Efficient construction of phase diagram by machine learning

Abstract

We developed a method called PDC (Phase Diagram Construction) that constructs a phase diagram using machine learning. In this method, based on uncertainty sampling, the most uncertain point in the phase diagram is selected as a candidate for the next experiments/simulations. Then, using experiments/simulations, the phase at the selected point is identified. Uncertainty sampling is performed again on the increased experimental data to select next candidates. By repeating this process, it is possible to draw a phase diagram quickly. By using PDC, an uninvestigated phase diagram for the deposition of Zn-Sn-P films by molecular beam epitaxy is constructed. In addition, we have released an application software for the phase diagram construction; the software can be executed on a Windows computer without any special settings.