Data-driven model for optical-force calculation in fiber-optic tweezers

Traditional numerical methods face significant challenges in three-dimensional optical force fields with high degrees of freedom: the point-by-point scanning computing paradigm leads to an exponential increase in computation time. In this paper, we propose a data-driven deep learning model to effectively calculate the optical force of particles in fiber-optic tweezers and quickly identify the locations of optical potential wells in three-dimensional space. The results show that the computing speed of the data-driven model is several orders of magnitude higher than that of the traditional ray optics method, and the mean square error (MAE) is approximately 0.04 pN. This method quickly generates a series of particle motion trajectories, and it can provide the dynamic process by which the particles are captured by the optical potential well. It clearly delineates regions that are easily captured, providing clear strategic guidance for optimizing the optical trapping efficiency. This method provides a general optical force calculation tool for fiber-optic tweezers and promotes the paradigm shift of optical force calculation from "physics-driven" to "data-driven". This method enables a rapid and intuitive understanding of the optical force distribution and dynamical behavior of particles in fiber-optic tweezers, thus showing a wide range of applications in the fields of fine optical manipulation and particle positioning.