Flexible InGaZnO (IGZO) thin-film transistor (TFT) with Al2O3 gate insulator (GI) which is deposited by low temperature (LT) atomic layer deposition (ALD) is proposed and its synaptic behavior and mechanical stability are demonstrated on a polyethylene terephthalate substrate. The change of threshold voltage under bending test is attributed to the generation of ionized oxygen vacancy resulting from the oxygen bond-breaking. In addition, the synaptic behavior is clearly observed and the convolutional neural network-based MNIST recognition rate of 87.2 % after 60,000 training is demonstrated by using the proposed IGZO TFTs. Stable synaptic behavior can be explained by the potentiation/depression pulse-dependent movement of hydrogens which the Al2O3 GI contains during LT ALD. Furthermore, it is found that the synaptic weight can be controlled and optimized by changing the thickness of Al2O3 GI.
展开▼