In the recently published ACS Photonics paper titled "Time Reversal Differentiation of FDTD for Photonic Inverse Design", co-first authors Jerry Tang Rui Jie and Soon Wei Daniel Lim present a novel approach for computing gradients in finite difference time domain (FDTD) simulations for photonic inverse design. Their technique surmounts the memory constraints inherent in automatic differentiation (AD) and preserves the computational efficiency of the state-of-the-art adjoint method. The method calculates gradients by replaying time-dependent fields at lossy boundaries, which significantly reduces memory usage—by as much as 98% compared to AD, as demonstrated in one 3D simulation scenario. The tool is employed to design intricate nanostructures, including a color sorter and a resonant array that enforces group delay, and can be generalized to other gridded simulators, potentially expanding the reach of topology optimization techniques into other domains.

Marcus Ossiander and Xinghui Yin also contributed to the paper.

Read the full paper in ACS Photonics here: https://doi.org/10.1021/acsphotonics.3c00694.