This repository provides a reference implementation and links related to the research paper:
"Physics-informed automated surface reconstructing via low-energy electron diffraction based on Bayesian optimization" (arXiv:2604.04578).
This work introduces a novel framework for automating the quantitative analysis of Low-Energy Electron Diffraction (LEED) I(V) curves. The method integrates physics-based multiple scattering forward models directly into a trust-region Bayesian Optimization (BO) loop. This approach simultaneously optimizes structural and experimental parameters, enabling efficient, autonomous exploration of complex, non-convex parameter spaces for accurate surface structure determination.
Key Components
- Citation of the Main Research Paper
If you use the methodology or findings from this research, please cite the following paper:
Physics-informed automated surface reconstructing via low-energy electron diffraction based on Bayesian optimization
Xiankang Tang, Ruiwen Xie, Jan P. Hofmann, Hongbin Zhang
arXiv:2604.04578 physics.comp-ph
https://arxiv.org/abs/2604.04578
- Code Source
The underlying computational package for quantitative LEED analysis is ViPErLEED:
Repository: viperleed/viperleed on GitHub
Description: ViPErLEED is an open-source Python package for quantitative LEED-I(V) analysis. It provides tools for measurement processing, calculation of theoretical I(V) curves, and optimization of structural models to fit experimental data.
Related Repositories:
https://github.com/viperleed/viperleed
- Source of Example Data
The input structures, experimental I(V) curves, and optimization parameters for the two benchmark systems are sourced from the official ViPErLEED documentation:
Source: viperleed.calc documentation – "Examples" section
Ag(100)-(1×1): Used as a simple test case for structure optimization.
α-Fe₂O₃(1-102)-(1×1): Used as a complex oxide surface example.
URL: https://www.viperleed.org/stable/content/viperleed_calc.html
The data includes POSCAR(structure), EXPBEAMS.csv(experimental I(V)), and DISPLACEMENTS(parameter search ranges) files, which are essential for reproducing the Bayesian optimization results.