Learning-based acoustic displacement field modeling and micro-particle control

Acoustic manipulation has emerged as a promising microtarget control method with potential applications in microchip assembly, noncontact control of chemical materials, and cell control. However, the precision of acoustic micro-/nano-manipulation is constrained by the ambiguity of the acoustic displacement field. The nonlinear properties of displacement fields cannot be appropriately simulated using conventional mathematical fitting techniques. In this study, considering the complexity of particle motion on a thin plate driven by acoustic waves, the original displacement dataset was collected based on an experimental platform and was then trained using a neural network to model the displacement of unknown points on the plate. The corresponding cost functions were designed based on a manipulation task and particle number to determine the frequency sequence and achieve motion control of micro/nano targets by selecting the driving frequency with the lowest cost during the controlling process. Based on the above predictions and control effects, the acoustic manipulation platform constructed in this study could be used to manipulate the linear motion of a single particle and control multiple particle encounters as well as to manipulate the complex trajectory "NK". This work opens up an accessible path for an acoustic displacement field and enables achievable microtarget motion control.