Reconfigurable Artificial Synapses between Excitatory and Inhibitory Modes Based on Single-Gate Graphene Transistors
Reconfigurable Artificial Synapses between Excitatory and Inhibitory Modes Based on Single-Gate Graphene Transistors
复制标题
基于单栅石墨烯晶体管的兴奋性和抑制性模式之间的可重构人工突触
DOI：
10.1002/aelm.201800887
复制
发表时间：
2019
影响因子：
6.2
通讯作者：
Jin Zhi
中科院分区：
文献类型：
--
作者：
Yao Yao;Huang Xinnan;Peng Songang;Zhang Dayong;Shi Jingyuan;Yu Guanghui;Liu Qi;Jin Zhi
可重构人工突触是构建人工智能系统的关键，其突触反应在兴奋和抑制模式之间进行调制。然而，用简单的单栅极晶体管实现这种可重构性仍然是一个挑战。在这里，采用富氢氮化硅膜作为栅极电介质来构建单栅极控制的基于石墨烯的人工突触，以实现可重构的突触响应。在这种介质中，陷阱和可移动的氢离子被引入，分别引起载流子捕获效应和电容门效应。相比之下，电容门控效应需要更强的电场激发，并且可以在更长的时间内显著地调制石墨烯沟道。利用石墨烯的载流子捕获效应和双极性特性，可以在每个响应模式中模拟基本的增强和抑制行为。然后，利用电容门控效应，可以实现兴奋性和抑制性反应模式之间的重构。所有的突触响应仅取决于通过背栅电极输入的信号，这与具有附加调制端子的先前动态设备明显不同。这种重构特性使人工突触能够在未来的人工智能系统中模拟一些复杂的生物行为，例如在不同条件下对不同外部刺激的可调节感知。
Reconfigurable artificial synapse with synaptic responses modulated between excitatory and inhibitory modes is critical for building artificial intelligence systems. However, it is still a challenge to realize such reconfigurability with a simple single‐gated transistor. Here, hydrogen‐rich silicon nitride film is employed as the gate dielectric to construct a single‐gate controlled graphene‐based artificial synapse to realize the reconfigurable synaptic responses. In this dielectric, both traps and movable hydrogen ions are introduced to induce the carrier trapping effect and the capacitive gating effect, respectively. Comparatively, the capacitive gating effect needs stronger electrical fields excitation and can significantly modulate the graphene channel in a longer time. Utilizing the carrier trapping effect and the ambipolar property of graphene, the fundamental potentiation and depression behaviors can be emulated in each response mode. Then, utilizing the capacitive gating effect, the reconfiguration between excitatory and inhibitory response modes can be achieved. All synaptic responses only depend on the signals inputted through the back‐gate electrode, which is distinctively different from previous dynamic devices with additional modulating terminals. Such reconfiguration feature provides the artificial synapse the ability to emulate some complicated biological behaviors in future artificial intelligence systems, such as the adjustable perception of different external stimuli under different conditions.