Machine learning analysis of the dynamic STEM data allows extracting the time trajectories of point defects, from which thermodynamics and kinetics of solid state reactions can elucidated on a single-defect level.
Scientific Achievement
Determined the kinetics of solid-state reactions by dynamic observations one defect at a time.
Significance and Impact
This work enables understanding solid state reactions in 2D materials, a key step towards developing applications in electronic and quantum materials.
Research Details
– Dynamic scanning transmission electron microscopy (STEM) visualizes the phase transformation process in a layered material at the atomic level.
– A robust machine learning method was developed to automatically convert STEM movies into atomic positions without any limitation on data volume.
– The unsupervised method allows building a library of point defects and further to quantitatively describe their dynamics and kinetics of transformation processes.
A. Maksov, O. Dyck, K. Wang, K. Xiao, D. B. Geohegan, B. G. Sumpter, R. K. Vasudevan, S. Jesse, S.V . Kalinin, and M. Ziatdinov, "Deep learning analysis of defect and phase evolution during electron beam-induced transformations in WS2," NPJ Comp. Mat. (2019). DOI: 10.1038/s41524-019-0152-9
