At least two or three times a week, we get a press release about some fundamental breakthrough in nanotechnology, silicon engineering, wireless or similar.
Normally, the story is rather less exciting than the PR would have us think: after a good twenty years of exciting fundamental breakthroughs in nanotech, and we've got accelerometers in our iPhones, mirrorchips in our projectors, and... well, not much else. In some cases, this is because it always takes a long time for new ideas to turn into products -- there's been steady progress, but how long have we been waiting for OLEDs? -- and in others, the idea just doesn't make it because it's impossible to produce economically, for technical reasons or because the market's moved on.
So I thought I'd leave HP's Memristor announcement of last week to... mature a bit. Some of the headlines - HP DIscovers Electronics God Particle - made me think it'd better to let that side of things burn itself out, and revisit it after cooler minds had taken a look.
I'm glad that I did, because now some of the smoke has cleared it looks a very compelling discovery - one, moreover, that has good potential for a relatively swift adoption by the industry..
Ignore all the stuff about 'a new fourth class of electronic component': there are loads of interesting weird electronic devices which aren't resistors, capacitors or inductors. What matters is what the memristor does, how it does it and whether it's going to be actually useful.
What it does is simply put: it has a resistance to electrical current, but as you put current through it that resistance changes. Take the current away, and it sticks. Come back some time later, and you can read the old state: it's an analogue memory circuit.
How it works is beguilingly simple. Titanium dioxide is a poor conductor of electricity, with one interesting twist: it changes its conductivity when it encounters oxygen - in fact, it's used in oxygen detectors. The more oxygen, the worse the conductivity.
Take a chunk of titanium dioxide - which has a crystal structure based around two oxygen atoms for every titanium. Arrange for some of the chunk to have holes in its crystal lattice where the oxygen should be. More holes - less oxygen - lower resistance.
The interesting thing is, when you pass a current through the substrate, the holes move across - reducing the overall resistance. Reverse the current flow, and they move back, bringing the resistance back up again. Not too dissimilar with the way that charges move around a semiconductor, but because the flow is ionic, the condition of the device stays constant when the motive force goes.
For a memory circuit, you pass current one way for a zero, the other for a one. That leaves the memristor in a high or low conductivity state. Come back later and measure the resistance, and you can read it back. (Yes, you at the back, measuring resistance does involve passing a current through the memristor and thus changing its state. Use alternating current, and you can easily leave it as you found it).
Sounds simple. So how come it took so long to find? Turns out that it only becomes a significant effect at nanotech scales: you need to get down to the nanometres to be able to spot it happening.
The really exciting thing -- assuming that I haven't missed anything: haven't seen the Nature paper yet -- is that this is pretty much a plug-in-and-go component for existing techniques. HP already has the nanowire crossbar technology that's necessary to turn the memristor into a memory array, and the business of putting down carefully tuned layers of chemicals; well, that's what the semiconductor industry's all about.
So not only does the memristor seem like a simple, effective and useful innovation that works in a reasonably clear way, it's within sight of the finishing line already.
Lots more on this to come.