3D stackable non-volatile RRAM enables 20GB arrays

The material stack starts with a metallic contact at the bottom (platinum in the proof of concept prototype, but nearly any metal will do, according to Tour) on top of silicon-dioxide insulating layer atop a standard silicon wafer. On top of the metallic contact is a pure tantalum with a nano-porous layer of tantalum-oxide atop the pure tantalum. Ten atomic layers of graphene (multi-layer graphene, or MLG, in the diagram) cap the tantalum-oxide, followed by a top metallic electrode (platinum in the diagram, but again any metal will do, according to Tour).

The memory cell switches on and off by starting out with all zeros--no direct connections between the crossbar electrodes. However, when a programming voltage is applied across two perpendicular crossbars, it causes the oxygen to migrate upward, allowing an all-tantalum filament to form across the porous-tantalum oxide toward the upper electrode. When it hits the graphene barrier layer current flows 10-times better than originally thus creating a "1" in the bit cell.

To reprogramme as a zero, a high-voltage signal is sent through the crossbar to erase the cell by knocking out a few atoms of tantalum from the filament. If a one is later desired the low programming voltage is reapplied and the filament reconnects.

The process is very fast, according to Tour, and can be repeated indefinitely, but best of all consumes 100-times less energy than the nearest competitive design.

In the paper, there are many more details about how the control voltage can switch between ohmic and Schottky behaviour, how the oxygen vacancies (holes in the tantalum-oxide arrays where oxygen atoms should be) migrate, how the tantalum/tantalum-oxide versus the tantalum-oxide/graphene interfaces work, and how the negatively charged oxygen ions create a diode-like barrier that nixes crosstalk.

Two hurdles to commercialisation remain, according to contributor to the paper, said Gunuk Wang, professor at Korea University (Seoul). Researchers need a way to control the size of the nanopores and the fabrication of a dense enough crossbar to prove the concept of addressing ultra-dense individual bits.

Other contributors to the work at Rice included former research scientist Jae-Hwang Lee (now at University of Massachusetts) and post-doctoral researchers Yang Yang, Gedung Ruan, Nam Dong Kim and Yongsung Ji.

- R. Colin Johnson
??EE Times

?? 1???2