| 气相掺杂有机忆阻器实现电导调谐范围优化与类脑计算应用 |
| Vapor⁃doped organic memristors for tunable conductance range optimization and neuromorphic computing applications |
| |
| DOI: |
| 中文关键词: 忆阻器;气相掺杂;电导调谐范围;突触可塑性;手写数字识别 |
| 英文关键词:memristor; vapor-phase doping; conductance tuning range; synaptic plasticity; handwritten digit recognition |
| 基金项目:国家自然科学基金(62174089,62275130)和江苏省自然科学基金(BK20240138)资助项目 |
| 作者 | 单位 | | 李雯 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 | | 曹国峰 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 | | 钱浩文 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 | | 雷梦奥 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 | | 俞豪 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 | | 仪明东 | 南京邮电大学 材料科学与工程学院,江苏 南京 210023 |
|
| 摘要点击次数: 177 |
| 全文下载次数: 41 |
| 中文摘要: |
| 神经形态计算通过模拟生物突触的可塑性实现高效并行信息处理,而忆阻器因其电导连续
可调的特性是实现人工突触的理想器件。然而,传统忆阻器较窄的电导调谐范围严重制约了其在
神经形态计算中的应用。针对这一挑战,提出了通过F4-TCNQ对P3HT进行可控气相掺杂的策略,
在保持 P3HT 本征结构完整性的同时优化其电学性能。掺杂忆阻器实现了两倍电导调谐范围扩
展,在人工突触模拟中毫秒级的时间响应与生物神经突触的典型时间尺度匹配。基于该器件构建
的神经网络在MNIST手写数字识别任务中准确率达97.67%,优于未掺杂器件(91.28%)。该气相掺
杂策略为开发宽电导调谐范围的神经形态器件提供了解决方案,在类脑计算和人工智能硬件领域
具有重要的应用价值。 |
| 英文摘要: |
| Neuromorphic computing enables efficient parallel information processing by emulating the
plasticity of biological synapses, and memristors have emerged as ideal artificial synaptic devices due to
their continuously tunable conductance. However, the narrow conductance tuning range of conventional
memristors significantly limits their application in neuromorphic systems. To address this challenge, this
paper proposes a strategy that holds controllable vapor-phase doping of P3HT through F4-TCNQ. It can
optimize P3HT’s electrical performance while maintaining its intrinsic structural integrity . The doped
memristor achieves a twofold increase in conductance tuning range, and its millisecond-level temporal responses in artificial synaptic simulations match the characteristic time scales of biological neural synapses. A neural network constructed with this device achieves a 97.67% accuracy on the MNIST handwritten digit recognition task, outperforming its undoped counterpart (91.28%). This vapor-phase doping
strategy offers a viable solution for developing neuromorphic devices with a wide conductance tuning range, and holds significant promise for applications in the fields of brain-inspired computing and artificial intelligence hardware. |
| 查看全文 查看/发表评论 附件 |