Last fall’s Intel Developer Forum in San Francisco included a section on technologies that helped extend the computing time of laptop computers. Medis Technologies demonstrated its fuel-cell-based, $1

PCM emerges as viable memory

MRAM, MirrotBit, FeRAM, ZRAM, PCM, XPM—the list of new and emerging memory technologies can go on and on. All that a designer needs to come out with an alternative memory technology is a material with hysteresis curve characteristics. Nature has provided many such materials. With Flash and DRAM technologies facing scaling problems in deep nanometrics, designers have been searching for a material that not only shows hsyteresis curve characteristics but can also scale down in accordance with Moore’s Law.

BASIC ISSUES SORTED OUT
Their search perhaps has come to an end with PCM, which makes use of chalcogenide, recently moving out of R&D stage. Many companies were working on memory designs using chalcogenide. Reports emerging from companies such as Samsung, IBM, Macronix, Qimonda, and others confirm that most of the basic issues in making chalcogenide a memory option have been sorted out.

For one thing, PCM is not a new technology. As long back as September 1970 Gordon Moore had written about it. Even earlier in the 1960s Stanford Ovshinsky first explored the properties of chalcogenide glasses and understood their memory potential. The crystalline and amorphous states of the chalcogenide glass have vastly different electrical resistivity. The high-resistant amorphous state represents binary 0, while the low-resistant crystalline state represents binary 1. However, issues such as material quality and power consumption restricted their use mainly to CD-RW and DVD-RW. These applications do not make use of electrical resistivity but of the material’s optical properties; chalcogenide’s refractive index changes with the state of the material.

According to a Samsung patent application concerning the technology, the phase transition process of chalcogenide is as fast as five nanoseconds. This is comparable with volatile memory devices at the leading edge today, but chalcogenide has the additional advantage of having the capability of being scaled down to 5nm and even beyond. Thus, while microprocessors and other semiconductor components seem to be hitting the wall at every node as they go beyond 65nm, including MOSFET-based memory products such as SRAM, DRAM, and Flash, PCM-based memory products technology can scale relatively easily, giving them an advantage not only over traditional memory products but also over new memory technologies such as MRAM and FeRAM. However, I do believe that even though various studies have convinced designers that theoretically PCM is scalable to less than 5nm, producing a device at this node will offer challenges that we cannot fully realize today. What would be the device structure? What would be the wordline/bit line configuration of the array? How will the designer get all of these to work in dimensions so small? Undoubtedly all such issues will be sorted out, but I would not hazard a guess in how much time.

PCM AS MAINSTREAM
Will PCM become the mainstream memory technology? If yes, when? With all its potential, it stands a good chance but nobody is willing to make a guess. Intel believes this may not happen at least in this decade or early next decade. Intel sees NAND Flash as the mainstream Flash memory in the foreseeable future. Some companies believe that PCM will not be taken up in a big way by a majority of memory makers, who would opt for the safer way of making incremental improvement in designs of NAND Flash rather than investing heavily in designing memories based on a yet-to-be-proved revolutionary technology. Many new technologies have gone through a similar phase of reluctant acceptance. Some of them fall by the wayside while others emerge from the shadows of doubts to storm the market. I have no doubt that PCM will belong to the second category.