SK Hynix Introduces AI-Based Semiconductor Measurement Achievements with Gauss Labs

Participation in International Conference 'SPIE AL' Held in the US
Two Papers Published Showcasing Joint Development Achievements

SK Hynix and Gauss Labs participated in the international conference 'SPIE AL 2024' held from the 25th to the 29th in San Jose, California, USA, and presented their achievements in developing AI-based semiconductor metrology technology. Gauss Labs is a Silicon Valley-based AI startup established in August 2020 with capital investment from SK Hynix.

SPIE AL (SPIE Advanced Lithography + Patterning) is a conference hosted by the International Society for Optics and Photonics (SPIE), an international society in the fields of optics and photonics founded in the United States in 1955. It is a venue where discussions on lithography technology for drawing semiconductor circuits take place.

Kim Young-han, CEO of Gauss Labs (center), is taking a commemorative photo with company members. /
[Photo by SK Hynix]

SK Hynix, aiming to improve semiconductor yield and productivity, collaborated with Gauss Labs in various areas and as a result, presented two papers related to metrology showcasing the development achievements of both companies. Semiconductor metrology refers to the process of verifying whether the physical and electrical characteristics of semiconductor devices meet the required standards at each production stage during chip manufacturing, thereby enhancing productivity.

Through this paper presentation, Gauss Labs introduced the AI-based virtual metrology solution 'Panoptes VM' and its improved algorithm 'Aggregated Adaptive Online Model (Aggregated AOM)' that enhances prediction accuracy. This algorithm is an upgraded version of the existing AOM.

Since December 2022, SK Hynix has been using Panoptes VM to perform virtual metrology on over 50 million wafers to date. Converted into time, this means virtual metrology was conducted on more than one wafer per second. Thanks to the performance of this software, the company was able to improve process variation by approximately 29%. Process variation refers to the quality variability of products produced in each process, and a reduction in variation decreases the likelihood of defects.

Gauss Labs also introduced the 'Universal Denoising' technology. This technology removes image noise (artifacts) seen when using a critical dimension scanning electron microscope (CD-SEM) during semiconductor metrology with AI in one step, thereby enhancing resolution. It reduces the image acquisition time to about one-quarter of the previous duration. Going forward, it is expected to help improve semiconductor metrology equipment productivity by 42%.

Kim Young-han, CEO of Gauss Labs, said, "We are committed to research and development to ensure that industrial AI software can be effectively used in semiconductor manufacturing sites," adding, "We will continue to launch various AI-based solution products to lead the 'AI transformation of manufacturing sites'."

This content was produced with the assistance of AI translation services.