Abstract: A reference optical image of a die is determined based on a design file with a deep convolutional neural network for image-to-image translation. The reference optical image is subtracted from the target image thereby generating a difference image. After applying a care area mask, the difference image can be binarized. The resulting binarized defective image can be used for optical inspection.

Abstract: A noise map is used for defect detection. One or more measurements of intensities at one or more pixels are received and an intensity statistic is determined for each measurement. The intensity statistics are grouped into at least one region and stored with at least one alignment target. A wafer can be inspected with a wafer inspection tool using the noise map. The noise map can be used as a segmentation mask to suppress noise.

Abstract: Techniques and systems to achieve more accurate design alignment to an image by improved pixel-to-design alignment (PDA) target selection are disclosed. PDA targets in an image frame of a die can be biased to include a hotspot location in one of the PDA targets. The PDA targets can be evaluated for repetitive patterns by analyzing the uniqueness of the points used as the PDA targets.

Abstract: Methods and systems for detecting defects on a specimen are provided. One system includes a first deep learning (DL) network configured for filtering nuisances from defect candidates detected on a specimen. Output of the first DL network includes a first subset of the defect candidates not filtered as the nuisances. The system also includes a second DL network configured for filtering nuisances from the first subset of the defect candidates. Computer subsystem(s) input high resolution images acquired for the first subset of the defect candidates into the second DL network. Output of the second DL network includes a final subset of the defect candidates not filtered as the nuisances. The computer subsystem(s) designate the defect candidates in the final subset as defects on the specimen and generate results for the defects.

Abstract: Methods and systems for detecting defects on a specimen are provided. One system includes a first deep learning (DL) network configured for filtering nuisances from defect candidates detected on a specimen. Output of the first DL network includes a first subset of the defect candidates not filtered as the nuisances. The system also includes a second DL network configured for filtering nuisances from the first subset of the defect candidates. Computer subsystem(s) input high resolution images acquired for the first subset of the defect candidates into the second DL network. Output of the second DL network includes a final subset of the defect candidates not filtered as the nuisances. The computer subsystem(s) designate the defect candidates in the final subset as defects on the specimen and generate results for the defects.

Abstract: Techniques and systems to achieve more accurate design alignment to an image by improved pixel-to-design alignment (PDA) target selection are disclosed. PDA targets in an image frame of a die can be biased to include a hotspot location in one of the PDA targets. The PDA targets can be evaluated for repetitive patterns by analyzing the uniqueness of the points used as the PDA targets.

Abstract: Systems and methods for removing nuisance data from a defect scan of a wafer are disclosed. A processor receives a design file corresponding to a wafer having one or more z-layers. The processor receives critical areas of the wafer and instructs a subsystem to capture corresponding images of the wafer. Defect locations are received and the design file is aligned with the defect locations. Nuisance data is identified using the potential defect location and the one or more z-layers of the aligned design file. The processor then removes the identified nuisance data from the one or more potential defect locations.

Abstract: A noise map is used for defect detection. One or more measurements of intensities at one or more pixels are received and an intensity statistic is determined for each measurement. The intensity statistics are grouped into at least one region and stored with at least one alignment target. A wafer can be inspected with a wafer inspection tool using the noise map. The noise map can be used as a segmentation mask to suppress noise.

Abstract: Systems and methods for removing nuisance data from a defect scan of a wafer are disclosed. A processor receives a design file corresponding to a wafer having one or more z-layers. The processor receives critical areas of the wafer and instructs a subsystem to capture corresponding images of the wafer. Defect locations are received and the design file is aligned with the defect locations. Nuisance data is identified using the potential defect location and the one or more z-layers of the aligned design file. The processor then removes the identified nuisance data from the one or more potential defect locations.