J. Chem. Soc., Faraday Trans. 1 , 1982, 78, 3545-3560 Electron Spectroscopic Studies of Galena and its Oxidation by Microwave-generated Oxygen Species and by Air BY STEPHEN EVANS* AND ELIZABETH RAFTERY Edward Davies Chemical Laboratories, University College of Wales, Aberystwyth, Dyfed SY23 1NE Received 5th March, 1982 The reaction of (100) surfaces of PbS crystals (natural galena) with reactive oxygen species generated by microwave discharge in NO has been compared with aerial oxidation in studies using angle-resolved X-ray and ultraviolet photoelectron spectroscopies. Reaction with the discharge products was much faster, facilitating the study of this process in a controlled environment. The final products of the two processes were similar. Initial attack by atomic oxygen at Pb sites led to the formation of oxide species with the liberation of free sulphur, most of which was lost from the surface.Some, however, remained trapped under an oxidized layer. Sulphate species were also formed, probably by direct attack on PbS. Surfaces of good crystallinity (assessed via their photoelectron diffraction patterns) and chemical purity could be regenerated from oxidized surfaces by 800 eV argon-ion bombardment followed by annealing at ca. 350 O C . Core-level binding energies (&) for bulk PbS, PbSO,, PbO, PbCO,, PbS,O, and 5Pb(OH), * Pb(NO,), were obtained via gold decoration and confirmed by reference to adventitious carbon. The Eb of the surface sulphate and oxide differed from those of the individual bulk lead salts, and they were inferred to coexist in an intimate mixture which probably also contained hydroxide. After the oxidized surfaces were heated to ca.350 O C only the sulphate and some of the oxide remained. The uptake of oxygen at surfaces of lead@) sulphide, a compound which occurs naturally in large single crystals as the mineral galena, has long been of interest. Galena is the principal ore of lead, and processes used in its extraction involve complex oxidation processes at PbS crystal surfaces.l This has led to a considerable body of research [briefly reviewed in ref. (I)] directed towards identification of the chemical species present in oxidized galena surfaces, from which, however, a general consensus has not become apparent. PbSO,, PbSO,, PbS,O,, PbO, elemental sulphur, PbS,O, - PbO, PbSO, - PbO and PbS,O, have all been reported as oxidation products under differing conditions [see ref.(l)], the first five resulting from aerial oxidation although no one group has reported more than three of these species. Two reports of thiosulphate formation (detected by infrared spectroscopy)2v were not supported by more recent X.P.S. data,' from which it was concluded that PbS,O, could not comprise > 5% of the total products. On the other hand, fundamental studies of PbS surfaces themselves, undertaken for their intrinsic interest and because of their technological importance in infrared emission and detection, have yielded a comprehensive picture of their electronic s t r u c t ~ r e . ~ - ~ In addition, these studies have shown that sulphur can desorb from the surface in vacuu, and subsequently be replaced reversibly by oxygen,, and that the adsorption of oxygen at room temperature is molecular, rather than dissociative.However, no discrete oxidized phase could be detected, even after 10l2 Lf expos~re.~ 1 L (Langmuir) = Torr s; 1 Torr = 133.3 Pa. 35453546 OXIDATION OF PbS STUDIED BY ELECTRON SPECTROSCOPY There have thus been no reports of the oxidation of galena through to ‘bulk’ products in an ultra-high vacuum system (where oxidation conditions can be more closely controlled than in external oxidation) : the reaction of pure dry ground-state molecular oxygen with PbS is negligibly slow. To bypass this problem, while retaining control of the oxidation environment, we have adopted a more active oxidant, microwave-excited nitric oxide, which contains excited molecular (l A) and atomic oxygen species.* Using this oxidant we have been able to form a variety of chemical species on galena surfaces, including several of those reported for aerial oxidation. Previous studies have used PbS surfaces prepared either by epitaxial deposition5 or by the cleavage in vacuum of natural galena.4q6 The latter is probably the better method, but it is difficult to prepare surfaces by this means which have sufficient area (ca.1 cm2) for X.P.S. studies. We have therefore investigated the effects of ion bombardment and annealing on air-cleaved crystals to ascertain whether chemically and structurally acceptable (100) surfaces can thereby be regenerated in vacuo as required.The effect of prior ion bombardment alone on the oxidation process has also been investigated. The techniques employed include angle-resolved X-ray and ultraviolet photoelectron spectroscopies (X.P.S., u.P.s.) and X-ray photoelectron diffraction (X.p.d.). EXPERIMENTAL Measurements were made using an AEI/Kratos ES200A electron spectrometer with Mg Kct and He I/He I1 photon sources, a PHI 04-131 ion gun, a rotatable probe and a preparation chamber (base pressure, low 1 0-lo Torr region), in which exposure to microwave-excited NO (2450 MHz; ca 100 W; pressure near sample ca 5 x Torr) was carried out as previously described.s During the course of this work (cu. 3 years) three excitation arrangements, of differing activity, have been used: most of the data were collected using the least efficient source (denoted 11) of ‘active oxygen’.Sources I and I1 were both nominally identical to the previous arrangement,8 but differed in their accuracy of alignment. The most recent source (111) had a shorter distance between plasma and sample, and was much more active than the others. Some results from source I1 and all the data from source I11 were collected digitally using a microcomputer-based data ~ y s t e m . ~ All the electron binding energies (Eb) reported here conform with the standards recom- mended by Bird and Swift.lo However, as a result of incidental instrument repairs and recalibrations, the kinetic energy (Ek) scale offset (the apparent spectrometer work function) varied considerably during the course of the study.Absolute Ek values from spectra relating to oxygen sources I, I1 and I11 are thus not directly equivalent. Six galena crystals were studied, each cleaved in air to ca. 12 x 7 x 2 mm from specimens supplied by Hilary Corke Minerals (from Ladywash mine, Eyam, Derbyshire; cleaves A-D) and by Gregory Bottley and Co. (cleaves E-F). Cleave E failed to yield a flat cleavage and was mechanically polished before examination. However, this very flat surface, ideal for angle- resolved X.P.S. work, cannot have been accurately (100) over its whole area. Heating a galena crystal attached directly to a copper probe tip resulted in a vigorous reaction, with rapid destruction of the crystal, migration of copper to the sample surface, dissolution of the probe tip and the formation of globules of metallic lead at the interface.A gold-foil separator was therefore inserted between tip and crystal in subsequent experiments. All samples initially suffered from carbonaceous contamination. Oxidation of one as-inserted sample was achieved with source I (cleave A; Pb 4flC 1s height ratio 36: 1). However, the presence of contamination often inhibited attack on the underlying sulphide when using source I1 because of its lower oxidizing activity. The most active oxidation source (111) was able rapidly to remove this contamination [cf. ref. (1 l)]. However, most specimens were oxidized only after ion bombardment (700 eV, 2-5 PA) and usually after subsequently annealing in uucuo for 0.5-2 h at the highest available temperature. This was estimated by infrared pyrometry as cu.350-400 OC at the sample surface. Cleave DS. EVANS AND E. RAFTERY 3 547 was first oxidized after annealingwithout prior bombardment. The initial Pb 4f: C is peak-height ratios for these surfaces lay in the range 189-390. Ion-bombarded surfaces and surfaces annealed following ion bombardment were examined before oxidation by both U.P.S. and X.P.S. (Pb 4f, S 23, Ar 2p, C Is, 0 Is), and the angular dependences of the Pb 4fand S 2s peaks were monitored to provide, uia X.p.d. effects, evidence relating to the extent of ion-induced disorder. The angular dependences of the Pb 4f signals were re-examined after prolonged oxidation. Oxidations were conducted stepwise, recording at least Pb 4f, 0 1s and S 2s X.P.S. after each oxidation period.Valence u.P.s., C 1s X.P.S., and/or S 2p X.P.S. were sometimes also recorded. However, few S 2p spectra were collected, because they coincided with energy-loss structure from the Pb 4f peaks (making background subtraction difficult) and their doublet character led to undue complexity in curve resolution (see fig. 3). After extensive oxidation some surfaces were heated and then re-examined by X.P.S. Six bulk compounds of lead were studied to provide Eb for comparison with those of the oxidized surfaces. Commercial samples of PbSO, and PbO were used, while PbSO,, PbS,O, and PbCO, were prepared from AnalaR lead(I1) acetate and the appropriate AnalaR sodium salt in stoichiometric quantities. As the existence of pure Pb(OH), is questionable,', a preparati~n'~ yielding SPb(OH), * Pb(NO,), (a suitable model for surface hydroxide) was adopted. The compounds were dried in a vacuum desiccator, and abraded surfaces of pressed pellet samples14 were examined by X.P.S.The PbO initially exhibited a pronounced doublet 0 1s signal : the high Eb component (OH) was removed by heating in UCLCUO to ca. 400 OC. The Pb 4f, S 2s and 2p, 0 Is, C 1s and N 1s peak areas were obtained to confirm the stoichiometries of the surfaces.14 Each sample was gold-decorated in the preparation chamber and the Eb calculated using Au 4fas reference.15 For the most important model compounds (PbSO, and PbO) and for PbS itself gold was evaporated stepwise to ensure that the reported Eb values were not coverage- dependent, while for the other compounds Au 4f: Pb 4fratios of 1-3 were achieved in a single evaporation.The peak profiles were often distorted, suggesting that decomposition might have accompanied the deposition of gold, although differential sample charging was thought more likely to have been the cause. Eb values for comparison were therefore also calculated from the original spectra using adventitious C 1s as a reference. Gold evaporation was continued on PbS until a visible layer had been deposited. The Au 4f peak intensity was then compared with the average Pb4f intensity before evaporation to estimateI6 the inelastic mean free path (i.m.f.p.) for ca. 11 13 eV electrons in PbS. RESULTS AND DISCUSSION GALENA SURFACES (I) CLEAVED I N AIR On cleave A, inserted into the vacuum within 15 min of cleavage ( i e .after the equivalent of > loll L oxygen exposure), the symmetrical Pb 4f and S 2s peaks had full widths at half maximum (f.w.h.m.) of 1.1 and 2.1 eV, respectively. The 0 1s peak area was < 0.3% that of Pb 4J reflecting well under a monolayer coverage. The absence of any evidence for oxidation confirms previous (11) &+-BOMBARDED SURFACES The Pb 4fX.p.s. peaks were skewed to higher Ek (f.w.h.m. now 1.35 eV), indicating partial reduction to elemental Pb: the broader S 2s peaks were unaffected. When the electron take-off angle was increased from 10 to 6 7 O , the high-E, Pb 4f component was roughly doubled in relative intensity, suggesting that the concentration of elemental Pb increased towards the surface. The angular dependence of the Pb 4fand S 2s peak intensities (fig. 1) showed marked diffraction structure, indicating that the crystal largely retained its structural integrity.3548 OXIDATION OF PbS STUDIED B Y ELECTRON SPECTROSCOPY annealed n 1.0 5 2 1.0 5 U - a 11 N .d - 0.5 - 0 30 60 * ''6' " ' ' ' ' L " 0 I" ion-bombarde d - I .. . l . I . I . . 0 30 60 0 I" FIG. 1 .-Normalized X.P.S. peak intensities from PbS (100) as a function of electron take-off angle, 8, for Pb 4f (0) and S 2s (a) signals from an ion-bombarded surface and a surface annealed after ion bombardment. All take-off angles are given with respect to the normal to the surface. However, recent work on GaAs" suggests that increasing the ion-beam power beyond a critical region would have resulted in more serious disorder. The angle- integrated Pb 4ffS 2s area ratios confirmed that the proportion of lead within the mean sampling depth was much higher than in stoichiometric PbS (see table 1 below).Some argon was retained by the crystal (Ar 2p w 0.35% of Pb 4farea). The Ar 2p E b (relative to the Fermi level) was 242.1 eV, cf. 248.6 eV in the gas phase.18 Approximating the difference between the reference levels in the two measurements as the work function, 3.7 eV,19 yielded the extra-atomic relaxation energy for the implanted Ar as 2.8 eV, close to those for monatomic Ar implanted in other solids.20* 21 No evidence for the formation of Ar clusters22 was found. The He I U.P.S. [fig. 2(a)] showed less fine structure than that of epitaxially grown (100) PbS surface^,^ suggesting that the extreme surface was appreciably disordered. No contribution from the free Pb metal to the He I valence band could be identified, because of the relatively featureless nature of the Pb valence band.ll However, the He II/? Pb 5d signals did show low E b shoulders at ca.0.9 eV separation [fig. 2(a)], confirming the presence of elemental Pb near the surface. Nevertheless, U.P.S. (from a smaller sampling depth) did not reflect a higher concentration of elemental Pb than the X.P.S. data. The elemental Pb was thus not localized at the vacuum interface, despite the increasing concentration of Pb in the near-surface region revealed by the FIG. 2.-He I and He I1 spectra from PbS. (a) Cleave F after ion bombardment (45 min, 800 eV, ca. 2 PA), showing the Pb 5d (metal) contribution at ca.0.9 eV to lower binding energy of Pb 5d (PbS), an identical shift to that seen on Pb 4f[136.9 eV1* cf: 137.8 eV (table l)]. (b) The same surface after annealing. Note the Pb 5d shift to lower binding energy, and the marked angular dependence of the He I spectrum. (c) Polished surface (E) after ion bombardment, annealing and oxidation by source I11 for 15 min. The He I spectra were 5-point quadratically smoothed 15 times, and the He I1 spectra in (a) and (6) 50 times [see ref. (23)]. The small residual undulations in these He I1 spectra are not significant. The He I1 spectrum in (c) was smoothed 200 times, and raw data are also shown for comparison. The Pb (metal) shoulders in (a) were still detectable after 200 smoothing cycles.S. EVANS A N D E. RAFTERY 3 549 .0.3 eV - . . 1 - . Pb 5d(Ht . ?. He I1 \* - *. . . ' . - . . . . . . . . *..J . - - . ". FIG. 2.-For legend see facing page. F A R 13550 OXIDATION OF PbS STUDIED BY ELECTRON SPECTROSCOPY X.P.S. angular dependence. The presence of excess Pb distributed below the surface would be expected to make the crystal strongly n-type throughout the X.P.S. sampling depth:4 this expectation was confirmed by the shift of 0.2kO.I eV to lower Ek found on ion bombardment for the Pb 45 S 2s and Pb Sd(He IID) peaks [Fig. 2(a) and (b); table 1 below]. The Au 4f-Pb 4f energy separation was smaller in the first stages of the stepwise gold decoration. This may have resulted from alloy formation with the free Pb released by the ion bombardment, or from changes in the Au 4f Eb with crystallite size.15 An appreciable Pb (metal) peak (ca.14% of the Au 4f area) remained even after deposition of a visible layer, due apparently to surface segregation of the free Pb. A correction for this was made when estimating the i.m.f.p. for I 1 13 eV electrons in PbS. A value of 21 A was obtained, less than the formula of Seah and D e n ~ h ~ ~ for compounds suggested, but close to that predicted for an elemental solid. (111) SURFACES ANNEALED AT 350-400 O C AFTER ION-BOMBARDMENT No metallic Pb was detected either by Pb 4fX.p.s. or He I1 u.P.s., which showed symmetrical Pb 5dpeaks of f.w.h.m. 0.8 eV, close to the value reported5 for epitaxially grown material. The Ar 2p peak had been eliminated, and the Pb 4fand S 2s X.P.S. peak widths and Eb values were the same as for cleaved PbS.The angle-integrated Pb 4f/S 2s intensity ratio (cf. table 1) was much lower than for the ion-bombarded surfaces, confirming that the excess Pb had been eliminated without serious loss of sulphur. The X.p.d. patterns (fig. I) were slightly enhanced relative to those from ion- bombarded surfaces, reflecting improved order. The patterns were also slightly superior to those from an air-oxidized cleaved surface (F). The Pb and S X.p.d. patterns were similar, but not identical: although the Pb and S sites have the same geometry and coordination they differ in the identity of the neighbouring atoms. As expected25 the patterns resembled closely those from (100) surfaces of the isostructural salts NaCl and KBr.26 The He I U.P.S. valence band [Fig.2(b)] resembled that from epitaxially grown PbS.5 Marked angular variations in the relative intensities of its several components were found, as expected for a well ordered single crystal. Ion bombardment followed by annealing thus yielded (100) PbS surfaces generally comparable with surfaces prepared by cleavage or epitaxial growth. BULK COMPOUNDS Stoichiometries derived from the X.P.S. peak areas are shown in table 1. All are in satisfactory agreement with the expected formulae, confirming the chemical integrity of the sample surfaces. The separations of the Au 4fand Pb 4f peaks were smaller for low Au coverages on both PbS (see above) and PbSO, although not on PbO. However, they became constant over the range of coverage preferred for energy calibration, 0.3 < (Au 4f/Pb 4f) < 3.Au 4f-derived E b values from this coverage range are compared in table 1 with E b values obtained using the adventitious C 1s peak as reference. The two methods gave results within 0.4eV for all but one compound (the hydroxide, with a discrepancy of 0.7eV). These E b values could therefore be applied within such error limits in comparisons with surface species. OXIDATION OF GALENA SURFACES (I) RESULTS The major features of the oxidation process were similar with all active oxygen sources and all methods of sample preparation. Chemical compositions estimated from the spectra are given in table 2.S . EVANS AND E. RAFTERY 3551 TABLE 1 .-STOICHIOMETRIES AND BINDING ENERGIES FOR LEAD COMPOUNDS FROM X.P.S. core-electron binding energiesa stoichiometry Au reference from peak C reference ‘fitted’ compound areas (k 0 1s S 2s Pb 4f,7/2 Pb 4-7,~ Pb 4f712 PbSC PbS(Ar+) PbSO, PbS203 PbSO, PbO 5Pb(OH), Pb(NO3), PbCO, Pbl.O0SLOld - 225.2 Pb1.22S1.ood - 225.3 Pbo~Q2Sl~o04~l 531.8 232.6 226.1 Pbl~03Sl~o03~l 531.2 231.0 N 1s c 1s Pbl.oCo.Q703.1 531.2 138.8 pb1.08s2.002.9 531’2 23 1.7 Pbl .Oo0.Q2 529.1 - pb5.8016.0N2.3 532’1e 405.9 137.6 137.6 - 137.8 139.5 139.3 138.7 138.4 138.8 139.2 139.1 139.0 - 137.8 138.0 138.3 138.0 138.7 138.3 - - 289.3 138.9 137.7 a Au 4f7,2 = 83.98 eV; C 1s = 284.7 eV.l0 ‘Fitted’ binding energies are the apparent values required to construct the best approximation, judged by eye, to the spectra of oxidized galena surfaces on the assumption that these phases might be present and possibly charged.They have no significance in any other context. Using Pb 4f, 0 Is, C Is, N Is, S 2s and S 2p peaks (as appropriate) except for PbS20, (S by S 2p only), and the method of ref. (14). Heated to ca. 380 O C , with or without prior ion bombardment: stoichiometry given is average of three surfaces, showing k 10% variation. The Au-referenced binding energies were derived from those for the ion-bombarded surfaces by the subtraction of the mean kinetic-energy shifts observed on annealing after bombardment. Peak intensities integrated over all accessible take-off angles (cf. fig. 1). Major peak; minor peak at ca. 529.2 eV. Only cleave A (unannealed) was oxidized with source I. Reaction had almost ceased after 60-120min oxidation, yielding the spectra shown in fig.3(a). No angular- dependence data were taken on this sample. In the earlier stages, the high-E, 0 1s feature was more prominent, as in the spectra reported below. The effect of heat on the final oxidized surface is shown in fig. 3(b). Several crystals were oxidized with source 11. Data from one typical cleaved surface, oxidized after (a) annealing, ( b ) argon-ion bombardment alone and ( c ) bombardment followed by annealing, and obtained in each case at two electron take-off angles, are shown in fig. 4. Reaction became too slow to monitor after ca. 120 min exposure. No major differences were observed in the rates of reaction following different surface preparation and the kinetic data were therefore plotted together (fig. 5 ) . No core-level photoelectron diffraction effects were detected for any of the oxidation products.Two crystals were oxidized with source 111, one cleaved (examined both before and after argon-ion bombardment) and one polished (studied after bombardment). As the reaction was much more rapid using this source, the earlier stages of oxidation could not be studied. Typical X.P.S. and U.P.S. are shown in fig. 6 and 2(c), respectively. Exposure of a crystal to air for several days before insertion into the vacuum system resulted in substantial oxygen uptake, with the development of chemically shifted Pb and S peaks similar to those produced by the excited oxygen sources. The oxidized layer was, however, very thin, and these shifted components were only obvious at high 115-2t b ) 0 Is c Is s 2s Pb &f A AA l .. . . 1 . . . . 1 . . -. 1 . . ' . ' . . L 715 720 725 960 965 970 FIG. 3.- indicate c P b 4f energy loss+ El m U W 4 rn r m 0 c c3 ? S2- kinetic energy/eV I . . . . l . . . . l . . . . l . . . . l . . . . l . . . . l . . . . I . . . . l . v 1080 1085 1090 1095 1100 1105 1110 1115 1120 z v1 cd kinetic energy/eV m -0 Is, C Is, S 2s, S 2p and Pb 4fX.p.s. (0 = 15O) from cleave A: (a) after 126 min oxidation using source I; (b) the same surface after heating to 350-400 O C . Assignments 0 the identities of species only: the correlation lines do not here accurately preserve either the Eb or Ek separations of the bulk compounds. In (b) there appears (uniquely) 2 0 0 cd .e to be an appreciable charging shift for the oxidized layer.2S. EVANS A N D E. RAFTERY 3553 1: PbO PbS kinetic energy/eV PbS,O, ‘ Pb(OH 1; PbCO PbSO, 1 I ......... ..PbCO, ........... ..PbS04 x 1003554 OXIDATION OF PbS STUDIED BY ELECTRON SPECTROSCOPY n c CI ." a 5 ." c ' 0 % a 0 60 120 I 1 0 0 0 I I200 I 1 --T-7--- 0 60 120 time/min time/min FIG. 5.-Plots of peak height against oxidation time (source 11) for the lowest (a) and highest (b) Ek S 2s components [assigned to sulphate and sulphide species, respectively: see section ($1, and the low-& (c) (largely sulphate and hydroxide) and high-E, (d) (oxide) 0 1s components. Bars indicate ordinate scales in counts per second. 0, Annealed after cleavage; 0 , ion-bombarded; 0, ., a, 0, annealed after ion bombardment . electron take-off angles (cleave F; see table 2 and fig.7). The substrate Pb 4fpeaks still gave strong X.p.d. patterns, comparable with those from other surfaces. (11) OUTLINE ASSIGNMENT OF MAJOR X.P.S. COMPONENTS The two new components of the S 2s signal seen in fig. 3(a), with roughly equal intensities at the one take-off angle used, might suggest the presence of thiosulphate, but such an assignment is not compatible with the later data. Similar spectra obtained with source 111 had an angular dependence (fig. 6 ) which demonstrated that the two components had different depth distributions: the species with the higher Eb was nearer the surface than the other. The lower Eb component was less obvious when using source 11, but attempts to fit the spectra of fig. 4 using a Du Pont curve-resolver confirmed its presence at low take-off angles.The S 2s spectra recorded at high take-off angles could, however, be fitted with only two components. Comparison of the spectra with those of the bulk compounds then indicated that higheSt-Eb S 2s component must be assigned to sulphate, and the other new component to elemental sulphur. None of the compounds other than thiosulphate had a S 2s component near the relevant energy, and the peak was close in energy to that expected from data for sulphur itse1f.l. 27* 28 The low-& 0 1s component was similarly assigned to oxide species, the sulphate 0 1s ionizing to higher Eb. However, the bulk Eb values were not an acceptable match to those of the oxidizeds 2s . ...: . .. ... .. . .. ~ 8 . i . .. . 1. c .% . * I . * * . . .. . . . . . . , J . . . ' l . ' " 1 " 15 720 725 . . . I , r , , l , , , l l . , . . l , . l . l r r . . l 1010 1015 1020 1025 1030 1035 .. . . .. . . . . \. , -. . . . . . . :*. . . . . .: . , . . *.: . . * p.::.,: .. ... :;*:-.*::- .. . . . I *;. '. . .:. . . . . . .* . .. .. . . * .: .* .*. :: .,*. . :. * . . ..:. - . :* .a- so:- . .- . z. - . . . . I t .. e =loo *----------..- I " ~ ~ ~ ~ ~ ~ ~ ' ~ ~ 1104 1109 1114 1119 1124 1129 . . . . ' . 9 = 67' FIG. 6.-0 Is, S 2s and Pb 4fX.p.s. recorded at two take-off angles from the polished PbS surface (E), after 3 min oxidation using source 111. Estimated curve resolutions are shown for the S 2s spectra, which still had rather poor signal-to-noise ratios even after data collection times of cu. 60 min.3556 OXIDATION OF PbS STUDIED BY ELECTRON SPECTROSCOPY .. . . . . ' . . '2: ,..#' . . . * -. . . 'I i , , p FS. EVANS A N D E. RAFTERY 3557 surfaces. The correlation lines on fig. 4 were drawn preserving the energy separations between peaks, but adjusting their absolute positions to give the best mean fit (estimated visually) to the spectra. The optimum shifts required varied considerably in both magnitude and direction (see table l), making it clear that charging of the oxidized species was not responsible. Although alignments for most of the compounds studied are shown, it must be stressed that conclusive evidence was available at this point only for elemental sulphur, sulphate and oxide species. (111) QUANTIFICATION OF X.P.S. INTENSITY DATA Accurate quantification was not possible, because the species produced had different depth distributions within a sampling depth which increased between the 0 1s and the higher-Ek S 2s and Pb 4fpeaks.Nevertheless, a useful indication of surface stoichiometry was achieved, using a model established for homogeneous solids14 but omitting the factor EE.5 allowing for the variation of sampling depth with Ek. This is accurate only for infinitely thin overlayers, and we may thereby have underestimated the oxygen content by up to 25%. Attempts were made to rationalize the resultant stoichiometries for a number of oxidized surfaces. The results are shown in table 2. Resolution of the S 2s spectra into three components yielded the PbSO,, elemental sulphur and PbS mole fractions, and estimation of the proportion of the 0 1s signal in the low-E, (oxide) shoulder provided the PbO mole fraction.If any detected Pb remained unassigned, the presence of Pb(OH),, formed from traces of water in the oxidant, was inferred. The formation of PbCO, from carbonaceous contamination cannot, however, be ruled out as an alternative explanation. Excess oxygen was frequently observed, and some may have been associated with the carbonaceous material itself. The discrepancies between the rationalized and observed oxygen contents must arise partly from the inadequacies of the quantification procedure, but resolution of selected 0 1s spectra using the Du Pont instrument confirmed that the high-Eb peak always consisted of at least two components. Unfortunately, the variety of reasonable fits precluded any quantitative use of the resolved spectra.(IV) INTERPRETATION A N D DISCUSSION The good general agreement between the Eb values determined independently by reference to C 1s and Au 4fstandards (table 1) suggested that a chemical explanation be sought for the poor correlation between the Eb values on the oxidized surfaces and those of the bulk compounds [see section (II)]. As the Eb of the end-products of aerial oxidation (fig. 7) were closely similar to those generated by excited oxygen, the formation of metastable oxidation products by excited-oxygen attack seemed unlikely. Experimental error was also discounted, especially as a similar discrepancy had already been reported' between bulk PbSO, and the surface sulphate species produced by aerial oxidation.The disappearance of the X.p.d. patterns on oxidation demonstrated that the oxidized layers were not epitaxial. Indeed, the basic sulphate formed was clearly not even a well characterized polycrystalline compound : the sulphate/oxide ratio (see table 2) varied with oxidation time, take-off angle (i.e. depth) and, especially, from one source to another. The only chemical shift which could be accurately determined (k0.2 eV) for the layer was S 2s (SO:-)-S 2s (PbS), and this too varied, from ca. 7.2 eV (source I) through ca. 6.8 eV (source 11) to ca. 6.6 eV (source 111), loosely paralleling changes in the sulphate/oxide ratio. The oxidized layers evidently possessed neither regular structure nor uniform composition. Their existence as intimate mixtures of species, rather than discrete phases, seems the best explanation of the3558 OXIDATION OF PbS STUDIED B Y ELECTRON SPECTROSCOPY TABLE 2.-sTOICHIOMETRIES OF OXIDIZED PbS SURFACES FROM X.P.S.mol xb (1 00) fig. no. surface treatment cleave A as inserted cleave D argon-ion bom barded cleave F argon-ion bombarded cleave E polished, argon-ion bombarded cleave D argon-ion bombarded, annealed cleave D annealed only cleave E polished, argon-ion bombarded cleave F argon-ion bombarded cleave F 126 rnin ox. source I Above heated 35 min ox. source I1 30 rnin ox. source I1 30 rnin ox. source I1 97 min ox. source I1 127 rnin ox. source I1 3 min ox. source 111 1 min ox. source I11 air for several days 15 15 15 67 10 67 10 67 15 67 15 67 10 67 10 67 67 in surface layer substrate 0 / O a PbSO, S PbO Pb(OH), PbS Oc(%) - 17 65 16 23 23 19 14 13 15 17 25 36 12 11 18 19 14 15 - 7 - 14 4 12 4 14 - 16 - 10 - 6 - 21 31 35 47 62 41 47 56 48 38 57 26 37 68 89 57 57 30 38 19 273 - 30 23 15 15 23 27 29 10 18 23 25 6 33 42 25 15 33 47 27 23 9 12 - - 19 55 25 23 35 96 105 103 98 74 134 153 113 159 69 71 63 61 100 127 128 146 86 a At a takeoff angle of 10-1 5 O 63 % of the signal comes from within a depth of 16 8, (0 1s)-21 8, (Pb 4f) : at 67 O these depths reduce to 6-8 A.From X.P.S. peak area ratios as described in the text. The figures for the substrate reflect the proportion of PbS detected, the sum of all the overlayer species being set to 100%. Percentage of detected oxygen included in formulation: if < 100, the 0 1s peak was larger in area than required for the rationalization given, and vice versa.The large deviations often found emphasize the limited accuracy of the procedures used. Resolution of the 0 1s (oxide) peak was particularly subjective, small changes in position and width permitting large changes in its peak area. Whenever excess oxygen was detected, the size of this component was minimized, to optimize the overall match to the total 0 1s intensity (i.e. maximizing the proportion of hydroxide). It is possible that in many entries oxide has been substantially underestimated, and hydroxide correspondingly overestimated.S. EVANS A N D E. RAFTERY 3559 differences in Eb values between the oxidized surfaces and the individual bulk compounds. Core-electron Eb values depend not only on the charge on an atom but also on the potential due to the surrounding ions, and the magnitude of their electronic relaxation as the electron leaves.The angular-dependence experiments [see section (11) above, fig. 4 and 6 and table 21 all showed (by the disappearance of its S 2s components at high 9) that there was a negligible proportion of elemental sulphur at the vacuum interface, while the 0 1s spectra (fig. 6) revealed a tendency for the proportion of oxide to increase towards that interface. The elemental sulphur was thus trapped at the galena surface by an overlayer of sulphate and oxide. The observation (e.g. fig. 4) of high-9 spectra showing a galena S 2s component but no elemental sulphur was probably a consequence of protection of part of the original surface by carbonaceous contamination [cf.ref. (1 l)]. With oxidation source 111, capable of eroding contamination rapidly, both components were lost at high angles (fig. 6). The large increase in the sulphate/oxide ratio on heating (table 2 and fig. 3) suggested that much of the oxide diffused into the bulk of the crystal at temperatures of 350-400 OC, while the elemental sulphur evaporated. The U.P.S. data confimed that the extreme surface was largely oxide, and that coverage of initially clean galena by oxidation products was complete. The lowest-& part of the He I valence band (largely S 3p) was totally lost early in the oxidation, and the higher-& structure (largely 0 2p), which developed concurrently, resembled that obtained from oxidized Pb.ll Moreover, the Pb 5d Eb value for an oxidized surface [fig.2(c)], close to that of annealed PbS, was much nearer to that expected for PbO (0.2-0.7 eV higher, assuming chemical shifts as in table 1) than to that for PbSO, (1.1 - 1.9 eV higher). The presence of oxide in all the oxidized layers, and its rapid development in the earliest stages of oxidation (fig. 9, suggested that oxide formation was a primary step, attack at Pb releasing elemental sulphur. The rapid reduction in the S/Pb ratio as the reaction proceeded implied that most of this sulphur was lost from the surface. Since elemental sulphur was never detected in the absence of sulphate, the sulphate was probably formed by direct attack on the galena rather than by oxidation of sulphur. The elemental sulphur peaks were much less prominent in the data obtained using the least active source (11), implying that more than one oxygen species was active in the oxidation, and that their concentration ratio varied from source to source.If recombination of 0 atoms were responsible for the reduction in activity in the sources with a long path from discharge to sample, source I1 would have been relatively rich in lA molecular oxygen. It could then be inferred that lA oxygen can oxidize elemental sulphur (with desorption) but not galena. Conversely, the oxide peaks were largest using source I11 (expected to yield the highest concentration of atomic oxygen species), and least obvious following aerial oxidation. These observations again suggested that oxygen atoms initiated disruption of the surface much more readily than did excited molecules.No marked increase in reactivity was observed for the ion-bombarded surfaces, which must have had a much higher concentration of surface defects, indicating that the initial attack of 0 atoms is not defect-dependent. CONCLUSIONS Large-area (100) galena surfaces for surface chemical studies can be prepared by low-energy argon-ion bombardment followed by annealing at 350-400 OC. Such surfaces exhibit good order and do not deviate by more than a few percent from the ideal s t oichiome try. Oxygen atoms attack (100) galena surfaces at Pb and probably also at S sites.3560 OXIDATION OF PbS STUDIED BY ELECTRON SPECTROSCOPY Reaction is not confined to defects. Pb(I1) oxide species form, and while substantial quantities of S are lost from the surface (probably as SO,), some elemental sulphur remains trapped at the interface between the semiconductor and its oxidation products.Sulphate species are also formed, and the sulphate and oxide coexist in an unstructured layer (which may also contain other species, principally hydroxide). Aerial oxidation yields similar products but at greatly reduced rates and in different proportions. Thiosulphate is not a significant oxidation product under any of the conditions studied. When an oxidized surface is heated strongly in uacuo, much of the oxide, the hydroxide and the elemental sulphur are lost (the former by diffusion into the bulk), and only an intimate mixture of lead(rI), sulphate and oxide groups remains.We thank the SERC for support, including an Advanced Fellowship (to S.E.). We also thank D. J. Keast and K. Downing of Gillette Industries Ltd for curve-resolving facilities and assistance, and Prof. J. S. Anderson, F.R.S. and Dr J. M. Adams for useful discussions. A. S. Manocha and R. L. Park, Appl. Surf. Sci., 1977, 1, 129. R. G. Greenler, J. Phys. Chem., 1962, 66, 879. G. W. Poling and J. Leja, J. Phys. Chem., 1963, 67, 2121. T. Grandke and M. Cardona, Surf. Sci., 1980, 92, 385. A. L. Hagstrom and A. Fahlman, Appl. Surf. Sci., 1978, 1, 455. T. Grandke, L. Ley and M. Cardona, Phys. Rev. B, 1978, 18,3847; Solid. State Commun., 1979,32, 353. F. R. McFeely, S. Kowalczyk, L. Ley, R. A. Pollak and D. A. Shirley, Phys. Reu. B, 1973, 7 , 5228. S. Evans, Proc. R. SOC. London, Ser. A, 1978, 360, 427. S. Evans and D. A. Elliott, Surf. Interface Anal., 1982, in press. lo R. J. Bird and P. Swift, J. Electron Spectrosc. Relat. Phenom., 1980, 21, 227. S. Evans and J. M. Thomas, J. Chem. SOC., Faraday Trans. 2, 1975, 71, 313. l2 D. Greninger, V. Kollonitsch, C. H. Kline, L. C. Willemsens and J. F. Cole, Lead Chemicals (International Lead Zinc Research Organization Inc., New York, 1975), p. 61. l3 J. L. Pauley and M. K. Testerman, J. Am. Chem. SOC., 1954, 76, 4220. l4 S. Evans, R. G. Pritchard and J. M. Thomas, J. Electron Spectrosc. Relat. Phenom. 1978, 14, 341. l5 S. Evans, in Handbook of X-ray and U. V. Photoelectron Spectroscopy, ed. D. Briggs (Heyden and Son, l6 S. Evans, R. G. Pritchard and J. M. Thomas J. Phys. C, 1977, 10, 2483. l7 I. L. Singer, J. S. Murday and L. R. Cooper, Surf. Sci., 1981, 108, 7. London, 1978), p. 121. G. Johansson, J. Hedman, A. Berndtsson, M. Klasson and R. Nilsson, J. Electron Spectrosc. Relat. Phenom., 1973, 2, 295. lS J. C. Riviere, Solid State Surface Science, ed. M. Green (Dekker, New York, 1969). 2o P. H. Citrin and D. R. Hamann, Chem. Phys. Lett, 1973, 22, 301. 21 C. D. Wagner, Faraday Discuss. Chem. SOC., 1975, 60, 291. 22 S. Evans, Proc. R. SOC. London, Ser. A, 1980, 370, 107. 23 A. Proctor and P. M. A. Sherwood, Anal. Chem., 1980, 52, 2315. 24 M. P. Seah and W. A. Dench, Surf. Interface And., 1979, 1, 1 . 25 S. M. Goldberg, R. J. Baird, S. Kono, N. F. T. Hall and C. S. Fadley, J. Electron Spectrosc. Relat. 26 S. Evans and E. Raftery, unpublished results. 27 R. K. Clifford, K. L. Purdy and J. D. Miller, AIChE. Symp. Ser., 1974, 71, 138. 2a D. M. Hercules, Anal. Chem., 1970, 42, 20A. Phenom. 1980, 21, 1 . (PAPER 2/392)