[16p-Z19-1~23] 10.1 新物質・新機能創成（作製・評価技術）

▲ [16p-Z19-22] Determination of the Dzyaloshinskii-Moriya interaction from a single magnetic domain image using image recognition

〇Masashi Kawaguchi¹、Kenji Tanabe²、Keisuke Yamada³、Takuya Sawa²、Shun Hasegawa¹、Masamitsu Hayashi¹、Yoshinobu Nakatani⁴ (1.The Univ. of Tokyo、2.TTI、3.Gifu Univ.、4.UEC)

キーワード：spintronics, Dzyaloshinskii-Moriya interaction, machine learning

The Dzyaloshinskii-Moriya interaction (DMI) , generating chiral magnetic orders in symmetry broken ferromagnetic multilayers, has been studied intensively. A great deal of effort has been spent on evaluation of DMI due to its difficulty. A simple and accurate evaluation method accelerate spintronics device development. In this study, we demonstrate determination of DMI strength from a single magnetic domain image employing image recognition supported by machine learning. A convolution neural network (CNN), trained with data sets prepared by micromagnetic simulations, can estimate the DMI strength of testing data sets with accuracy ~ 0.05mJ/m². Image recognition of magnetic domain patterns using CNN is applied to experimental systems that consist of Pt/Co based ferromagnetic multilayers, and returns consistent DMI strength.