Multiscale Modeling of Metal/Oxide/Metal Conductive Bridging Random Access Memory Cells: from Ab Initio to Finite Element Calculations

We present a multiscale simulation framework to compute the current vs. voltage (I-V ) characteristics of metal/oxide/metal structures building the core of conductive bridging random access memory (CBRAM) cells and to shed light on their resistance switching properties. The approach relies on a finite element model whose input material parameters are extracted either from ab initio or from machine-learned empirical calculations. The applied techniques range from molecular dynamics and nudged elastic band to electronic and thermal quantum transport. Such an approach drastically reduces the number of fitting parameters needed and makes the resulting modeling environment more accurate than traditional ones. The developed computational framework is then applied to the investigation of an Ag/a-SiO2/Pt CBRAM, reproducing experimental data very well. Moreover, the relevance of Joule heating is assessed by considering various cell geometries. It is found that self-heating manifests itself in devices with thin conductive filaments with few-nanometer diameters and at current concentrations in the 10s-microampere range. With the proposed methodology it is now possible to explore the potential of not-yet fabricated memory cells and to reliably optimize their design.

💡 Research Summary

The paper introduces a comprehensive multiscale simulation framework designed to predict the current‑voltage (I‑V) characteristics of metal/oxide/metal structures that form the core of conductive‑bridging random‑access memory (CBRAM) cells. Recognizing the limitations of existing approaches—namely, the heavy reliance on experimentally fitted parameters and the difficulty of simultaneously capturing electronic, ionic, and thermal phenomena—the authors combine finite‑element method (FEM) modeling with material parameters derived directly from first‑principles calculations and machine‑learned empirical models.

The workflow proceeds in several tightly coupled stages. First, density‑functional theory (DFT) and ab‑initio molecular dynamics (AIMD) are employed to compute the electronic structure of Ag atoms, the oxidation‑reduction potentials at Ag/a‑SiO₂ interfaces, and the diffusion barriers for Ag⁺ ions within the amorphous SiO₂ matrix. Nudged elastic band (NEB) calculations provide accurate transition‑state energies for ion migration pathways. These atomistic results feed into a Butler‑Volmer description of interfacial redox reactions, supplying reaction rate constants without any empirical fitting.

Ion transport across the solid electrolyte is modeled with the Nernst‑Planck equation, while electron transport through the growing metallic filament is treated using a tunneling model based on a spatially varying potential (WKB approximation). The FEM framework solves the coupled electrostatic, ionic drift‑diffusion, and heat‑conduction equations in a quasi‑2D axisymmetric geometry that captures the essential features of an Ag/a‑SiO₂/Pt cell, including a truncated‑cone seed filament attached to the inert Pt electrode.

Thermal effects are incorporated by solving the heat equation with material‑specific electronic and phononic conductivities derived from the same first‑principles data. The simulations reveal that when the filament diameter shrinks below ~5 nm, even modest currents in the 10 µA range generate significant Joule heating, raising the filament temperature by tens of kelvin. This self‑heating accelerates filament dissolution during the RESET operation, directly influencing the hysteresis shape of the I‑V curve. Conversely, thicker filaments dissipate heat efficiently, making thermal effects negligible.

Applying the framework to an experimentally characterized Ag/a‑SiO₂/Pt device, the authors achieve excellent agreement with measured I‑V curves, reproducing the SET voltage, compliance current behavior, and resistance ratios without any post‑hoc parameter tuning. The only inputs required are the device geometry and the applied voltage sweep; all material‑specific quantities are obtained from the multiscale calculations.

The study demonstrates that integrating atomistic insights into continuum FEM models dramatically reduces the number of free parameters—typically from over ten in conventional kinetic Monte Carlo or continuum approaches—to essentially zero. This enables reliable prediction of novel material combinations and device architectures before fabrication. The authors suggest future extensions to fully three‑dimensional geometries, alternative solid electrolytes (e.g., HfO₂, Ta₂O₅), and coupling with circuit‑level simulators for system‑wide design optimization. Overall, the work provides a powerful, physics‑based tool for advancing CBRAM technology toward ultra‑low‑power, high‑density memory applications.