Automated, quantitative electron tomography of dislocation morphology combined with deep learning technology

The three-dimensional morphology of dislocations significantly influences the mechanical properties of crystalline materials. Achieving high-accuracy dislocation tomography through transmission electron microscopy remains a formidable challenge, primarily due to the complex nonlinear diffraction effects required for dislocation observation. In this article, we propose an automated electron tomography approach for reconstructing dislocation morphology by integrating a powerful deep learning-based detection technique. The goal is to address the various challenges involved in reconstructing tomography from diffraction contrast images and to accurately determine dislocation positions. Experimental images of dislocations were acquired and reconstructed to validate the proposed approach. It has been demonstrated that this approach effectively reduces the skill barriers associated with dislocation tomography reconstruction. •We propose a quantitative dislocation electron tomography approach by integrating a deep learning-based detection technique.•The influence of diffraction contrast on reconstruction is minimized to the greatest extent.•Dislocations in a nickel-based superalloy were imaged and accurately reconstructed by the proposed approach.