Neural pattern recognition applied to AES depth profiling

Neural pattern recognition was used to analyse the low‐energy Auger spectra of a thermally annealed Si/Ni/Si layered structure measured during the acquisition of a depth profile. The purpose was to gain information about the chemical state of the elements at the interfaces by processing the data in a way quite similar to conventional target factor analysis (TFA). The new approach, however, has some important advantages: no standards are required, it is extremely fast and it is fully automatic. In principle, there is only one arbitrary parameter, the vigilance parameter ρ, which sets a threshold for the level of similarity required for assuming two spectra as belonging to the same class of data. However, the requirement that the optimal value for ρ should correspond to the maximal correlation between the experimental data set and the recalculated spectra makes the system also robust against misconclusions based on subjective interpretation of the data set, which is not always the case in TFA.