A transmission electron microscope micro-area diffraction crystal direction analysis method combined with a graph neural network
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- UNIV OF SCI & TECH BEIJING
- Filing Date
- 2026-04-29
- Publication Date
- 2026-07-24
AI Technical Summary
Existing methods for determining crystal orientation in micro-area diffraction during transmission electron microscopy rely on manual measurement, which is highly dependent on the operator's experience. Furthermore, when dealing with high-index weak spots, lattice distortions, or complex orientation relationships, it is difficult to balance calibration efficiency and accuracy. Existing graph structure learning methods have failed to effectively integrate crystallographic physical laws, resulting in significant errors.
By constructing a constrained topology graph and utilizing a graph neural network in combination with reciprocal space vectors and normalized intensity, the edge weights and topology construction parameters are dynamically adjusted, and the crystal orientation analysis is iteratively optimized, thereby improving the overall robustness of calibration for complex diffraction patterns.
It improves the overall calibration accuracy and robustness of complex diffraction patterns, enhances the ability to detect weak signal spots, reduces errors, and improves the interpretability and accuracy of crystal plane index analysis.
Smart Images

Figure CN122453641A_ABST