Atomic view of chalcogenide-based resistance switching memories

Abstract

Resistive random access memories (ReRAM) form a new class of emerging non-volatile memory device foreseen to replace the current Flash technology due to their excellent scalability and low power consumption. ReRAM consist of a solid electrolyte such as chalcogenide glasses or transition metal oxides sandwiched between two metallic electrodes. The operation principle of ReRAM is based on the resistance change of the cell under an applied voltage and the corresponding mechanism often involves the electrochemical formation and dissolution of conductive metallic filaments embedded in the electrolyte. Interestingly, copper-doped germanium-based chalcogenide glasses exhibit two distinct I-V characteristics including bi-polar or uni-polar filamentary switching and threshold switching, depending on the concentration of copper and the composition of the chalcogenide glass. However, the working mechanisms that control switching are not fully understood and are highly dependent on the electrolyte materials. In this study, we propose to explore the optimal composition of the ternary glassy materials Ge-S-Cu for resistance switching by performing molecular dynamics simulations. To achieve this goal, we developed a reactive force field based on a training set of first principle calculations to describe germanium sulfide glass and its interaction with copper. We performed high throughput calculations to generate amorphous structures at various compositions following a melt-and-quench procedure and, we evaluated the corresponding mobility of Cu. Our procedure led to compositions with optimal Cu diffusivity and we studied in details their structural and electronic properties to reveal the atomistic mechanisms of copper diffusion in the glass matrix. We found that high sulfur and copper contents usually lead to high Cu mobility. Moreover, the glassy mixture is found to be a semiconductor for Cu and S contents lower than 40% and ranging from 30% to 60%, respectively. The highest Cu mobility is predicted for the composition Ge0.1S0.8Cu0.1 with an activation energy as low as 0.21 eV. We believe that this work contributes to the general knowledge of metal diffusion in chalcogenide glasses and, can be used to design ultrafast resistance switching devices.