Design of surface microstructured substrate for optical trapping using machine learning.

Abstract

The development of a light trapping substrate is being pursued. This technology captures microparticles by the optical tweezer effect, utilizing the phenomenon where an electric field concentrates when light is illuminated on micro-structured metal surfaces. The electric field distribution generated near the micro-structures by the incident light is analyzed using the finite element method for electromagnetic fields. Then, Bayesian optimization, a type of machine learning method, is used to identify the micro-structure that maximizes electric field enhancement. Once identified, this micro-structure is created on an aluminum plate, with the possibility of optical trapping demonstrated by the velocity change of microparticles in a solution that was dropped onto the surface.