Emulating Bilingual Synaptic Response Using a Junction-Based Artificial Synaptic Device.
He TianXi CaoYujun XieXiaodong YanAndrew KostelecDon DiMarzioCheng ChangLi-Dong ZhaoWei WuJesse TiceJudy J ChaJing GuoHan WangPublished in: ACS nano (2017)
Excitatory and inhibitory postsynaptic potentials are the two fundamental categories of synaptic responses underlying the diverse functionalities of the mammalian nervous system. Recent advances in neuroscience have revealed the co-release of both glutamate and GABA neurotransmitters from a single axon terminal in neurons at the ventral tegmental area that can result in the reconfiguration of the postsynaptic potentials between excitatory and inhibitory effects. The ability to mimic such features of the biological synapses in semiconductor devices, which is lacking in the conventional field effect transistor-type and memristor-type artificial synaptic devices, can enhance the functionalities and versatility of neuromorphic electronic systems in performing tasks such as image recognition, learning, and cognition. Here, we demonstrate an artificial synaptic device concept, an ambipolar junction synaptic devices, which utilizes the tunable electronic properties of the heterojunction between two layered semiconductor materials black phosphorus and tin selenide to mimic the different states of the synaptic connection and, hence, realize the dynamic reconfigurability between excitatory and inhibitory postsynaptic effects. The resulting device relies only on the electrical biases at either the presynaptic or the postsynaptic terminal to facilitate such dynamic reconfigurability. It is distinctively different from the conventional heterosynaptic device in terms of both its operational characteristics and biological equivalence. Key properties of the synapses such as potentiation and depression and spike-timing-dependent plasticity are mimicked in the device for both the excitatory and inhibitory response modes. The device offers reconfiguration properties with the potential to enable useful functionalities in hardware-based artificial neural network.