Melting-Free Phase-Change Memory for Associative Learning

Abstract

RRAMs as artificial synapses have shown their prospects in associative learning tasks. However, the circuit design based on input signal overlapping is contradictory with the characteristics of conventional phase-change memory (PCM), resulting in limited scalability of RRAM-based associative learning at network level. Here, we present a concise circuit utilizing unconventional PCM and realize the complete associative learning function. By employing MoTe <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf> non-melting phase transition of the fabricated PCM, corresponding circuit and waveform design can be matched, thus provides a promising solution for the array implementation of associative learning. To demonstrate the associative learning ability of this design, a dynamic model of electrical characteristics of the PCM is built, and complete function of the Pavlov’s dog is simulated with HSpice.