Method and System for High-Throughput Defect Inspection Using the Contrast in the Reduced Spatial Frequency Domain

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention describes a high-throughput method and system for inspecting micro- and nano-patterned surfaces, such as semiconductor wafers or photomasks, for defects by analyzing how light is scattered (diffraction) from repeated pattern blocks. By focusing only on the most sensitive features in the spatial frequency domain—ignoring irrelevant data—the invention enables rapid detection and mapping of pattern deviations (defects) compared to a known ideal or learned pattern. The system works with various light types, including visible, UV, EUV, or X-rays, and can automatically learn normal patterns or use a library of known possible defects for targeted inspection.

Use CasesContent extracted from patent full text and abstract with AI.

• Rapid quality control and inspection of semiconductor wafers during manufacturing to detect pattern defects.
• Inspection of photomasks used in photolithography for defect detection before they are used in mass production.
• Quality assessment of optical or nanostructured surfaces in industries such as data storage, MEMS, or display manufacturing.
• Automated inspection routines in high-volume production environments where pattern fidelity is critical.
• Research or failure analysis of nanoscale devices where fine defects can impact performance.

BenefitsContent extracted from patent full text and abstract with AI.

• Significantly higher throughput than traditional optical inspection methods, enabling faster production quality assessment.
• High sensitivity to both phase and amplitude defects, important for advanced semiconductor nodes (e.g., sub-22 nm).
• Reduced data collection and processing requirements by focusing only on the most informative parts of the spatial frequency domain (regions of interest).
• Lensless, high-resolution imaging that is not limited by optical aberrations, depth of focus, or the need for expensive high-NA optics.
• Enables simultaneous detection of multiple defect types (phase and amplitude) in a single 2D scan, improving inspection efficiency.
• Self-learning capability allows the system to adapt to recurring defect patterns or to function without a complete pre-existing defect database.
• Supports both rapid screening (defect mapping) and more detailed analysis using advanced processing or imaging techniques (e.g., ptychography).
• Applicable across different wavelengths (from visible to X-ray), making it versatile for various materials and industry requirements.

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The present invention discloses methods and a system for scanning scattering contrast inspection for the identification of defects (4) in an actual pattern block (6) on a sample (2) as compared to the desired pattern block on the sample (2) hereby offering a novel solution in order to achieve defect sensitivity and throughput. The method omit the most of the information in reciprocal space (spatial frequency domain) in order to increase the throughput and it captures on the information in the reciprocal space that gives the highest defect information, i.e. contrast signal between the defected and defect-free structure. Subject of the present invention is that the investigation for only deviations from the expected diffraction pattern will allow rapid identification of the defects on the actual pattern of the sample. While the first method describes a method that learns the correct reconstructed diffraction image by the comparison of the repetitive pattern blocks, the second method focusses on the appearance of predictable defects in the spatial frequency domain of the reconstructed diffraction image thereby defining regions of interest where the defects materialize and speeds up the investigations since only those regions of interest have to be considered and compared to the reconstruction diffraction image of a defect-free pattern block.