Abstract

Associative memory, considered as the ability to link unrelated objects, is the most appealing behaviour found in both biological and artificial neural networks. Here we for the first time proposed a three terminal phase-change based nanoscale device where associative memory can be theoretically achieved according to the developed electro-thermal model. Compared with the conventional neural networks circuit, the designed device allows for an exemption from a large number of circuit elements, and thus exhibits an exciting potential for ultra-high integration density and super-fast response speed. The simulation results opens up a novel route to imitate the complex mechanism of the biological neural networks, thus rendering phase-change materials a promising contender for the key components in next-generation neuromorphic system.

Keywords: Associative memory, model, neural, phase-change.

Graphical Abstract