An analysis is presented of issues involved in peak location for the calibration of the binding energy (BE) scales of x-ray photoelectron spectrometers. These issues include the e †ects of peak asymmetry, the surface core-level shift, and the avoidance of a sloping background when Ðtting spectra for energy calibration purposes. Examples of uncertainty budgets for BE measurements are then presented in which illustrative values are shown for the repeatability standard deviation (for repeated BE measurements of the same calibration peak), the expanded uncertainty (at the 95% conÐdence level) for BE measurements following calibrations based on di †erent numbers of peak measurements, and the tolerance for BE-scale drift and non-linearity for two chosen values (»0.1 and »0.2 eV) of the total expanded uncertainty for a BE measurement (at the 95% conÐdence level). It is recommended that a user prepare an uncertainty budget of this type to show clearly the sources of random and systematic error in BE measurements following a calibration.
The reference data published by the UK National Physical Laboratory for BE-scale calibration were obtained from Ðts with a quadratic function to a group of points comprising the top 5% of each peak. Most commercial x-ray photoelectron spectrometers have software available for spectrum synthesis, and we consider here the use of the commonly available Lorentzian, Gaussian, and asymmetric Gaussian functions for peak location. Illustrative Ðts with Cu spectra (measured with unmonochromated Al x-rays) showed that comparable accuracy and 2p 3¿2 precision could be obtained with Lorentzian and Gaussian functions as with the quadratic-equation method when di †erent fractions of the peak were Ðtted. For this asymmetrical line, the asymmetric Gaussian function allowed better accuracy and precision to be obtained with a greater fraction of the line than was possible with the symmetrical functions.
1997 by