The K13E, or "Nanoscale Devices" group at Almaden, works on the design of memory arrays for Storage Class Memory, or SCM. Our research is aimed at creating a next-generation Solid-State Non-Volatile memory. This work could have an impact on the Flash memory market, embedded memory applications, and even the Hard Disk drive market.

While Flash memory is expected to be a $25-70 Billion market by 2012, the scaling of such devices to the tiny dimensions that are also expected in this time-frame has become a significant concern for the industry. Meanwhile, hard disk drives continue to steadily improve in areal density - but the reliability, power consumption, and access time of such storage devices is not improving, nor is it expected to improve significantly.

In contrast, solid-state memories are much more reliable, and consume 100x less power than hard-drives. And while Flash memory can offer little in the way of write speed over hard drives, alternative technologies (such as one which we work with, called phase-change memory) could potentially offer access times nearly 5 orders of magnitude faster (<100 nanoseconds). In terms of scaling, a simple SLC NAND flash in 2010 is expected to offer an effective areal density of 48 Gb/sq. inch (with more complex MLC devices offering commensurately more). In contrast, the sublithographic Storage class Memory schemes we work on could potentially offer densities of 100-600 Gb/sq. inch. Thus SCM has dramatic implications across the memory industry, and in particular for IBM, within enterprise systems. The target specification of an SCM Drive (one that might look like a hard disk drive externally, but which has a high-density, high-performance solid-state memory inside) might include:

• a cost no more than 3-5X larger than a conventional hard drive (<$1.15/GB in 2012)
• read/write/erase times < 1ms and > 20,000 Read I/Os per second
• Lifetime of 10⁹ - 10¹² write/erase cycles
• 10x lower power than a conventional hard disk drive

To enable this vision, in our group we work on three interlocking questions:

• How are we going to store the data? This comes down to which alternative non-volatile memory technology to work on. Although we continue to explore and consider a number of candidates, most of our work to date has been on Phase Change Memory.
  
• How are we going to access the data? In addition to the underlying physics of the storage device, we are interested in memory accessing schemes that finesse the fabrication difficulties inherent in making devices with nanoscale features, yet which could practically implement densities that would be high enough for SCM technology to actually be interesting (and thus financially viable to manufacture).
  
• How are we going to build these memory arrays? In this topic, we collaborate closely with our colleagues throughout S&T, as well as with external partners, in order to understand the inevitable interaction between the lithographic scheme used to fabricate the device, the array architecture used to access the device, and the memory device itself.