Unsupervised Deep Learning for Limited-Angle STEM-EDX Tomography -- Application to 3D Chemical Analysis of Phase-Change Memory Devices
Researchers have developed a new unsupervised deep learning method, DIP-TV, to improve 3D chemical analysis in STEM-EDX tomography. This technique addresses limitations caused by restricted tilt ranges and low-dose imaging, which typically lead to missing-wedge artifacts and degraded reconstruction quality. The enhanced multi-channel version, DIPm-TV, reconstructs multiple elemental maps simultaneously by leveraging spatial correlations, outperforming existing methods in compensating for severe angular limitations and noise. AI
IMPACT Enhances 3D chemical analysis capabilities in microscopy, potentially improving materials science research and device characterization.