A floating gate memristor

Abstract

The invention relates to a floating gate memristor, the basic structure of which comprises a front electrode, a front dielectric layer, a front floating gate layer, a nano battery anode, a nano battery electrolyte, a nano battery cathode, a rear floating gate layer, a rear dielectric layer, and a rear floating gate layer. Electrodes; where the front electrode, front dielectric layer, front floating gate layer and nanobattery anode simulate the presynaptic membrane, use electron tunneling and field effects to convert electronic signals into ion signals, and nanobattery electrolytes act as ion channels to simulate synapses The gap, nanobattery cathode, rear floating gate, rear dielectric layer, and rear electrode mimic the postsynaptic membrane that converts ionic signals into electrical ones. The floating gate memristor of the present invention is stable in reading and writing, good in controllability, simple in structure, compatible with CMOS, easy to integrate, capable of large-scale production and commercialization, and can promote the development of neuromorphic computing and brain-inspired computing.

Description

technical field [0001] The invention relates to a floating gate memristor, which relates to the field of neuromorphic computing based on the memristor, in particular to an ionic synaptic memristor based on a nano battery and a transistor floating gate. Background technique [0002] Neural network-based intelligent learning systems have made remarkable achievements, such as Google DeepMind's Go artificial intelligence AlphaGo, Carnegie Mellon University's Texas Hold'em program, and Stanford University's skin cancer recognition program. Artificial intelligence systems based on neural networks have been involved in all aspects of finance, medical care, transportation and the environment. The basic building blocks of neural networks are synapses and neurons. For a long time, people have been trying to develop electronic synapses. In 1971, Professor Cai Shaotang, Department of Electronic Engineering, University of California, Berkeley first proposed the memristor model. In 2008, ...

AI Technical Summary

Problems solved by technology

However, it is very difficult to use a two-terminal device to read the change in conductance caused by the change of ion distribution, because electrons basically cannot pass through the dielectric of the nanobattery, and the migration history of ions is only expressed as the ion concentration in the positive and negative materials. It is necessary to introduce a new structure to record and read the ion migration history in the nanobattery to simulate the synaptic behavior

Images