A morphologically self-consistent phase field model for the computational study of memristive thin film <i>current</i> – <i>voltage</i> hysteresis

Abstract

A multiphysics phase field model is used for the computational study of memristive thin film current–voltage hysteresis and conducting filament morphology. In contrast to previous computational methods, no requirements are made on conducting filament geometry. Our method correctly predicts conducting filaments evolve on thermodynamic paths that are energetically favored due to stochastic structural and chemical variations naturally occurring at the atomic level, due to both latent and intentional fabrication imperfections, precluding the need for direct local density of states computation. Less than ten physical parameters are required, in contrast to continuum, molecular dynamics, and first-principle methods that each require substantially more parameters. Computational efficiency and scalability of our model are a significant advantage, enabling practical wafer-scale mapping, uniformity, and endurance analysis and optimization.