Multi-plane acoustic hologram generation with a physics-enhanced neural network for micro-particle manipulation

Abstract

A physics-enhanced multi-plane acoustic hologram deep neural network (PhysenNet_MPAH) approach is proposed for generating multi-plane acoustic hologram. By combining a convolutional neural network with a physical model, the PhysenNet_MPAH approach can generate high-quality acoustic holograms for holographic rendering of targeted acoustic intensity fields at multiple planes. This approach is capable of reconstructing both strong and weak-targeted multi-plane fields, with superior quality compared to traditional iterative angular spectrum approach. The reconstructed multi-plane acoustic fields are also demonstrated to be useful for three-dimensional particles manipulation, indicating potential applications in dynamic particles manipulation and volumetric display.