Physical reservoir operation utilizing polarization change in a liquid crystal film

Abstract In recent years, there has been growing interest in physical reservoir computing, which operates with low power consumption and low computational cost, utilizing various phenomena for their operations. In this study, we developed a liquid crystal-based reservoir that utilizes dielectric polarization as a new state variable for the physical reservoir operation. Nonlinear polarization change in a liquid crystal film was confirmed both by observation using a polarized optical microscope and by electrical measurements. We also evaluated the performance of the liquid crystal reservoir by basic methods such as short-term memory tasks. In the MNIST classification task, we achieved an accuracy of approximately 88%, suggesting the potential of the liquid crystal reservoir for future applications.