Physics-based modeling of valence change mechanism memristor

Abstract

As conventional semiconductor devices encounter limitations, such as the memory wall and power wall, as the Moore’s Law comes to an end soon. A promising semiconductor device known as the “memristor” has emerged to overcome these limitations. The word "memristor" comes from two words which are "memory" and "resistor". Memristors have several key features including high density, low power consumption, excellent scalability, and fast operation speed, which making them highly desirable and feasible as a promising device for the future AI hardware and novel memory device. More importantly, the memristor array's neuromorphic property makes it analogous to a single synapse in the human brain. This neuromorphic potential positions memristors to play a significant role in neuromorphic integrated circuits. Nevertheless, memristors still has some challenges for further investigation and research, particularly in developing simulation models for these devices. Simulation methods serve as essential bridges, enabling the translation of continuous real-world physical phenomenon into the numerical realm. The numerical models in this dissertation enable the discretization of continuous equations, such as the Poisson and continuity equations of charge carriers, effectively rendering them interpretable by computers while ensuring an accurate fit with the experimental data. In this dissertation, we present a finite volume method simulation model implemented in C++ for the SrTiO3 VCM memristor. This research strives to enhance our understanding of memristors and the physical mechanism and simulation algorithm of it. Keywords: Memristor, VCM, Simulation model.