Bayesian inference of perovskite oxide interface structure based on surface x-ray diffraction data

Crystal truncation rod (CTR) scattering method is one of the approaches to determine the structure of ultrathin films. To perform structure refinement from CTR scattering data, a good initial model is needed to obtain the correct structure. Any surface electron density analysis requires to extrapolate the scattering data.
We developed a robust structure refinement method for experimentally obtained CTR scattering data by using Bayesian inference with an exchange Monte-Carlo method.
The exchange Monte Carlo method [2] enables us to search a wider parameter area in the parameter hyperspace, resulting in a better acceptance for the initial model. Our test analysis shows the structure of LaNiO₃ ultrathin film on LaAlO₃ was successfully refined from a simple LaAlO₃ structure as the initial model.
The merit of using Bayesian inference is providing the degree of parameter uncertainty with a background of informatics. Proper estimation of uncertainty will be discussed.
[1] M. Anada et al., J. Appl. Cryst. 50, 1611-1616 (2017).
[2] K. Hukushima and K. Nemoto, J. Phys. Soc. Jpn. 65, 1604 (1996).