Abstract
Manual TEM analysis with elemental analysis through EELS or EDS has long been a major throughput bottleneck in physical failure analysis. The novel non-commercial technique presented in this paper is a historical and industry-first milestone of developments of combining elemental analysis and raw TEM data to automate data reporting. The approach uses a layered architecture in which “specific skills” are created on top of so-called "generic skills" (i.e., generic robot behaviors like cursor motions). With this skill-oriented robot programming approach, time and cost-efficient data production has been established with high accuracy, speed, and comprehension. The speed and productivity of images and maps, obtained from both collected raw data and jobs in the queue, cannot be achieved by manual operation, and the accuracy and comprehension cannot be practically obtained routinely on each job by human interaction because it is too time consuming.
