We present a ferroelectric-based data reservoir designed for neuromorphic computing applications. The reservoir consists of an ensemble of nonlinear ferroelectric nanodots capable of storing information through their stable polarization states. These nanodots are confined between electrodes and form a network of parallel connected capacitive elements, which can either be addressed individually or integrated between common conducting plates. By leveraging the intrinsic nonlinearity and memory of ferroelectric polarization switching, the device maps incoming signals into a high-dimensional space of polarization configurations. This physical transformation enables efficient temporal information encoding and provides a rich dynamic representation for subsequent processing in a readout layer of a neuromorphic circuit.
Razumnaya et al. (Fri,) studied this question.