講演情報
[10a-N106-10]Polymer Wire Synaptic Device with Enhanced Endurance for Neuromorphic Wetware Computing
〇(D)Adrian Dy Go1, Seiya Watanabe1, Hiroyuki S. Kato1, Megumi Akai-Kasaya1 (1.Osaka Univ.)
キーワード：
Synaptic Device、Neuromorphic Computing、Polymer Wire
The rapid advancement of Artificial Intelligence (AI) has driven demand for energy-efficient and high-performance computing architectures beyond the traditional Von Neumann model. Neuromorphic computing offers a brain-inspired alternative by integrating memory and processing. Organic-based synaptic devices are promising neuromorphic platforms due to their biocompatibility, flexibility, and tunability. However, challenges remain regarding structural degradation and operational stability. In this study, we report a two-terminal synaptic wetware device based on hydroxymethyl PEDOT:sodium dodecyl benzene sulfonate (PEDOT-MeOH:SDBS) polymer wires. Repeated voltage pulses on one side of the wire induced structural asymmetry and enabled short-term potentiation and depression via uneven ionic doping. Additionally, the wire showed excellent endurance over 100 cycles when suspended in ethylene glycol. To demonstrate neuromorphic functionality, the conductance states of the wires were used as physical kernels for a convolutional neural network digit classification task. During training, the conductance states of the wires were tuned via voltage application. The system achieved 96% accuracy over 450 epochs for classifying 5 x 5 digits from 0-9, showcasing the potential of polymer-based wetware for neuromorphic computing.