AI Applications for Failure Analysis - Grain size analysis by differential phase contrast (DPC)- scanning transmission electron microscopy (STEM) and artificial intelligence (AI) segmentation
AI Applications for Failure Analysis - Grain size analysis by differential phase contrast (DPC)- scanning transmission electron microscopy (STEM) and artificial intelligence (AI) segmentation
Wednesday, November 19, 2025
Summary:
Grain size significantly influences the performance of semiconductor devices, affecting their electrical, mechanical, and thermal properties. As devices shrink, efficient nanometer-scale grain resolution methods are essential. Current grain analysis methods like SEM-based EBSD and TKD have spatial resolution limitations, while TEM-based techniques such as PED and 4D-STEM, though higher in resolution, are time-consuming and data-intensive. This paper introduces differential phase contrast (DPC) imaging in STEM for polycrystalline materials, combining high resolution with fast acquisition speeds. By adjusting DPC conditions, signals from heavy and light elements can be separated, demonstrated in 3DNAND devices containing Si and W grains. Utilizing AI and computer vision, an automatic image analysis workflow examines DPC images based on polarization, texture, and shape geometry. This method enables detailed grain structure analysis, revealing differences in grain shape and size between materials. The results highlight the potential of combining DPC imaging and AI analysis to develop a TEM-based method that offers high resolution and efficiency, promising significant integration into the semiconductor industry.
Grain size significantly influences the performance of semiconductor devices, affecting their electrical, mechanical, and thermal properties. As devices shrink, efficient nanometer-scale grain resolution methods are essential. Current grain analysis methods like SEM-based EBSD and TKD have spatial resolution limitations, while TEM-based techniques such as PED and 4D-STEM, though higher in resolution, are time-consuming and data-intensive. This paper introduces differential phase contrast (DPC) imaging in STEM for polycrystalline materials, combining high resolution with fast acquisition speeds. By adjusting DPC conditions, signals from heavy and light elements can be separated, demonstrated in 3DNAND devices containing Si and W grains. Utilizing AI and computer vision, an automatic image analysis workflow examines DPC images based on polarization, texture, and shape geometry. This method enables detailed grain structure analysis, revealing differences in grain shape and size between materials. The results highlight the potential of combining DPC imaging and AI analysis to develop a TEM-based method that offers high resolution and efficiency, promising significant integration into the semiconductor industry.