In the past few decades, the application of electronics and computers has dramatically changed the way that we learn, work, and live. Advances in semiconductor technology have enabled computers to perform increasingly well. The reduction in chip size has led to lower preparation costs, faster computer computing rates, and less power consumption. More than 50 years have passed since the development of Moore's law, and the characteristic length of silicon-based transistors is getting closer to the physical limit. On the other hand, conventional computer systems based on the von Neumann Architecture have additional latency and power consumption due to the separation of the computation and storage centers, leading to the emergence of the von Neumann Bottleneck, which limits the further improvement of the computational speed. In this background, neuromorphic computing inspired by the human brain has attracted a lot of attention in recent years. This integrated computation and storage center architecture can provide highly parallel, fast, efficient, and low-energy computing and storage to overcome the von Neumann bottleneck. Unlike the current mainstream neuromorphic computing that relies on software algorithms, memristors can mimic the synaptic function of biological nerves and thus build electronic devices that implement neuromorphic computing at the physical level. Nowadays, computer systems on silicon-based chip have great limitations and face new challenges for further development. The memristor-based neuromorphic computing system provides an alternative way to realize the integration of memory and computation, bio-inspired parallel computing and efficient reconfigurable memory computer system under the development needs of big data and internet of everything. To date, several materials have been applied to the development of memristors, such as oxides, two-dimensional materials, organic materials. However, these materials have certain limitations, such as complicated preparation processes, poor stability and reliability, and insufficient bionic properties. While perovskites have excellent electronic properties, which is widely used in light-emitting diodes, solar cells, and other fields. Perovskites are suitable for the preparation of memristive devices because of the advantages of high defect tolerance, high carrier mobility, and electron capture behavior. In recent years, perovskite-based artificial synapses have also been gradually developed, but they still have the problems of lack of stability and performance. Herein, we reported a highly stable memristor prepared with cesium lead bromide (CsPbBr3) perovskite and polyvinylpyrrolidone (PVP) by a one-step method with the structure of ITO/CsPbBr3:PVP/Au. We used CsPbBr3 perovskite as the resistive switching layer material and added PVP to the precursor solution to prepare memristors that can be used for neuromorphic computing. After different annealing temperatures, CsPbBr3 crystals with different sizes were formed, and the device annealed at 250 degrees C had the most excellent performance. The ITO/CsPbBr3:PVP/Au memristor device still maintains excellent performance when exposed to air for 9 d, exhibiting excellent stability. Meanwhile, our device successfully simulates biological synaptic behavior, including synaptic plasticity, long-term potentiation (LTP) and short-term potentiation (STP), amplitude-dependent plasticity, paired-pulse facilitation (PPF), short-term and long-term memory. The excellent performance means that our devices can be widely used in neuromorphic computing systems, which is a step forward for the development of next-generation computer systems.
