LONDON Resistive RAMs in various forms are making their way on to the research roadmaps for memory at the IMEC research institute. But while non-volatile memory may provide a replacement for NAND flash and DRAM as we know them today, that does not mean that a single "universal" memory is going to sweep the field.
That might be attractive from the manufacturers' point of view but application cases make specifications which in turn define device requirements. And they are becoming increasingly diverse, suggesting that a number of different types of RRAM may make it into manufacture.
As a result the advent of a "universal memory" is unlikely, according to Laith Altimime, program director for CMOS technology at IMEC. However, the technical and economic challenges facing scaling memories down to 10-nm and the variety of fundamental switching mechanisms, provides IMEC with a broad landscape in which to try and add value.
IMEC began work on resistive RAM (RRAM) in 2008. It began work with the then major manufacturers Samsung Electronics Co. Ltd., Hynix Semiconductor inc., Elpida Inc., Micron Technology Inc. and Qimonda AG., At that time Altimime was with Qimonda, based in Dresden, and where he was responsible for development of advanced node memory technologies beyond 3X-nm. Those companies, but with the exception of the now-defunct Qimonda, are still participating with IMEC on the RRAM memory research, Altimime said.
IMEC is now looking at two classes of materials for use in RRAMs metal oxides and metal-chalcogenides mixtures of a metal with germanium, antimony and tellurium. In general there is a trend to investigate the metal-oxides, for feasibility and rewritability as a replacement for NAND flash and chalcogenide-based material systems for DRAM replacement, Altimime said.
When asked which metal-oxide is favored, Altimime said nickel-oxide is the popular option amongst researchers around the world, but the materials is very hard and this makes it very difficult to etch. Part of IMEC's research is therefore into finding a patterning solution for nickel-oxide. But other materials such as copper oxide and hafnium oxide have possibilities and then there are mixed systems which may show up enhanced properties.
For chalcogenide switching copper is one possible ion source, potentially provided by copper electrodes, although silver is an alternative, Altimime said.
IMEC's researcher have been doing a series of fundemantal cross-bar studies at 50-nm to try and better understand a broad range of RRAM materials. In the second-half of 2010 the team is scheduled to produce one-transistor, one-resistor (1T1R) memory cell studies to prepare the way for memory arrays made using the incoming extreme ultraviolet (EUV) lithography. Altimime said that in early 2011 he expects to be making arrays at 10-nm minimum dimensions.
Altimime said IMEC research teams are hoping to present the results of some RRAM work at IEDM in December, for which the submission deadline is fast approaching.
Related links and articles:
Superlattices enable small, fast, low-power RRAM
Applied-backed firm emerges in resistive RAM
Australian-backed memory startup Qs Semi emerges