Scholarlink

<p>This paper proposes a tunable terahertz filter based on graphene and black phosphorene (BP), with machine learning-aided analysis. The design consists of five square elements fabricated from graphene, BP, or a hybrid of both, integrated with a metallic waveguide on an Arlon substrate. The study explores three structural setups. In the first, the squares are made entirely of graphene; in the second, they are composed of black phosphorus (BP); and in the third, a graphene layer is placed on top of a BP layer. Modeling and simulation of the structure were performed using the Wave Concept Iterative Process method alongside CST Studio Suite. By leveraging the complementary properties of graphene and BP, the hybrid configuration demonstrates enhanced performance compared to filters based on either material alone. Specifically, the resonant frequency shifts toward higher values, accompanied by a broader tuning range than in the other configurations. In the hybrid case, the resonance frequency <InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(f_r\)</EquationSource> </InlineEquation> increases from 7.5 to 8.45 THz, while the bandwidth <InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(\Delta f_r\)</EquationSource> </InlineEquation> varies from 0.13 to 0.35 THz, indicating a notable improvement in tunability and spectral selectivity. Machine learning models, including Random Forest and XGBoost, were trained on simulation data to predict resonance frequency and bandwidth based on key material features which are graphene’s chemical potential, temperature, relaxation time, and BP’s electron doping level. Both models demonstrated high accuracy with <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(R^2\)</EquationSource> </InlineEquation> values above 0.998, enabling rapid and reliable tuning of the filter. The Random Forest model showed slightly better predictive performance, making it a valuable tool for efficient filter design optimization in reconfigurable terahertz applications.</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

A tunable terahertz filter based on graphene and black phosphorene with machine learning-aided analysis

  • 期刊论文

  • 发表日期：

    2026年4月29日

查看全文
  • Hamza Ben Krid,
  • Achref Gadhgadhi,
  • Aymen Hlali,
  • Yassine Hachaïchi,
  • Hassen Zairi

摘要

This paper proposes a tunable terahertz filter based on graphene and black phosphorene (BP), with machine learning-aided analysis. The design consists of five square elements fabricated from graphene, BP, or a hybrid of both, integrated with a metallic waveguide on an Arlon substrate. The study explores three structural setups. In the first, the squares are made entirely of graphene; in the second, they are composed of black phosphorus (BP); and in the third, a graphene layer is placed on top of a BP layer. Modeling and simulation of the structure were performed using the Wave Concept Iterative Process method alongside CST Studio Suite. By leveraging the complementary properties of graphene and BP, the hybrid configuration demonstrates enhanced performance compared to filters based on either material alone. Specifically, the resonant frequency shifts toward higher values, accompanied by a broader tuning range than in the other configurations. In the hybrid case, the resonance frequency \(f_r\) increases from 7.5 to 8.45 THz, while the bandwidth \(\Delta f_r\) varies from 0.13 to 0.35 THz, indicating a notable improvement in tunability and spectral selectivity. Machine learning models, including Random Forest and XGBoost, were trained on simulation data to predict resonance frequency and bandwidth based on key material features which are graphene’s chemical potential, temperature, relaxation time, and BP’s electron doping level. Both models demonstrated high accuracy with \(R^2\) values above 0.998, enabling rapid and reliable tuning of the filter. The Random Forest model showed slightly better predictive performance, making it a valuable tool for efficient filter design optimization in reconfigurable terahertz applications.