Testing commonly used background removal procedures for XPS quantitation of homogeneous material: Fe2O3, a transition metal oxide example

We have examined, for Fe2O3, the three commonly used background removal procedures: Shirley (iterated), S; Tougaard (a modified U2 version), T; and Linear, L, for the Fe2p and O1s spectra. The high BE end point, which is where it is assumed that there is no further intrinsic signal contribution, is most important. It is quite easy, but arbitrary, to adjust the Fe2p and O1s end points to return an “expected” answer for Fe2O3, 40% atomic Fe, while assuming 100% accuracy for cross sections and for inelastic mean free path length parameters for Fe2p and O1s. This further assumes (a) that the transmission function of the instrument is accurate and (b) that O1s intensity from surface OH is accounted for correctly. We determine the variation in the composition with the variation in the end point used. Ideally, one should include all observable substructures, while keeping the eV range similar for Fe2p and O1s. T and S then give quite similar compositions, despite the fact that T removes far less intensity. The reason is clear. The scattered electron backgrounds are proportional to the photoemission signals generating them, so provided the same procedure over the same eV range is used for both Fe2p and O1s, the functional form of the subtraction is of lesser importance. Owing to this, it may even be acceptable to use a method ratioing total intensities, signal plus background. Absolute compositional accuracy is not the primary subject of this paper, but we do address this in the discussion.