Speaker
Description
The goal of specular neutron and x-ray reflectometry is to infer a material's scattering length density (SLD) profile from its experimental reflectivity curves. In this talk I will quickly describe the ill-posed non-invertible problem and an approach to some solutions via the use of artificial neural networks (ANNs). In particular, I will describe a set of tailored numerical experiments with the aim of assessing the applicability of data science and machine learning to the analysis of data generated at large-scale neutron scattering facilities. For this purpose, sample physical models are described under a new paradigm: layer-by-layer descriptions --in terms of SLDs, thicknesses and roughnesses, are replaced by parameter-free curves ρ(z), moving the focus of a priori assumptions towards the sample family to which a given sample belongs (e.g., 'thin film,' 'lamellar structure',etc.). The proposed methodology, when implemented correctly would enable quicker batch analyses of large datasets.
