An almost knowledge-free approach to XPS intensity evaluation where use of atomic photoemission cross sections suffices for yielding material-specific inelastic background

Abstract

A general procedure for peak intensity evaluation on the inelastic background of X-ray photoelectron spectra (XPS) has been developed, which is made up of tasks in three nested levels, level I, II and III. The task in level I is a simple background optimization for the case when the inelastic mean free path (IMFP, denoted λ(E)) and other constraints are given in advance [M. Jo, Surf. Interface Anal., 2003, 35, 729–737]. In order to find appropriate constraints, a batch job, which is a series of single jobs with systematically varied constraints, is constructed and iteratively solved in level II by improving the constraints from the viewpoint of how successfully the background is subtracted. It is found that, by the task in level II, for polycrystal Cu metal, the spectrum after inelastic background subtraction still contains a baseline component proportional to E−k, where E (>50eV) is the electron kinetic energy and the exponent k around 2.4. This baseline is attributed to the high-energy tail of primary excitation spectrum of true secondary electron emission. The relative core intensities thus determined are in good agreement with the Scofield’s photoexcitation cross sections. The relation of core and Auger peak intensities is consistent as demonstrated by the fact that the sum of Cu 2p and 2s equals that of all Cu LMM peaks. The level III calculation is designed for the general case where λ is not available before the analysis. The relative energy dependence of λ except for the absolute value can be estimated by iterating the level II tasks with updated λ(E), based on the assumption that the peak intensities should be proportional to Scofield’s values. The relative deviation of thus determined λ(E) in level III from those by the well-known calculation (Tanuma, Powell, and Penn, TPP) is less than 1.7% between E=600 and 2000eV if their values at 800eV are assumed to coincide. The normalized peak shapes and relative intensities after background subtraction in this level are also very similar to those obtained in level II starting with λ(E) by TPP that is currently known to be reliable. These findings indicate that an almost knowledge-free analysis, which is free from the detail of the solid’s properties such as chemical composition and nevertheless retains the material-dependent results, became possible for the first time, as long as one keeps away from the discussion of absolute intensity.