Artificial neural network techniques for analysis of ion backscattering spectra

Ion backscattering spectrometry is an analysis technology that is dedicated to the compositional analysis of samples with the thickness of μm level. The problem of spectral data analysis, which is to determine the sample structure from the measured spectra, is generally ill-posed. In this study, artificial neural network (ANN) techniques have been developed for spectral data analysis. A multilayer feedforward neural network was constructed and applied to the specific case of SiGe thin films on a silicon substrate. The network was trained by the resilient backpropagation algorithm with hundreds of simulated spectra of samples for which the structures are known. Then the trained network was applied to analyse spectra with unknown structure of samples. The ANN prediction results are excellent. The constructed neural network can handle properly redundancies, which were caused by the constraint of output variables.