Beschreibung
Small-angle neutron scattering (SANS) is one of the most important techniques for microstructure determination, which is used in a wide range of scientific disciplines such as materials science, physics, chemistry, and biology. Conventional SAS can probe microstructural (density and composition) inhomogeneities in the bulk and on a mesoscopic length scale between a few and a few hundred nanometers. Being sensitive to magnetism, small-angle neutron scattering (SANS) also provides a unique magnetic contrast.
Despite drastic improvements over the last decades, SANS is inherently flux limited, similar to any other neutron scattering technique, caused by the limited brilliance of todays neutron sources, that is essentially given by the properties of the target or core materials.
We show first results of a recently popular approach to optimize the usage of SANS beamtime. In this project, we use algorithms based on machine learning to optimize and automatize the measurement strategy of a pinhole SANS instrument, based on a set of exemplary standard SANS samples. Our model includes the desired statistical resolution, intensity and Q-resolution for the different geometrical setting of the instrument and is able to provide reduced I(Q) data of a set of samples as an output. In combination with machine-learning-based analysis of the measurement-data we work on an AI-based optimization of the measurement strategy.
Therefore our project may form an important contribution to developing a fully autonomous SANS experiment.
