Ultraflexible and Energy‐Efficient Artificial Synapses Based on Molecular/Atomic Layer Deposited Organic–Inorganic Hybrid Thin Films
Chang Liu,Lin Zhu,Jing‐Kai Weng,Yu‐Chen Li,Zhuo Chen,Jin Lei,Tao‐Qing Zi,Di Wu,Wei Fa,Shuang Chen,Bin Liang,Ai‐Dong Li
DOI: https://doi.org/10.1002/aelm.202200821
IF: 6.2
2022-12-11
Advanced Electronic Materials
Abstract:Molecular/atomic layer deposited ultraflexible organic–inorganic hybrid synaptic devices are developed and evaluated for energy‐efficient neural network applications. These devices exhibit exceptional flexibility and robustness with extremely low energy consumption. The simple bilayer‐architecture and manufacture technology compatible with semiconductor processing enable hybrid devices to have great prospects in future flexible neuromorphic computing and wearable devices. Flexible artificial synapses, a conjunctive product of brain‐inspired neuromorphic computing and wearable electronics, arouse enormous interest in highly connected and energy‐efficient neural networks. The organic–inorganic hybrid materials hold great potential in flexible devices due to versatile properties. Here, an organic–inorganic hybrid synaptic device consisting of 2 nm Al2O3 and 22 nm Al‐based hydroquinone (Al‐HQ) sandwiched between Pt and poly(3,4‐ethylenedioxythiophene):polystyrene sulfonate (PEDOT:PSS) electrodes is prepared on highly flexible cellophane by molecular/atomic layer deposition (MLD/ALD). The vertically integrated Pt/Al2O3/Al‐HQ/PEDOT:PSS device exhibits reliable resistive switching with an ON/OFF ratio greater than 103. Several important bio‐synaptic functions, such as long‐term potentiation, long‐term depression, paired‐pulse facilitation, and spike‐time‐dependent plasticity, are realized in this device with the extremely low energy consumption of ≈25.2 fJ per reset operation, which is ascribed to the unique electron trapping/detrapping and tunneling mechanism. Remarkably, the excellent flexibility and robustness of this hybrid synaptic device is confirmed under the minimum curvature radius of ≈0.7 mm after 104 bending cycles. A pattern recognition computation based on these hybrid synapse devices shows a 90.2% learning accuracy. This research paves a way for the MLD/ALD‐derived organic–inorganic‐hybrid‐based artificial synapse applications in flexible energy‐efficient neuron network systems toward system‐on‐plastics.
materials science, multidisciplinary,physics, applied,nanoscience & nanotechnology