PIC2O-Sim

PIC2O is a physics-inspired CNN-based neural operator that serves as a surrogate for the FDTD simulation in solving Maxwell's equations.

Prerequisites

To run the code, please install the open-source FDTD simulator MEEP, pyutils and other necessary packages in requirements.txt

Data Generation

To run the code, first, you need to generate data for three different types of photonic devices:

1. Change directory to the data folder for FDTD simulations:

   cd ./data/fdtd

2. Run the following scripts to generate data for each device:

   python simulation_mmi.py
   python simulation_mrr.py
   python simulation_metaline.py

Training

After generating the dataset, the training could be launched using the following command:

python ./scripts/fdtd/cnn/train_PICCO.py