Enhancing RRAM Device Performance: A Design of Experiments Approach

Oxide-based Resistive Switching Random Access Memory (RRAM) devices hold significant promise as candidates for next-generation memory technology and novel brain-inspired computing applications. Among these, TaO_x (1<x<2.5)-based RRAMs have achieved substantial attention due to their promising performance in terms of durability and low energy consumption. However, these devices still suffer from instability and endurance degradation issues which hinder their large-scale adoption. One critical parameter is the forming voltage, which significantly influences resistive switching activation, device energy efficiency, and overall RRAM reliability but high forming voltage can increase the parasitic capacitance discharge which accelerates the degradation in the higher number of consecutive ON/OFF cycles. Intrinsic properties of TaO_x switching films such as film stoichiometry and oxygen vacancy profile play a pivotal role in shaping RRAM device performance metrics. The stoichiometry engineering of oxide films, enabled by varying the deposition parameters of a sputtering system, offers substantial advantages in optimizing RRAM behavior. However, the multitude of controllable and uncontrollable experimental parameters and potential interdependencies render this task challenging. In this work, we study the optimal processing conditions for sputtered TaO_x films that will yield the desired device metrics compatible with the algorithm requirements of neuromorphic computing. We employ a 2-level design of experiments (DOE) approach to examine the interactive impact of oxygen content on TaO_x films by varying the oxygen pressure and deposition power. Furthermore, we demonstrate how this set of parameters impact critical device performance metrics such as forming and set/reset voltages, the type of resistive switching (bipolar or unipolar, among others), initial resistance, memory window, and the endurance variation coefficient.