Advancement Towards FIB Active Auto Thinning (AAT) Process

Summary:

The feature size at the targeted cut face during the FIB process has been reduced to sub-ten nanometer scale in memory devices. Even with high-quality SEM resolution level scanned cut-face image data, determining the stopping point of the process remains a significant challenge. To address this, Micron has successfully implemented cut-face probability calculation and sliced image data interpolation methods using machine vision image AIPs (artificial intelligence platform) for the latest memory devices. This innovation enables machines to render a device’s three-dimensional geometry, infer and predict the upcoming cut face in SEM data points, and determine the stopping point in the FIB thinning process using detector’s raw data signal extractions. This approach avoids signal post-processing and transformation into the visible light range for humans, thereby preventing meaningful data loss. The new methodology surpasses human eye detection limits, maximizes SEM resolution utilization, and converts an automation system into an autonomous system capable of deciding where to stop the FIB milling at each cut-face, based on the sensory input of Micron wafer’s cut-face analysis and learning algorithms. This autonomous FIB thinning is defined as active auto thinning (AAT), distinguishing it from the conventional fiducial-based FIB auto thinning, known as passive auto thinning (PAT).