Scientific Achievement
A machine learning (ML) inversion approach is used to determine effective interactions in condensed matter phases from scattering data that fully circumvents the efficiency-accuracy trade-off.
Significance and Impact
This ML approach does not suffer from the drawbacks of explicit modeling and therefore allows efficient quantitative extraction of relevant parameters, independent of the condensed matter system, providing a new toolbox to facilitate the study of condensed matter systems using experimental scattering.
Research Details
•Gaussian process (GP) regression is used to determine the interaction potential from scattering data
•A digital simulation environment based on molecular dynamics is used to train the GP for inverting the scattering structure factor
M-C. Chang, et al., Nature Communications Physics, 5 46 (2022). https://www.nature.com/articles/s42005-021-00778-y
