By Protochips
Deep learning, combined with the Fusion AX heating capabilities, enabled automated analysis of large in situ TEM datasets to identify interlayer stacking changes in bilayer phosphorene. Applied across extensive time-series data, the model captured how these displacements evolved both across the sample and over time, particularly during edge reconstruction. This allowed the researchers to directly see how the structure changes during heating, highlighting localized distortions that quickly fade into the material. Here, deep learning is used in the Poseidon AX system to enhance contrast and resolution when analyzing nanoparticles forming in liquid environments. Using a simulation-trained model, noisy LP-STEM videos were significantly improved, with image clarity increased up to fifteen-fold. This enabled automatic detection and tracking of nanoparticle nucleation and growth across large datasets, achieving expert-level precision at much higher speed. In this research, Fusion, AXON and deep learning algorithms were combined to analyze advanced orientation and grain boundary characterization. Automated analysis of thin films revealed which grain boundaries appeared most frequently. Thin films provided a robust platform for studying grain growth, enabling seamless correlative TEM imaging with crystal orientation data, high spatial and temporal resolution, and large-scale automated data analysis.
