Computer vision assisted decomposition analysis of atom probe tomography data

Abstract While atom probe tomography (APT) offers 3D spatially-resolved compositional characterization at the nanometer scale, individual inspection of large datasets is time demanding and interpretation thereof highly operator-dependent. Recently, decomposition behavior (V,Al)N thin films was investigated by APT. The conventional analysis, employing standard tools from commercial software, laborious only a small fraction (below 5%) total acquired data has been used. In present work, analysis workflow automated, taking into account more than 90% data. As first step, products, including matrix enriched phases, were identified via deep learning-based semantic image segmentation. Performance ‘matrix’ product ‘aluminum-rich’ very good, achieving F1-scores 0.994 0.943, respectively, while performance for ‘vanadium-rich’ with 0.854 slightly lower. segmentation enables investigation phase formation evolution over temperature. next early stages spinodal probed on neighborhood analysis. results consistent prior in this work statistically robust leveraging an order magnitude larger dataset. end, transferability presented (V,Al)N-based model shown discussed isostructural (Ti,Al)N.