Transparent ferroelectric 0.8Bi0.5Na0.5TiO3-0.2SrTiO3 resistors for neural synapse sensors and image recognition

Abstract

Ferroelectric memristors possess advantages such as low operating voltage, high storage density and sub-micron-scale multi-state storage, making them widely used in fields such as information storage and mimicking biological synaptic functions. Here, the ferroelectric resistive switching (RS) devices with an Ag/(1-x)Bi 0.5 Na 0.5 TiO 3 -xSrTiO 3 /FTO (x = 0.1-0.3) structure are fabricated by the sol-gel method. The structure, memory resistance characteristics, RS mechanism and neural state application simulation of the memristors have been thoroughly studied. The optimal memristor characteristics with 10 3 storage window, 300 resistive switching cycles and 2 × 10 3 s retention time is obtained in the Ag/0.8Bi 0.5 Na 0.5 TiO 3 -0.2SrTiO 3 /FTO (BNT-ST) sample. The RS mechanism of the BNT-ST sample is mainly attributed to the combined effect of ferroelectric polarization and conductive filaments. Moreover, the BNT-ST memristor possesses outstanding synaptic functions, including excitatory/inhibitory postsynaptic current (EPSC/IPSC), paired-pulse facilitation/depression (PPF/PPD), spike-timing-dependent plasticity (STDP) and long-term potentiation/depression (LTP/LTD), as well as the characteristics of nociceptors and pain sensitization behavior. It also simulated image memory related to human brain learning-forgetting and the variations in body water levels across distinct operational states. Furthermore, through an artificial neural network model, the BNT-ST memristor achieved a recognition rate of 93.45% on the MNIST handwritten digit dataset, demonstrating excellent accuracy.