特集「透過電子顕微鏡の最新の動向」
機械学習によるデータ駆動型の液中透過型電子顕微鏡“その場”観察
2023 年 66 巻 12 号 p. 700-705
詳細
2023 年 66 巻 12 号 p. 700-705
Non-equilibrium processes such as nucleation are difficult to observe in situ using transmission electron microscopy (TEM) because of spatiotemporal stochastic process. Therefore, we have been developing a method to predict/detect nucleation events and observe under low electron doses conditions in real time with the support of machine learning. Low electron dose observation is important to avoid radiolysis of water in the observation of aqueous solutions using liquid-cell TEM. Our data-driven TEM that can suggest observation points to the operator by processing in situ observation data in real time. In other words, it is data-driven TEM in which the TEM helps the operator, rather than the TEM being driven by the data. By incorporating the two codes into the TEM's software, nucleation can now be observed efficiently.
