Quantitative Surface Analysis by X-Ray Photoelectron Spectroscopy (ESCA) Robert S. Swingle I1 Central Research Department, Experimental Station, E. 1. du Pont de Nemours and Company, Wiimington, Del. 19898
Energy loss (EL) mechanisms for free electrons in solids are discussed in relation to their effects on quantitative accuracy in ESCA. Changes in measured Nal,/F1, intensity ratios for a series of stoichiometric inorganic compounds are qualitatively explained from structure differences in the energy loss regions of the photoelectrons. Both extrinsic and intrinsic EL mechanisms are operative in these systems. The effect of surface contamination on ESCA intensities is illustrated for NaF. A first attempt at calculation of electron-electron scattering coefficients of the C,, photoelectron in graphite and polyethylene is described.
Since the pioneering work of Seigbahn and his colleagues ( I , Y “ ) , X-ray photoelectron spectroscopy (ESCA) has become a powerful tool for the analysis of solid surfaces. T h e technique has demonstrated the ability to solve surface related problems in materials ranging in composition from organic polymers to “clean” single crystal metals. Consequently, it has become important for the practicing analytical chemist to understand the factors which presently limit quantitative accuracy in ESCA, and to understand when the technique can (or cannot) be applied in a quantitative manner. A number of papers have appeared in the literature which have successfully used relative photoelectron intensities as quantitative measures of atomic surface concentrations (3-10). These studies, as well as applications in our own laboratory, indicate that ESCA can be used as a semiquantitative (< f50% relative error) or even quantitative (< f10% relative error) tool when comparing surfaces of chemically similar samples. In these cases, the same mechanisms should be responsible in each sample for inelastic electron scattering, and the mean escape-depths of the photoelectrons should have nearly the same kinetic energy dependence from sample to sample. Unfortunately, the relative photoelectron intensities from chemically different samples can show wide variations. An illustration of this effect can be found in Wagner’s data on seven sodium and fluorine containing compounds ( I 1 ), where the Nals/FI, intensity ratios, corrected for stoichiometry, were found to vary from I .60 to 3.14. In this paper, X-ray photoelectron energy-loss spectra are compared for some selected Na-F containing compounds. From this experimental evidence, it is shown t h a t the variations in Nals/F1, intensity ratios found by Wagner can be qualitatively understood from different energy-loss processes which are characteristic of each compound. Both extrinsic and intrinsic ( 2 2 ) loss processes are evident in these spectra. Also in this paper, the available theory for calculating relative inelastic electron scattering coefficients is briefly reviewed, and a first attempt in applying calculated coefficients to measured, relative CIS intensities from polyethylene and graphite is described. Last, the effects of surface contamination on quantitative results are demonstrated for NaF.
EXPERIMENTAL Apparatus. Two ESCA spectrometers, a Varian IEE-15 with magnesium anode, and a specially designed spectrometer (CKDEPL) with retarding field analyzer and aluminum anode were used in this study. The CRD-EPL instrument and analyzer have been described in detail (13, 1 4 ) . A Nicolet Model 1072 signal averager was added to this instrument which provided for coherent addition of spectral regions from 20 to 200 eV. Reagents and Procedures. The graphite and Na-F compounds were used as received from the manufacturers. Samples were mounted on double-sided adhesive tape (Minnesota Mining and Manufacturing Co., No. 665), which completely covered the surfaces of the respective sample holders. Calculations. Several expressions have been developed which relate measured photoelectron intensity to fundamental parameters (15, 16). These expressions are more or less complex depending upon which variables (sample density, photon flux, sourcesample-analyzer angular geometry, surface “shading” effects, instrument kinetic energy response function, etc.) are taken into account. For this paper, the simplified expression first proposed by R. C. Wendt ( 1 7 ) will be used. For an infinitely thick sample, the intensity for photoelectron j , Ij(-),can be written 3s
where (Y, is the emissivity of atom j, approximated by atomic photoelectric cross-sections, N j is the number of such atoms per cubic centimeter, K is an instrument response function dependent upon the kinetic energy of the photoelectron, and S, is the total photoelectron scattering coefficient in cm-’ discussed below. If the element of interest is in a layer of thickness X, then the photoelectron intensity I j ( X ) is given as
and when the sample is covered by a layer of thickness l’
where S,* is the total scattering coefficient of the photoelectron in the covering layer. These equations assume no attenuation of Xray flux within the sampled volume. S, can he written as
Sj = E a S j ( M , )
+
bSj(M?)
+ ..
.
(4)
m
Here the coefficients a, b, . . . represent the fractional contribution to the total coefficient of each scattering mechanism S,(M) operative in the matrix of interest. For cases where single electron-electron (e-e) scattering is considered dominant, Equation 4 simplifies to Sj = Sj(e-e)
(5)
Wendt has proposed an equation relating Sj(e-e) to the kinetic energy of the photoelectron, Ej, and the binding energies of all electrons, i, in the scattering matrix which can be written as
where 2 is the number of electrons i in the matrix with binding energy E b , and C is a normalization constant.
