[19p-Z15-1~11] 6.4 Thin films and New materials

[19p-Z15-8] Realization of closed-loop epitaxial thin-film growth optimization of superconducting TiN via machine learning

〇Isao Ohkubo¹, Zhufeng Hou¹, Jiyeon N. Lee², Takashi Aizawa¹, Mikk Lippmaa², Toyohiro Chikyow¹, Koji Tsuda³, Takao Mori¹ (1.NIMS, 2.ISSP, U. of Tokyo, 3.U. of Tokyo)

Keywords:machine learning, molecular beam epitaxy, transition-metal nitrides

Closed-loop optimization of epitaxial titanium nitride (TiN) thin-film growth was accomplished using metal-organic molecular beam epitaxy (MO-MBE) combined with the machine learning technique of a Bayesian approach, successfully reducing the required number of thin-film growth experiments. Epitaxial TiN thin films grown under the conditions optimized by a Bayesian approach exhibited abrupt metal–superconductor transitions above 5 K, demonstrating a new approach to the efficient development of undeveloped materials such as transition-metal nitrides. The demonstrated combination of thin-film growth technique and Bayesian approach is expected to open the door to accelerating the development of automated operation of thin-film growth apparatuses.