Linearity Improvement of TiOx‐Based Flexible Memristor Synapses Even Under Bending

Abstract

The flexible memristor synapses based on binary metal oxides have significant potential in building artificial neural networks to meet people's demands. To obtain flexible memristor synapses with high linearity of conductance update, herein, the effect of Al doping and SiOx layer adding on the flexible TiOx‐based memristor synapse even under bending is studied. The results show that the doping Al and adding SiOx layer improves the linearity of conductance update by 3 times. The bending test results of the SiOx/Al:TiOx‐based memristor synapse show that the linearity decreases slightly with the decrease of the bending radius. When the bending radius is reduced to 5 mm, the total nonlinearity (NL) value of the device is about 0.42. It is still much smaller than the total NL value of TiOx‐based memristor synapse in flat state (about 0.76). The test results of conducting atomic force microscopy show that the improvement of linearity is due to the optimization of conductive filaments. The results provide a feasible program for improving linearity of TiOx‐based flexible memristor synapse even under bending, which can support for the development of high‐performance binary metal oxide flexible memristor synapse.