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

Nanometer-to-micron multi-step complex structures are widely used in precision manufacturing. Accurate and rapid measurement of these large-height complex step samples is crucial for product quality control and the development of manufacturing technologies. However, conventional optical interferometry methods face challenges, such as time consumption, high noise levels, or system complexity, due to the simultaneous requirement for high accuracy and large measurement range when applied to such samples. To solve these problems, we propose a learning-enhanced dual-wavelength digital holography (LeDWDH) based on the Y-Net. Dual-wavelength digital holography provides a sufficient measurement range for the maximum step height of the sample using only a single dual-wavelength multiplexed interferogram. The two wrapped phases obtained from dual-wavelength digital holography serve as the input of the Y-Net, while the 3D morphology obtained from scanning white light interferometry acts as the label. A two-to-one Y-Net is trained to establish the relationship between them, resulting in a LeDWDH that overcomes the trade-off between high accuracy and a large measurement range. When measuring complex step samples with step heights of 75 nm, 1 μm, and 5 μm, the measurement errors for each step are all less than 10 nm, demonstrating that the proposed method achieves both high precision and a large measurement range for large-height complex step samples. In particular, the proposed method uses only one frame dual-wavelength multiplexed interferogram, which greatly simplifies the measurement system and provides a new strategy for dynamic measurement of such samples.