A Physical Model of Filamentary TaOx/HfO2 ReRAM Devices

The advancement of information technology has increased exponentially the amount of data that is gathered and that requires processing, and the latter is predicted to become one of the most energy demanding operations in the very near future. Current data processing relies on Complementary Metal-Oxide-Semiconductor (CMOS) technology, but even the most energy efficient implementations suffer from an intrinsic power and speed inefficiency known as the von Neumann bottleneck, caused by the data transfer between memory and processing unit. It is therefore necessary to transition to new computing architectures and paradigms such as bio-inspired computing. The key component of these new solutions is the memristor, an element with multiple resistance levels that can be modified in a non-volatile manner. One of the most promising memristive technologies is filamentary valence-change mechanism (VCM) Resistive Random Access Memory (ReRAM), due to its scalability, ease of integration with current CMOS technologies and resistance range. The working principle of these devices is the creation of a conductive filament through an oxide layer sandwiched between two electrodes, followed by the modulation of the filament size to change the resistance. The main challenges in the development of ReRAM technologies are symmetry, linearity and stochasticity. A possible solution to these problems is the addition of a Conductive Metal Oxide (CMO) layer in the stack, though the exact role played by the CMO in the filament forming and the resistance switching is not yet fully understood. This master thesis work is focused on the modeling of a CMOS-compatible bilayer ReRAM composed of substochiometric tantalum oxide (TaOx) and hafnium oxide (HfO2) sandwiched between two titanium nitride (TiN) electrodes. DC characterization, TEM analysis and CTLM measurements have been previously performed on devices with different TaOx thickness and/or stochiometry and are used in this work as support for building the model. The selected simulation tool for this task is Ginestra®, a Kinetic Monte Carlo simulator developed by Applied Materials capable of emulating the behavior of multiple kinds of electronic devices. It allows to study how each parameter of the layers in the device stack affects the behaviour of the cell and extract the parameters from experimental data. Capacitance-voltage measurements were performed to identify the metallic nature of the substoichiometric TaOx and to build an effective model capable of capturing the effect of the CMO on both the forming and the set simulations. Different modeling approaches were needed to obtain a structure that could match the experimental data. The analysis of the effect of different parameters on pristine state devices allowed to identify possible improvements of the fabrication process that could be implemented to fabricate devices with better performance. The oxidation and reduction interplay between the metallic filament and the effective interface in this proposed model is demonstrated to be the phenomenon that governs the behavior of the devices by using bond breakage of the metal-oxide molecules, positive feedback Joule heating and recombination of oxygen ions and oxygen vacancies