[17p-A405-1~14] 3.6 Laser processing (formerly 3.7)

△ [17p-A405-4] Prediction of contact resistance of 4H-SiC electrode by machine learning using optical microscope images after laser doping

〇Yuki Iwaizumi¹, Takuma Yasunami¹, Keita Katayama¹, Yoshiaki Kakimoto², Daisuke Nakamura¹, Tetsuya Goto³, Hiroshi Ikenoue^1,2 (1.Grad. Sch. of ISEE, Kyushu Univ., 2.Department of Gigaphoton Next GLP, Kyushu Univ., 3.New Industry Creation Hatchery Center Tohoku Univ.)

Keywords:laser doping, machine learning

There are many parameters in laser doping, and determining the optimum conditions can be very costly. In particular, many costs are spent on device fabrication to evaluate whether the conditions were optimal. In this study, we use machine learning to predict doping results from the surface image of the SiC sample after laser irradiation. Prediction accuracy of doping results was more than 85%. It is now possible to classify with more accuracy than a person in a short time.