A memristor based on bismuth oxyiodide film, its preparation method and application

Abstract

The invention provides a memristor based on bismuth oxyiodide film, its preparation method and application. The memristor in the present invention includes a bottom electrode, on which a resistive switch function layer made of bismuth oxyiodide thin film is formed, and a top electrode is formed on the resistive switch function layer. The present invention uses a bismuth oxyiodide thin film as the resistive switching functional layer of the memristor. Bismuth oxyiodide can increase the number of oxygen vacancies in the device, and these oxygen vacancies can be used as a stable conductive channel between the top electrode and the bottom electrode, thereby making the device There will be no abrupt changes in the resistance of the resistance. Because the memristor in the present invention adopts bismuth oxyiodide film as the resistive function layer of the device, it realizes the simulation of the biological synapse function of the device, including short-term plasticity, long-term plasticity, double-pulse facilitation, and peak-time plasticity , learning-forgetting-relearning, etc., which means that the powerful computing power of the human brain and high-efficiency artificial intelligence chips have an optimistic development prospect.

Description

technical field [0001] The invention relates to a memory, in particular to a memristor based on a bismuth oxyiodide thin film, its preparation method and application. Background technique [0002] The ability of the human brain to perform highly parallel information processing at very low power has been shown to be overall superior and efficient compared to traditional von Neumann architectures. As a result, scientists around the world are working hard to develop the building blocks of electronic brains. The basic element in the human brain is the synapse, which is huge in number (up to 10 15 individual), and is a key factor in learning and recognition. Traditional neuromorphic computing hardware with a von Neumann architecture consumes a lot of power and is less efficient overall. Recently, IBM developed a hardware method to implement the function of simulating millions of synapses using static random access memory (SRAM). However, since each SRAM cell contains six tran...

AI Technical Summary

Problems solved by technology

Although the emergence of memristors has great appeal for neuromorphic applications, the resistance of general RRAMs will suddenly change from high to low during the "SET process", and there is a sudden "SET" characteristic that limits the number of conductance states.

It makes it difficult for RRAMs to mimic the learning behavior of synapses, thus requiring a device with a gradual change in resistance

Images