Learning enhanced dual-wavelength digital holography for measuring large-height complex step samples.

Area of Science:

• Metrology and Measurement Science
• Optical Engineering
• Artificial Intelligence in Manufacturing

Background:

• Precision manufacturing relies on accurate measurement of nanometer-to-micron complex structures.
• Conventional optical interferometry struggles with simultaneous high accuracy and large measurement range for these samples.
• Existing methods are often time-consuming, noisy, or system-complex.

Purpose of the Study:

• To develop a novel measurement technique for accurate and rapid assessment of large-height complex step samples.
• To overcome the accuracy-range trade-off inherent in conventional metrology methods.
• To simplify measurement systems for dynamic monitoring in precision manufacturing.

Main Methods:

• Proposed a learning-enhanced dual-wavelength digital holography (LeDWDH) approach utilizing a Y-Net architecture.
• Employed dual-wavelength digital holography to acquire a single multiplexed interferogram, ensuring a sufficient measurement range.
• Trained a two-to-one Y-Net using wrapped phases as input and scanning white light interferometry data as labels to reconstruct 3D morphology.

Main Results:

• The LeDWDH method achieved measurement errors less than 10 nm for step heights of 75 nm, 1 μm, and 5 μm.
• Demonstrated simultaneous high precision and a large measurement range for complex step samples.
• The system uses a single interferogram frame, significantly simplifying the setup.

Conclusions:

• LeDWDH successfully addresses the limitations of conventional methods for measuring large-height complex step samples.
• The proposed technique offers a new strategy for dynamic measurement in precision manufacturing.
• Achieved high accuracy (<10 nm error) and large measurement range in a simplified system.