Accelerating Finite-Difference Frequency-Domain Simulations for Inverse Design Problems in Nanophotonics using Deep Learning

Inverse design of nanophotonic devices becomes increasingly relevant for the development of complex photonic integrated circuits. Electromagnetic first-order simulations contribute the overwhelming computational cost to the optimization routines in established inverse design algorithms, requiring more efficient methods for enabling improved and more complex design process flows. Here we present such a method to predict the electromagnetic field distribution for pixel-discrete planar inverse designed structures using deep learning. Our model is able to infer accurate predictions used to initialize a conventional Finite Difference Frequency-Domain-algorithm and thus lowers the time required for simulating the electromagnetic response of nanophotonic device layouts by about 50 %. We demonstrate the applicability of our deep learning method for inverse design of photonic integrated powersplitters and mode converters and we highlight the possibility of exploiting previous learning results in subsequent design tasks of novel functionalities via finetuning on reduced data sets, thus improving computational speed further.