Autonomous Thin-Film Materials Synthesis by Machine Learning and Robotics

Accelerating the development of new materials is indispensable for a sustainable society. However, new materials development in laboratories involves repeated cycles of conception, synthesis, and characterization, manually performed by researchers. Under such circumstances, the inclusion of machine learning, robotics, and big data into these cycles promises to revolutionize materials research.<br/>Recently, we have designed and built a closed-loop system that combines Bayesian optimization, automatic synthesis, and automatic physical property evaluation for inorganic thin-film materials. This system implements a concept; lab equipment and instruments should be Connected, Autonomous, Shared, and operate in a High-throughput manner. With this approach, we have already achieved a proof-of-concept of autonomous material synthesis, which was demonstrated by autonomously optimizing the synthesis conditions specified by Bayesian optimization to minimize the electrical resistance for Nb-doped TiO₂ thin films fabricated via the sputtering method [1]. We present an expanded system to evaluate various physical properties: X-ray diffraction, scanning electron microscope/energy dispersive X-ray spectroscopy, Raman spectroscopy, and UV-Vis spectroscopy. Coevolution with such technologies, researchers work on more creative research, accelerating materials science research.<br/><br/>[1] R. Shimizu, and T. Hitosugi <i>et al</i>., APL Mater. 8, 11110 (2020).