近日，西安交通大学何刚团队报道了基于硫代紫罗兰素的电化学离子门控实现生物启发的高性能神经形态器件。相关论文发表在2025年12月23日出版的《德国应用化学》杂志上。

神经形态计算是一种受生物启发的计算范式，通过模拟生物神经网络的结构与功能实现高效信息处理，这一领域对前沿材料与器件架构提出了迫切需求。

研究组展示了一种基于噻吩紫精电解质的仿生电化学神经形态器件。在紫精结构中引入噻吩基团形成的噻吩紫精二价阳离子，不仅降低了能隙，还提升了自由基稳定性与电化学活性。该器件展现出优异的环境稳定性，并能通过电压脉冲刺激实现电导率的连续调节。

将其集成于卷积神经网络进行图像识别任务时，在Fashion-MNIST数据集上取得了接近80%的准确率。此外，该器件成功模拟了包括时序依赖可塑性、巴甫洛夫联想学习在内的关键突触功能，并支持双端逻辑门运算。这些研究成果显著拓展了紫精类材料在神经形态电子学中的功能维度，为新一代电化学人工突触的设计提供了新思路。

附：英文原文

Title: Bioinspired High-Performance Neuromorphic Devices Enabled by Thienoviologen-Based Electrochemical Ion Gating

Author: Siyu Sun, Yueyan Zhang, Zhikang Han, Chengjing Liu, Bai Sun, Wei Zhang, Gang He

Issue&Volume: 2025-12-23

Abstract: Neuromorphic computing is a bioinspired paradigm that emulates the structure and functionality of biological neural networks, demanding cutting-edge materials and device architectures. In this work, we present a bioinspired electrochemical neuromorphic device (BEND) utilizing a thienoviologen-based electrolyte. The incorporation of thiophene groups into the viologen structure (ThV2+) leads to a reduced energy gap, improved radical stability, and enhanced electrochemical activity. The device exhibits excellent ambient stability and continuously tunable conductivity in response to voltage pulse stimulation. When integrated into a convolutional neural network (CNN) for image recognition, BEND achieves an accuracy of nearly 80% on the Fashion-MNIST dataset. Moreover, the device successfully mimics essential synaptic functions such as spike-timing-dependent plasticity (STDP), Pavlovian learning, and supports dual-terminal logic gate operations. These results significantly expand the functional versatility of viologen-based materials in neuromorphic electronics and offer new insights into the design of next-generation electrochemical artificial synapses.

DOI: 10.1002/anie.202523345

Source: https://onlinelibrary.wiley.com/doi/10.1002/anie.202523345