Surface analysis: x-ray photoelectron spectroscopy ... - ACS Publications

Apr 1, 1986 - Surface analysis: x-ray photoelectron spectroscopy and Auger electron spectroscopy. Noel H. Turner. Anal. Chem. , 1986, 58 (5), pp 153â€...
0 downloads 0 Views 3MB Size
Anal. Chem. 1986, 58, 153R-165R

Surface Analysis: X-ray Photoelectron Spectroscopy and Auger Electron Spectroscopy Noel H. Turner Chemistry Division, Naval Research Laboratory, Washington, D.C. 20375-5000

This fundamental review is on the subject of X-ray photoelectron spectroscopy (XPS) and Auger electron spectroscopy (AES) for the period of 1983-1985. The review will cover the literature abstracted in Chemical Abstracts between November 16, 1983, and November 4, 1985. The review is written in three separate parts for the convenience of the reader: section A, XPS; section B, AES; and section C, combined XPS-AES topics. XPS and AES are used widely for the analysis of surfaces. From about 1970 to the present time, these techniques have grown in acceptance by the scientific community. Much of this activity has been documented in earlier Fundamental and Amlication Reviews in Analvtical Chemistry (1-12). While this review is lengthy, it is not an all-inclusive bibliography of XPS and AES during the review period. The articles have been selected with the idea of the imtxovement in the "state-of-the-art" of these techniques. The goal of this review is to help analysts solve the problems that are encountered in using XPS and AES in a regular laboratory with commerically available equipment. Finally, the names of the authors of the papers cited could not be included in text due to space limitations.

__

X-RAY PHOTOELECTRON SPECTROSCOPY Introduction

X-ray photoelectron spectroscopy or electron spectroscopy for chemical analysis (ESCA) is one of the most widely used techniques for the elemental analysis of the near surface region. Also, the technique gives information about the chemical environment of the observed atoms. Much useful information can be obtained from XPS, even though a complete understanding of binding energies and intensities has not been achieved. There have been a number of general reviews of XPS during the past 2 years (AI-A5). Liquid-phase XPS has been reviewed (A6),and a compilation of gas-phase core level binding energies has been made (A7). Neither of these two topics will be covered in this review. Also, work that has involved synchrotron radiation will not be considered, but several reviews have been written (A8-A11). More specific reviews (but with some general information) in the areas of catalysis and chemisorption (A12,A13), corrosion(A14),coatings (A14,A15), electrode surfaces (A16),biomaterials (A17, A18), Si and Si compounds (A19), and the light rare-earth elements (A20) have appeared also in the reviewing period. Inelastic mean free paths (IMFP) will be considered in the combined AESXPS portion of this review. Factors Affecting XPS Analysis

Binding Energies. One of the most important measurements in XPS is that of the binding energies (BE) of observed elements on a surface. Two approaches to solve this problem have been proposed. A detailed account (with the various sources of errors) about the use of field-emmitter referenced ESCA (FRECSA) for the determination of BE and the kinetic energies of Auger electrons has been given (A21). Several BE for various metals were given; it was concluded that the use of the Fermi edge for calibration purposes is very difficult. Both nonmonochromatic and monochromatic X-ray sources were employed and two different analyzers were used. The other approach involved BE calibrations with several metals (A22). Extremely careful methods for obtaining the data were used and errors of 0.02 eV were observed. In tests at several laboratories with different instruments, the largest sources of errors were in the voltage scaling and zero points. Another

group has proposed the use of noble-metal core levels for the calibration of BE from the published results with two instruments; recommended values were given (A23). Several approaches have been tried to overcome the effects of charging in determination of BE. Changes in the BE of adventitious C-on Pd and A1203have been observed to be a function of coverage, but no such change was found on SiOz (A24). The changes have been ascribed to extra-atomic relaxation for Pd and Al; the lack of change in Si02has been suggested to be due to rearrangement of the SiO, surface. Differences between the main XPS peak and an Au line have been found to be dependent on the Au coverage for P d and BN, but not for NaCl (A25). Extra-atomic effects were suggested in the case of Pd and BN, and a reaction between the Au and NaCl was proposed for NaCl to explain the results. However, Ar+ ion bombardment of the surfaces may havh had an effect on these observations with the insulators. A combination of a Au dot and a negative bias flood gun has been used in an attempt to overcome sample charging (A26). Examples with a series of Ca-containing minerals appeared to give reasonable results. Changes in BE with the formation at monolayer coverages for hydroxyapatite from a NaF solution have been noted (A27). It has been found that heating ash particles can lower the It has been shown for small MnO particles observed BE (M). that the valence band region became narrower and the BE of Mn had shifted higher (A29). Lack of extra-atomic relaxation for the small particles has been offered to explain the results. The increase in BE with smaller clusters of Sn on C has been suggested to be due to Coulombic effects in the final state (A30). However, the shift was reduced by substrate screening, and then became metallic-like as cluster size increased. The same effect has been proposed in a study of Sn and Pb clusters on C (A31). Changes in reference level and a different number of d valence electrons have not been considered important. An investigation of Pd and Pt clusters on graphite has shown that with smaller cluster size, the core level peaks were more symmetric, but shifted to higher BE (A32). It was postulated that smaller clusters tend to reduce screening effects due to the boundary and that final state effects and changes in the work function accounted for the changes in BE. Similar findings have been made with Pt clusters on A1203and TiO,; extra-atomic effects again have been attributed to these observations (A33). Both shifts to higher BE and a broadening have been found for Pt clusters on Si02and poly(tetrafluoroethy1ene) (PTFE) (A34). However, these results have been ascribed to initial state effects, and a theoretical model has given similar findings. The controversy in the explanation of the shifts to higher BE with smaller particle size has not been settled. For certain systems procedures have been suggested for the prediction of BE. The equivalent-core approximation when used with a Born-Haber cycle can give information about chemisorption-induced core level shifts with substrates (A35, A36). Also, heats of desorption and the effect of adsorbed gases on surface segregation can be determined. Examples have been given with CO and N2 adsorbed on Ni(100). A previously developed procedure that utilizes a Born-Haber cycle approach to determine shifts in BE has been applied to Nzf implanted into stainless steel (A37). The results indicated a charge transfer from the N to the metal. A model with partial charges of a diatomic species for IR absorption has been developed that correlates to shifts in BE (A38). Examples with Cu and Ni dithiocarbamates and Ni xanthates have been given. A method that is based upon measured IR stretches of C-H bonds has been used to compute Si BE for various compounds (A39). A correlation of the inductive-

This article not subject to US. Copyright. Publlshed 1986 by the American Chemical Society

153 R

SURFACE ANALYSTS

substituent constant, signal, and shifts in BE with polymers has been found (A40). Differences in the BE of surface and bulk atoms for Au, Cu, and Ag have been reviewed, with both experimental and theoretical points covered (A41). Surface core level shifts in the Cu 2p3 2 level have been found with the use of monochromatic k-rays and high resolution for Cu(100) (A42). The results are in fair agreement with earlier work with synchrotron sources. Although XPS cannot detect H directly, in some materials the effect of its presence can be found. Examples have been demonstrated with various Zr, Nb, and Y hydrides in both the core and valence band regions (A43-A45). There have been some unusual BE reported. An 0 Is peak at 526.6 eV has been ascribed to 0 in a OPb4 environment in solid nPb0-PbSO, (n = I, 2, or 4) (A.46). This shift a pears to be in the wrong direction for such a structure base upon numerous other compounds. Peaks observed at BE of 172-179 eV for CdS have been ascribed to “ghost” Cd peaks when Mg X-ray excitation has been used (A47). The Ughostsnhave been due to stray A1 X-ray lines from a dual anode source. Anomalous chemical shifts in Ba compounds have been suggested to be due to occupied d levels in the initial state ( A B ) . The effect is smaller with Sr and Ca.

B

Factors In Quantltatlve Analysis by XPS

Various procedures that have been employed to quantify XPS have been reviewed (A49). The intensity ratios for ion-bombardment-cleaned noble-metal core-level peaks as a function of kinetic energy (KE) for different analyzers have been determined (A50). From these data the instrument response as a function of energy for different types of analyzers can be derived. In addition, the energy response for these samples has been determined for several different, commerically available XPS spectrometers (A51). A quantitative procedure for a cylindrical mirror analyzer (CMA) has been given for uniform, thin layers (A52). The angular distribut,ion of the emitted electrons, escape depth, and acceptance cone of the analyzer have been taken into account. Examples with etched Si and SiOz were used. Several studies have been made on the problem of collisions of emitted photoelectrons. The effect of elastic collisions on observed core level intensities has been calculated by a Monte Carlo approach (A53). The results indicated that the effect is larger as the atomic number increases with electrons approximately the same KE. The same method has been employed for both clean and covered surfaces (A54). However, several questions remained unsettled from this study. The Monte Carlo approach has been employed to investigate the effect of elastic scattering with nonideal surfaces on quantitative XPS (A55). The results indicated that if the matrix and/or overlayer do not differ from standards of a similar nature, then elastic scattering is not significant. The effect of elastic scattering has been determined from an A1 foil by measuring photoelectrons emitted from the side opposite the X-ray source (A56). At small takeoff angles, the results differ from the simple free atom equation. A procedure to remove background contributions from XPS spectra has been developed (A57). A previously derived function for electron emission from solids that does not have any adjustable parameters has been used. Good results were obtained for the Ag spectrum for a kinetic energy range of 200--900 eV. The use of linear and integral corrections for inelastic background has indicated the latter approach better in most cases (A58). Deconvolution using an electron beam for the removal of background and loss features near core level eak ener ies has been demonstrated (A59). Another deconvoktion metaod has been proposed that employs previously suggested A1 and Mg line shapes has given similar spectra to those from a monochromatic A1 X-ray source. However, the results were better if the highest resolution was used and if the analyzer resolution also can be taken into account (A59). The deconvolution of loss features has been treated with a model that assumes an exponential decrease in intensity from a solid surface of electron emitters (A6O). Both a homogeneous distribution and an attenuated emitter source have been considered. It has been shown for area measurements that the combined Gaussian-Lorentzian function for curve fitting must be done 154R

ANALYTICAL CHEMISTRY, VOL. 58, NO. 5, APRIL 1986

with care (A58). Also, conduction band effects have to be included in area determinations. A nonlinear least-squares approach to fit the 2p region of Ni, Ni--0 compounds, and a Ni/A1203catalyst has been investigated (A61). However, a large number of parameters were needed for good fits, and this makes some interpretations difficult. Elemental sensitivity factors are most important in the application of quantitative XPS. Cross sections have been measured on a commercial XPS spectrometer for 55 elements and compounds at survey scan resolution (A62). In addition, a procedure for peak identification has been proposed. A comparison of quantitative XPS with the use of computed vs. experimental sensitivity factors has been made with Cr-0 containing compounds (A6*3).In most cases the experimental values (compared to the actual composition) were better. On occasion, care has to be used when evaluating intensity measurements. The 3d transitions of As in untreated oxides of GaAs have been observed to have different intensity ratios that depend on many-electron interactions (A64). Fair agreement between a theoretical model and experimental results has been found. An analytical internal calibration procedure has been developed to st,udy dilute Cu-Sn and Ag-Sn alloys (A65). Stoichiometric compounds of these elements were found to have a linear variation in peak heights. In general, area measurements should be used, since changes in peak widths can affect relative peak height determinations. Surface vs. bulk measurements are often important in surface analysis. From previous bulk sensitivity factors, surface atom sensitivity factors have been computed with the use of an assumed expression for IMFP (A66). A method to determine the surface vs. bulk composition of mixed oxide solid solutions has been developed (A67). The procedure uses the ratio of the metal to 0 Is signal; deviation from linearity indicates differences in the surface vs. bulk composition. Quantitative XPS results for Fe-A1203 granules have been found to be in good agreement with electron microprobe analysis and Mossbauer spectroscopy (A68). A method to determine the thickness of an adventitious layer of C on a sample has been proposed (A69). The measurement is based upon a knowledge of the intensities of the C Is and C KVV transitions for a layer of infinite thickness. A procedure to measure overlayers of film growth has been derived and tested successfully with Si02/Si and Si3N,/Si films (A70). The method can be used for most analyzers except the CMA. Irregular surfaces can affect observed XPS spectra. Signal ratios for random samples (i.e. irregular surfaces) have been modeled by layers, hemispheres, and spheres (A71). The derived expressions are similar to those for ideal surfaces. The distribution of elements in small particles (50 pxn) by XPS analysis has been examined (A72). Examples with different carbon blacks were given. Additlonal Toplca

X-ray Induced Auger Transitions. In many analyses X-ray Auger electron spectroscopy (XAES) transitions can give information that can not be obtained from core level lines alone. Quantitative analysis by XPS and XAES without the use of reference materials has been outlined (A73). Results with alloys have been shown to vary by about 2% from bulk composition. XAES KVV transitions have shown narrower lines for ionic materials vs. those from more covalent materials (A74). This is in agreement with theory in which there is electronic rearrangement before Auger electron emission with covalent compounds. The use of the Auger parameter, loss satellite structure, and peak area changes for AI and Al. compounds and alloys has indicated that more information about bonding can be obtained than from just conventional XPS and Auger electron spectroscopy (AES) (A75). The determination of the Coulomb repulsion from XAES can be made by referencing to ionization potentials and BE; benzene was used as an example (A76). For some molecules identification of the various levels involved in the observed spectra can make this determination difficult. XAES and XPS spectra of various Al- 0 compounds have demonstrated differences in energy between the A1 KLqL1KL2,3L2,3and 0 Is transitions (A77). These results were related to the ionicity of the oxygen by a simple model. Different Auger parameters have been observed with various

SURFACE ANALYSIS

thiclmegses of oxide layers and crystallinity on Si (A78). With the use of intensity ratios and two different anodes, electron escape depths gave reasonable agreement with previous results. The use of the electron-beam-induced second derivative lowenergy Auger transitions for comparison to the same X-ray induced transitions (where the large number of secondary electrons makes the observation difficult) has been suggested (A79). An example was given with U and UO,;however, the effect of differences in observed KE if charging occurs,requires care in the use of this procedure. XAES M,,5Nz,,V transitions for Y. Zr, and Nb implanted with H have displayed larger shifts compared to those of the metal's 3 d5,, lines (A80). Differences of up to 2.7 eV have been found. Angle Resolved X-ray Photoelectron Spectroscopy (ARXPS). The use of ARXPS can enhance the information available from regular XPS. This area has been reviewed extensively (A81). The BE of 0 and Ni have been investigated for NiO as a function of various treatments (A82). Surface species of 0-and Ni(II1) have been identified, and ARXPS has indicated that 0 is in surface excess, which is consistent with other observations. ARXPS has been used to determine the distribution of various oxygen-containing layers on GaAs (A83). The van Cittert method was used to improve the resolution, but noise effects might affect the conclusions of some of the depth assignments. ARXPS has been used to obtain chemical state information on oxidized GaAs (A84). Theoretical IMFP and sensitivity factors supplied by an instrument manufacturer were employed. The analysis of the contamination layer on an anodic oxide of Nh,O, on Nb has been made by ARXPS with a CMA (A85). This procedure does not involve moving the sample. The measurement of the orientation, thickness, and coverage of a fluorosurfactant adsorbed on various substrates has been determined by ARXPS (A86). ARXPS has been used to investigate the coverage of thin protein films on fluorocarbon polymen (A87). A simple model has been derived and has been able to estimate the coverage of the adsorbed film, which appeared to be in the form of islands. A model based on the electron takeoff anele for XPS sDectra of Lanemuir-Blodeett films of nocGdecylamine o; Si has suggested that thesk films were not well-ordered (A88). .~~~ Channeling of photoelectrons along internuclear axes has been ohserved hv ARXPS. Enhanced ARXPS spectra of overlayers (monolayer and above coverages) have been observed in several systems (A89). The determination of differences between more uniform layers and islanding bas been suggested. The changes found in ARXPS (and angle resolved AES) have been analyzed and shown to be due to two effects (A90). Intensity is increased along internuclear axes, and a secondary effect is ascribed to structural factors. Similar results have been found from single scattering calculations, which have been compared to ARXPS measurements of e p itaxial layers of Cu or Co on Ni(100) (A9Z). ARXPS of self-intercalated Ti~(I)iV(I),.~V~~II)Sz has shown minima in peak intensity along low Miller index planes for intersitital atoms on the crystal plane of a major component axis (A92). V has two oxidization states that could be resolved in the angular distribution curves. ARXPS curves have been determined for different azimuthal angles for 1T- and 2H-TaSe2 (A93). The variations in intensity with polar angles were roughly correlated with high-energy electron diffraction theory.

With 0 Ka radiation. the effect was much weaker. ARXPS intensity ratios of c(2 X 2)s on Ni(100) have been found to agree well with a previous theory ( A B ) . However, the results had to be averaged over several angles to account for diffraction, and inelastic effects were found for small takeoff angles. Sample Damage. Changes during an XPS analysis due to X-ray radiation or as a result of ion bombardment are possible in some systems. Heating of a sample due to the power being dissipated by the X-ray source also can give rise to changes in the material under study. I t has been found that radiation damage due to Bremsstrahlung produced with 15-keV electrons from an AI anode is about 3% of the total potential damage due to X-rays (A95). Nitric acid treated carbon fibers have been found to undergo changes under X-ray radiation (A#). I t was not determined if the X-ray radiation or heat from the X-ray source caused the changes. Reduction of six Cu compounds during XPS analysis has been investigated on three different XPS spectrometers (A97). Heating of the sample was found to be the largest cause of the observed degradation. Sputtering also has been found to alter materials during surface analysis. In an analysis to determine the surface vs. bulk composition of 1-54 pm environmental particles, it has been noted that sputtering may have altered the chemical state of some of the surface species (A98). The effect of bombardment of HOPG graphite with D+ ions has been observed by XPS (A99). C-D bond formation gave a 0.2 eV increase in BE, while damage caused a negative shift. Broadening of the XPS peak was ascribed to changes in vibrational broadening; saturation models have been correlated with the experimental results. Sputtering of Standard Reference Material 470 (Glass K-411) has shown that the Si and Ca did not change significantly, hut some changes did occur (A100). Damage effects for Fe were found, and 0 appeared to sputtered selectively. The effect of fast neutral or ion bombardment has been investigated for SnOz, PbO,, Si, and polystyrene (AZOZ). The oxides showed little effect, while there was the loas of the shakeup structure in the polystyrene. Several different species were found for Si. The effect of Art (