Our machine-learning approach for solving the atomic structures of nanomaterials, published open access in Advanced Materials in January, has been selected for the cover of issue 13/2023.

Determining the structure of nanocrystalline or disordered materials is a major challenge, because pair distribution function (PDF) analysis of X-ray or neutron scattering data usually needs a good starting model. Our algorithm does not: an on-the-fly trained machine learning model combines DFT calculations with the comparison of calculated and measured PDFs, and searches for the structure that fits both. This way, even metastable configurations and stacking disorder can be identified.

The work is a collaboration with the groups of Bo B. Iversen and Bjørk Hammer at Aarhus University. Congratulations to Magnus Kløve and Sanna Sommer! We are now extending the method to other experimental techniques, such as surface X-ray diffraction.