Thermodynamic Equilibrium‐Nonequilibrium Competition in Nanofilaments for Achieving Complementary Electronic Synapses
Thermodynamic Equilibrium‐Nonequilibrium Competition in Nanofilaments for Achieving Complementary Electronic Synapses
复制标题
DOI：
10.1002/adfm.202207885
复制
发表时间：
2022-10
影响因子：
19
通讯作者：
Jiawei Guo;Wenhao Li;Yitao Liao;Yiwei Shen;Kun Wang;Chan Hee Suk;Xiongtu Zhou;Yongai Zhang
中科院分区：
文献类型：
--
作者：
Jiawei Guo;Wenhao Li;Yitao Liao;Yiwei Shen;Kun Wang;Chan Hee Suk;Xiongtu Zhou;Yongai Zhang
在人脑中，突触连接的自然互补行为通过星形胶质细胞和小胶质细胞的参与而增强和减弱，在训练和记忆过程中发挥着重要作用。尽管基于电化学金属化机制（ECM）的忆阻器在实现电子突触（e-synapses）方面具有巨大潜力，但缺乏互补的ECM e-synapses在一定程度上阻碍了人工神经网络的构建。在这项研究中，通过修改金属离子纳米通道来精心设计纳米丝中的热力学平衡-非平衡竞争（TC），以实现互补的 ECM e-突触。 TC诱导的纳米丝形态的演变包括电刺激引起的纳米丝细化行为和金属离子螯合引起的粗化行为。 TC 效应导致正脉冲刺激下电导降低，而休息时电导增加，这些行为与传统 ECM 电子突触的行为完全互补。最后，使用行为级建模验证了基于互补 ECM 电子突触制造具有学习能力的逻辑门电路的可行性。这些具有互补功能的 ECM 电子突触预计对大型神经网络系统领域的研究人员具有重要意义。
In the human brain, the natural complementary behaviors of synapse connections, which are enhanced and weakened through the participation of astrocytes and microglias, play an important role in the process of training and memory. Even though memristors based on the electrochemical metallization mechanism (ECM) have great potential in realizing electronic synapses (e‐synapses), the lack of complementary ECM e‐synapses has, to a certain extent, hindered the construction of artificial neural networks. In this study, the thermodynamic equilibrium‐nonequilibrium competition (TC) in nanofilaments is meticulously designed by modifying the metal‐ion nanochannels to achieve complementary ECM e‐synapses. The evolution in the morphology of a TC‐induced nanofilament includes a thinning behavior of the nanofilaments caused by electrical stimulation and a coarsening behavior caused by metal‐ion chelation. The TC effect leads to a decrease in the conductance under positive pulse stimulation and to an increase in the conductance while resting, these behaviors being exactly complementary to those of a conventional ECM e‐synapse. Finally, the feasibility of fabricating logic gate circuits with learning capability based on complementary ECM e‐synapses is verified using behavioral‐level modeling. These ECM e‐synapses with complementary functions are expected to be significant to researchers in the field of large neural network systems.