インフォマティクスを利用したフラックス法研究の新展開

抄録

  Flux method is one of effective crystal growth technique to provide high-quality crystals, which can be applied to various functional materials. Usually, flux method takes long time to develop crystal materials, because of its complicated growth mechanism and various experimental parameters. This disadvantage prevents wide uses of this method for material applications. To solve these issues, we are incorporating informatics into the flux method. Possible prediction of crystal growth by machine learning would drastically change the workflow of material development. This report introduces our recent challenge to develop flux growth study using informatics. In detail, we show 3 topics. Firstly, we developed data-driven flux crystal growth system, based on Bayesian optimization. Here, we show machine-leaning design of experiment customized for flux method. Secondly, data format for flux method was statistically studied. Here, importance of raw material information and their description are proposed. Thirdly, we developed experimental automation system using robot. Here, application of the system to crystal purification process was carried out.