Separation of pure elemental and oxygen influenced signal in ELNES

Abstract

The energy loss near edge structure (ELNES) of many elements is strongly influenced by the presence of oxygen or other elements at surfaces, grain boundaries, or in the bulk material. The presented investigation deals mainly with the influence of oxygen at the surface. A method for the separation of both, the pure bulk signal and the oxidized surface signal, was evaluated and tested on Al, Cu, Mg, and Si. A comparison of experimental data with ab initio bandstructure calculations and other proofs of the accuracy of ELNES separation are presented. Influences of error propagations were tested and are exemplarily given for Al and Si.

Introduction

The fine structure of ionization edges in electron energy-loss spectrometry (EELS) yields information on the oxidation state of the probed atoms. Due to the fact that many specimens oxidize rapidly when in contact with air, one often has signals from the bulk and the oxidized surface in an EEL spectrum. When investigating ionization edge fine structures in alloys or other compounds, it is obvious that one must take into account the influence of the oxide. Furthermore, ELNES can be used to separate grain boundaries and interfaces in spatially resolved analysis [1], [2], [3], [4], [5], [6], [7], [8], [9].

The method is build on the “spatial difference method” introduced by Berger and Pennycook [10] and Bruley [11]. The main difficulty of the spatial difference method is the scaling factor to be used when subtracting the two spectra. Bruley et al. [5] suggests to normalize the spectra to the same intensity in a $\text{50}\text{eV}$ wide window located $\text{50}\text{eV}$ above the edge threshold. This can only be done if the spectra have been acquired at specimen positions of the same thicknesses because of the background signal whereas Scheu et al. [6] reports that these scaling factors can be determined by trial and error using several guidelines reported in [7], [8], [9]. All these methods are well working and important techniques for a qualitative and sometimes semi-quantitative distinction of the ELNES of the same atomic species in different chemical environments such as bulk and interface. On the other hand, they are not sufficiently reliable for an accurate determination of the ELNES, according to the uncertainty in the scaling factor.

The present method relies on experimentally determined scaling factors, allowing separation of the ELNES signal of oxidized and unoxidized material.

The basic idea, as described in Ref. [12], utilizes the fact that the detected signal is a superposition of ELNES of oxidized surface signal and unoxidized bulk signal. Therefore, two spectra with different oxygen contributions are needed. Applying the scaling factor and calculating the difference spectrum results in a spectrum containing only the bulk signal or only the surface signal. (We have chosen the expressions bulk signal and surface signal under the assumption that all oxygen is situated at the surface.) But it is irrelevant, if the oxygen, other elements, or crystal defects appear at the surface only, because the energy-loss spectrum is caused by a two-dimensional projection of ionized atoms contained in the irradiated volume, such as in solid solutions and dispersions.