Multiphysics Computation for Resistive Random Access Memories With Different Metal Oxides

In this article, we use an in-house finite element method (FEM)-based parallel computing simulator to model multiphysics processes of resistive random access memories (RRAMs). These RRAMs are based on six different metal oxides, including TiO
_x
, NiO
_x
, HfO
_x
, WO
_x
, ZrO
_x
, and TaO
_x
. In a single RRAM cell, the resistance ratio, reset voltage, reset current, reset power, and temperature are numerically compared. In the RRAM array, the resistance ratio, the minimum feature size (FS) to avoid bit loss, temperature, and oxygen vacancy density are obtained and discussed. The advantages of the six metal oxides-based RRAMs are analyzed based on their physical parameters, including diffusivity, electrical conductivity, and thermal conductivity. Numerical results reveal that thermal crosstalk effects are more severe when the FSs decrease, even the inactive cell can be transferred from a low-resistance state (LRS) into a high-resistance state (HRS). A crossbar array with a diamond heat sink is proposed to suppress the thermal crosstalk effect.