Numerical processing of wavelength-dispersive X-ray fluorescence spectra

A software package has been developed for the quick processing of wavelength-dispersive X-ray fluorescence spectra. Given the original scanning intensity data and the working conditions of the instrument, the software automatically completes a series of tasks, namely: identifying background, recognizing peaks, performing single-peak fitting, decomposing overlapped peak groups, and finally reporting the positions, net intensities, background intensities, and widths at half maximum of all recognized peaks. The main contributions made by this study are: (1) a criterion called average amplitude of noise (AAN) was proposed and employed to support the completion of all tasks; (2) an algorithm for effective recognition of shoulder peaks was developed by combining first- and second-order derivatives; (3) the Newton-Raphson iteration was modified by introducing a new constraint to accelerate the process of peak fitting and profile decomposition. Extensive computation tests proved that the software developed in this study was robust, fast and precise in the sense that it always gave results very close to and consistent with human experts.