Inference of process variations in silicon photonics from characterization measurements

Area of Science:

• Materials Science
• Electrical Engineering
• Optics

Background:

• Manufacturing silicon photonic integrated circuits (PICs) is hindered by challenges in understanding and quantifying process variations.
• Accurate characterization of geometric variations is crucial for achieving high-yield PIC fabrication.
• Traditional methods rely on replicated test structures, which can be resource-intensive.

Purpose of the Study:

• To develop a novel Bayesian-based method for inferring systematic geometric variations in silicon photonics.
• To enable the extraction of variation distributions without the need for replicating identical test structures.
• To facilitate efficient characterization and study of process variations for improved PIC design.

Main Methods:

• A Bayesian inference framework was developed to analyze characterization data.
• The method was applied to silicon nitride ring resonators with varying design parameters.
• Systematic geometric variations (width, thickness, partial etch depth) were extracted.

Main Results:

• The method successfully inferred distributions for width (28 nm std dev), thickness (0.8 nm std dev), and partial etch depth (3.8 nm std dev).
• Spatial maps illustrating the distribution of these variations were generated.
• The approach demonstrated efficiency in characterizing process variations.

Conclusions:

• The developed Bayesian method provides an efficient approach for studying process variations in silicon photonics.
• This technique aids in understanding the magnitude and spatial distribution of geometric variations.
• The findings will contribute to the future design of high-yield silicon photonic circuits.