Multiobjective optimization in integrated photonics design

TL;DR

This paper introduces the parallel tabu search algorithm as an efficient alternative to genetic algorithms for inverse design in integrated photonics, demonstrating its effectiveness in beam shaping tasks.

Contribution

The paper presents the application of parallel tabu search to integrated photonics design, showing it outperforms genetic algorithms in efficiency and solution quality.

Findings

PTS achieves comparable or better solutions than GA

PTS requires less computation time

PTS handles multiobjective optimization effectively

Abstract

We propose the use of the parallel tabu search algorithm (PTS) to solve combinatorial inverse design problems in integrated photonics. To assess the potential of this algorithm, we consider the problem of beam shaping using a two-dimensional arrangement of dielectric scatterers. The performance of PTS is compared to one of the most widely used optimization algorithms in photonics design, the genetic algorithm (GA). We find that PTS can produce comparable or better solutions than the GA, while requiring less computation time and fewer adjustable parameters. For the coherent beam shaping problem as a case study, we demonstrate how PTS can tackle multiobjective optimization problems and represent a robust and efficient alternative to GA.