Opinion

posted on 11 April 2008 07:58

If HP were to announce it was researching chemical bonding processes in blue gas giant stars and one of its researchers was a Nobel prize winner we would yawn, vaguely think 'that was impressive', and then ask why the heck HP was wasting shareholders' money looking into something that was nothing whatever to do with its business. Quite right too.

Yet IBM can announce it is making progress researching into spintronics memory and no-one asks, 'Why?'. What on earth is it to do with IBM's business of selling computing services, mainframe and other computer servers, lots of computing software, and storage products? IBM doesn't make memory. It isn't in the semi-conductor foundry business. It doesn't develop and licence memory IP like Rambus does. So why is it paying someone to research spintronics memory?

A company researcher, Stuart Parkin, an IBM Fellow, says IBM could develop new intellectual property that it could license. Yes it 'could' but is this a realistic business outcome?

What is this new potential IP?

It's to do with storing information with entities called domain walls, the areas between differently charged magnetic regions, in microscopically thin nanowires, colloquially termed race tracks. Electrical currents are used to move the domain walls along the wire to a sensing device which reads them as a one or zero. The domain walls are read as magnetic moments in Parkin's phrase.

A paper in todays issue of the Science journal describe the long, hard, complex and very difficult road the IBM researchers have trodden to get to a point where they can create these magnetic region walls, then move them along the nanowire and detect their state in a reader as they pass by it.

The nanowires could be stacked in layers to make denser memory chips that might, might, be 100 times denser than flash memory and cost 100 times less. They might also have a longer working life than flash chips.

Actually it seems that IBM researchers are creating tiny and stable magnetised regions on a nanowire. Then they move them without decay or damage along the wire and the reader detects the edges between the regions and interprets the regions as ones or zeros.

When a tape head reads a tape it detects the changes between the magnetised regions, not the regions themselves, ditto a disk drive head. A change in polarity defines the edge of a region. If there is no edge then there is no polarity change. The magic that the IBM researchers have achieved is to find a way of replicating this edge detection ability at the nanowire level and, in a brilliant demonstration of technological prowess, move the magnetised regions along the nanowire without compromising their integrity.

Imagine moving magnetised areas along a tape ribbon or around a disk drive track. It is a brilliant and awesome piece of work. Yet the chances of IBM successfully commericialising this are so very remote. Remember Millipede and its nanoscale hammers making readable impressions in some substrate? That was magic too, and about as realistic as the Terminator in the Arnie films. Great special effects but they stop as soon as you leave the movie theatre.

IBM researchers hope to develop a prototype racetrack chip with the capability of storing data as walled domains on a nanowire, moving them along the wire to a reader and detecting the information. We might see this in a couple of years and we might see real nanowire racetrack memory chips in the 2016-2020 period.

We might see also Jon Toigo become the SNIA chairman, or Steve Ballmer might even marry Oprah Winfrey. Who knows what 'might' happen? But these things are unlikely, as is, I fear, the likelihood of IBM successfully commercialising racetrack memory.

As awesome, brilliant, mind-boggling scientific and technical prowess as it undoubtedly is, why does IBM bother?

^{[Chris Mellor.]}