Integrating Machine Learning with Neutron Scattering

S. A. GRIGERA

Amsterdam

Congreso; International Conference on Strongly Correlated Electron Systems (SCES) 2022; 2022

Quantum materials research requires co-design of theory with experiments and involvesdemanding simulations and the analysis of vast quantities of data, usually including patternrecognition and clustering. Artificial intelligence is a natural route to optimise theseprocesses and bring theory and experiments together. This talk will discuss a scheme thatintegrates machine learning (ML) with high-performance simulations and scatteringmeasurements, covering the pipeline of typical neutron experiment [1]. This approach usesnonlinear autoencoders trained on realistic simulations along with a fast surrogate for thecalculation of scattering in the form of a generative model. As an example ofimplementation of these techniques and of the approach, I will discuss how ML can be usedto extract an effective Hamiltonian for the highly frustrated magnet Dy2Ti2O7 and how thescheme was used to guide neutron scattering experiment under hydrostatic pressure, extractmaterial parameters and construct a phase diagram.[1] A. M. Samarakoon, D. A. Tennant, F. Ye, Q. Zhang , and S. A. Grigera, ”Integration of MachineLearning with Neutron Scattering: Hamiltonian Tuning in Spin Ice with Pressure”. arXiv preprintarXiv:2110.15817 (2021)[2]A. M. Samarakoon et al. "Machine-learning-assisted insight into spin ice Dy2Ti2O7." Naturecommunications, 11 1-9 (2020)