Data-driven design and optimization of ultra-tunable acoustic metamaterials

This paper presents a data-driven design and optimization of acoustic metamaterials with three-phase materials for highly tunable wave transmission. The geometry of representative unitcell is defined by the trigonometric series function to describe an arbitrary shape with symmetry, which enables the unitcell to achieve a large sub-wavelength bandgap. We propose a lightweight and efficient algorithm, ‘decoupled gradient decent (DGD)’, to search for the optimal design and uncover the ‘best’ shape features—the interface curvature—in tuning the wave transmission. As a result, the host composite can partly overlap the individual cell’s bandgap and achieve a wide frequency gap that forbids wave transmission, namely a passive tunability. Another advantage of the trigonometric series designed shape is the high flexibility. A slight surface pressure obviously deforms the unitcell and shifts its band structure. Our simulation shows that a moderate pressure dramatically changes the frequency forbidding gap for both traversal and longitudinal wave transmissions, which indicates an active tunability. The surface deformation can be applied by either a mechanical pressure or external electric field if the composite uses a dielectric substrate. Therefore, this study opens a sandbox of manipulating wave transmission through the topology and structure optimization in applications such as seismic damping (Hz), noise insulating (kHz) and ultrasound imaging (MHz).