Complex-amplitude design of dynamic acoustic holography via physics-assisted neural network

Acoustic holography promises diverse applications, including high-resolution imaging, water–air communication, and in-vivo tissue/organ printing. While complex-amplitude design of acoustic holography yields superior performance in resolution and information content for sophisticated acoustic fields, but most holograms are spatially and temporally static once fabricated. This dilemma may be addressed by introducing the incident wavefront modulation as an additional degree of freedom, and dynamic holography would be realized by tuning the incident wavefronts via electrically switching a transducer array. Here, we propose a physics-assisted neural network optimization framework for the complex-amplitude design of dynamic acoustic holography, which enables joint optimization of multiple incident amplitude and phase profile. The framework embeds an angular-spectrum propagation model into neural network and introduces amplitude-binarization regularization to enable dynamic control through an electrically switched partitioned transducer. Furthermore, the proposed method exhibits remarkable versatility, supporting the phase modulations in either continuous or binary manner and even cascaded configuration. As a proof-of-concept demonstration, it successfully achieves the dynamic display of three-frame animations and 26 alphabetic letters, creating compelling “acoustic movies”. Quantitatively, the proposed method outperforms traditional designs both numerically and experimentally, which reduces mean squared error by 24.1% and 32.1%, respectively. We believe this work would greatly advance acoustic holography and imply unprecedented opportunity for broad acoustics.