Machine-Learning-Assisted Acoustic Consecutive Fano Resonances: Application to a Tunable Broadband Low-Frequency Metasilencer

The Fano resonance is a widespread wave-scattering phenomenon associated with an ultrasharp line shape, which just serves a narrow working frequency range around the interference frequency, rendering the realization of Fano-based applications extremely challenging. Here, we present and experimentally verify a mechanism of acoustic consecutive Fano resonances (ACFRs) with a symmetric profile for broadband sound attenuation, and extend to a practical implementation of a tunable low-frequency double-helix metasilencer. Based on the ACFRs' dependence on material parameters in the bilayer metamaterial model, we employ an inverse design using Bayesian machine learning to search the optimal broadband insulating performance with a rapid convergence speed (15 iterations). For practical requirement, we extend the ACFRs' prototype to a continuously tunable double-helix metastructure for broadband low-frequency sound attenuation. This broadband effect can be interpreted by the dual-band single-negativity property. A good agreement between numerical simulation and experiment evidences the effectiveness of the proposed metasilencer with tunable sound attenuation (>90%) in 425-865 Hz and high ventilation (>80%) at various double-helix combinations. Our proposed ACFRs' mechanism and its associated metastructure would open routes to promising acoustic metamaterial-based applications, such as filtering, switching, and sensing, and beyond.

Automated summary

The study introduces a mechanism of acoustic consecutive Fano resonances (ACFRs) for broadband sound attenuation and implements a tunable low-frequency double-helix metasilencer. The optimal broadband insulating performance of ACFRs is obtained using Bayesian machine learning, and the effectiveness of the metasilencer is confirmed through numerical simulation and experiments.