J. Chem. SOC., Faraday Trans. I , 1986,82, 1721-1731 Surface-charging Effects in the X-Ray Photoelectron Spectra of some Semiconducting Oxides Simon J. Cochran Department of Chemistry, Monash University, Clayton, Australia 3168 Frank P. Larkins* Department of Chemistry, The University of Tasmania, Hobart, Australia 7001 The irradiation of poor electrical conductors in X-ray photoelectron spectroscopy induces an equilibrium charge at the sample surface. The resulting shifts in the spectra have been measured for a number of semi- conducting metal oxides (NiO, Co,O, and ZnO) with a variety of pretreat- ment conditions. Heating the samples, or the adsorption of electron-donating or-withdrawing gases, causes potential shifts which could be related to the electrical properties of the substrate.These effects were investigated for nickel, cobalt and zinc oxides with the adsorbates NO, NO,, NH,, CO and 0,. The reactive system NO + 0, was also investigated with the nickel oxide catalyst. It is found that, for a given sample configuration, this effect is reproducible, and can provide important information on the nature of adsorbed species. Electricalconductivity measurement is a frequently used method for obtaining information on the electronic properties of a solid.l-, Absorption may cause changes in the bulk conductivity of the substrate. Measurement of such changes is, however, subject to a number of experimental difficulties. In particular, sample fragility and poor reproduci- bility have limited the accuracy which can be obtained from the conductivity.Conductivity changes have also recently been shown to be reflected in surface-charging effects in photoelectron This effect is evident in earlier work,’ although its significance was apparently not realised at the time. Most of the earlier studies using classical techniques have employed relatively poor vacua, and it has been demonstrated in a previous paper8 that surface cleanliness is important in the electronic behaviour of transition metal oxides. Therefore, a study of the effects of impurities under u.h.v. conditions may provide further insight into earlier work. Irradiation of insulators and semiconductors with X-rays results in the presence of an equilibrium surface charge as a consequence of the loss of photoelectrons during irradiation.This changes the effective surface potential, resulting in shifts in the kinetic energies of the photoelectron lines. The origins and problems associated with this effect have been discussed by Ebel and Ebel.9 The magnitude of the charge depends primarily upon (a) the rate of loss of electrons by photoionization, i,, (b) attraction of stray electrons to the positively charge sample surface, i,, and (c) conduction through the sample from the earthed holder, i,. This is illustrated diagrammatically in fig. 1. There have been several investigations into the surface charge on insulators and insulated con- ducting films9-12 where the contribution from (i,) has been shown to be very small. However, there have been few attempts to study the charging effects on polycrystalline semiconducting sample^.^ Ley et al.13 did however establish a broad relationship between sample charging and band gap for a series of freshly cleaved single crystals of semi- conductor.For a fixed X-ray flux and sample configuration and assuming the thermal electron 17211722 Surface-charging Efiects - I Fig. 1. Schematic diagram of electron flow under irradiation by X-rays, i, is the current due to ejected photoelectrons, i, represents stray thermal electron and i, conduction through the sample. current (iz) to be constant, the relative charge depends only on i,, thus giving an indirect measure of sample conductivity. It has recently been shown5 that X.P.S. can provide useful information on the donation or withdrawal of electrons by adsorbates.Thus, measurement of sample charge offers additional evidence for the assignment of the spectra of adsorbed species. Sample conductivity depends on the following factors : (a) temperature ; (b) adsorbate coverage ; (c) impurities, e.g. contamination, doping; (d) non-stoichiometry ; and (e) contact resistance between particles of the polycrystalline material and between the sample and the holder. The doping of semiconducting oxides with foreign ions has been the subject of many studies of conductivity.14~ l5 The doping of a p-type oxide such as nickel oxide with univalent ions such as lithium or potassium produces a corresponding increase in the number of NilIJ ions in the NiII lattice. This is equivalent to a downward bending of the conduction band and results in an increase in the conductivity of the material.Likewise, the doping with trivalent iron or chromium ions reduces the number of NiIII charge carriers and therefore the conductivity. It has also been shown that adsorption of a gas capable of accepting electrons from the lattice of a p-type semiconductor also increases the number of charge carriers, while for an electron donating species, the reverse is true.16 By contrast, in an n-type oxide such as zinc oxide, where electrical conduction is by an excess of electrons in the lattice, doping or adsorption with a similar species will have the reverse effect. For example, electron- donating gases such as CO decrease the bulk conductivity of NiO, whereas CO and H, increase the conductivity of Zn0.l' Gases which are only weakly adsorbed as a molecular species, such as CO,, do not change the substrate conductivity.Two potential applications of the variations in surface charge are (1) as an aid in the identification of adsorbed species and (2) in understanding the differences in catalytic properties of materials. The aim of this work is to analyse systematically the data obtained for selected transition-metal oxides, namely nickel oxide (NiO), cobalt oxide (Co,04 ) and zinc oxide (ZnO) with a range of adsorbates, NO, NO,, NH,, CO and 0,. Our goal is to examine the feasibility of quantifying the charging phenomenon and to examine the various factors which contribute to the effect.S. J. Cochran and F. P . Larkins 1723 Experiment a1 X.p. spectra were recorded on a modified AEI (Kratos) ESlOO equipped with a sample preparation chamber as described previously,s using A1 K, radiation operating at 15 kV and 15 mA.Base pressui ,s in the analyser and sample preparation chambers were better than 2 x 10-lo Torr (1 Torr z 133.3 Pa) and 1 x Torr, respectively. Samples were either pressed directly into an earthed copper holder which was attached to a variable- temperature probe or first pressed into a pellet using a 10 tonne press and then mounted on a stainless-steel holder. The samples were rectangular (5 mm x 16 mm) with a thickness of < 0.5 mm when pressed directly into the holder and a thickness of ca. 1 mm when prepared as pellets. The area of the sample irradiated by the X-ray beam was ca. 25% of the exposed surface. Spectra were calibrated using the Au(4f,,,) line binding energy = 83.98 eV).18 Pure oxide samples of nickel, cobalt and zinc were obtained from Johnson Matthey (Specpure grade).A high-surface-area nickel oxide (140 m2 g-l) was obtained from Univar. Samples were normally pretreated in situ in the sample preparation chamber by heating at 300 "C in 100 Torr oxygen for 1 h, followed by evacuating and heating at 200 "C in vacuo to remove excess oxygen. C.P. grade nitric oxide from Matheson was purified by multiple freeze-pumpthaw cycles on a separate vacuum line and passed through a dry-ice-acetone bath before admission to the spectrometer. Gas purity was verified by use of a mass spectrometer. Ammonia and NO, were similarly purified. C.P. grade CO from Matheson and high-purity 0, from CIG were passed through a dry-ice bath before use.The adsorption conditions involved exposure of 0.1 Torr of gas for 15 min at room temperature in the sample preparation chamber for nickel and cobalt oxides prior to evacuation and recording of spectra. For zinc oxide the adsorptions were at 100 K because at room temperature no marked charging effect was observed. Potential shifts were measured as an effective shift in the binding energy of the 0 (1s) photoelectron line of the lattice oxygen from that of the uncharged oxide. This is generally the sharpest line in the spectra of metal oxides and is not sensitive to minor changes in the chemical environment of the surface layer. The lattice 0 (1s) binding energies for the uncharged oxides were 529.6, 529.6 and 530.1 eV for nickel, cobalt and zinc oxides, respectively.The shifts of several other lines in the spectrum in addition to the lattice 0 (1s) line were also measured in a number of experiments, including Auger lines, and all shifts were identical within experimental error. This result was also obtained by Wagner. l1 Peak areas were determined by deconvolution of spectra using Gaussian lineshapes after suitable smoothing, background subtraction and satellite removal. Spectral inten- sities were corrected using theoretical cross-sectionslg and an experimental analyser sensi tivi ty function.20 Results Temperature-dependent Charging Effects The effect of temperature on the surface charge of the nickel, cobalt and zinc oxide samples pressed directly into an earthed copper holder is shown in fig.2. These samples were pretreated in oxygen as outlined in the previous section. After the pretreatment zinc oxide will be oxygen deficient (n-type semiconductor) and therefore conductive throughout the temperature range investigated. Nickel and cobalt oxides behave as p-type semiconductors and are almost stoichiometric after pretreatment. The p-type oxides have relatively low conductivity under high-vacuum conditions and the surface potential depends strongly on temperature. Semiconductor conductivity is known to obey the:1724 Surface-charging Eflects T/K Fig. 2. Effect of temperature on the surface potential of (a) high surface area NiO, (b) Specpure Co,O, and ( c ) Specpure ZnO. equation log (c) = - k / T , where c is the bulk conductivity and k is a constant.21 Because of the variation in conductivity values reported in the literature for these oxides, which demonstrates that the sample history is of great importance, it was not practical to use such values in analysing the present data.Smart22 found that a plot of log (Kh) US. T1 was linear over part of the temperature range investigated. The results of the present study, where the temperature range of interest is below room temperature, are not consistent with a linear plot over a significant temperature range. Charging was found to be completely reversible in all cases, and reproducible to within k0.2 eV between samples with the same configuration and pretreatment. The major uncertainty was in the temperature of the sample surface.This uncertainty poses a restriction on the quantitative information which may be obtained from these results, amounting to ca. & 5% in typical cases. Adsorbatedependent Charging Effects The effect of adsorption on the surface charge for a number of adsorbate-substrate systems is shown in table 1. The oxide samples for this work were pelletised before mounting on the metal holder. Adsorbate coverages, estimated either from signal attenuation or by the method of Carley and were in good mutual agreement and were generally in the range 0.1-0.5 mon01ayers.~~ The measured oxygen adsorption on freshly prepared nickel and cobalt oxides was, however, more than one monolayer. This result was due to oxygen incorporation into the lattice. Comparison of the results for NO and CO on NiO with those of Roberts and Smart5 shows that the present results are consistent in sign, but vary in magnitude.Roberts and Smart recorded shifts of - 2.2 and 0.7 eV for NO and CO adsorbed on a sample calcined at 1450 "C. However, the potential shift on the clean sample was only 3.6 eV, and values obtained on samplesS. J. Cochran and F. P. Larkins 1725 Table 1. Effect of adsorbate on surface charge at room temperaturea substrate NO -0.7 -0.3 1 .o NO2 -2.1 -1.2 2.9 co 0.6 0.2 -2.4 -6.7 -3.5 - NH3 -1.6 -0.7 1.1 0 2 NO+02 - 7.4 - - a Values in the table are potential shifts (eV) relative to the potential on clean surfaces of 8.6, 4.2 and 3.1 eV for NiO, Co304 and ZnO, respectively. ZnO recorded at 100 K, NiO and Co30, at 300 K. Gas exposures were 0.1 Torr for 15 min followed by evacuation at the temperatures indicated.calcined at lower temperatures were less. By comparison, shifts of up to 8.6 eV were observed in this study. Large shifts are by no means uncommon, as shifts of 10 eV or more have been reported by others.2s The differences are attributed to somewhat cleaner surfaces resulting from the pretreatment procedure, variations in contact resistance and a different spectrometer configuration used in the present study: Several of the spectra presented by Roberts and Smart5 also show evidence for metallic nickel due to surface reduction.8 The adsorption of ammonia on NiO provides an illustration of the use of charging effects in the assignment of absorbate spectra. It can be seen from table 1 that ammonia adsorbed at room temperature is electron-withdrawing overall, producing a decrease in surface potential of 1.6 V.However, exposure of NiO to 0.1 Torr of ammonia for 15 min at 100 K in a separate experiment produced an increase in the equilibrium surface potential of 0.4V. This is consistent with the proposition that ammonia absorbs molecularly at 100 K with electron donation into the substrate through the lone pair. A single N (1s) peak is observed at 399.7 eV. As the temperature is raised above 200 K a second N (1s) peak is observed at 398.0 eV and the surface potential decreases. This observation is assigned to the formation of the NH2-species2* at the higher temperature. The effect of raising the temperature on the surface charge of a nickel oxide sample pressed directly into the holder and exposed to 0.1 Torr of NO for 15 min at 100 K followed by evacuation is shown in fig.3(b). Fig. 3(a) shows the curve produced by a clean NiO surface [fig. 2(a)] which was included for comparison. The range over which the two curves diverge (150-200 K) may be associated with the formation of the electron-withdrawing species on the sample surface. At 100 K the dominant adsorbate species has an N (1s) peak at 402 eV, attributed to the molecular species, but on warming the sample a new N (1s) peak at 400.5 eV is observed, which increases in intensity with temperature relative to the 402 eV peak. From the binding energy and surface potential changes the adsorbate species is assigned as NO-.24 It is the relative, not the absolute, potential changes for a particular sample with various treatments which should be reproducible for different spectrometers and experimental conditions.The relative changes provide the information to interpret behaviour at the oxide catalyst surface.1726 Surface-charging Efects Effect of Contamination on Charging The change in the potential shift for a fresh sample pelletised and mounted on the stainless steel holder as a function of pretreatment temperature is shown in fig. 4. In these experiments, samples were heated in vacuo in the spectrometer for 1 h at the temperature indicated and then cooled to room temperature and the spectrum recorded. They were not pretreated with oxygen, as was the case for the results presented in fig. 2. In the case of nickel oxide [curve (a)], as the preheating temperature is increased the surface charge, as measured by the shift in the 0 (1s) line, increases, reaching a maximum at ca.200 "C. This corresponds to the loss of loosely bound water, some of the carbon contamination and other species. Further heating results in the reduction of the oxides, but does not alter significantly the surface potential.* Thus, following reduction, Ni atoms must form small isolated clusters on the sample surface. Only after very significant reduction of the highly active Univar oxide does the charge begin to reduce, indicating some metallic conduction. A similar result is seen for cobalt oxide [fig. 4(b)]. However, zinc oxide [fig. 4 (c)] displays a different characteristic. As the temperature is raised the potential goes through a maximum before reducing to zero.This may be attributed in the first instance to the loss of mobile hydroxyls on the surface, which decreases conductivity, followed by the loss of lattice oxygen to produce the oxygen-deficient non-stoichiometric oxide with increased conductivity. The surface oxygen can also be correlated with the change in surface potential of theS. J. Cochran and F, P. Larkins ' O n 1727 0 100 200 300 pretreatment temperature/"C Fig. 4. Effect of preheating temperature on the surface potential of (a) high surface area NiO, (b) Specpure Co,O, and (c) Specpure ZnO. high-surface-area nickel oxide pellet mounted on a metal holder. Fig. 5 shows the effect of the ratio of the surface oxygen peak at 53 1.4 eV to the lattice oxygen peak at 529.6 eV as measured by the relative spectral intensities as described earlier.The surface/lattice oxygen ratio was varied by preheating the sample for an appropriate time in order to remove surface oxygen species. There is clearly a considerable electron-withdrawing effect by such species. With a large amount of surface oxygen, the number of charge carriers is greater, and the potential shift due to chxging is lower. Effect of X-Ray Flux on Surface Charge Fig. 6 illustrates the variations of surface charge with X-ray power for high-surface-area nickel oxide and alumina samples pressed into a copper holder. The result for alumina, obtained at 298 K, is consistent with previous studies9* 22 It is obvious that the characteristics of the two materials are quite different.Whereas the charge on alumina saturates at a power well below that normally used in X.P.S. experiments, for high- surface-area nickel oxide studied at 100 K the charge does not saturate until a substantially higher flux is used. It is clear that in this case sample photoconductivity does reduce the surface charge on nickel oxide at low temperature. For samples held at low temperature, the thermal effect of the X-rays may also be important. However, on exposure of a sample with an adsorbed gas to the X-ray beam, only a small increase in system pressure was observed as a result of sample heating. Ebel and Ebelg have investigated the relationship between sample charge and X-ray photo flux for a number of insulators. They found that for their system charging was saturated at ca.4 kV and 160 W. The present work is primarily concerned with the effect on charge as a function of sample treatment on semiconductors. However, fig. 6 shows1728 Surface-charging Efects 10 8. > 2 6 .A Y 5 Y 0 a 0 g 4 v) 2 0 o( 1s)531.6/0( lS)529.6 Fig. 5. Variation of surface potential with non-stoichiometric oxygen measured by the spectral intensity ratio on high surface area NiO. 10- 210 1 1 1 0 5 10 15 X-ray genera tor volt age/ kV Fig. 6. Effect of X-ray flux on surface potential: (a) alumina at 298 K, (b) high surface area NiO at 100 K.S. J. Cochran and F. P. Larkins 1729 that under normal operating conditions in our spectrometer (15 kV, 15 mA) the semiconductors can be charged further by increasing the X-ray flux.Discussion Correct potentials are, in general, non-ohmic and will vary with crystallite size, the contact resistance between the particles of the polycrystalline material and the contact resistance between the sample and the holder. In this work, samples which were pressed into a pellet and then mounted on a stainless steel holder were in poor electrical contact with the spectrometer and exhibited larger charging shifts and line-broadening effects than those which were pressed directly into a copper holder. For example, the high-surface-area nickel oxide mounted as a pellet and used for the adsorbate study reported in table 1 had a potential shift of 8.6 eV at 300 K for the clean surface, while for a sample of the same oxide directly pressed into the holder and used for the temperature-dependence study presented in fig.2 and 3, the potential shift at 300 K was only 0.1 eV. An initial relatively high surface potential was an advantage for adsorbate studies because changes due to the presence of different adsorbate species were more easily observed at room temperature. The data in table 1 indicate, for example, that when NO is adsorbed on a high-surface-area nickel oxide pellet at 300 K a potential shift of - 0.7 eV was recorded relative to the potential on the clean surface; however, for a sample pressed directly into the holder and used to obtain the data presented in fig. 3 the relative potential shift was only - 0.1 eV at 300 K with the NO adsorbate present. Although the magnitudes of the energy shifts are different, they have the same sign in both cases and one can conclude that the adsorbate species is electron withdrawing from both experiments.Similar circumstances prevailed for the study of the effect of surface contamination and the charging study presented in fig. 4 and 5. For pressed samples relatively good electrical contact is established between the thin sample (< 0.5 mm thickness) and the metal support; consequently, the major conduction path for such polycrystalline oxides is in this direction and not across the irradiated surface. This does not imply, however, that bulk conductivity is higher than conductivity on the surface of the particles. This observation is the consequence of a geometric effect related to the shape and size of the pellets and the fact that polycrystalline materials are used.Measurement of surface charging can aid interpretation of adsorption results. The temperature dependence of the charging associated with ammonia adsorption on nickel oxide mentioned earlier and that for nitric oxide presented in fig. 3 are interesting examples. In the latter case the potential shift relative to the clean oxide surface is -4.8 eV at 150 K, compared with only -0.1 eV at 300 K. The change can be related to the concentration and nature of the adsorbed species. Another interesting finding is that the coadsorption of NO and 0, results presented in table 1 shows an additive effect on the surface charge. This finding is consistent with the independent adsorption of these gases. NO may adsorb onto surface nickel atoms, while 0, may dissociate into 0,- ions which are incorporated into the lattice.From the data in table 1 it is concluded that the adsorption of oxygen on NiO has a more dramatic effect on the surface potential (- 6.7 eV) than either NO (-0.7 eV) or CO, (0.6 eV), consistent with the incorporation as 02-. Previous did not observe such changes in the spectra on exposure to oxygen below 350 "C. The difference is attributed to the higher surface area and stoichiometry of the oxide used in the present study. The absence of significant spectral broadening effects in our work due to differential charging indicates that the part of the catalyst surface being analysed acquires an effectively uniform potential. For example, in the case of NiO, the lattice 0 (1s) line was broadened from 1.4 to 1.8 eV when the surface charge increased from 0.0 to 8.6 V at1730 Surface-charging Efects room temperature.However, more severe broadening effects of up to 2.8 eV, probably due to thermal effects, were observed on samples cooled to 100 K. Smart2, has investigated the effect of temperature on the surface charge on nickel oxide in the range 20-400 "C and has interpreted the change in terms of a band model and calculated an activation energy of 0.23 V. However, it was assumed in that paper that the bulk conductivity (a,) was independent of temperature. This is clearly not a good assumption, as demonstrated by the present work, and it may be that about half of this activation energy can be accounted for by variations in oo.For adsorbate studies clearly the relationship between surface charge in X.P.S. and bulk conductivity is not a simple one, and further work is necessary for quantitative interpretation. In the case of a p-type semiconductor the adsorption of an electron-withdrawing species effectively increases the number of charge-carrying positive holes and thus the conductivity. The reverse is the case for an n-type semiconductor. This may be explained in band theory by a bending of the conduction bands downwards, thus they will have a higher occupancy and the material will become more conductive. It has generally been assumed in previous r e ~ e a r c h ~ ? ~ that the current from stray electrons (i, in fig. I) is independent of the sample charge. This is not necessarily a valid assumption, particularly where large potential changes are observed.Unfortunately, this assumption is not easy to test experimentally. Smart2, has proposed a mechanism such as Zener breakdown to account for the surface charge reaching a maximum as conductivity decreases. However, an increasingly positive surface should become a more effective sink for stray electrons, and this dependence cannot be neglected. Conclusions Changes in the effective surface potential arising from temperature and adsorption effects can provide important information on the electronic properties of semiconducting surfaces. Electron donation and withdrawal by adsorbates can readily be detected, even when spectra of the adsorbates are very weakly bound. NO, NO, and NH, adsorb as electron withdrawing species, while CO is electron donating. Although a comprehensive quantitative theory has yet to be developed, semi-quantitative data can provide a basis for comparison between different adsorption systems.Surface charging is a complex phenomenon depending upon several factors, but with careful work the effects are reproducible. By suitable choice of sample geometry and temperature; the effects of adsorption can be studied for a range of systems more conveniently than by traditional techniques. Provided the X.P.S. technique can be adequately calibrated, it offers a rapid and simple means of obtaining further information about the adsorption process which will assist in the assignment of adsorbed species. Financial assistance from the Australian Research Grants Scheme for this research is gratefully acknowledged. References 1 S.P. Mitoff, J. Chem. Phys., 1961, 35, 882. 2 R. W. Wright and J. P. Andrews, Proc. Phys. Soc., 1949,62A, 446. 3 E. Heiland, E. Mollwo and F. Stockman, Solid State Phys., 1959,8, 191. 4 M. W. Roberts and R. St. C. Smart, Chem. Phys. Lett., 1980, 69,234. 5 M. W. Roberts and R. St. C. Smart, Surf. Sci., 1980, 100, 590. 6 M. W. Roberts and R. St. C. Smart, J. Chem. Soc., Faraday Trans. I , 1984,80,2957. 7 W. P. Dianis, Ph.D. Thesis (Northwestern University, 1974). 8 S. J. Cochran and F. P. Larkins, J. Chem. Soc., Faraday Trans. I , 1985,81,2179. 9 M. F. Ebel and H. Ebel, J. Electron Spectrosc. Relat. Phenom., 1974, 3, 169. 10 H. Ganska, H. J. Freund and G. Hohlneicher, J. Electron Spectrosc. Relat. Phenom., 1977, 12, 435.S. J. Cochran and F. P. Larkins 1731 1 1 C. D. Wagner, J. Electron Spectrosc. Relat. Phenom., 1980, 18, 345. 12 R. B. Bjorklund and J. Lundstrom, J. Catal., 1983, 79, 314. 13 L. Ley, R. A. Pollak, F. R. McFeely, S. P. Kowalczyk and D. A. Shirley, Phys. Rev. B, 1974, 9, 600. 14 A. Bielanski, J. Deren, J. Haber and J. Sloczynski, Trans. Faraduy Soc., 1962, 58, 166. 15 J. Notwotny and J. B. Wagner, Bull. Pol. Acad. Sci., Ser. Sci. Chem., 1973, 21, 931. 16 I. A. Myasnikov, Russ. J Phys. Chem. (Engl. Transl.), 1960, 34, 183. 17 Y. Kobokawa, Bull. Chem. SOC. Jpn, 1960,33,739. 18 R. J. Bird and P. Swift, J. Electron Spectrosc. Relat. Phenom., 1980, 21, 227. 19 J. H. Scofield, J. Electron Spectrosc. Relat. Phenom., 1976, 8, 129. 20 A. Lubenfeld, MSc. Thesis (Monash University, 1977). 21 J. B. Goodenough, Prog. Solid State Chem., 1971,5, 145. 22 R. St. C. Smart, Surf: Sci., 1982, 122, L643. 23 A. F. Carley and M. W. Roberts, Proc. R. SOC. London., Ser. A, 1978,363,403. 24 S. J. Cochran and F. P. Larkins, unpublished results. 25 M. W. Roberts and R. St. C. Smart, Surf. Sci., 1981, 108, 271. 26 K. S. Kim and N. Winograd, Surf. Sci., 1974,43,625. Paper 51945; Received 3rd June, 1985