AI technology is being used across industries. Semiconductor measurement technology is converging with AI to improve productivity in manufacturing sites.

SK Hynix announced on the 29th that it participated in 'SPIE AL 2024', an international conference held in San Jose, California, USA, together with Gauss Labs, and presented the results of the development of AI-based semiconductor measurement technology.

SK Hynix stated, “We have been collaborating with Gauss Labs in various fields to increase semiconductor yield and productivity, and this time, we have presented two papers containing the development achievements of both companies at a prestigious international academic conference,” and “We will continue to cooperate with Gauss Labs in the future to secure technological superiority.”

Through this paper, Gauss Labs introduced the 'Aggregated Adaptive Online Model (AOM),' an algorithm that improves the prediction accuracy of the AI-based virtual metrology solution 'Panoptes VM (Virtual Metrology).'

SK Hynix has implemented Panoptes VM since December 2022 and has performed virtual metrology on more than 50 million wafers to date. This translates to more than one virtual wafer per second, enabling the company to improve process variation by approximately 29 percent thanks to the software's performance.

The dispersion, which refers to the size of the quality variation in products produced in a given process, should be managed so that it does not exceed an appropriate level, as the possibility of defects decreases as the dispersion decreases.
The new algorithm that Gauss Labs unveiled at the conference is an upgraded version of the existing AOM, solving the problem of data shortage by integrating and modeling data from equipment that share the same pattern, while also improving prediction accuracy. Gauss Labs explains that applying this algorithm increases the process dispersion improvement rate.

Gauss Labs also introduced ‘Universal Denoising’ technology in its conference presentation.

Some semiconductor metrology work is based on scanning electron microscope (CD-SEM) images for semiconductor structural inspection. In order to accurately measure extremely small nanometer units, it is important to remove noise (noise) from the electron microscope images to increase the resolution.

This technology, developed by Gauss Labs, uses AI to remove noise from various types of images at once. The company said, “As a result of testing with SK Hynix, we confirmed that the image acquisition time was shortened to 1/4 of the existing technology,” and predicted, “This technology will improve the productivity of semiconductor measuring equipment by 42% in the future.”

Gauss Labs CEO Kim Young-han said, “Our company is focusing on research and development to ensure that industrial AI software can be effectively used in semiconductor manufacturing sites,” and added, “We will continue to launch various AI-based solution products to lead the ‘artificial intelligence in manufacturing sites.’”