Surface Science Insight Note: A Good Figure‐Of‐Merit Does Not Necessarily Yield Accurate Chemical Composition: Outcome of Mixed‐Metal Sulfate XPS Data Peak Fitting Using Nonlinear Least‐Squares Fitting of Peak Models

ABSTRACT The use of peak models and nonlinear least squares optimization, when applied to highly correlated signals without proper context, is shown to be inadequate for chemical state analysis. However, the use of peak models with physically motivated constraints and sufficient contextual information permits fitting of such models to spectra by nonlinear optimization, the outcome of which is an enhanced understanding of sample composition. A sample containing sulfur in at least two chemical states is used to show that a peak model constructed for S 2p alone, when fitted to a spectrum using nonlinear least squares optimization, is incapable of recovering quantitative information. Extending the peak model for S 2p to include spectra with signal from Mo 3d and S 2s and applying constraints to optimization parameters based on the assumption that molybdenum is solely bonded to sulfate ions recovers a feasible relationship within S 2p that was absent when the focus for optimization was S 2p only.