Summary

We built the world’s most advanced differentiable GPU-accelerated FDTD for electromagnetic simulation and inverse design for photonics, terahertz, microwave and RF.

Features

Powerful

• Electromagnetic inverse design and simulation in just few lines of Python code!

• Broadband and multimode S-parameters

• Embedded mode solver for modal sources and monitors

• .gds and gdsfactory integration

• .stl / .step 3D geometry import

Fast

• GPU acceleration on NVIDIA, AMD, and Apple Silicon

• Adaptive graded mesh reduces cell count

• Tensor subpixel smoothing boosts accuracy

Smart

• Fully differentiable (native automatic differentiation in Julia)

• Simultaneous inverse design of multiple 2D and 3D structures

• Length scale controlled geometry optimizer with fabrication constraints

Comprehensive (some features require additional dev)

• Modal sources, plane waves, Gaussian beams, custom sources

• Oblique sources and monitors

• PML, periodic, Bloch, PEC boundaries

• Near and far field radiation patterns

• Nonlinear, dispersive and anisotropic materials