Machine-learning based data reduction and analysis of single crystal neutron crystallography

Single crystal neutron diffraction is a unique technique in crystallography, with the capability to probe challenging structural properties which are difficult for other methods, such as finding the accurate structural information of hydrogen atoms, determining magnetic structures, and solving complex modulated structures from nuclear and magnetic phase transitions. This makes single crystal neutron diffraction an essential tool in various fields, including structural chemistry, magnetism, and quantum materials. We have developed a suite of machine-learning-based software tools to efficiently perform data reduction and analysis on neutron wavelength-resolved time-of-flight Laue diffraction data collected at the Oak Ridge National Laboratory Spallation Neutron Source using the TOPAZ single crystal diffractometer. Our model demonstrates strong performance in key data processing tasks such as background removal and peak integration. The suite of software tools greatly improves the automation of data processing and offers a streamlined and user-friendly platform for neutron crystallography.