Showing 1–1 of 1 results for author: Zorba, V

AI Driven Laser Parameter Search: Inverse Design of Photonic Surfaces using Greedy Surrogate-based Optimization

Authors: Luka Grbcic, Minok Park, Juliane Müller, Vassilia Zorba, Wibe Albert de Jong

Abstract: Photonic surfaces designed with specific optical characteristics are becoming increasingly important for use in in various energy harvesting and storage systems. , In this study, we develop a surrogate-based optimization approach for designing such surfaces. The surrogate-based optimization framework employs the Random Forest algorithm and uses a greedy, prediction-based exploration strategy to id… ▽ More Photonic surfaces designed with specific optical characteristics are becoming increasingly important for use in in various energy harvesting and storage systems. , In this study, we develop a surrogate-based optimization approach for designing such surfaces. The surrogate-based optimization framework employs the Random Forest algorithm and uses a greedy, prediction-based exploration strategy to identify the laser fabrication parameters that minimize the discrepancy relative to a user-defined target optical characteristics. We demonstrate the approach on two synthetic benchmarks and two specific cases of photonic surface inverse design targets. It exhibits superior performance when compared to other optimization algorithms across all benchmarks. Additionally, we demonstrate a technique of inverse design warm starting for changed target optical characteristics which enhances the performance of the introduced approach. △ Less

Submitted 20 June, 2024; originally announced July 2024.