Reservoir computing, as a rapid and low-power neuromorphic computing technology, has garnered significant attention for processing temporal signals due to its low training cost, which stems from the need to train only a simple readout layer. However, traditional reservoir computing is limited in feature dimensionality for spatiotemporal signal extraction, and its output characteristics are restricted to parameters such as voltage or light intensity. Circularly polarized light (CPL), an optical excitation with chiral selectivity, enables multidimensional information processing through left-handed circularly polarized light (L-CPL) and right-handed circularly polarized light (R-CPL), thereby introducing a new physical dimension into reservoir computing. Current reservoir computing systems for temporal signal processing typically rely on separate sensing, memory, and computing units, resulting in inefficiency, high energy consumption, and bulky architectures. To address these challenges, we developed a CPL-based synaptic transistor using a chiral organic semiconductor and its corresponding reservoir computing system. The device employs the chiral organic semiconductor (S,S,R,R)-DPP6T, which exhibits strongly differentiated electrical responses to different polarization states of light, providing a novel mechanism for optical regulation in reservoir computing systems and filling the gap in polarization-dimension control for light-input electrical-response reservoir computing. This system achieved handwritten digit recognition under both L-CPL and R-CPL illumination, with high accuracies of 95.52% and 96.06%, respectively. This capability enables the interpretation of the same dataset under different polarization states, offering a new solution for efficient, low-power, and multidimensional reservoir computing.
LIN et al. (Wed,) studied this question.