Fast Simulation and Inverse Design of Nanophotonic Lasers

Student’s name: Peter Mugaba Noertoft Home Institution: Stanford University NNCI Site: CNS @ Harvard University REU Principal Investigator: Dr. Kiyoul Yang - School of Engineering and Applied Sciences, Harvard University REU Mentors: Aditya Paul - EECS, MIT and Rui Jiang - Applied Physics, Harvard University Abstract: The past decade has seen a push to miniaturize lasers, leveraging nanophotonics to shrink bulky tabletop systems to chip-scale devices, compact enough for a new suite of applications. Harnessing the full power of the nonlinear dynamics, machine learning and fast numerical methods promise increased flexibility in the design and control of ultrafast lasers[1]. The goals of this project are twofold. The first stage is to develop fast GPU-accelerated differential equation solvers for models describing multimode and solid-state gain dynamics in nonlinear nanophotonic waveguides. Inspired by algorithms for similar problems in optical fibers [2], we seek to leverage parallel computing methods so the solvers are efficient for use on large scale design problems. The second stage is to develop tools for ‘learning' of laser parameters and designs, such as for applications in supercontinuum lasers with uniform output spectrum in the visible regime[3]. Rather than manual sweeping of parameters, test data paired with machine learning and mathematical optimization inform design and characterization of devices in a massive parameter space. This project complements existing work on photonic inverse design by expanding the tools available for design with optical nonlinearity.