Physics-informed automated surface reconstruction via low-energy electron diffraction based on Bayesian optimization

Abstract Low-energy electron diffraction (LEED) is a cornerstone technique for determining surface atomic structures 1 , yet the quantitative analysis of electron diffraction intensity as a function of incident electron energy—that is, LEED- I(V) analysis—remains a complex inverse problem. In this work, we tackle quantitative LEED- I(V) analysis based on physics-informed Bayesian optimization (BO). By embedding multiple scattering LEED forward models and crystallographic symmetry for structural parameter pruning directly into a trust-region BO loop, our approach simultaneously optimizes both structural and experimental parameters, adaptively adjusting trust regions for efficient exploration of complex non-convex parameter spaces while substantially reducing manual intervention. The robustness and scalability of the approach are demonstrated using the Ag(100)-(1 × 1) and Fe 2 O 3 (1 $$\overline{1}$$ 1 ¯ 02)-(1 × 1) surfaces as examples. Under the same number of samples, the BO workflow achieves faster convergence and lower R-factors than random walk, fast simulated annealing, and genetic algorithm baselines. These results provide a practical and sample-efficient route toward automated LEED- I ( V ) structure reconstruction.