact that Steinbach and Harborth detect photolysis and desorption of atomic oxygen on ultra-high vacuum treated (zinc rich) zinc oxide layers (where we do not), is probably due to higher sensitivity (lower background level) and to the higher intensity of the u.v.-lamp used by Steinbach (1000 W compared to 200 W in our experiments).Prof. F. Steinback and Mr. R. Harborth (Hamburg) said: The following results confirm the understanding of the desorption of 0 and C 0 2 as true photo-processes and answer the questions raised. In lock-in experiments using modulated light of the full mercury arc, only 0, C 0 2 and Zn were photo-generated; no response was found for CO, H20, H2 and €3. Also, no response was found for 02, though, by increasing the sensitivity of the system, very weak photo-desorption, at the present state of sensitivity hidden in the noise of the background, might be found. The photo-generation of 0, C 0 2 and Zn was observed most clearly in the frequency range between 1 and 20 Hz, the best signal-to-noise ratio being obtained in the region of 2 to 5 Hz.The photo-generation of 0 and C 0 2 was most distinctly observed at temperatures above 400°C ; the photo-generation of Zn was observed only at temperatures of about 500°C. The emission of Zn from the surface into the vacuum was about lo3 times smaller than the generation of 0 ; however, Zn emission was investigated in a vacuum of 1 x 1 O-lo Torr, whereas 5 x 1 O-'O Torr of flowing O2 was present during the photo generation of 0. 1. Lock-in experiments with the full mercury arc. However, this seems to be due mainly to the lock-in system used. E. V. Bolshun and 1. A. Myasnikov, Russ.J. Phys. Clrem., 1973, 47, 263.180 GENERAL DISCUSSION The conditions for the observation of the various effects were chosen to allow the most significant observation, i.e., to obtain the highest signal-to-background ratio. With increasing sensitivity of the experiments, even at temperatures lower than 400°C photo-desorption of 0 might be observed. For equal chopping frequencies in the range from 1 to 20 Hz, the phase shift of the photo-desorption of 0 was different from the phase shift of the photo-generation of CO, ; this finding indicates different rates for the two processes. While the absence of photo-desorption of 0 in the presence of CO indicates that the photo-breaking of the ZnO bond is a step involved in both processes (photo-desorption of 0 as well as photo-generation of CO,) the different reaction times show that the photo-breaking of the ZnO bond is influenced by the further fate of the 0 atom, whether it is emitted into the vacuum or incorporated into CO,.So far, only the formation of CO, has been investigated using various sections of the spectrum of the mercury arc for illumination. Setting the slit width of the mono- chromator at 200 nm, three sections of the mercury spectrum were separated: (a) the lines from 275 to 366 nm, (b) the three lines at 366, 407 and 435 nm and (c) the two lines at 546 and 580nm. Using one of the two energy-rich line groups for illumination of the ZnO crystal, photo-generation of C 0 2 was observed as with the full arc. No photo-generation of CO, was observed when the ZnO crystal was illuminated with the two lines at 546 and 580 nm only.By these experiments, de- sorption or reaction due to adsorption of infrared radiation or due to phonon processes is clearly ruled out. The threshold energy for the formation of C 0 2 must be higher than the energy of the line at 546 nm, i.e., 2.28 eV. 2. Lock-in experiments with different photon energies. Dr. W. Bauer and Prof. A. Hausmann (Aachen) said : Most of the results are obtained with Li-doped ZnO using a rather high dopant concentration of about 1 %. Incor- poration of a Li-ion in ZnO gives rise to a strong local field, resulting in a highly distorted crystal field for the Li-concentrations used, to which e.s.r. is very sensitive. We doubt whether the results can be related to pure ZnO.ZnO-crystals exhibit a metal-excess nonstoichiometry, leading to well-conducting samples. This conductivity may be due to Zni (interstitial Zn) donors, as mentioned in the paper. However, the Vo-centre, an oxygen vacancy, may also act as a donor There is even more evidence for the latter : e.g., chanelling measurements and oxygen influence on electrical conductivity.2 Is it possible to take the Vo-centre into con- sideration for the interpretation ? The interpretation of the results obtained with u.v.-irradiation is the formation of Vz, centres. This model is not consistent with the known facts of u.v.-influence on surface-conductivity. The g = 2.0130 centre has been given a different interpretation by Galland et aL4 Dr. A. J. Ten& (AERE, Harwell) (communicated) : In the paper by Haber et al., e.s.r.signals at g = 2.0130 and 2. 0145 are attributed to hole centres involving an oxide ion adjacent to a cation vacancy, i.e. an 0- ion. A consideration of the energy levels of the 0- ion indicates that the degeneracy of the p functions (p:p;pz) will be lifted by the presence of a tetragonal or lower symmetry crystal field so that p z lies B. R. Appleton and L. C. Feldman, J. Phys. Chem. Solids, 1972,33, 507. C . Bogner, J. Phys. Chem. Solids, 1961, 19, 235. W. Heiland, E. Mollwo and F. Stockman, Solid State Phys., 1959, 8, 191. D. Galland and A. Herve, Solid State Comm., 1974, 14, 953.GENERAL DISCUSSION 181 above p x and pY.l Such an effect would be expected from 8 neighbouring vacancy or from the surface and would give rise to an e.s.r.spectrum of axial or lower symmetry. With Kibblewhite we have carried out ultra-violet irradiation of nominally pure ZnO powders in uacuu and obtained an e.s.r. signal with gl = 2.0246 and 911 = 2.0043. This g tensor agrees well with previous work on MgO where 0- has been positively identified and agrees with similar work by Wong et aL3 The signal could be destroyed by oxygen or CO indicating that it was near the surface. A signal at g = 2.0139 was also observed in agreement with the results of Haber et al. but the symmetric single line suggests that it is probably not associated with 0- but with some other species, for example, 0 3 , with some rotational freedom. Prof. W. Hirschwald (Berlin) said: Evidence for the existence of zinc vacancies in zinc oxide has been reported by Seitz, Pinter and Hirthe from thermo-luminescence investigations who applied e.s.r.techniques. The latter authors identify their signals, with multiplicity three, one as a hole trapped on one of the three basal 02- ions in the oxygen tetrahedron surrounding the vacancy and one as a hole trapped on the 02- ion adjacent to the vacancy along the c-axis. This offers an energetically favourable pathway for activation of surface oxygen. Regarding (Zn2+ 02-),,1id+ 0- @+(Zn+O-)solid as the first step of bond-breaking, requiring an energy of 3.2 eV (band gap). This energy can be split up in to smaller amounts by intermediate levels as, for instance, the zinc vacancy, located in the forbidden gap : and recently by Galland and Herve, creating a hole in the valence band of oxygen by electron trapping6 This could give rise to a low activation energy of decomposition of the solid if oxygen pretreatment is able to create zinc vacancies (see our comment to Harborth and Steinbach’s paper).Dr. W. Bauer (Aachen) said: Photo-assisted release of O2 from a ZnO-surface is reported by Cunningham et al., whereas Steinbach and Harborth could not detect 02, but atomic oxygen only, as Heiland also reported in his contribution to the dis- cussion of Steinbach’s paper. Is there an explanation for this difference? Prof. J. Cunningham (Cork) said: I would make two points relating to oxygen evolution from ZnO surfaces. The first point relates to Steinbach’s results and to the valuable comments by Hirschwald which support Steinbach’s conclusion that atomic A.J. Tench, T. Lawson and J . Kibblewhite, J.C.S. Faraday I, 1972, 68, 1169. N. B. Wong and 5. H. Lunsford, J. Chern. Phys., 1971,55,3007. N. B. Wong, Y . Ben Taarit and J. H. Lunsford, J. Chern. Phys., 1974, 60,2148. M. A. Seitz, W. F. Pinter and W. M. Hirthe Materials Res. Bull., 1971, 6, 275. D. Galland and A. Herve, Solid State Comm., 1974, 14,953. F. Schelfaut and W. Maenhout van der Vorst, Photogr. Korresp. 8. Sonderhef, “ Elektro- photographie ”, 1966, 19.182 GENERAL DISCUSSION oxygen, but not molecular oxygen, is evolved from ZnO surfaces at temperatures > 350°C. All these observations emphasise the necessity of taking into account possible reactions of “ oxygen-atom like ” intermediates on ZnO surfaces at elevated temperatures.Their observation of atomic oxygen as the major species released to the gas phase in these conditions contrasts with results of our dynamic mass spectro- meter studies demonstrating photo-assisted release of molecular oxygen from flash- illuminated ZnO surfaces at room temperature. I would suggest that the apparent contrast between these results arises because “ oxygen-atom ” species produced thermally or photolytically on the ZnO surface do not escape efficiently except at temperatures > 350°C. Heiland has already commented on earlier work by Hecht indicating that photo-assisted release of atomic oxygen from ZnO requires an acti- vation energy, albeit a rather small one of 0.24 eV. Existence of an activation energy for oxygen-atom desorption would be expected to affect our observations at room temperature rather more strongly than observations at > 350°C by other workers.Retention of photolytically-produced atomic oxygen on the ZnO surface at room temperature (but not at > 350OC) and its reaction thereon to produce molecular oxygen which then desorbed, could account for the appearance of molecular oxygen as a product in our system. The second point concerns the large changes in surface processes on ZnO which apparently result from U.V. photolysis in vacuo. In our studies this was indicated by the large and progressive decline in amount of molecular oxygen released into the gas phase by successive flashes delivered at 1 min intervals (see fig. 3 of our paper). Likewise Heiland has remarked that the mobility of electrons is adversely affected by photolysis of the ZnO surface.It appears that a central question concerning such effects is the relative importance of various surface defects, such as zinc interstitials, zinc vacancies or oxygen vacancies, in bringing about these changes. I, therefore, welcome the new insight which Haber’s e.s.r. results gives into conditions favouring formation of zinc vacancies at the surface of lithium doped ZnO. Prof. M. W. Roberts (Bradford) said: It is interesting that Cunningham reports CH, as one of the products of the interaction of CH,OH with Ti02. It may be useful to investigate the decomposition of partially deuterated methanol, e.g., CDsOH (rather than CD30D as reported) since there is evidence (with Ni for example) that methane can arise entirely from the CH3 part of the methanol molecule. Prof.J. Cunningham (Cork) said: An examination of the decomposition of partially deuterated methanol, CD30H, over Ti02 as Roberts suggest, should indeed be valuable in checking whether CD, can form entirely from the CD3 part of CD30H. In published studies of the production of methanes from CD31 over ZnO surfaces 2(a) our detection of CD, in comparable amounts to CD3H showed that methane for- mation was possible over ZnO using only methyl hydrogen atoms. We were, however, unsuccessful in attempts to produce CD4 from CD3T/TiQ2 systems treated in similar manner. 2(h) Dr. T. B. Grimley (Liverpool) said : Photo-desorption is important theoretically because it is one of the simplest molecular rate processes resulting from the interaction of photons with surfaces.There are apparently some interesting theoretical problems to be solved before an expression for the cross section can be derived, problems Roberts and Stewart in Clzernisorptiorz and Catalysis, ed. P. Hepple (Institute of Petroleum, J. Cunningham and A. L. Penny, J. Phys. Cliem., (a) 1972, 76,2353 ; (6) 1974,78,870. 1970).GENERAL DISCUSSION 183 connected with the continuous spectrum of potential energy curves for adsorption by metals, or with the existence of a reaction coordinate for example, but assuming that these problems can be solved, progress will hinge on the efforts now being made to develop the theory of chemisorption on extended substrates to a point where meaning- ful potential energy curves can be calculated.On the other hand, there is very little experimental information to provide the theoreticians with guidance. CO on W (perhaps also on Ni) is, I believe, the only molecule known to be photo-desorbed from any metal. Prof. D. Menzel (Munich) said: The papers presented on photodesorption all concern adsorbates on semiconductors, where the primary step of light absorption probably takes place predominantly in a layer of a few hundred A inside the solid, and the transport of the excitation to the surface occurs in a second step. A more direct correlation with the discussion presented in section A would be possible for photodesorption from metals where it is most likely that the primary process of light absorption must take place in the adsorption layer proper.Several authors have argued that all observed effects are either due to thermal effects or can be traced back to the existence of a thin oxide layer on the metal surface. There is at least one case, however, of true photodesorption from a metal, namely, for CO on W ; the cross sections found were very small, though. There seem to be two main reasons for this: first, the light absorption cross sections of adsorbate complexes must be small, and second the same recapture mechanism of excited species which has been shown to operate in electron impact desorption 9 most probably reduces the effect further by up to several orders of magnitude. It is difficult to study, therefore, although it would be of considerable interest in connection with other surface excitation pro- cesses, as became obvious from the introductory remarks of Grimley. In one respect the small cross sections are consoling: it appears unlikely that any disturbances of photoelectron spectroscopic investigations of adsorbates by photodesorption would have to be expected, at least for metals, unlike the conditions found for electron spectroscopy using electrons as primary particles. Under high-flux situations, however, such as in synchrotrons or electron storage rings, or in fusion reactors, considerable effects must be expected.Photodesorption from metals is a controversial topic.' Prof. H. P. Boehm (Munich) said: Ti02 is extensively dehydroxylated after out- gassing at 250-350°C; alcohols are chemisorbed on such surfaces with formation of alkoxide and hydroxide groups (ions).Hydroxide ions on TiO, surfaces react as traps for photo-produced holes. The resulting OH radicals have been detected by e.s.r. ~pectroscopy.~ In contact with illuminated TiO, (anatase), acetanilide is hydroxylated in the 0- and p-position exclusively;' this behaviour is typical of attack by OH radicals, whereas reaction with 0,-ions results in equal quantities of o-, m- and p-hydroxyacetanilide. Liquid methanol in contact with illuminated TiO, is oxidised to formaldehyde even in the absence of oxygen, and alcohols seem to be see e.g., D. Menzel, Desorption, in Surfnee Physics, ed. R. Gomer (Springer Verlag, in press), and references given therein. P. Kronauer and D. Menzel, in Adsorption-Desorption Phenomena, ed.F. Ricca (Academic Press, London, 1972), p. 313. D. Menzel and R. Gomer, J. Chem. Phys., 1964,41,3311. H. G. Volz, G. Kampf and H.-G. Fitzky, Proc. 10th Fatipec Congr. (Verlag Chemie, Weinheim, 1970), p. 107. J. Steinle, Diplomarbeif (Munich, 1972). H. J. Staudinger and V. Ullrich, 2. Nafurforsch., 1964, 19b, 409.184 GENERAL DISCUSSION oxidised more easily than hydrocarbons. I think that in Cunningham’s experiments *OH or -OR radicals were the primary reaction products. Does he think that organic radicals, produced either by hole capture or by reaction with an -OH radical, can diffuse on the surface until two such radicals might disproportionate to alcohol and aldehyde ? Prof, 3. Cunningham (Cork) said: In response to the questions of Boehm as to the possible role of surface hydroxyl in photoeffects on Ti02 surfaces, I would first report results of experiments we made in an effort to check this point and second to urge caution before attempting to correlate results of our dynamic mass spectrometer technique with results obtained under continuous low-intensity illumination.Our experiments involved measurement of the magnitude of photodesorption of anhydrous C2D50D and of its photo-assisted conversion to CDSCDO and CzD4 all with t+- 0.1 s over Ti02 or ZaO surfaces previously outgassed for 16 h at 300°C. Having thus determined the magnitude of these effects in conditions which minimised the number of surface hydroxyl groups, the photoeffects were remeasured with C2D50D con- taining 5% D20. Contrary to the effects expected if surface hydroxyl groups deter- mined the efficiency of the photoeffects, we found that presence of D20 at this level decreased the extent of these photoeffects by 60% over ZnO but did not significantly affect them over TiOa. We have not, as yet, carried out similar experiments to test the influence of surface hydroxyl groups on photoeffects involving oxygen at the TiOz surfaces. Data obtained with the dynamic mass spectrometer technique on photoeffects involving oxygen at TiOz surfaces having residual hydroxyl groups reduced to small coverage (see table 1 in our paper) point, however, to the need for caution in attempt- ing a comparison between our results and results, such as those of Teichner, which also involve oxygen but were obtained under continuous, low-intensity illumination. The “ half-lives ” t+ of 1-10 min for the approach to 50% reaction in these latter conditions may be contrasted with t3 values of 0.1 s measured by the dynamic mass spectrometer for O2 photodesorption or of 5 s for photo-oxidation of the TiO, surface after flash illumination. The widely different time scales make it appear probable that the dynamic mass spectrometer technique preferentially observes and time-resolves fast processes, whereas results under continuous illumination may reveal only the net effect of various slower process at the illuminated interface. U. Kaluza, Diss. (Heidelberg, 1969).

ISSN:0301-7249    

DOI:10.1039/DC9745800175

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

Photointeraction on the Surface of Titanium Dioxide between Oxygen and Alkanes N. DJEGHRI, M. FORMENTI, F. JUILLET AND s. J. TEICHNER" Institut de Recherches sur la Catalyse (C.N.R.S.) and Universit6 Claude Bernard, 69626 Villeurbanne, France Received 20th May, 1974 Under the influence of ultra-violet illumination (210-390 nm) alkanes, both normal and branched from ethane to octane, are photo-oxidized at room temperature, mainly into ketones and aldehydes, on the surface of non porous particles of titanium dioxide (anatase). Alkanes CnH2n+2 form : ketones C,H2,0 and other aldehydes C,H2,0 with 2 < m < II, ketones CmH2m0 with 3 c m < n if the alkane is branched and finally COz and HzO. The mean selectivity into products of partial oxidation (ketones and aldehydes) is 76 % for n-alkanes, 80 % for isoalkanes, and 57% for neoalkanes.The reactivity of different types of carbon atom follows the sequence Ctert > Cquat > GC > Cprim. The carbon atom which is preferentially attacked by oxygen is that which presents the highest electron density together with the least steric obstruction. A proposed mechanism of the oxidation of alkanes implies the formation of an alcoholic intermediate, on primary, secondary and tertiary carbon atoms, which in turn is oxidized into aldehyde or ketone. On secondary and tertiary carbon atoms this intermediate may be first dehydrated into an olefin which is finally oxidized to an aldehyde or /and ketone. The photo-oxidation of various organic molecules (except alkanes) in the presence of ultra-violet irradiation and a solid oxide surface is a phenomenon which has been observed for many ~ e a r s , l - ~ the partial photocatalytic oxidation of alkanes (propane, isobutane, n-butane) was recorded only recently.6-10 In the present paper the previous work is extended to alkanes from C, to C8 and the reactivities of various types of carbon atom are considered.A tentative mechanism for this photo-oxidation is proposed. EXPERIMENTAL The catalyst is a non-porous anatase prepared by the flame reactor method l1 which has already been described in detail.' The differential photocatalytic fixed bed reactor was also described previously ** lo as well as the gas chromatographic methods for the analysis of effluenfs.l2 A mercury vapour lamp (Philips HKP 125 W) was used throughout. RESULTS AND DISCUSSION The photocatalytic activity of Ti02 powder at room temperature in the oxidation of propane (in the mixture 30 % propane, 20 % oxygen and 50 % helium at atmospheric pressure, with a flow rate of 20 cm3 min-l) as a function of the mass of the catalyst, spread out in the reactor as a thin homogeneous layer, is represented in fig.1. The areas of chromatographic peaks for acetone, propanal and ethanal and the height of the peak for C02 first increase with the mass of the catalyst and then level off, showing that with the increasing thickness (or the mass) of the catalytic bed, only catalyst layers irradiated by U.V. light are active. The same relationship was observed for the photo-oxidation of isobutane on this catalyst. 185186 a 5c E 4o 2 30 20 10 0 - - - PHOTOINTERACTION ON THE SURFACE OF TITANIUM DIOXIDE - 200 -150 wE E P; --.100 50 I0 20 30 40 50 '0 mass of Ti02/mg FIG. 1.-The photocatalytic activity in the oxidation of propane as a function of the mass of catalyst in the reactor. 0, acetone ; 0, propanal ( x 10) ; A, ethanal ( x 5) ; 0, COz Photocatalytic activity in the oxidation of propane as a function of time is re- presented in fig. 2. The stationary state for the production of C 0 2 is observed after 10 min whereas other reaction products require about 30 min to reach this limit, probably because of their physisorption in the tubes leading from the reactor to the analytical system. The stationary activity is constant for many hours, but when the irradiation is stopped the reaction ceases.In the absence of oxygen but still under irradiation no reaction takes place. Similarly, with oxygen but without the t/min 200 ~ 5 0 % E P \ 100 50 I FIG. 2.-The photocatalytic activity in the oxidation of propane as a function of time. 0, acetone ; 0, propanal ( x 10); A, ethanal (x5); e, COz.N. DJEGHRI, M . FORMENTI, F . J U ~ L L E T A N D s. J . TEICHNER 187 catalyst no photo-oxidation is recorded. Therefore the conditions which have to be fulfilled for reaction are the simultaneous presence of titania, oxygen and U.V. irradiation. The quantum yield in the photo-oxidation of isobutane, as determined previously lo by potassium ferrioxalate or uranyl oxalate actinometry, or directly by microcalori- rnetry, has the value 0.1 for the wavelength 320 nm.This result is of the same order of magnitude as the quantum efficiency in the photocatalytic oxidation of isopropanol on r ~ t i l e . ~ The results concerning the nature of the products of photo-oxidation of linear alkanes (from Cl to C,) are summarized in table 1. They show that every one of the TABLE 1 .-PRODUCTS OF THE PARTIAL PHOTO-OXIDATION OF n-ALKANES alkanes methane ethane propane butane pentane hexane heptane octane partial oxidation products ketones aldehydes acetone but an one ethanal propanal, ethanal butanal, propanal, ethanal pentan-2-one pent anal, but anal, pent an-3-one propanal, ethanal liexan-2-one hexanal, pentanal, hexan-3-one butanal, propanal, ethanal heptan-Zone heptanal, hexanal, hep t an-3-one pentanal, butanal, hep t an-4-one propanal, ethanal octan-Zone octanal, heptanal, octan-3-one hexanal, pentanal, octan-4-one butanal, propanal, acetone (trace) ethanal carbon atoms of a paraffin molecule may be attacked. For instance, in the case of n-pentane, all the bonds C-H and C-C are attacked because pentan-2-one and -%one are formed as well as all the possible aldehydes, from C5 to C2.Formaldehyde was never detected but its photo-oxidation yields CO, and H20. It is therefore probable that this intermediate would be entirely photo-oxidized and remain un- detected. In some cases alcohols also were detected (see below) but in relatively small quantities. Though they do not figure among the partial oxidation products in the tables listed below, their role is essential in the oxidation mechanisms.The results concerning the nature of the products of photo-oxidation of alkanes with one or two tertiary carbon atoms (all isoalkanes + 2,3-dimethylbutane and 3- methylpentane) are summarized in table 2. Products termed X and Y are unidentified. Again it can be seen that every carbon atom of the molecule may be attacked. In the case of isoalkanes acetone is the major product showing the high reactivity of the tertiary carbon atom. Alkanes containing a quaternary carbon atom such as neo- pentane (2,2-dimethylpropane) are, in comparison to previous alkanes, more inert. Table 3 summarizes the results for these alkanes. The quantitative results concerning the distribution of reaction products are188 PHOTOINTERACTION ON THE SURFACE OF TITANIUM DIOXIDE TABLE 2.-&ODUCTS OF THE PARTIAL PHOTO-OXIDATION OF ALKANES CONTAINING ONE OR TWO TERTIARY CARBON ATOMS alkanes 2-me t hylpropane acetone partial oxidation products ketones aldehydes 2-methyl propanal (+ t-butyl alcohol) 2-methylbu tane 3 -methylbutan-2-one, 2-methylbutanal, butanone, acetone 2-methylpropanal, ethanal, (X) 2-me thy 1 pentane 4methyl pent an-2-one, 2-met h ylpen t anal, 2-methylpentan-3-one, 2-methyl butanal, pent an-2-one, acetone 2-methylpropanal, propanal, ethanal 3-methylpentane 3-methylpentan-2-one, (3-met hylpen t anal), pentan-3-one, 2-et hyl but anal, butanone, acetone propanal, ethanal, (Y) 2-methylhexane 5-methylhexan-2-one, (2-met hylhexanal), 5-me t h yl hexan-3 -one, 2-methylpentanal, 2-methylhexan-3-one, 2-met hyl butanal, hexan-2-one, acetone 2-methylpropanal, butanal, propanal, ethanal 2,3-dimethyl butane 3 -met hylbut an-2-one, acetone 2-methylpropanal, (2,3-dimethyl but anal), ethanal expressed in terms of the selectivity S toward a particular reaction product.S represents the ratio of the number of moles of the particular product formed to the number of moles of the alkane consumed during the reaction. This definition of selectivity may exceed 100 % if a hydrocarbon molecule gives two or more molecules of the reaction product (like COz). For this reason, the selectivity S, is introduced TABLE 3.-mODUCTS OF THE PARTIAL PHOTO-OXIDATION OF ALKANES CONTAINING ONE QUATERNARY CARBON ATOM partial oxidation products alkanes ketones aldehydes 2,2-dimethylpropane acetone (2,2-dirnethylpropanal), (X) 2,Zdimethylbutane 3,3-dimethyl butan-2-oneY (3,3-dimethyl but anal), butanone, acetone propanal, ethanal, (X), (Y) and defined as a selectivity related to the number of carbon atoms of the reaction product.It represents the ratio of the number of moles of the alkane required to form the reaction product to the total number of moles of the alkane consumed during the reaction. In this way the sum of all S, for a given paraffin should equal 100 %. It was observed that the difference between this theoretical value and the experimental value never exceeded 8 %, for any paraffin from C2 to C8.N. DJEGHRI, M . FORMENTI, F . JUILLET AND s. J . TEICHNER 189 Table 4 displays the selectivities for photo-oxidation of 2-methylpentane (iso- hexane). The formation of ketones is clearly favoured in comparison with the forma- tion of aldehydes.The presence of a tertiary carbon atom in the reagent accounts also for a preferential formation of acetone. This is not the case in the photo- oxidation of n-octane as shown in table 5. Again ketones are favoured in comparison with aldehydes. TABLE 4.-sELECTIVITIES FOR KEACTION PRODUCTS OF THE PHOTO-OXDATION OF ISOHEXANE products S Sc! % co2 2-met hylpentan-3-one 4-methylpentan-2-one pentan-2-one acetone 4-methylpentanal 3-met hyl but anal 2-methylpropanal propanal ethanal 60 18 16 3 33 2 3 16 14 13 10 18 16 2 17 2 2 10 7 8 In table 6 the selectivities S, for ketones and aldehydes are given for all the alkanes used in this work and also the total selectivities into the products of partial oxidation, i.e.all the products except C02 and H20. When the sum of S, for ketones and aldehydes is smaller than the S, for all the products of the partial oxidation (e.g. for 2-methylpropane) then we conclude that alcohols or some other unidentified products TABLE 5.-sELECTIVITIES FOR REACTION PRODUCTS OF THE PHOTO-OXIDATION OF n-OCTANE products S Sol % co2 octan-Zone octan-3-one octan-4-one oct anal heptanal hexanal pent anal butanal propanal ethanal (not measurable) 184 20 18 14 1 5 7 7 6 9 < 1 23 20 18 14 1 5 5 5 3 3 < 1 are formed. They are expressed in equivalent of moles of butan-2-01. These results show that the formation of ketones is favoured for all the alkanes. More generally, photocatalytic oxidation of alkanes would seem to be particularly useful if the partial oxidation of these compounds into aldehydes and ketones is required, as shown by the last column of table 6.Table 7 summarizes the mean values of selectivities S, for decomposition to ketones and aldehydes for n-alkanes, isoalkanes and alkanes containing one or two tertiary carbon atoms and finally for neoalkanes. Table 8 shows the reactivity of n-alkanes as a function of the length of the hydrocarbon chain. The experimental conditions were the same for all reagents as190 PHOTOINTERACTION O N THE SURFACE OF TITANIUM DIOXIDE TABLE 6.-sELECTIVFTIES FOR KETONES, ALDEHYDES AND FOR ALL PRODUCTS OF THE PARTIAL OXIDATION OF ALKANES USED IN THE3 WORK alkanes methane ethane propane butane 2-met hylpropane pentane 2-methylbutane 2,2-dime t hylpropane hexane 2-methylpentane 3 -meth ylpentaiie 2,2-dimethyl butane 2,3-dimethyl butane heptane 2-methylhexane octane total Sc/ % total A‘,/% for partial oxidation ketones aldehydes products 0 0 57 49 61 51 54 42 73 53 53 44 78 49 65 52 0 18 1 1 19 7 22 26 8 14 29 20 19 12 33 17 22 0 18 68 68 77 73 80 59 87 52 80 70 90 82 82 74 indicated previously, and therefore the overall conversion percentages, given in this table, representing the sum of conversions into different reaction products, are the measure of the reactivity of various alkanes.It may be concluded from these results that in general the reactivity increases as the number of carbon atoms in the chain increases. This behaviour indicates that any carbon atom of the chain may be attacked and the longer is the chain, the higher is the conversion.TABLE 7.-sELECTIVITIES FOR KETONES AND ALDEHYDES FOR THE THREE GROUPS OF ALKANES alkanes Sc/ ”/, ketones Sc/ % aldehydes n-alkanes 56 20 alkanes possessing one or two tertiary carbon atoms 61 19 neoalkanes 43 14 For isoalkanes and neoalkanes the same behaviour is observed, as shown in table 8. Finally the reactivities of various types of carbon atoms follow the sequence : The reactivity of an alkane in its photo-oxidation seems to depend only on its structure and is in particular completely independent of its ionization potential as it can be seen by comparing the over-all conversions with the ionization potentials of alkanes l3 used in this work. TABLE 8.-OVERALL CONVERSION OF ALKANES n-alkanes conversion/ ?(, iso- and neo-alkanes conversion/ % ethane 0.59 propane 0.74 butane 0.87 pentane 1 hexane 1.85 heptane 1.15 oc t alle I .25 2-methylpropane 1.20 2-methylbutane 1.60 2-methylpentane 1.97 2-me t h y 1 hexane 2.42 2,2-dimethylpropane 0.66 2,2-dimetliyl butane I .76N .DJEGHRI, M . FORMENTI, F . JUILLET AND S . J . TElCHNER 191 The reaction mechanism was studied with particular emphasis on the photo- oxidation of isobutane and propane. In the case of isobutane the various selectivities S, are : C 0 2 23 %, acetone 61 %, 2-methylpropanal (isobutanal) 7 %, t-butanol 9 %. Isobutanal is produced by attack on a primary carbon atom (methyl groups) of the isobutane, whereas acetone and t-butanol result from attack on a tertiary carbon ($€I). The last two reaction products could result either from two parallel reactions : (scheme 1) or from the consecutive reactions : isobutane -+ t-butanol -+ acetone.(scheme 2) A common intermediate in scheme 1 could be a hydroperoxide of isobutane, in the same way as in the photochemical oxidation of cyclohexane into cyclohexanol and cyclohexanone in the liquid phase.14 However, no hydroperoxide of isobutane was detected in this work by gas chromatography and moreover if the photocat?lytic reactor is fed with the hydroperoxide of isobutane only minor amounts of acetone are detected with no trace of t-butanol. In the same way t-butyl peroxide is inactive on Ti02 under U.V. irradiation. It is therefore difficult to extend the results for the photo-oxidation of cyclohexane and to consider a common precursor of a hydro- peroxide type for two parallel reactions in scheme 1.For scheme 2 the following experimental evidence was recorded. The rate of production of acetone (V,) by photo-oxidation of t-butanol on Ti02 during U.V. irradiation, is equal to the rate of production of acetone from isobutane (V,) under the same experimental conditions. Therefore in scheme 2 : isobutane --+ t-butanol -+ acetone V. vb t _I_- v c I the rate determining step in consecutive reactions would be the oxidation of t-butanol. As the selectivity in the photo-oxidation of isobutane into t-butanol is not high (9 %) the rates V, and v b are not very different and in particular V, is not much higher than the rate Vb of the rate determining step. Evidence for the formation of acetone from t-butanol was also obtained from i.r.absorption spectroscopy on a sample of TiO, U.V. irradiated in the presence of t-butanol, or of t-butanol and oxygen. An experiment of the same type was carried out with propane and isopropanol. However, isopropanol gives acetone with a rate Vi which is almost 30 times faster than the rate Vd for formation of acetone from propane : v,: v; propane -+ isopropanol -+ acetone. v: t I This shows that isopropanol as the intermediate will not be detected during photo-oxidation of propane. Now under the same experimental conditions the rate of formation of acetone from isopropanol is 50 times faster than the rate of formation of acetone from t-butanol. This result is in agreement with the relative ease of oxidation of secondary and tertiary alcohols.For isopropanol the conventional mechanism involves the attack by a nucleophile (oxygen species) on the hydrogen atom attached to the same carbon atom as the OH group :192 PHOTOINTERACTION ON THE SURFACE OF TITANIUM DIOXIDE ys, 0-H .o 13 /,” CH3-C-H 0 -+CH3--C +HOH. I 1 CH3 CH3 For tertiary alcohols which do not have such a hydrogen atom this scheme However, tertiary alcohols are easily dehydrated to olefins cannot be applied. which in turn are oxidized into aldehydes and ketones, according to the scheme : R R R -H2O \ 0 2 \ \ / I / / C-CH2-R 4 C=CH-R + CO+R-CHO. R- I R OH R The photo-oxidation of isobutene showed that the rate of formation of acetone from this olefin is higher than the rate of formation of acetone from t-butanol.The intermediate olefin, formed by dehydration of t-butanol in the first step, will not therefore be detected during the photo-oxidation of this alcohol. The photocatalytic oxidation of isobutane could be therefore represented by the following steps : H OH I 0 2 I -- H20 0 2 I I I I CH3 CH3 ( 3 - 4 3 CH3 CH3-C-CH3 + CH3-C-CH3 -+ CHj-C=CH;! + CHS-C=O. detected undetected intermediate intermediate The simultaneous formation of isobutanal would result from the conventional attack on a primary carbon atom : H H OH H CH3-C-CH3 I 0 2 -+ CH3-C-CHZ I I 0 2 + CH3-C-C ‘ //O I I 1 ‘H CH3 CH3 CH3 In the case of propane, oxidation at the carbon atom to which the OH group is attached gives acetone directly : H OH 0 I 0 2 I 0 2 II I I CH3-C-CH3 -+ CH3-C-CH3 + CH3-C-CH3. H H For higher alkanes with secondary carbon atoms, the secondary alcohol formed in the first step can either be directly oxidized to ketones (table 5 ) or it can be de- hydrated into an olefin.Indeed, only this intermediate can explain the formation of lower aldehydes from the alkanes (tables 1,2,4, 5). In the preceeding schemes it was assumed that the first step is the formation of an alcohol intermediate at the primary, secondary or tertiary carbon atom. This intermediate was observed in the photo-oxidation of isobutane.N . DJEGHRI, M. FORMENTI, F . JUILLET AND s. J . TEICHNER 193 Now this initial formation of the alcohol from the paraffin is in principle possible with an atomic oxygen species.16 However, the only oxygen species detected on TiOz under U.V. irradiation are 0; molecular ions.** However, the photocatalytic oxidation of alkanes requires an amount of U.V.energy (3.5 eV) compatible with the formation of positive holes.8* lo It is then quite possible that 0; precursor is able to combine with positive holes, to form free or trapped, neutral or ionic atomic oxygen species. Finally it should be mentioned that photo-oxidation of isopentane gives the same results, as far as the intermediate alcoholic species is concerned, as the photo-oxidation of isobutane. In conclusion, the consecutive reaction scheme can be represented by the following sequence of steps : 0 2 primary C 4 secondary C 4 0 2 0 2 tertiary C -+ 0 2 primary alcohol + aldehyde 3 ketone o2 alcohol-l+ olefin + aldehydes +ketones - HzO - € I 2 0 0 2 tertiary alcohol olefin 3 ketone +aldehyde.For quaternary carbon atoms (2,2-dimethylpropane, 2,2-dimethylbutane) the lack of information concerning the nature of the species termed X and Y (table 3) pre- cludes the establishment of a reaction scheme. In conclusion, the photocatalytic oxidation of alkanes on titania at room tempera- ture, which is very selective towards the formation of products of partial oxidation (aldehydes, ketones), proceeds via an alcohol intermediate and not via a hydroperoxide or peroxide intermediate. It is probably this characteristic, together with the low temperature range used, which accounts for this high selectivity. The financial support of D.G.R.S.T. is gratefully acknowledged. C. Renz, Helv. Chim. Ada, 1921, 4, 961. V. N. Filimonov, Doklady Akad. Nauk S.S.S.R., 1964,154,922; 1964, 158, 1408 ; Kinetika i Katalyz, 1966, 7 , 512. S. Kata and F. Mashio, Kogyo Kagaku Zasshi, 1960,63,745 ; 1964,67,8. T. S. McLintock and M. Ritchie, Trans. Faraday SOC., 1965, 61, 1009. R. I. Bickley, G. Munuera and F. S. Stone, J. Catalysis, 1973, 31, 398. M. Formenti, F. Juillet and S. J. Teichner, Compt. rend. C, 1970,270,138. ’ M. Formenti, F. Juillet, P. Meriaudeau and S. J. Teichner, Bull. SOC. chim. France, 1972, 69. M. Formenti, F. Juillet, P. Meriaudeau and S. J. Teichner, Chem. Tech., 1971,1,680. P. C. Gravelle, F. Juillet, P. Meriaudeau and S. J. Teichner, Disc. Faraday SOC., 1971,52, 140. l o M. Formenti, F. Juillet, P. Meriaudeau and S. J. Teichner, Proc. 5th Int. Congr. CaZalysis, 1972, ed. J. W. Hightower (North Holland, Amsterdam, 1973), p. 1101. l 1 J. Long and S. J. Teichner, Rev. Hautes Temp., 1965, 2, 47. l 2 N. Djeghri, Thesis (University Claude Bernard, Lyon I, 1973). R. W. Kiser, Tables of Ionisation Potentials, U.S.A.E.C. Office of Technical Information, T.I.D. 6142, 1960 and 1962. N. Kulewsky, P. V. Sneeringer, L. 0. Grina and V. I. Stenberg, Photochem. Photobiol., 1970,12, 395. l 5 P. Arnaud, Cours de Chimie Urgarzique, ed. Gauthier (Villars, Paris). l6 H. Yamazaki and R. J. Cvetanovic, J. Chem. Phys., 1964, 41, 12, 3703 ; G. Paraskevopoulos and R. J. Cvetanovic, J. Chem. Phys., 1970,52, 11, 5821. G. Paraskevopoulos and R. J. Cvet- anovic, J. Chem. Phys., 1969, 50, 2, 590. 5 8-G

ISSN:0301-7249    

DOI:10.1039/DC9745800185

出版商:RSC    

年代:1974

数据来源: RSC

摘要:

Photo-adsorption and Photo-catalysis on Titanium Dioxide Surfaces Photo-adsorption of Oxygen and the Photocatalyzed Oxidation of Isopropanol BY R. I. BICKLEY" AND R. K. M. JAYANTY? School of Chemistry, University of Bradford BD7 1DP Received 17th May, 1974 A study of the photoadsorption of oxygen by temperature-programmed desorytion reveals that, on fully oxidised rutile surfaces, adsorbed hydroxyl groups act as traps for the photo-holes which are produced by irradiating the specimen with ultra-violet light A > 330 nni. The photo-electrons, which are simultaneously produced, are then free to participate in the chemisorption of oxygen. On reduced rutile surfaces it has been established that the photoadsorption of oxygen is enhanced by the additional presence of Ti3+ ions which also fulfil the role of traps for photo-holes.The photo-oxidation of isopropanol has been shown to be quite complex. The initial product is acetone, but the subsequent photo-oxidation of the acetone produces formic acid. Acetaldehyde is also produced, but probably arises from the photo-oxidation of adsorbed propene which is formed by the thermal dehydration of isopropanol. Ultimately the products of prolonged photo-oxidation are carbon dioxide and water. A number of previous studies 1-4 of the surface photochemistry of titanium dioxide have reported that the photoactivity of powdered specimens was enhanced by the presence of adsorbed water. The importance of adsorbed water in the photo- adsorption of oxygen on rutile surfaces has recently been confirmed; it was also demonstrated that the photocatalysed oxidation of isopropanol would occur only to a small extent on well-outgassed rutile but that the photoactivity could be partially regenerated by treating the specimen with water vapour.Several studies have been made of the adsorption of water on titanium dioxide using infra-red spectroscopy. 7-1 O These studies clearly show that molecularly adsorbed water is bound to the surface, in addition to hydroxyl groups which arise from the dissociative adsorption of water molecules. Adsorption isotherms of water vapour on rutile at 273 and 291 K show that only a part of the adsorbed water can be removed by evacuation at the adsorption temperature but that a dehydroxy- lated surface appears to form at temperatures in excess of 600 K. Recently, tempera- ture programmed desorption experiments have demonstrated that water can exist in at least two strongly adsorbed states on rutile surfaces ; a third strongly adsorbed state may also be produced if rehydroxylation of the specimen is effected in the presence of air or oxygen.I2 The present work investigates further the role of adsorbed water in the photo- adsorption of oxygen and examines the photocatalysed oxidation of isopropanol.In contrast to previous studies,5* which have examined these processes by mano- metric methods and by gas-liquid chromatography, the present study applies temperature-programmed desorption, using a mass spectrometer as analyser. Additional experiments involving the use of electron spin resonance were also carried out.1- present address : Ionosphere Research Laboratory, The Pennsylvania State University, Univer- sity Park, Pa. 16802 U.S.A. 194R . 1. BlCKLEY A N D R . K . M. JAYANTY 195 The adsorption of water, isopropanol, acetone and acetic acid on rutile has been examined before and after irradiation of the specimen with ultra-violet light A 3 330nni. The water desorption spectra have been used to interpret the effect of progressively increasing the outgassing temperature upon the reactivity of the specimen towards the photoadsorption of oxygen. Spectra obtained at various stages of the room temperature photocatalysed oxidation of isopropanol has enabled a reaction path to be deduced. EXPERIMENTAL The specimen preparation and adsorption-desorptioii experiments were performed in a vacuum system of conventional design in which ultimate vacua of the order of lW4 N m-2 were attainable.The system was constructed in two sections, a gas-handling section containing greased stopcocks and the adsorption-desorption section (volume 490 cm3) which contained metal valves and a mass spectrometer, and was capable of being baked to 450 K. The silica reaction cell was identical to that used in earlier work and was connected to the system by a graded seal. The temperature-programmed experiments were made with a Perkin-Elmer linear programming device in conjunction with a Sirect proportional controller (model LT. C.N.S. Ltd). This arrangement enabled heating rates of 2,4, 6 and 8 K min-' to be attained from 300 to 800 K. Heating rates as large as 20 K min-' were obtained with a second propor- tional controller, the helical potentiometer of which was rotated by a geared electric motor ; using this system temperatures in the region of 1300 K were attained.kg of rutile specimen was continuously pumped through a calibrated conductance of m bore and 5 x m length. A second capillary, the conductance of which was approximately 5 % of the value of the main conductance and which operated in parallel with it, enabled desorbing gases to enter the head of a small mass spectrometer (V.G. Micronlass 2A) which scanned the mass ranges 12-60 and 48-240, each scan taking 1 min to complete. Electron spin resonance experiments were made with a Varian spectrometer model V4502 operating at 9.3 Gc s-' with 100 kc s-l field modulation.The specimens, in triplicate, were attached in silica tubes to the greaseless part of the vacuum system. After suitable prepara- tion, the specimens were detached by collapsing the tubing with a gasfoxygen flame. Spectra were measured at 75 and 298 K. Irradiation of the specimens with ultra-violet light was made with a water-cooled medium-pressure mercury arc (Hanovia, 500 W). Rutile TiOz was kindly supplied by Tioxide International Ltd., Billingham, and had been calcined in air at approximately 1100 K. The B.E.T. surface area, as measured by krypton adsorption at 78 I<, was 4.2 m2 8-l. Oxygen and Krypton were supplied in Pyrex bulbs by B.O.C. Ltd Wembley. Conductivity water and B.D.H. reagent-grade isopropanol, acetone and acetic acid were each subjected to several freeze-pump-thaw cycles before use.Mass spectra of the pure compounds under study were determined in order to faciliiate analyses; only the three most significant mass nuinbcrs which appeared in the cracking patterns were used to analyse the desorption spectra. During the desorption experiments 1.5 x RESULTS OXYGEN PHOTOADSORPTION The effect of increasingly severe outgassing conditions on the kinetics of oxygen photoadsorption at 273 K is shown in fig. I . The conditions were chosen by measuring the temperature-programmed desorption profile of water desorbing from rutile and then selecting temperatures which would ensure the selective removal of the corresponding water peaks. " As received '' specimens desorbed carbon monoxide, carbon dioxide and sulphur dioxide in addition to water and a small quantity of oxygen (6 x 1OI6 molecule m-2) (fig.2). After outgassing and exposuer to oxygen at 873 K, specimens which were rehydroxylated in water vapour at roomsignal intensity/arbitrary units f"' v1 b ul n 0 - - _ - - - - oxygen uptake x 108/molm-aR . 1. BICKLEY A N D R . K . M. JAYANTY 197 temperature, or in boiling water, only desorbed water and oxygen; a typical profile is shown in fig. 3. The coverage of water corresponds to -2x lo1' molecule m-2 of which 65 % exists as undissociated molecules. p8 I I I 373 473 573 673 773 temperature/K FIG, 3.-The effect of oxygen photo-adsorption on the desorption profile of water. photo-adsorption ; 0, H20 after photo-adsorption ; 0, oxygen.0 , H20 before Following prolonged oxygen photoadsorption (20 h) the specimen became saturated with oxygen (8.1 x 10l6 molecule m-2) and produced a modification in the desorption profile (fig. 3) suggesting that the more strongly bound form of water becomes even more strongly bound as a result of the uptake of oxygen. The kinetics of the photoadsorption process are adequately described by plotting q, the quantity of oxygen adsorbed, against t*, the square root of the illumination time. The slope k of the parabolic plot shows an unusual temperature dependence which is illustrated in fig. 4. Furthermore, the pressure dependence of the reaction is also complex, the reaction being independent of pressure below 298 K but showing a marked dependence on oxygen pressure above 333 K.As a result of outgassing the rutile at various temperatures, two e.s.r. signals were observed ; a small singlet g = 2.001 & 0.001, observable at room temperature after198 PHOTO-ADSORPTION AND PHOTOCATALYSIS 1.0 c L E E 0.5 H -.I 0 * 0 5 1 M W d 0.0 I I 1 I I1 2.8 3 . 0 3.2 3.4 3.6 lo3 Kltemperature FIG. 4.-The effect of temperature on the “parabolic” rate constant k for oxygen photoadsorption. 9- 1.946 g-I 9 5 8 673 773 873 973 I073 temperature/K FIG. %-The temperature-programmed desorption of oxygen from rutile and the associated changes in paramagnetism, +, oxygen ; A, Ti3+R . I . BICKLEY AND R . K . M . JAYANTY 199 outgassing at 473 K and higher, arid a triplet g1 = 1.967, y2 = 1.958, g3 = 1.946k 0.001 which was only observed at 78 K.This signal has previously been observed in reduced rutile and has been ascribed to Ti3+ ions.13 The triplet appeared at 620 K, progressively increasing in intensity initially as the outgassing temperature increased, but finally broadening and becoming less distinct. These results are shown in fig. 5. PHOTO-OXIDATION OF ISOPROPANOL The desorption profiles arising from isopropanol adsorption upon well-outgassed rutile, and upon fully rehydroxylated rutile are shown in fig. 6. The isopropanol is held in two strongly bound states, the weaker of which forms acetone and a trace of propene, and from which some unreacted isopropanol also can be desorbed. The 323 373 423 473 523 5 7 3 temperature/K FIG. 6.-The temperature-programmed desorption profile of rutile surfaces saturated with isopropanol.Open symbols, a fully hydrated surface ; dark symbols, a dehydroxylated surface. 0, isopropanol ; A A, acetone ; W 0, propene. stronger of the two states is irreversibly held, most of the isopropanol being de- hydrated to form propene but a trace of acetone is also produced from a dehydroxy- lated surface. The size of the propene peak is significantly influenced by the state of hydroxylation of the specimen emphasizing the importance of competitive adsorp- tion l1 between water and isopropanol. The strongly adsorbed isopropanol was subjected to photo-oxidation under ca. 100 N m-2 of oxygen. After 2 h, the composition of the gas phase was analysed and the desorption profile was determined (fig. 7). There is a marked decrease in the quantity of unreacted isopropanol desorbed from the weaker state and also a small200 PHOTO-ADSORPTION AND PHOTOCATALYSIS decrease in the quantity of acetone desorbed. A slight diminution is also observed in the quantity of propene released from the stronger state but a small increase occurs in the amount of propene liberated from the weaker state (see fig.6 and 7). 323 373 423 473 523 573 temperature/K FIG. 7.-The temperature-programmed desorption profile of a rutile surface saturated with isopropanol following 2 h of photo-oxidation. 0, isopropanol ; A, acetone ; 0, propene. A similar experiment was conducted for 6 h after which time the analytical procedures were repeated. Fig. 8 shows that only a trace of unreacted isopropanol exists after the reaction.A small quantity of acetone is desorbed from the weaker I 323 373 423 473 523 573 temperature/K FIG. 8.-The temperature-programmed desorption profile of a rutile surface saturated with isopropanol following 6 h of photo-oxidation. 0, isopropanol ; A, acetone ; 0, propene; 0, formic acid.R . I . BICKLEY AND R . K . M. JAYANTY 20 1 state but a large quantity is desorbed from the stronger state together with a greatly reduced quantity of propene. Most significantly, a new species of molecular mass 46 appears in easily detectable amounts. The composition of the gas phase after photo-oxidation was determined by condensing a small volume of the gas at 78 K ; the residual gas was oxygen. Warming the trap to 193 K yielded principally C 0 2 , but on further warming to 298 K a mass spectrum was produced which was consistent with the presence of a mixture of acetone, acetaldehyde and water.Additional experiments were carried out to determine the desorption profiles of acetone, and acetic acid before and after photo-oxidation ; these results are shown in fig. 9. Acetone is desorbed from one state which is less strongly bound than the strongest state of isopropanol. Acetic acid is very strongly held and can only be partially desorbed without decomposition. 20 i- 323 373 423 473 523 573 623 temperature/][( FIG. 9.-The temperature-programmed desorption profiles of a rutile surface (i) saturated with acetone ; A, before illumination ; A, following 12 h of photo-oxidation, and (ii) saturated with acetic acid ; 0, before illumination ; 0, following 12 h of photo-oxidation. DISCUSSION OXYGEN PHOTO-ADSORPTION Below 473 K, the outgassing of rutile surfaces has only minor effects on the ability of the specimen to photo-adsorb oxygen, whereas above 473 K a marked decrease in activity is observed. This decrease is associated with the progressive removal of the second peak in the water desorption profile.Infra-red studies of rutile surfaces indicate that the first peak in the desorption profile arises from chemi- sorbed molecular water and that the second peak corresponds to water resulting from the condensation of hydroxyl groups according to a mechanism of the type : OH, +OH; + H20(g)+0z- + OL(02-). It appears that a direct relationship has now been established between the photo- activity of the specimen and the presence of hydroxyl groups on fully-oxidised rutile surfaces.The fact that the water desorption profile becomes modified following oxygen photoadsorption also gives support for this suggestion. Moreover, the activity of specimens which contain molecular water strongly suggests that water202 P H 0 T 0 - A D SORP TI 0 N AND P HOT 0 CAT A L Y SI S molecules do not play an important part in the photo-adsorption process. All of these observations lend support to the proposals that the hydroxyl groups act as traps for photo-holes produced in the fundamental adsorption process. TiOz + hv + h-e (exciton) + h +e h + OH- + *OH trapping OZads + e Ozads chemisorption -OH + Ozaas -+ HOz- + O;, surface diffusion The loss of hydroxyl groups from the surface also results in the surface diffusion step being more difficult to achieve and will therefore also lead to a lowering of the activity.During the course of dehydroxylation there does not appear to be a large build-up of paramagnetism. It is suggested that the electron spin resonance signal (g = 2.001) is associated with an electron trapped at a surface anion vacancy (1x1 ;(02-)). Above 673 K, the appearance of Ti3+ ions is detected and appears to be related to the loss of oxygen from the specimen at this temperature and above. The associated rise in photoactivity is ascribed to the ability of the Ti3+ ions also to act as traps for photo- holes. Ti3+ + h + Ti4+. The work of Gravelle et aZ.16 on reduced rutile also shows that this type of specimen exhibits photocatalytic activity.These authors distinguish two distinct types of Ti3+ site, each possessing axial symmetry. Under the conditions which prevail in the present experiments, the triplet signal which is observed is most likely due to Ti3+ ions in sites of lower than axial symmetry.13 The dependence of the photoadsorption kinetics upon temperature and pressure suggests that at lower temperatures a surface oxygen species is important, whereas at the higher temperatures the concentration of surface oxygen is directly related to the oxygen pressure. This might be reflected in the competition between the chemi- sorption and surface diffusion processes. P H 0 T 0 - 0 XI D A TI 0 N 0 F I S 0 P R 0 PAN OL The desorption profiles of isopropanol on rutile before photo-oxidation (fig.6 ) indicates that the weaker of the two adsorbed states of isopropanol gives rise mainly to dehydrogenation to form acetone, whereas the stronger state undergoes dehydration to form propene. Since some unreacted isopropanol can be desorbed from the weaker of the two states it is concluded that the molecules are adsorbed mostly in an un- dissociated form and are bound to the surface by hydrogen bonding forces in a manner similar to adsorbed molecular water, which desorbs in approximately the same region of temperature. It seems that water and isopropanol molecules compete for the same adsorption sites since a distinct difference exists in the profiles arising from the hydroxylated and dehydroxylated surfaces. The degree of hydration of the surface does not appear to have much effect on the quantities of acetone or propene which are desorbed from the weaker state.The strongly-bound isopropanol is probably adsorbed dissociatively since it is only possible to detect a trace of unreacted isopropanol desorbing from this state. A marked difference in the quantity of propene desorbing from the specimen is observed between the hydroxylated and fully dehydroxylated specimens indicating that the hydroxyl groups and the analogous species deriving from isopropanol (PrO or Pro-) are competing for the same sites.R. I . BICKLEY AND R. K. M. JAYANTY 203 The illumination of adsorbed isopropanol in the presence of oxygen initially produces acetone, a portion of which is displaced into the gas phase. By comparison of fig.6 and 7, even after two hours of photo-oxidation a marked reduction in the quantity of unreacted isopropanol has occurred. In addition, a significant reduction is observed in the amount of propene arising from the more strongly adsorbed state of isopropanol indicating that the species remaining on the surface after the dissociative adsorption of isopropanol also has undergone photo-oxidation. FIG. 10.-A possible mechanism for the overall photo-oxidation process. The quantity of acetone remaining adsorbed is also reduced, presumably because it is easily displaced by the water l 1 or hydroxyl groups l7 which are also produced during the reaction. The increased quantity of propene appearing from the weaker FIG. 11.-A mechanism to account for the appearance of acetaldehyde in the reaction sequence.adsorbed state of isopropanol is unexpected since the increasing degree of hydration of the surface would probably favour the dehydrogenation reaction. However, its appearance is presumably due to the thermal dehydration reaction occurring during204 PHOTO-ADSORPTION AND PHOTOCATALYSIS the desorption experiment and may arise from a second dissociated state of isopropanol which is less strongly held than the strongest state. The longer exposure of 6 h was selected so as largely to complete the photo- oxidation of the isopropanol mono1ayer.6 After this period of time, only traces of unreacted isopropanol were detected and a significant reduction in the amount of propene desorbing from both adsorbed states (fig. 7 and 8) was observed. Acetone now appears from the stronger of the two adsorption states of isopropanol indicating that the dissociated fragment arising from the adsorption of isopropanol can either be photo-oxidised or be thermally dehydrated.The appearance of adsorbed formic acid is considered to arise from the photo- oxidation of acetone which occurs at a slower rate than the photo-oxidation of isopropanol, fig. 9. It is also suggested that acetate groups are produced on the surface. The fact that only traces of acetic acid are detected in the gas phase is due to it being very strongly adsorbed at the rutile surface (fig. 9.) Acetic acid itself is slowly photo-oxidised with approximately 50 % of a monolayer disappearing in 12 h (fig. 9). Support for the presence of acetate groups appears in work by Fili- monov et al., who detected the presence of acetone and acetate groups by infra-red spectroscopy during earlier studies of the adsorption l4 and photo-oxidation of isopropanol on titanium dioxide.In the absence of cold traps, which prevents re-adsorption of molecules following their displacement from the surface, the ultimate products of the prolonged photo- oxidation of a monolayer of adsorbed isopropanol, in the presence of an excess of oxygen, are carbon dioxide and water. During the course of the process, a complex sequence of reactions occurs and these are summarised in fig. 10. The reaction path appears to be quite specific and this is considered to be due to the involvement of charged species rather than free radicals from which a larger range of products might have been expected through a more random abstraction of hydrogen atoms, from the adsorbed species. Acetaldehyde which was observed during the course of the longer photo-oxidation experiment probably originates from the photo-oxidation of adsorbed propene (fig. 11). The authors gratefully acknowledge the financial support given to this project by Tioxide International Ltd. C. Renz, Helv. Chim Acta, 1921, 4,961. F. McTaggart and J. Bear J. Appl. Chem. 1955, 5, 643. F. McTaggart and J. Bear, J. Appl. Chem., 1958, 8,72. H. B. Boehm and 0. Kaluza, J. Catalysis, 1971, 22, 347. R. I. Bickley and F. S. Stone, J. Catalysis, 1973, 31, 389. R. I. Bickley, G. Munuera and F. S. Stone, J. Catalysis, 1973, 31, 398. ’ P. Jones and J. A. Hockey, Trans. Farahy SOC., 1971, 67, 2669. * P. Jackson and G. D. Parfitt, Trans. Faraday Soc., 1971, 67,2469. K. Lewis and G. D. Parfitt, Trans. Faraday SOC., 1966, 62,204. l o M. Primet, P. Pichat and M. Mathieu, J. Phys. Chem., 1971, 75, 1221 G. Munera and F. S. Stone, Disc. Faraduy. Soc., 1971, 52, 205, l2 R. I. Bickley and R. K. M. Jayanty, Disc. Faraday SOC., 1971, 52,226. l 3 V. A. Shvets and V. B. Kazanskii, Kinetica i Kataliz, 1971, 12, 935. l4 Y. M. Shchekochikhin, V. N. Filirnonov, N. P. Keier and A. N. Terenin, Kinetica i Kataliz, l5 V. N. Filimonov, Doklady Akad. Nauk S.S.S.R., 1964,158, 1408. l 6 P. C. Gravelle, F. Juillet, P. Meriaudeau and S. J. Teichner, Disc. Furaday SOC., 1971, 52, 140. 1964,5,113. R. I. Bickley and M. J. Sidorowicz, to be published.

ISSN:0301-7249    

DOI:10.1039/DC9745800194

出版商:RSC    

年代:1974

数据来源: RSC