From Lab to Fab: AI-Powered Metrology for High-Volume Manufacturing

M.Y.-H. Kim
Gauss Labs Inc.,
United States

Keywords: virtual metrology, image metrology, high-volume manufacturing, advanced process control, yield management


Metrology plays a crucial role in continuing and accelerating Moore's law. Despite tremendous progress in advanced sensor technology, physical metrology is facing numerous challenges for high-volume manufacturing, such as the high capex and opex of tools and sensors, the space and time required before and during measurement processes, and the overall impact on sustainability. Due to these financial, spatiotemporal, and environmental costs, the sampling rate of physical metrology is often kept under an adequate level and each measurement is also limited to checking already known variables. This talk discusses how such limitations of physical metrology can be overcome by leveraging advanced software technology in general and artificial intelligence (AI) in particular. Utilizing the vast amount of data generated from high-volume manufacturing processes, we can augment physical metrology without adding new hardware and extract more information from each measurement. Under this broad vision of making manufacturing processes more visible and predictable with data and AI, we will discuss mainly two technological directions — machine learning-based virtual metrology and computer vision-based image metrology. Although our focus will be on recent progress in AI components underlying these technologies, we will also present how they can transform real high-volume manufacturing through concrete use cases.