The human brain achieves remarkable efficiency in learning and memory by precisely regulating ionic flux through nanoscale protein channels. Inspired by this principle, we present droplet interface membranes incorporating ion channels as artificial synapses for neuromorphic computing. In this platform, aqueous droplets are connected by lipid bilayers embedded with ion channels, creating minimal yet functional iontronic synapses. These devices exhibit memristive ion transport with hysteretic current-voltage characteristics, originating from the interplay between voltage-dependent channel dynamics and ionic conduction. Beyond fundamental synaptic plasticity, they reproduce higher order functions, including short-term memory and associative learning. Leveraging these capabilities, we implement reservoir computing with droplet-based synapses, demonstrating accurate handwritten digit classification and adaptive tic-tac-toe gameplay at low energy cost. Finally, we show that droplet-ion channel synaptomimetic elements can be scaled into networks with potential applications in neuromorphic sensing. Together, these results establish droplet membranes with ion channels as a scalable, soft, and biomimetic platform for next-generation neuromorphic computing.
Li et al. (Sun,) studied this question.