betaFeedback

WorksPhysics and Astronomy

Gradient-Free Inverse Design of 3D Nanophotonic Devices Using the Level-Set-Method

Abstract

Advances in 3D nanophotonic additive manufacturing enable increasingly complex, almost free-form devices for manipulating light at sub-wavelength scales. Extending nanophotonic designs into three dimensions introduces additional degrees of freedom. Exploiting this freedom requires inverse-design methods that can efficiently navigate this large design space while accommodating fabrication constraints. Here, we combine gradient-free optimization with the level-set method (LSM) to provide such an approach for 3D additive manufacturing. Representing the 3D geometry as a level set constructed from Gaussian radial basis functions retains substantial geometric freedom with only a few hundred design variables. This compact representation makes gradient-free optimization practical for fully 3D nanophotonic devices, eliminating the need for adjoint calculations. This unlocks the use of any Maxwell solver or auxiliary algorithms that should express fabrication constraints for which the required derivatives are unavailable or impractical to obtain, and enables embarrassingly parallel workflows. Consequently, binary fabrication requirements, such as structural connectivity, can be rigorously enforced, avoiding the differentiable approximations commonly used in gradient-based optimization. Here, we demonstrate the combination of the gradient-free LSM with a connected-component analysis algorithm to enforce the structural integrity of the design. We demonstrate the applicability of the gradient-free LSM by optimizing a Y-splitter while strictly enforcing the structural integrity of the material and the void in the design.

Paper details

vlnc:W_uaa3w7m4
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC-BY-NC-ND-4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…