Challenges and Opportunities of Machine Learning on Neutron and X-ray Scattering

Vol. 35, No. 4, 2022, Synchrotron radiation newS Feature article Challenges and Opportunities of Machine Learning on Neutron and X-ray Scattering NathaN C. DruCker,1,2 toNgtoNg Liu,1,3 ZhaNtao CheN,1,4 ryotaro okabe,1,5 abhijatmeDhi ChotrattaNapituk1,6 thaNh NguyeN,1,7 yao WaNg,8 aND miNgDa Li1,7 1Quantum Measurement Group, MIT, Cambridge, Massachusetts, USA 2School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts, USA 3Department of Physics, MIT, Cambridge, Massachusetts, USA 4Department of Mechanical Engineering, MIT, Cambridge, Massachusetts, USA 5Department of Chemistry, MIT, Cambridge, Massachusetts, USA 6Department of Electrical Engineering and Computer Science, MIT, Cambridge, Massachusetts, USA 7Department of Nuclear Science and Engineering, MIT, Cambridge, MA, USA 8Department of Physics and Astronomy, Clemson University, Clemson, South Carolina, USA Introduction Machine learning has been highly successful in boosting the research for neutron and X-ray scattering in the past few years [1, 2]. For diffraction, machine learning has shown great promise in phase mapping [3, 4] and crystallographic information determination [5, 6]. In small-angle scattering, machine learning shows the power in reaching super-resolution [7, 8], reconstructing structures for macromolecules [9], and building structure-property relations [10]. As for absorption spectroscopy, machine learning has enabled the rapid inverse search for optimized structures [11, 12] with improved spectral interpretability [13, 14]. Overall, as a data-driven approach, the success of the machinelearning-based scattering analysis depends on a few criteria, including: