(Gauss Labs: Seoul, South Korea) -- Gauss Labs has released the second version of its AI-based virtual metrology solution, Panoptes VM 2.0.

Panoptes VM provides all-wafer measurement data in real time by predicting process outcomes from sensor data. By applying this solution, it’s possible to predict the process results of all products without physical full-scale measurement, significantly reducing time and resources.

Panoptes VM 2.0 is making a new leap forward with modeling features, including multistep modeling, operation-group modeling, and automatic model selection. With these new features, the prediction accuracy and usability have been greatly improved compared to the previous version.

Gauss launched Panoptes VM 1.0 in November 2022. The system has been deployed to the thin-film deposition process at SK hynix’s high-volume manufacturing fabs since December 2022. By integrating the virtual measurement results from Panoptes VM with APC, SK hynix improved process variability by approximately 29% and also enhanced yield rate.

Thin film deposition: The process of creating and depositing thin film coatings onto a substrate material

APC (advanced process control): A solution that adjusts process condition for equipment adaptively during a manufacturing process

Variability: The scale of the quality variation of the products from a certain process. It’s important to maintain a low process variability to keep the product quality consistent.

SK hynix plans to expand the application to APC use case. With the new multistep modeling feature, it expects to cover the etching process effectively. This feature creates prediction models using data from both current and previous steps to improve the accuracy of VM predictions, which is especially critical for etching process.

Etching: The process of removing unnecessary parts on the exterior of the circuit engraved on the wafer through photolithography.

Panoptes VM 2.0 also provides operation-group modeling that enables users to group data from similar operations for more effective modeling, even when the data are scarce. The automatic model selection, which supports multi-algorithm architecture, automatically selects the best-performing model based on data characteristics to further improve the accuracy and usability of VM.

Gauss Labs, headquartered in Silicon Valley, was founded in 2020 with the vision of “revolutionizing manufacturing with AI” and has focused on “developing AI solutions for manufacturing data intelligence (MDI),” extracting valuable information and insights from a vast amount of raw manufacturing data that can’t be easily analyzed by humans.

Mike Kim, CEO of Gauss Labs, says, “Our concerted efforts over the past four years are now creating meaningful impact in the semiconductor industry, the most advanced manufacturing sector. Equipped with our state-of-the-art industrial AI technology, we will continue our march toward the global industrial AI market.”