Theoretical Simulation of the Conductive Filament in the Resistive Switching Memory

Abstract

In this thesis, resistive switching properties of resistive memory (RRAM) is extensively modeled and investigated. The RRAM is a promising candidate to replace the Flash ram nowadays because of its performance, endurance and cost. The resistive switching behavior in RRAM is explored by solving a set of equations that represent the important elements of the device physics. The non-equilibrium Green’s function method is used to model charge transport for system size scale at atomistic scale. The Cu dopant in the α-alumina are calculated using ab initio density functional theory to give the Hamiltonian. The current saturation and negative differential resistance can be found in the system. The current saturation occurs due to the narrow bandwidth on the transmission coefficient. The negative differential resistance is analyzed based on the match and mismatch of local density of states. The current difference for different filament configuration makes multi-level memory cell possible. Prior work has had difficulty in modeling the processes of set, reset and retention in a comprehensive manner. In this thesis, a more unified picture of these processes using Kinetic Monte Carlo simulations is provided. The Kinetic Monte Carlo simulations are based on atom/vacancy motion to model filament growth and dissolution in an oxide, in multiple scenarios. The kinetic Monte Carlo simulations are based on the processes of formation, diffusion and recombination. The classical heat and Poisson equations are solved to give the local electric field and temperature. In the surface generation model, oxygen vacancy is generated in Hafnia near the interface, with the corresponding oxygen atom entering the metal electrode. These oxygen atoms form a thin insulating oxide layer at the Hafnia-active electrode interface. This interfacial layer can change the direction of the electric field and help to thicken the filament. This thickening of the conducting filament is captured by my model and it explains the trend of resistance decrease with an increase in compliance current found in some experiments. In RRAM with negative thermophoresis materials, and two inert electrodes, the forming and unipolar reset processes are simulated. The negative thermophoresis precludes the traditional mechanism of RESET where the vacancy moves away from the filament as a result of high filament temperature. Simulations reveal a new RESET mechanism which involves the diffusion of oxygen interstitials to break the vacancy-based filament in a region close to the top electrode. With the mechanism of oxygen interstitial diffusion, the filament can be RESET at a higher temperature than in SET.