English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Optical computing is an emerging technology for next-generation efficientartificial intelligence (AI) due to its ultra-high speed and efficiency.Electromagnetic field simulation is critical to the design, optimization, andvalidation of photonic devices and circuits. However, costly numericalsimulation significantly hinders the scalability and turn-around time in thephotonic circuit design loop. Recently, physics-informed neural networks havebeen proposed to predict the optical field solution of a single instance of apartial differential equation (PDE) with predefined parameters. Theircomplicated PDE formulation and lack of efficient parametrization mechanismslimit their flexibility and generalization in practical simulation scenarios.In this work, for the first time, a physics-agnostic neural operator-basedframework, dubbed NeurOLight, is proposed to learn a family of frequency-domainMaxwell PDEs for ultra-fast parametric photonic device simulation. We balancethe efficiency and generalization of NeurOLight via several novel techniques.Specifically, we discretize different devices into a unified domain, representparametric PDEs with a compact wave prior, and encode the incident light viamasked source modeling. We design our model with parameter-efficientcross-shaped NeurOLight blocks and adopt superposition-based augmentation fordata-efficient learning. With these synergistic approaches, NeurOLightgeneralizes to a large space of unseen simulation settings, demonstrates2-orders-of-magnitude faster simulation speed than numerical solvers, andoutperforms prior neural network models by ~54% lower prediction error with~44% fewer parameters. Our code is available athttps://github.com/JeremieMelo/NeurOLight.

NeurOLight: A Physics-Agnostic Neural Operator Enabling Parametric Photonic Device Simulation