Dopant selection rules for extrinsic tunability of HfOx RRAM characteristics: A systematic study

Recently, resistive random access memory (RRAM) has been explored as a promising technology for applications including on-chip memories, 3D-FPGA and neuromorphic computing systems. For systems with RRAM integration, it is beneficial to have better control of device characteristics to provide large design space and advanced functionalities. For example, in order to mimic the human nervous system with a complex hierarchy, it is desirable to develop multiple, well-separated RRAM characteristics on a single chip. However, intrinsic tuning of device parameters is subject to process restrictions and performance trade-off. In this paper, we present the first systematic approach to understand and utilize the doping effects for tuning RRAM characteristics. HfOx-based RRAM devices with 5 different dopants are fabricated to demonstrate the substantial tunability of key parameters including forming voltage, SET voltage and ON/OFF ratio. To explain the experimental observations, detailed ab initio calculations are performed to model the dopant-vacancy and dopant-filament interactions. The effects of 12 candidate dopants on forming and switching processes are analyzed to provide universal guidelines for dopant selection. Based on the improved tunability, novel design methodologies are further proposed to enable high-performance and multi-level RRAM devices.