End-to-end inverse design for programmable acoustic tweezers with simultaneous force and torque control by metasurfaces.

Acoustic tweezers leverage acoustic radiation forces for noncontact manipulation. One of the core bottlenecks in multidimensional manipulation is the lack of a systematic design methodology, which prevents the generation of an acoustic field that simultaneously meets the collaborative control requirements of multi-degree-of-freedom forces and torques, making it difficult to achieve precise control under conditions of stable suspension, high-frequency rotation, and complex spatial constraints. To address this challenge, we develop an end-to-end inverse design methodology for acoustic tweezers based on coding metasurfaces, establishing a dual-objective, dual-scale optimization paradigm. At the microscale, the phase modulation and transmission efficiency are co-optimized through coupled physical models. While at the mesoscale, the particle suspension and rotation dynamics are considered. Based on the inverse design framework constructed with a finite-bit element library, we successfully optimized the metasurface configuration with specific acoustic response characteristics and achieved noncontact, multi-degree-of-freedom customized manipulation of individual particles. This approach provides implementation pathways for adaptive multiscale strategies in precision engineering applications.