FDTDX is an efficient open-source Python package for the inverse design of three-dimensional photonic nanostructures using the Finite-Difference Time-Domain (FDTD) method. Built on JAX, it provides native GPU support and automatic differentiation capabilities, making it ideal for large-scale 3D design in nanophotonics.

Key Features

• High Performance: GPU-accelerated FDTD simulations with multi-GPU scaling capabilities
• Memory Efficient: Leverages time-reversibility in Maxwell's equations for efficient gradient computation
• Automatic Differentiation: Built-in gradient-based optimization for complex 3D structures
• User-Friendly API: Intuitive positioning and sizing of objects in absolute or relative coordinates
• Large-Scale Design: Capable of handling simulations with billions of grid cells
• Open Source: Freely available for research and development

Documentation

Visit our documentation for:

• Detailed API reference
• Tutorial guides
• Best practices

Also check out our whitepaper for some examples and a comparison to other popular FDTD-frameworks.

Installation

Install FDTDX using pip:

pip install fdtdx

For development installation, clone the repository and install in editable mode:

git clone https://github.com/ymahlau/fdtdx
cd fdtdx
pip install -e .

Multi-GPU

# The following lines often lead to better memory usage in JAX
# when using multiple GPU.
export XLA_PYTHON_CLIENT_ALLOCATOR="platform"
export XLA_PYTHON_CLIENT_PREALLOCATE="false"
export NCCL_LL128_BUFFSIZE="-2"
export NCCL_LL_BUFFSIZE="-2"
export NCCL_PROTO="SIMPLE,LL,LL128"

Citation

If you find this repository helpful for you work, please consider citing:

@article{schubertmahlau2025quantized,
  title={Quantized Inverse Design for Photonic Integrated Circuits},
  author={Schubert, Frederik and Mahlau, Yannik and Bethmann, Konrad and Hartmann, Fabian and Caspary, Reinhard and Munderloh, Marco and Ostermann, J{\"o}rn and Rosenhahn, Bodo},
  journal={ACS Omega},
  doi={10.1021/acsomega.4c10958},
  year={2025}
}

Acknowedgement

This project was developed at the Institute of Information Processing at Leibniz University Hannover, Germany and sponsored by the cluster of excellence PhoenixD (Photonics, Optics, Engineering, Innovation across Disciplines).