Guide to XPS data analysis: Applying appropriate constraints to synthetic peaks in XPS peak fitting

Abstract

Peak fitting of x-ray photoelectron spectroscopy (XPS) data is the primary method for identifying and quantifying the chemical states of the atoms near the surface of a sample. Peak fitting is typically based on the minimization of a figure-of-merit, such as the residual standard deviation (RSD). Here, we show that optimal XPS peak fitting is obtained when the peak shape (the synthetic mathematical function that represents the chemical states of the material) best matches the physics and chemistry of the underlying data. However, because this ideal peak shape is often unknown, constraints on the components of a fit are usually necessary to obtain good fits to data. These constraints may include fixing the relative full width at half maxima (peak widths), area ratios, and/or the relative positions of fit components. As shown in multiple examples, while unconstrained, less-than-optimal peak shapes may produce lower RSDs, they often lead to incorrect results. Thus, the “suboptimal” results (somewhat higher RSDs) that are obtained when constraints are applied to less-than-perfect peak shapes are often preferable because they prevent a fit from yielding unphysical or unchemical results. XPS peak fitting is best performed when all the information available about a sample is used, including its expected chemical and physical composition, information from other XPS narrow and survey scans from the same material, and information from other analytical techniques.

Insight note: X‐ray photoelectron spectroscopy (XPS) peak fitting of the Al 2p peak from electrically isolated aluminum foil with an oxide layer

“…An additional chemical complication of fitting aluminum oxide peaks is that the signals from aluminum oxide and aluminum hydroxides overlap, and both are expected to be present in the type of oxide analyzed in this work 25 . We emphasize that the successful use of two synthetic peaks to model the aluminum oxide and aluminum signals depends on appropriately constraining them; for example, the two synthetic peaks in each case have the same widths 10 . Finally, while we are trying to account for the underlying physics of spin‐orbit splitting in our fitting of the oxide signal, the existence of one or more doublets is not implied by our analysis of the oxide peak.…”

“…25 We emphasize that the successful use of two synthetic peaks to model the aluminum oxide and aluminum signals depends on appropriately constraining them; for example, the two synthetic peaks in each case have the same widths. 10 Finally, while we are trying to account for the underlying physics of spin-orbit splitting in our fitting of the oxide signal, the existence of one or more doublets is not implied by our analysis of the oxide peak.…”

“…Peaks with a significant amount of Gaussian and/or Lorentzian character, such as Gaussian–Lorentzian sum and product functions and also Voigt functions, are typically used for XPS peak fitting 6–9 . Appropriate constraints must often be applied to fit components 10 . In some cases, asymmetry must be added to these peaks.…”

“…[6][7][8][9] Appropriate constraints must often be applied to fit components. 10 In some cases, asymmetry must be added to these peaks. The residuals of peak fits and a figure-of-merit like the residual standard deviation of the fit are often used to assess the quality of a peak fit.…”

Surface analysis insight note: Analysis of X‐ray photoelectron spectroscopy images with summary statistics

“…7,8 In response to this concern, guides that cover multiple aspects of the technique have recently been written. 1,3,5,[8][9][10][11][12][13][14] Another recent effort to provide highquality tutorial information to XPS users is the Insight Notes in Surface and Interface Analysis. 15 This article is one of these Notes.…”