Fabrication-aware inverse design with shape optimization for photonic integrated circuits

Area of Science:

• Photonics
• Computational Design
• Nanofabrication

Background:

• Inverse design (ID) in silicon photonics optimizes device geometry but often results in designs sensitive to fabrication imperfections.
• Process variations during manufacturing, such as those from deep-ultraviolet (DUV) and electron-beam lithography (EBL), significantly degrade device performance.

Purpose of the Study:

• To develop a fabrication-aware inverse design (FAID) method that accounts for lithography process variations.
• To improve the performance and manufacturability of silicon photonic devices designed using inverse design.

Main Methods:

• Integrated lithography models for DUV and EBL into the inverse design optimization workflow.
• Generated and fabricated a Y-branch coupler and a subwavelength grating (SWG)-to-strip converter using the FAID approach.

Main Results:

• FAID successfully generated devices that compensate for predicted process bias.
• Fabricated devices using FAID exhibited up to 0.7 dB lower insertion loss per device compared to conventional ID.

Conclusions:

• Fabrication-aware inverse design is a viable strategy to mitigate performance degradation in scalable silicon photonics fabrication.
• The FAID workflow enables the creation of robust photonic devices that are resilient to manufacturing variations.