Fig. 5: Application of the Neuroshaper framework to various nanophotonic devices.
From: Shaping freeform nanophotonic devices with geometric neural parameterization

a 2D mode converter design demonstrating efficient conversion between different waveguide modes. The optimized structure maintains smooth connection with both input and output waveguides and is disconnected from the boundary boarders. b 2D bandpass filter showing improved contrast ratio compared to conventional approaches while maintaining structural connectivity across all waveguide interfaces. c 2D wavelength multiplexer design exhibiting superior transmission efficiencies compared to conventional designs while ensuring robust waveguide connectivity. d 3D grating coupler optimization incorporating multiple constraints: connectivity with the output waveguide, boundary disconnection, and symmetry-preserving feature size limits. The conventional design violates feature size limits. e 3D plasmonic nanoantenna optimization with periodic boundary conditions and feature size constraints, demonstrating enhanced field intensity compared to conventional designs. Simulated field distributions for these devices can be found in Supplementary Figs. 3 and 4.