Artificial Intelligence tools are starting to impact reflectometry (1-3) as they are
- faster at analysing reflectometry data
- handle noisy data very well
- produce fewer outliers than genetic algorithm fits
As reflectometry is an inverse problem - it is easy to calculate the reflectivity given a sample structure, but hard in the inverse direction - supervised learning with synthetic data yields good results.
Ready made AI models that can be used directly on e.g. a laptop without need for AI training are available under the following link. They work for x-ray and neutron reflectometry curves for film samples with up to three layers:
https://github.com/schreiber-lab/reflectorch
(1) Greco, A.; Starostin, V.; Karapanagiotis, C.; Hinderhofer, A.; Gerlach, A.; Pithan, L.; Liehr, S.; Schreiber, F.; Kowarik, S. Fast Fitting of Reflectivity Data of Growing Thin Films Using Neural Networks. J Appl Crystallogr 2019, 52 (Pt 6), 1342–1347. https://doi.org/10.1107/S1600576719013311.
(2) Starostin, V.; Dax, M.; Gerlach, A.; Hinderhofer, A.; Tejero-Cantero, Á.; Schreiber, F. Fast and Reliable Probabilistic Reflectometry Inversion with Prior-Amortized Neural Posterior Estimation. Science Advances 2025, 11 (11), eadr9668. https://doi.org/10.1126/sciadv.adr9668.
(3) Marecek, D.; Oberreiter, J.; Nelson, A.; Kowarik, S. Faster and Lower-Dose X-Ray Reflectivity Measurements Enabled by Physics-Informed Modeling and Artificial Intelligence Co-Refinement. JOURNAL OF APPLIED CRYSTALLOGRAPHY 2022, 55, 1305–1313. https://doi.org/10.1107/S2053273322008051.