Accelerated Small Angle Neutron Scattering Algorithm for Polymeric Materials

This work presents an algorithm to accelerate SANS experiments by estimating the minimum number of counts needed for parameter estimation and model differentiation at a specified level of uncertainty. Three classes of model polymer materials were examined and analyzed, and time slices of SANS data were used to model a reduced number of counts. Accurate parameter estimations within 5% of full counts were achieved using only 1-50% of full counts, depending on the sample. Robust error quantification methods are essential for projecting uncertainties at lower counts. Monte Carlo bootstrapping method tends to overestimate fitting errors compared to experimental replicates. While weighted least squares estimators are generally unbiased, some models yield biased estimators. To differentiate between models, both the Akaike Information Criterion and Bayesian Information Criterion can be used, reduced numbers of counts can still identify the best model for our samples from a group of related candidate models for each material. The proposed algorithm can help SANS users optimize valuable beamtime and accelerate the use of SANS for structural characterization of libraries of materials while obtaining reasonable parameter estimation and model differentiation.