Automation of deconvolution in X-ray fluorescence

X-ray fluorescence spectra are distorted by various smearings such as the natural width, the window function of a spectrometer, etc. The removing of distortions from measured data numerically by using the deconvolution method is necessary to obtain true information. Many methods are proposed to avoid such problems in the deconvolution procedure as noise enhancement, sidelobes and solution with negative portions. Another important procedure is judging of convergence. Resolution is more enhanced as deconvolution proceeds but at the same time artifacts become conspicuous. We adopt the Minimum AKAIKE's Information Criterion Estimation (MAICE) procedure applied on several deconvolution methods such as the derivative expansion method and the Hilbert-Schmidt eigenfunction expansion method. As a result it is recognized that the MAICE procedure is effective for termination of successive deconvolution. (author)