National Research Council of Canada. Advanced Electronics and Photonics
National Research Council of Canada. Digital Technologies
2022 Photonics North (PN), May 24-26, 2022, Niagara Falls, ON, Canada
silicon photonics; machine learning; deep learning; convolutional neural networks; fabrication process variations; fabrication; performance evaluation; degradation; scanning electron microscopy; electric potential; predictive models
Next-generation silicon photonic devices often contain complex, nanoscale features to enhance their performance; however, these features experience significant variations from fabrication imperfections, which cause significant performance degradation in practical implementation. We present a machine learning model that learns from a modest set of scanning electron microscope images to quickly and accurately predict the fabrication variations. Additionally, we present the potential use of this model in the automated correction of fabrication-sensitive device features.