Two-terminal resistive switches can change their conductance values depending on the input history. Physically, the device can be either based on the redistribution and associated redox processes of metal cations (so-called conductive-bridge memory, CBRAM) or oxygen anions (so-called oxide-RRAM). Such a device is essentially a resistor with inherent memory and is sometimes termed a memristor (memory + resistor), or a more generally defined memristive system. Particularly, due to their simple two-terminal structure and the ionic nature that causes the memory effect (which leads to excellent scalability and non-volatility), these devices are promising candidates for ultra-high density, non-volatile memories that may potentially replace flash-based memory in data storage market, as well as in the fast-growing embedded memory market.
Cation-based RRAM fundamental studies and modeling.
Conductive-bridge memory, based on the redistribution of metal cations that leads to the formation and rupture of a conducting filament inside an otherwise insulating dielectric, offers high on/off resistance ratio, excellent scalability and low programming current. In the last few years, we performed systematical experimental and modeling studies that unambiguously revealed the dynamic filament growth process and the fundamental thermodynamic and kinetic factors that determine the filament growth direction, shape and morphologies (Yang Nat. Comms. 2014, Yang Nat. Comms. 2012, Sheridan Nanoscale 2011). These studies provided critical insight into the understanding and continued optimization of CBRAM devices. Additionally, we focused on atomic-scale control of the filament growth process and showed that through fundamental physics understanding and optimized device engineering, the programming current can be reduced to be below nA while still maintaining an on/off ratio > 10e3 (Gaba EDL 2014). We are currently employing such devices for very low power memory and logic applications. Additional efforts have been made to develop 3D integrated RRAM devices, including stacking of crossbar arrays layer by layer, and fabricating vertical devices with the sidewall active device structures. We also perform atomic-scale dynamic Monte Carlo simulations to support the experimental findings and extract important device and material parameters, examine the requirements for successful memory array operations during read and write and study how the native stochastic resistance switching process can lead to novel computing approaches.
Anion-based RRAM fundamental studies and modeling.
Besides CBRAM, we also perform extensive studies on oxide-RRAM, which relies on the redistribution of oxygen anions in oxides that leads to the change in local conductivity. Compared to CBRAM devices, oxide-RRAM devices offer lower high on/off resistance ratio and typically requires higher programming current. However, they typically offer longer write/erase endurance since the resistance change involves only native species (oxygen ions). Additionally, incremental resistance changes can be obtained by controlling the amount of oxygen ions that are distributed. These properties make such device excellent candidates for logic applications where endurance and analog conductance change are oftentimes desired. Like the CBRAM device project, we aim to perform systematic experimental and theoretical studies on oxide-RRAM devices, to identify the thermodynamic and kinetic factors governing the dynamic resistance switching processes, study critical noise and retention characteristics, optimize device performance including the desired non-linear I-V in the on-state and improve the desired analog switching characteristics. New device concepts such as complementary resistive switching obtained in a single device and 3D device integration are also of interests.
New multifunctional devices enabled by ionic processes.
Based on the fundamental understanding we developed from two-terminal resistive switching devices, we recently started applying the same principles in other structures to explore new device concepts. For example, in materials science research, the lanthanum aluminate/strontium titanate (LAO/STO) heterojunction has attracted significant interest after the discovery of a spontaneous two-dimensional electron gas (2DEG) formation at the LAO/STO interface. The exact mechanism behind the 2DEG formation is still being debated although many believe the formation and location of oxygen vacancies (VOs) play a critical role. Coincidentally, we have gained extensive knowledge of VO generation and migration in our studies in oxide-based RRAM devices. In this project, we aim to control the location and density of VOs in the LAO layer in situ using an external electric field. This study will not only shed light into the roles of VO in the 2DEG formation, but also can lead to novel devices where the density and threshold of the 2DEG layer can be reconfigured in a non-volatile fashion on the fly.