Tailoring robust photonic components using stochastic topology optimization

Nanophotonic systems that rely on strong resonances or near-unity transmission through individual components are highly sensitive to geometric imperfections introduced during fabrication. To address this challenge, we present a stochastic topology optimization framework for designing photonic devices that are inherently robust to such imperfections. Manufacturing errors are modeled as Gaussian random fields and incorporated into a double-filtering scheme that perturbs the design geometry in a controlled manner during optimization. The underlying physics is modeled using the scalar Helmholtz equation in two dimensions, solved via a custom finite element implementation in MATLAB. Using Monte Carlo sampling and a p-mean objective function, the method consistently produces designs that outperform traditional deterministic and robust approaches under both ideal and perturbed conditions. The optimization yields simpler and more resilient designs to both frequency changes and geometric perturbations. Notably, a subset of designs optimized for correlation lengths close to the effective wavelength exhibit robustness across a wide geometric range, in turn offering a promising route toward scalable and reliable photonic device manufacturing.