2D material property characterizations by machine-learning-assisted microscopies

Microscopy characterization techniques can provide intuitive images of 2D materials with certain spatial resolutions. At the same time, machine-learning algorithms, which have experienced tremendous advancement in image processing over passed decades, are able to extract comprehensive information directly from a large scale of the images. Combining microscopy characterization techniques with machine-learning algorithms can offer insight into the structures and properties of 2D materials with the advantages of high automation, high accuracy, and high throughput. Herein, we will give a review of this interdisciplinary area, from foundations and progress to challenges and potential opportunities. The developments in this field are first overviewed according to its characterization techniques. Then, this review focuses on the theoretical and practical foundations of machine-learning-assisted microscopies for 2D material property characterizations, followed by two case studies to illustrate the implementation details. Finally, challenges and opportunities are addressed for future research and industrialized applications. We hope this review article can provide a clear guideline for both the academic society and general readers and inspire researchers for further explorations of this promising area.