Deep-learning-assisted scattering structured-illumination confocal microscopy for industrial super-resolution...

Area of Science:

• Optics and Photonics
• Materials Science
• Artificial Intelligence

Background:

• Laser scanning confocal microscopy (LSCM) is crucial for industrial inspection but faces limitations like the edge shadow effect and Abbe diffraction limit.
• These limitations hinder precise nanostructure inspection in semiconductor manufacturing.
• Existing methods struggle with distorted structured light patterns in complex industrial samples.

Purpose of the Study:

• To propose a novel deep learning-driven scattering laser scanning structural illumination microscope (DL-sLSSIM).
• To achieve super-resolution inspection of semiconductor nanostructures.
• To overcome the limitations of conventional microscopy in industrial metrology.

Main Methods:

• Developed a polarization-compensated scattering confocal configuration to enhance edge contrast and suppress background.
• Analyzed the imaging mechanism of point-scanning structured illumination and identified stripe distortion issues.
• Introduced a residual channel attention network (RCAN) for nonlinear demodulation and high-frequency information extraction.

Main Results:

• Achieved a lateral resolution improvement of approximately 1.8×, reaching 480 nm under low-magnification inspection.
• Significantly improved the signal-to-noise ratio in imaging semiconductor wafers.
• Demonstrated the effectiveness of DL-sLSSIM for super-resolution metrology in advanced manufacturing.

Conclusions:

• DL-sLSSIM provides a high-fidelity, non-destructive super-resolution metrology solution for semiconductor manufacturing.
• The integration of deep learning with scattering structured illumination overcomes conventional imaging barriers.
• This approach enables enhanced inspection of nanostructures with improved resolution and clarity.