用于神经形态计算的基于软生物材料的柔性人工突触,Advanced Electronic Materials

当前位置： X-MOL 学术 › Adv. Electron. Mater. › 论文详情

Our official English website, www.x-mol.net, welcomes your feedback! (Note: you will need to create a separate account there.)

用于神经形态计算的基于软生物材料的柔性人工突触
Advanced Electronic Materials ( IF 5.3 ) Pub Date : 2022-07-10 , DOI: 10.1002/aelm.202200449
Tao Guo ₁ , Jiawei Ge ₂ , Bai Sun ₃ , Kangqiang Pan ₄ , Zhao Pan ₁ , Lan Wei ₄ , Yong Yan ₅ , Y. Norman Zhou ₁ , Yimin A. Wu ₁

Affiliation

人工突触对于可能彻底改变冯诺依曼系统的神经形态计算芯片至关重要。基于生物材料的生物忆阻器由于其生物相容性、可降解性、灵活性和低成本而被研究作为突触仿真器来开发神经形态计算芯片。然而，现有的基于生物材料的人工突触存在生物突触功能有限、可靠性不足、耐久性差等问题。特别是，用于神经形态计算的具有长期增强/抑制（LTP/LTD）稳定突触性能的基于蛋白质的人工突触具有挑战性。在这里，卵蛋白@CuO 的软材料被用来开发人工突触。该设备可以模拟生物突触功能，包括兴奋性突触后电流 (EPSC)，尖峰数依赖性可塑性 (SNDP)、配对脉冲促进 (PPF) 和 LTP/LTD。用于模式识别的神经形态计算模拟已经获得了95%的高精度。结合密度泛函理论计算、多物理场模拟和电气测量，模拟电阻开关机制归因于电子跳跃。此外，该设备具有灵活性，可用于开发可穿戴系统。结果揭示了生物相容和可穿戴的神经形态计算芯片。模拟电阻开关机制归因于电子跳跃。此外，该设备具有灵活性，可用于开发可穿戴系统。结果揭示了生物相容和可穿戴的神经形态计算芯片。模拟电阻开关机制归因于电子跳跃。此外，该设备具有灵活性，可用于开发可穿戴系统。结果揭示了生物相容和可穿戴的神经形态计算芯片。

"点击查看英文标题和摘要"

Soft Biomaterials Based Flexible Artificial Synapse for Neuromorphic Computing

Artificial synapses are vital for neuromorphic computing chips that can potentially revolutionize von Neumann systems. Biomaterials-based bio-memristors have been investigated as synaptic emulators to develop neuromorphic computing chips due to their biocompatibility, degradability, flexibility, and low costs. However, the existing biomaterials-based artificial synapses suffer from limited biological synapse functions, insufficient reliability, and poor endurance. Particularly, protein-based artificial synapse with stable synaptic performances of long-term potentiation/depression (LTP/LTD) for neuromorphic computing is challenging. Here, the soft material of egg albumen@CuO are employed to develop an artificial synapse. The device can mimic bio-synaptic functionalities, including the excitatory postsynaptic current (EPSC), spike-number-dependent plasticity (SNDP), paired-pulse facilitation (PPF), and LTP/LTD. High accuracy of 95% has been obtained by the neuromorphic computing simulation for pattern recognition. Combining with density functional theory calculations, multiphysics simulations, and electrical measurements, the analog resistive switching mechanism is attributed to electron hopping. In addition, the device is flexible and can be used to develop wearable systems. The results shed light on the biocompatible and wearable neuromorphic computing chips.

更新日期：2022-07-10

点击分享查看原文

点击收藏

阅读更多本刊新发论文本刊介绍/投稿指南