For Resistive RAM (RRAM), imec built a model to understand the properties of the filaments that result in a stable RRAM operation. Such fundamental understanding of the filament properties is key to bridge the gap in the development of RRAM as a successor memory technology.
One of the major technical challenges for the DRAM industry is the difficulty to maintain target leakage currents at lower Effective Oxide Thicknesses (EOT) to meet the DRAM capacitor scaling roadmap. Recently, imec demonstrated a novel RuOx/STO/TiN stack that showed a 100x reduction in leakage with DRAM MIMcap compatible dielectrics at 0.4nm equivalent oxide thickness (EOT).
Now, imec has achieved a further 10x improvement by optimizing the stack, resulting in a record leakage current density (JG) of 2x10-8A/cm² at 0.4nm EOT. In addition, imec was able to explain the mechanism of leakage reduction. This allowed a path to demonstrate a further potential through lowering the trap density, to a theoretical leakage current density (JG) limit for trap-free STO of 10-15A/cm² at ~0.4nm EOT. These results demonstrate that the STO-based stack is a promising technology for DRAM scaling.
RRAM is a promising concept for future non-volatile memories. In RRAM, a dielectric, which is normally insulating, can be made conductive through a filament or conduction path formed by applying a sufficiently high voltage. Imec now has made breakthroughs in understanding the properties of the filaments. Imec established a.o. that the minimal achievable current after reset depends on the physical nature of the filaments, resulting in a direct method to predict that current from the filament properties. With these results, it is now possible to choose the desired properties of the filaments to ensure a stable RRAM operation.
These results were obtained in cooperation with imecs key partners in its core CMOS programmes: Globalfoundries, INTEL, Micron, Panasonic, Samsung, TSMC, Elpida, Hynix, Fujitsu and Sony.