International Conference on Solid State Devices and Materials

Metal halide perovskites have drawn, in recent years, great attention for neuromorphic electronic devices. However, the toxicity of lead as well as the variabilities and energy consumption of devices, still pose great challenges for the further progress of neuromorphic computing and memory applications. Here, we prepared a two-dimensional (2D) metal halide of composition BA2Pb0.5Sn0.5I4 that exhibits outstanding resistive switching memory performance after cesium carbonate (Cs2CO3) surface treatment. In particular, the device exhibited excellent switching characteristics (endurance of 5×105cycles, ON/OFF ratio ~105) and achieves 90.1% accuracy in pattern recognition (MNIST). More importantly, we proposed a novel energy-efficient content addressable memory (CAM) architecture building on perovskite memristive devices for neuromorphic applications, called nCAM, which has a minimum energy consumption of around 0.025 fJ/bit/cell. A mechanism involving defects induced charge trapping and de-trapping is proposed to explain the resistive switching behavior of the memristive device.