Chinese semiconductor thread II
- Thread starter vincent
- Start date
Non-contact defect detection system for large-size gratings based on high-precision displacement stage.
Abstract
Two-dimensional gratings serve as critical components in high-precision optical systems, such as inertial confinement fusion (ICF) drivers and extreme ultraviolet (EUV) lithography machines. However, surface defects on gratings can degrade their diffraction efficiency and induce wavefront phase distortion. Traditional inspection methods, such as atomic force microscopy (AFM), although capable of sub-micron accuracy, suffer from low throughput and heavy reliance on manual operation, making them unsuitable for automated industrial applications. To address these challenges, this study proposes a defect detection system integrating laser-interferometry-based nanometer-level motion control, dark-field bilateral illumination imaging, and multi-level morphological processing. The system employs a path planning algorithm within the motion control module to achieve full-area scanning imaging of the grating surface, followed by data optimization through image stitching and multi-scale filtering preprocessing, subsequently integrates morphological segmentation with connected component analysis to extract multi-dimensional defect features—including type, positional coordinates (X-Y), and geometric dimensions (length-width)—thereby constructing a high-precision defect database based on a four-tier classification framework (L1-L4), which facilitates process traceability and provides quantitative criteria for quality assessment in grating manufacturing. Experimental validation on a 300 × 70 mm2 gold-coated grating demonstrates that the system achieves a detection throughput of 30 mm2/min. It has been successfully applied to the small-batch process validation of terminal grating components in EUV lithography machines. This work provides a non-destructive and noncontact measurement solution that enables coordinated optimization of precision, efficiency, and reliability for defect control in precision instruments. |
Yeah it is surprising how fast the catchup in DRAM and NAND has been. Realistically the gap is pretty minimal at this point, they're basically on the world leading edge. I've been very surprised how CXMT has been able to make such competitive memory without EUV. I wonder if their yields are significantly lower though.
Still while DRAM and NAND has been a major bright spot, independent lithography that we know of (due to huge secrecy) is still far behind and CPU/GPU (for non AI) situation isn't much better. Hopefully in next few years, by 2030, we will see actual commercial rollout of a more comprehensive block of competitive parts. The extremely fast catchup in DRAM/NAND at least shows that rapid catchup to roughly leading edge level is possible.
"It has been successfully applied to the small-batch process validation of terminal grating components in EUV lithography machines"
Woah. (sorry for low quality comment but that line is pretty significant)
Maybe the chaebols should have treated their employees better.
Then it most likely entered risk production in 2024.