2020/10/15 by Juan Manuel Carmona-Loaiza, Carmona-Loaiza, Juan Manuel, Zamaan Raza +1
Physics and Astronomy · Materials Science · #Nuclear Physics and Applications #X-ray Diffraction in Crystallography #Machine Learning in Materials Science
paper · pdf · doi:10.48550/arxiv.2010.07634
The goal of Specular Neutron and X-ray Reflectometry is to infer materials\nScattering Length Density (SLD) profiles from experimental reflectivity curves.\nThis paper focuses on investigating an original approach to the ill-posed\nnon-invertible problem which involves the use of Artificial Neural Networks\n(ANN). In particular, the numerical experiments described here deal with large\ndata sets of simulated reflectivity curves and SLD profiles, and aim to assess\nthe applicability of Data Science and Machine Learning technology to the\nanalysis of data generated at neutron scattering large scale facilities. It is\ndemonstrated that, under certain circumstances, properly trained Deep Neural\nNetworks are capable of correctly recovering plausible SLD profiles when\npresented with never-seen-before simulated reflectivity curves. When the\nnecessary conditions are met, a proper implementation of the described approach\nwould offer two main advantages over traditional fitting methods when dealing\nwith real experiments, namely, 1. sample physical models are described under a\nnew paradigm: detailed layer-by-layer descriptions (SLDs, thicknesses,\nroughnesses) are replaced by parameter free curves \ρ(z), allowing a-priori\nassumptions to be fed in terms of the sample family to which a given sample\nbelongs (e.g. "thin film", "lamellar structure", etc.) 2. the time-to-solution\nis shrunk by orders of magnitude, enabling faster batch analyses for large\ndatasets.\n