The National Science and Technology Council (NSTC) aims to make Taiwan a world leader in advanced manufacturing and chip processes by supporting basic research and putting in place key science and technology policies.
Professor Liang-Chia Chen of National Taiwan University’s Department of Mechanical Engineering oversees the cross-disciplinary research team that the NSTC has set up to combine cutting-edge technologies like automated optical inspection (AOI), high-precision measurement, new optical instrumentation, and AI algorithms.
Because of this, the next generation of AOI systems is being made to meet the needs of modern manufacturing and semiconductors. The NSTC has also set up the Advanced Optical Inspection Equipment Association (AOIEC), which has 43 companies as members. The AOIEC is made up of university labs, equipment makers, and people who use the equipment.
The AOIEC is a specialised research and development organisation that makes inline AOI equipment for advanced automated optical inspection and precision engineering businesses.
It brings together research teams from colleges and works with top semiconductor vendors and Taiwan AOI makers. The goal is to create cutting-edge AOI technologies that are driven by artificial intelligence to fill important technological gaps in optical critical dimension (OCD) metrology for 3D advanced packaging.
The system has a depth-to-width ratio of 15 times and can measure openings as small as a submicron. This is a new technological breakthrough for advanced semiconductor packing technology as it won the 2022 NSTC Future Technology Award.
The miniaturisation of ICs and improvements in 3-D packaging technology brings benefits like shorter conductor lengths, better thermal conduction, higher signal bandwidth, lower power usage, and smaller package volumes, but they also make it harder to measure.
Silicon vias are an important part of advanced 3D packaging for chip assembly in the vertical direction. But as the size of silicon vias gets smaller and the ratio of depth to width goes up, it gets harder to find good ways to measure them.
Cross-sectional critical dimension measurements are mostly done with scanning electron microscopes (SEM), but this method is damaging and takes a long time. Atomic force microscopy could be used to read profiles, but it isn’t very good and can’t get into the hole very well.
To solve these problems, a new optical measurement device that uses AI was suggested. The approach uses physical optical simulation and AI deep learning optimisation to penetrate high-depth-ratio blind holes and quickly find precise critical dimension information, meeting strict inline process time demands.
Advanced reflectometry and scatterometry were used to build the visual measurement technology that uses AI. It uses new numerically controlled laser broadband light illumination for optical sensing and AI deep learning optimisation of inverse reasoning through new algorithms.
It has gotten past the technical problems of having a minimum aperture and the highest depth-to-width ratio that can be measured. Several new optical critical dimension (OCD) measurement systems and techniques have been made using this technology. These systems and techniques can measure world-leading micron-level silicon via critical dimensions with industrially confirmed results.
With a signal-to-noise ratio (SNR) 28 times higher than non-coherent light sources, the system can measure multiple key dimensions in real-time. This is a world record. This technology is currently the best in the world, and it can be used in many difficult measure situations.
This new idea has been given a Taiwan design patent (No. 111106641) and a US patent (No. 06A-210153US). The technology is also being developed with the help of equipment manufacturers. Several domestic AOI equipment manufacturers have helped with technology transfer and study.