Controllable tailoring and understanding the phase-structure relationship of the 1T phase two-dimensional (2D) materials are critical for their applications in nanodevices. Thein situtransmission electron microscope (TEM) could regulate and monitor the evolution process of the nanostructure of 2D material with atomic resolution. In this work, a controllably tailoring 1T-CrTe2nanopore is carried out by thein situTEM. A preferred formation of the 1T-CrTe2border structure and nanopore healing process are studied at the atomic scale. The controllable tailoring of the 1T phase nanopore could be achieved by regulating the transformation of two types of low indices of crystal faces {101¯0} and {112¯0} at the nanopore border. Machine learning is applied to automatically process the TEM images with high efficiency. By adopting the deep-learning-based image segmentation method and augmenting the TEM images specifically, the nanopore of the TEM image could be automatically identified and the evaluation result of DICE metric reaches 93.17% on test set. This work presents the unique structure evolution of 1T phase 2D material and the computer aided high efficiency TEM data analysis based on deep learning. The techniques applied in this work could be generalized to other materials for controlled nanostructure regulation and automatic TEM image analyzation.
Keywords: deep learning; image segmentation; in situ transmission electron microscope; nanopore; two-dimensional material.
© 2021 IOP Publishing Ltd.