Magnetic neutron scattering experiments take days of scarce beam-time, even with high neutron fluxes. Because neutron beamtime is so valuable, extracting the most physics information from the available neutrons is critical. One way to extract more information without hardware improvements is autonomous experimental design. These are algorithmic approaches that control a neutron scattering experiment to reduce the time needed to characterize S(q, ω). Programmatically steering neutron experiments requires a quantitative understanding of the efficiency of an experiment and on-line analysis of data, both of which can be interesting even for an expert user.
I will present a method for autonomous experiments that aims to free users from the minutia of tracking a signal’s evolution, which provides time for live analysis of their experiment. This method exploits the smoothness of the dynamic structure factor S(Q, ω) to build a surrogate S(Q, ω) model that is used to sequentially choose (Q,ω) points. The resulting method has been simulated to effectively track a dispersion curve, even in low signal with poor signal to noise. Such efforts are required to reach our goal of fully autonomous and efficient neutron experiments.
Conditions of access for people from outside CEA-Saclay Center
To access to the CEA-Saclay center, a request of visit has to be done before coming. To do so, please, send an email to indicating why you need to enter (attend seminar), and giving your : gender, name, surname, date and location of birth, citizenship and employer institution.