Modelling the reliability of valence change mechanism devices

The steadily increasing demands on modern memory technologies have sparked interest in resistive memory solutions, particularly valence change mechanism-based (VCM) memory cells. VCM is a promising subcategory of redox-based resistive switching random access memory (ReRAM) and offers properties such as high scalability, low power consumption, and non-volatility, making it suitable for high-density applications. However, the widespread adoption of VCM ReRAM faces significant challenges related to reliability, particularly regarding variability, data retention, and endurance. These issues are primarily due to the stochastic dynamics of oxygen vacancies within the metal oxide layer, leading to fluctuations in resistance states, switching voltages, and read currents.This dissertation addresses these challenges through the development of two Kinetic Monte Carlo (KMC) models: a one-dimensional (1D) model focusing on the study of endurance and a three-dimensional (3D) model that addresses variability and retention. The 1D KMC model enables the simulation of numerous switching cycles with high computational efficiency and accurately captures the movement of oxygen vacancies in a simplified filament structure. This allows for a detailed analysis of ageing mechanisms such as the “stuck-bit” phenomenon, which is caused by interactions between cell and peripheral resistances. The 3D KMC model extends the analysis by simulating the spatially resolved dynamics of oxygen vacancies, enabling a differentiated investigation of variability and retention under realistic operating conditions. Both models show a strong agreement with experimental observations, precisely reproducing the measured resistance distributions, retention behaviour, and read noise characteristics. This agreement confirms the predictive power of the models for reliability analysis of VCM ReRAM.The combined modelling approach allows for a comprehensive statistical assessment of VCM ReRAM reliability across numerous cells and cycles. Based on these results, targeted optimisations, such as programming adjustments and doping strategies, can be proposed to significantly improve the reliability and commercial viability of VCM ReRAM. Future work could aim to further refine these models by incorporating additional physical mechanisms and expanding the focus to surface effects, thereby enhancing the suitability of VCM ReRAM for large-scale applications.