Computational Study on Filament Growth Dynamics in Microstructure-Controlled Storage Media of Resistive Switching Memories

Abstract

The filament growth processes, crucial to the performance of nanodevices like resistive switching memories, have been widely investigated to realize the device optimization. With the combination of kinetic Monte Carlo (KMC) simulations and the restrictive percolation model, three different growth modes in electrochemical metallization (ECM) cells were dynamically reproduced, and an important parameter, the relative nucleation distance, was theoretically defined to measure different growth modes quantitatively; hence their transition can be well described. In our KMC simulations, the inhomogeneity of storage medium is realized through introducing evolutionary void versus non-void sites within it to mimic the real nucleation during filament growth. Finally, the renormalization group method was used in the percolation model to analytically illustrate void-concentration-dependent growth mode transition, fitting KMC simulation results quite well. Our study found that the nanostructure of the medium can dominate the filament growth dynamics, as the simulation images as well as the analytical results are consistent with experiments results. Our study spotlights a vital and intrinsic factor, void concentration (relative to defects, grains, or nanopores) of a storage medium, in inducing filament growth mode transition within ECM cells. This theoretically proves a mechanism to tune performance of ECM systems that controlling microstructures of the storage media can dominate the filament growth dynamics, indicating an accessible strategy, nanostructure processing, for device optimization of ECM memristors.

Kinetic Monte Carlo simulations on electroforming in nanomanipulated conductive bridge random access memory devices

“…Based on our previously developed KMC modeling, 10 which successfully reproduced three filament growth modes in the ECM cells observed in experiments, some modifications were made to mimic filament growth processes in ECM cells by insertion of an ion-blocking layer (IBL), a nanopore-containing graphene monolayer, between the anode and the dielectric layer as shown in Fig. 1a to illustrate the cooperative effect of the ion-blocking layer and the inhomogeneity of the dielectric layer: (1) the IBL was treated as 0.2 nm-thick inactive grids, and it keeps unchanged in the whole KMC simulations.…”

“…In our previous kinetic Monte Carlo (KMC) simulations, we found that the filament growth dynamics in active metal electrode/dielectric/inert metal electrode ECM cells can be well tuned by only changing the microstructures of dielectric layers with sufficient consideration of their inhomogeneity. 10 This nanomanipulation was not considered in the previous KMC simulations on CF growth in ECM cells, by using either two-dimensional (2D) grids 11,12 or three-dimensional (3D) grids. 13,14 We believe that the more complex nanomanipulation here by both introducing a nanopore-containing IBL and adopting the inhomogeneous RS layer in ECM cells can affect filament growth further to finally fulfill the memristor performance enhancement.…”

“…13,14 We believe that the more complex nanomanipulation here by both introducing a nanopore-containing IBL and adopting the inhomogeneous RS layer in ECM cells can affect filament growth further to finally fulfill the memristor performance enhancement. Therefore, the KMC modeling developed previously 10 was employed here to first make some modifications and then carry out simulations on electroforming ( i.e. , the initial formation of filaments connecting both electrodes) in nanomanipulated ECM cells to finally illustrate the influence of our proposed nanomanipulation on filament growth and device performance for ECM systems.…”

“…Pt). In order to express the inhomogeneity of the dielectric layer, the innovative treatments learned from our previous KMC simulations 10 were also adopted here. The dielectric layer can be distinguished as void (white) versus nonvoid (light grey) sites in Fig.…”

“…Just increasing the void concentration in our previous KMC simulations induces the filament growth modes to change from reversed growth mode, cluster aggregation mode to cathode growth mode. 10 This growth mode transition corresponds to the initial nucleation sites of growing filaments varying from near the anode, in the middle region of the dielectric layer, to near the cathode. We are really curious about how microstructure tuning combined with nanopore size manipulation affects filament growth in the studied ECM cells.…”