Bayesian optimization with experimental failure for high-throughput materials growth
A crucial problem in achieving innovative high-throughput materials growth with machine learning and automation techniques, such as Bayesian optimization (BO) and robotic experimentation, has been a lack of an appropriate way to handle missing data due to experimental failures. Here, we propose a new BO algorithm that complements the missing data in the optimization of materials growth parameters. The proposed method provides a flexible optimization algorithm capable of searching a wide multi-dimensional parameter space. We demonstrate the effectiveness of the method with simulated data as well as in its implementation for actual materials growth, namely machine-learning-assisted molecular beam epitaxy (ML-MBE) of SrRuO3, which is widely used as a metallic electrode in oxide electronics. Through the exploitation and exploration in a wide three-dimensional parameter space, while complementing the missing data, we attained tensile-strained SrRuO3 film with a high residual resistivity ratio of 80.1, the highest among tensile-strained SrRuO3 films ever reported, in only 35 MBE growth runs.