Abstract

Photonic devices are known to be sensitive to perturbations of the design geometry, and obtaining devices that perform sufficiently requires either trial and error or specialized control of the manufacturing. We present a density based stochastic topology optimization method that obtain photonic devices that theoretically are much more robust to non-uniform geometric perturbations compared to traditional optimization methods. The scalar Helmholtz equation, describing out-of-plane polarized light with time-harmonic behavior, is discretized using Q4 elements, and eigenmodes and effective indices are computed to accurately control the input and output of guided modes and attenuation in the PML layers. A mode conversion from an even to an odd mode is used for demonstration. The results show that the stochastic approach provide simpler design geometries while still being much more robust to perturbations of both input frequency and geometry. More efficient stochastic optimization methods will also be demonstrated, although the results are not final. More efficient methods makes it feasible to solve more complex problems, such as the full 3D Maxwells equations.