“Through next-generation CFE (cold field emission) technology and deep learning AI image models, 3D advanced semiconductor inspections such as GAA (Gate-All-Around) can be performed three times faster and more accurately than before.”

Applied Materials Korea held a press conference at the Park Hyatt Hotel in Samseong-dong on the 20th and announced a new defect review system, 'SEM Vision H20'.

Jang Man-soo, Director of Imaging and Process Control Technology, who was in charge of the presentation, expressed confidence, saying that the SEM Vision H20 system combines the industry's most sensitive electron beam technology with advanced AI image recognition capabilities to analyze nanometer-sized defects inside cutting-edge semiconductors more accurately and quickly.

Director Jang Man-soo mentioned that as we have recently entered the era of advanced semiconductors, many technologies that were not important in the past have become necessary. He also mentioned that as semiconductor patterns have become very small and changed to 3D structures such as GAA, the inspection field, which has to find defects between very small patterns and detect defects in invisible areas of the lower layers of the 3D structure, is faced with difficult challenges.

3∼4 years ago, there were no problems with the node, but now there are many defective factors, and it has become difficult for people to distinguish between defective and non-defective parts. He said that when viewed through a traditional process, the number of defective candidates has increased by more than 100 times compared to the past, but when examined again with SEM, there are cases where there are only 1 or 2 defects, and in the past few years, there have been many difficulties due to the additional cost, time, and manpower required to distinguish real from fake data during the inspection process.

The solution to address these issues is the SEM vision H20, which satisfies engineers’ needs for a defect review system that can quickly and accurately analyze the explosive growth in samples while maintaining the speed and sensitivity required for semiconductor mass production.

The SEM vision H20’s Cold Field Emission (CFE) technology delivers the sub-nanometer resolution needed to identify the finest embedded defects, operates at room temperature, and uses more electrons to create a narrower beam, delivering up to 50 percent better nanoscale image resolution and up to 10 times faster imaging speed than conventional Thermal Field Emission (TFE) technology.

In particular, the fast imaging speed increases coverage for each wafer, reducing the time it takes semiconductor manufacturers to collect the same information by one-third, he said.

In addition, the existing defective judgment took at least two images and overlapped them to determine a defective product if the difference exceeds a certain score, but experienced engineers introduced AI that can make judgments like a human with the idea of being able to know by looking at just one image.

As you take each image, the AI learns what is normal and what is not, so you can quickly inspect the image by taking just one image instead of two.

You can also input a pattern design and draw an outline based on it. By comparing it to a photo of the pattern, even parts that are difficult for humans to distinguish can be judged as defective. By utilizing AI, not only does it become faster, but it can also distinguish between real and fake defects, which allows for an innovative reduction in fake defective data.

Director Jang Man-soo said, “This enables us to catch a huge number of defects with much less effort than before, and improve productivity. Through this product, semiconductor manufacturers can speed up semiconductor development and widely utilize electron beam technology in mass production.”