Machine-learning-assisted growth of ultrahigh-quality epitaxial 4<i>d</i> ferromagnetic Weyl semimetal SrRuO₃

Although Weyl fermions have been predicted to exist in various oxides [1], evidence for their existence remains elusive. SrRuO₃, a 4dferromagnetic metal often serves as an epitaxial conducting layer in oxide heterostructures, provides a promising opportunity to seek their existence. State-of-the-art oxide thin film growth technologies, augmented by machine learning techniques, may allow access to such topological matter, which we have actually accomplished [2] using machine-learning-assisted molecular beam epitaxy (ML-MBE)[3]. To simplify the intricate search space of three entangled growth conditions (growth temperature, Ru/Sr ratio, oxidation strength), we ran the Bayesian optimization for a single growth condition while keeping the other growth conditions fixed. As a result, an ultrahigh-crystalline-quality SrRuO₃film exhibiting the highest residual resistivity ratio (RRR) of 84 was developed. With the highest RRR ever reported, our SrRuO₃thin films are superior to those by any other method and have enabled to probe the intrinsic quantum transport properties of Weyl fermions [2].
[1] X. Wan, et al., Phys. Rev. B 83, 205101 (2011).
[2] K. Takiguchi, Y. K. Wakabayashi, et al., Nat. Commun. 11, 4969(2020).
[3] Y. K. Wakabayashi, et al., APL Mater. 7, 101114 (2019).