A Machine Learning Approach to the Analysis of X-ray Diffraction Patterns From Multilayered Thin Film Diffusion Couples

A machine learning approach based on Neural Network and Gaussian Process Regression algorithms has been developed to extract structural information from the x-ray diffraction (XRD) patterns of multilayered Au/Pt diffusion couples. These algorithms have been trained on a subset of about 100,000 simulated XRD patterns computed at various stages of diffusional mixing in multilayers containing a fixed number of bilayers but a fluctuating number of atomic planes in each (unreacted) Au and Pt layer. When used to analyze the simulated diffraction patterns, this approach has been successful in reconstructing the known composition profile and the number of unreacted Au and Pt planes in each layer.