Speaker
Description
Quantitative low-energy electron diffraction [LEED $I(V)$] is a powerful method for surface-structure determination, based on the comparison of experimentally observed diffraction intensities $I$ with computations for structural models. As the diffraction intensities are highly sensitive to subtle structural changes, local structure optimization is essential for assessing the validity of a structure model and finding the best-fit structure. The ViPErLEED project (Vienna Package for Erlangen LEED) drastically reduces the user effort required for LEED $I(V)$ studies [1,2]. The talk will focus on two recent developments. A new implementation of structure search reformulates the optimization problem in a way that allows the use of standard optimization algorithms, including gradient-based methods [3]. This new code is based on JAX and can make use of graphics processing units (GPUs), accelerating structure search by more than an order of magnitude. Structure optimization requires a measure of agreement between the calculated and experimental data, a so-called R factor. We show that the previously used R factor $R_\text P$ (introduced by J. Pendry) has several deficiencies and is ill-suited for gradient-based optimization. We present an improved R factor $R_\text S$ that avoids these problems and is as good as $R_\text P$ or better in steering the optimization to the correct result [4]. As an example for the application of these new developments, a new structure model of the Fe$_2$O$_3$$(1\bar{1}02)$-$(2\times 1)$ surface will be presented.
[1] Kraushofer et al., Phys. Rev. Res. 7, 013005 (2025)
[2] Schmid et al., Phys. Rev. Res. 7, 013006 (2025)
[3] Imre et al., arXiv:2512.09737
[4] Imre et al., J. Phys.: Condens. Matter 38, 105001 (2026)