Neutron scattering is a powerful but expensive technique to study materials and discover new matters. However, the operation of the neutron facilities relies heavily on beamline scientists. Some experiments can take one or two days with experts making decisions along the way. Leveraging the computing power of DOE’s HPC platforms and recent AI advances, we demonstrate an autonomous workflow for the single-crystal neutron diffraction experiments. Our method overcomes two major challenges, i.e., weak peak detection and complex background removal, in autonomous neutron experimentation.
Our AI algorithm has been deployed for automated data analysis at the Dimensional Extreme Magnetic Neutron Diffractometer (DEMAND) beamline of High Flux Isotope Reactor (HFIR) at Oak Ridge National Laboratory. The required beam time for each experiment could be reduced by 30% to 50% once the entire workflow is integrated into the neutron instrument. Also, the workflow can be extended to other instruments at HFIR or the Spallation Neutron Source (SNS) as a AI module for the smart neutron facility.
(Left) Our method can remove complex environment background without affecting the intensity of the Bragg peak; (Right) A prototype workflow that connects the neutron facility and OLCF.
[1] Y. Hao, E. Feng, L. Zimmer, Z. Morgan, B. Chakoumakos, G. Zhang and H. Cao, Machine Learning Assisted Automation of Single Crystal Neutron Diffraction, Journal of Applied Crystallography, Vol. 56 (4), pp. 1229-1241, 2023 (https://doi.org/10.1107/S1600576723001516)
[2] J. Yin, G. Zhang, H. Cao, S. Dash, B. Chakoumakos, F. Wang, Toward an Autonomous Workflow for Single Crystal Neutron Diffraction, Springer Lecture Notes on Computer Science, Vol. 1690, pp. 244-256, 2022. (https://doi.org/10.1007/978-3-031-23606-8_15).