Scholarlink

<p>Machine learning (ML) has become an important tool in nanophotonics for predicting the optical response of plasmonic nanostructures based on parameters, such as geometry and material. Traditional modeling approaches often involve high computational costs and long simulation times, which hinder rapid optimization and design. This work investigates the plasmonic resonance behavior of cylindrical gold nanostructures and evaluates several ML regression models for direct and inverse design. Among the tested methods, the Extra Trees Regressor (ET) achieved the best performance in predicting scattering spectra, with MAE of 0.097, MSE of 0.052, RMSE of 0.224, and <i>R</i><sup>2</sup> of 0.999. For inverse design, which maps spectral inputs to physical parameters, the model demonstrated high accuracy and strong generalization, recovering nanocylinder radius with errors below 2.6%. These results highlight the potential of ML-based strategies to accelerate nanophotonic design, reduce dependence on high-cost simulations, and support efficient characterization workflows.</p> Graphical abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Resonance spectrum prediction and inverse design of plasmonic gold nanostructures based on regression

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月3日

查看全文
  • Luana S. P. Maia,
  • Darlan A. Barroso,
  • Tiago B. Marinho,
  • Matheus R. Araújo,
  • João V. A. Pereira,
  • Carlos Alexandre R. Fernandes,
  • Renato J. Martins,
  • Benoit Cluzel,
  • Glendo F. Guimarães

摘要

Machine learning (ML) has become an important tool in nanophotonics for predicting the optical response of plasmonic nanostructures based on parameters, such as geometry and material. Traditional modeling approaches often involve high computational costs and long simulation times, which hinder rapid optimization and design. This work investigates the plasmonic resonance behavior of cylindrical gold nanostructures and evaluates several ML regression models for direct and inverse design. Among the tested methods, the Extra Trees Regressor (ET) achieved the best performance in predicting scattering spectra, with MAE of 0.097, MSE of 0.052, RMSE of 0.224, and R2 of 0.999. For inverse design, which maps spectral inputs to physical parameters, the model demonstrated high accuracy and strong generalization, recovering nanocylinder radius with errors below 2.6%. These results highlight the potential of ML-based strategies to accelerate nanophotonic design, reduce dependence on high-cost simulations, and support efficient characterization workflows.

Graphical abstract