Multiscale Modeling of Conductive Bridging Random Access Memory

Abstract

Conductive bridging random access memory (CBRAM) cells are typically realized as metal/oxide/metal structures. The technology relies on nonvolatile resistive switching between a high- (HRS) and a low-resistance state (LRS), which can be used as the logical 0 and 1 in information storage. The core of the switching operation is the reversible growth and rupture of a metallic filament through the oxide layer upon application of an external voltage. To drive the technology towards its limits, the device sizes have been scaled down such that only a few atoms are involved in the HRS/LRS switching process enabling high-speed operation at low power consumption. However, these appealing characteristics are often counterbalanced by poor retention and reliability, capabilities that a deeper understanding of the underlying physics and chemistry governing the switching mechanism could help improve. The goal of this thesis is to design a versatile simulation platform to evaluate the operation performance of novel CBRAM cells, provide insight into their switching characteristics, and support the formulation of guidelines for next-generation CBRAM devices. To satisfy the vast amount of different material and operational aspects and to enable a comparison to experimental data, a multiscale approach is chosen. It involves ab initio calculations based on density-functional theory, force-field-based techniques, and continuum modeling relying on the finite element method (FEM). In the first part of this thesis, a combined force-field/ab initio scheme is presented to study the switching behavior of Ag/a-SiO2/Ag CBRAM at the atomic level. Initially, the creation of realistic atomic CBRAM structures is thoroughly discussed, emphasizing the necessity of high-quality oxide layers for the evaluation of the resistance states. The dynamics of the switching process is then investigated with classical molecular dynamics simulations using a machine-learned moment tensor potential (MTP), while the electrical properties are determined through ab initio quantum transport calculations. With a structural analysis of the oxide, we shed light on the switching mechanism, which takes advantage of preferred channels containing wide SiO2 rings through which Ag+ ions migrate. Also, it is demonstrated that moving only few atoms in such channels can change the resistance state by several orders of magnitude within a few picoseconds. The second part of the thesis is dedicated to a multiscale simulation framework that enables the computation of the "current vs. voltage" (I-V) characteristics of CBRAM and to shed light on self-heating effects. The approach relies on a FEM model whose input material parameters are extracted either from ab initio or from semiempirical force-field 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 that are typically needed by continuum FEM models 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 tens-microampere range. The proposed methodology enables the study of the switching characteristics of CBRAM cells from the atomic to the continuum level. It is worth to emphasize that the developed framework could be extended to most filamentary-type memory technologies relying on metal/oxide/metal stacks. This includes even not-yet-fabricated memory cells whose performance could be conveniently predicted and optimized.