Abstract: Journal of Material Sciences & Applied Engineering

Abstract:
In order to overcome the memory bottleneck encountered in the digital computing von-Neumann architecture, non-volatile memory arrays are developed as synaspes in neural networks, and to build an energy-efficient AI accelerator surpassing GPUs. This paper describes the material science and engiofneering requested for producing energy-efficient non-volatile memories such as resistive random access memory (ReRAM), phase-change memories (PCRAM), ferroelectric memories (FeRAM), magnetic memories (MRAM) and electro-chemical Random-Access Memory (ECRAM).