Li-ion-gated Synaptic Transistors Based On Graphdiyne/MoS₂ Heterostructure

In the era of the“big data”,massive unstructured data including images,sound,video and text need to be handled efficiently,which is becoming a severe challenge for conventional von Neumann computers with physically separate processing and memory units.Human brain-inspired neuromorphic computing,which can parallelly process unstructured information in an energy-efficiency way,has attracted great attention in recent years.Various artificial synapses,such as two-terminal memristors and multiterminal neuromorphic transistors,have been proposed to construct hardware artificial neural networks(ANNs)for neuromorphic computing.Especially,ion-gated synaptic transistors using electrolyte to modulate the conductance state of channel,have demonstrated excellent linearity and symmetricity for weight update,as well as extremely low switching voltages and long retention time,enabling a near-ideal recognition accuracy with ultra-low energy consumption.However,in these devices,the ions(e.g.Li⁺)are directly intercalated into the conductive channels.The frequent intercalation/deintercalation of ions between the electrolyte and the channel,which is responsible for the long-term potentiation(LTP)and long-term depression(LTD),will inevitably destroy the lattice structure of the channels,leading to an irreversible performance degradation after many cycles.Graphdiyne(GDY),as a two-dimensional carbon allotrope consisting of sp-and sp²-hybridized carbon atoms,exhibits extraordinary features for the storage and diffusion of Li ions.For instance,the evenly distributed sp carbon atoms provide sufficient space for the storage of metal atoms,enabling a Li storage ratio of 1:3(i.e.Li C₃),which is twice larger than the commonly used graphite(Li C₆).Moreover,the large pores(0.542 nm)and wide layer distance(0.365 nm)of GDY enable both the in-plane and out-of-plane diffusion of Li ions with a quite low barrier(0.17–0.84 e V).Therefore,GDY is considered to be a promising candidate for an anode material in energy-storage devices,such as Li-ion batteries and supercapacitors.However,no effort has been payed to demonstrate the advantage of GDY as a Li-ion storage material in ion-gated synaptic transistors.In this work,an all-solid Li-ion-gated synaptic transistor based on GDY/MoS₂vertical heterostructure was proposed,where MoS₂ acted as the conductive channel and GDY was used as a Li-ion storage layer to modulate the conductance of MoS₂.A volatile change of channel conductance is induced by the accumulated Li ions at GDY surface,while a non-volatile one is associated to the intercalated Li ions in GDY interlayers,which correspond to the short-term and long-term plasticity,respectively.Spiking-timing-dependent plasticity(STDP),such as Hebbian and anti-Hebbian learning rules,as well as Pavlovian learning have been demonstrated by the device.The excellent flexibility of GDY/MoS₂ heterostructure and polymer electrolyte endows the device with robust bending stability and reliability for wearable applications.Since the amount of the intercalated Li ions in GDY can be precisely controlled by the applied current pulses,the conductance-update trajectory features high linearity and low noise under a current-pulsing mode,which facilitates the achievement of high classification accuracy and ultralow energy consumption in neural network simulations.Moreover,logic-in-memory functions have been demonstrated by the synaptic transistor,where the logic results can be in situ stored in the device during logic operations...