부산대학교 도서관

In this work, we investigate the performance of two-dimensional graphene oxide (GO)-based memristor and explore their advantages over the most popular metal oxide (TaOx)-based memristor for non-volatile memory and neuromorphic computing applications. The performance analysis of memristors is done using a self-developed numerical modeling framework, based on self-consistent solutions of the continuity, Joule's heating, and current continuity equations. GO-based memristor has demonstrated excellent switching performance with significantly lower sneak current (5.41 µA), higher ON/OFF resistance $(R_{on}/R_{off}$ = 200), higher read window (11.6), and higher non-linearity (106.4) than that for TaOx-based memristor. Further, a higher ON/OFF resistance ratio in a GO-based memristor promises a larger crossbar array size (2 6 ) over TaOx-based memristor counterpart with a minimum readout margin. Over GO-based memristor, the TaOx-based memristor exhibits linear and symmetrical conductance modulation with an identical applied pulse train, which makes them a more promising candidate for neuromorphic computing with probability of higher reading accuracy.