SPring-8, the large synchrotron radiation facility

Speaker: Prof. Mingda Li

Language: English

Affiliation: Department of Nuclear Science & Engineering Massachusetts Institute of Technology (MIT)

Title: Machine learning: A data-driven spectrometer for X-ray and neutron scattering

Abstract:
Scattering techniques have made remarkable progress in the past decades, but the understanding of the microscopic interaction mechanisms in quantum materials remains challenging. With ever-increasing scattering data, machine learning brings new hope and can serve as a new probe to perform data analysis. In this seminar, we first provide some basics on machine learning and, how machine learning can be used to reveal the hidden information in scattering data and elucidate the quantum materials. For elastic scattering, we introduce an improved identification of the nuanced interface effect from reflectometry with super-resolution ^[1]. For inelastic scattering, we then introduce symmetry-preserved neural network and the capability to predict phonon density-of-states for vibrational spectroscopies ^[2]. For spectroscopy, we provide another example to see how X-ray absorption can be used to classify the materials’ topology ^[3]. Finally, for time-resolved techniques, we will introduce our recent work aiming to extract hidden thermal transport from time-resolved diffraction ^[4]. We highlight the importance of the representations and envision a few scattering problems that can benefit from machine learning ^[5].

[1] https://aip.scitation.org/doi/10.1063/5.0078814
[2] https://onlinelibrary.wiley.com/doi/10.1002/advs.202004214
[3] https://arxiv.org/abs/2003.00994
[4] https://arxiv.org/abs/2202.06199
[5] https://aip.scitation.org/doi/10.1063/5.0049111