If you buy something using links in our stories, we may earn a commission. Learn more.
In the spring of 2008, HP told the world it had done something that computer wonks had dreamed about for nearly four decades. The hardware giant announced that researchers inside HP Labs had built a memristor -- a new type of electrical building-block -- and it said this hardware missing link would reinvent computer memory as we know it.
"To find something new and yet so fundamental in the mature field of electrical engineering is a big surprise, and one that has significant implications for the future of computer science," said HP's Stan Williams, the man who oversaw the project.
Now, HP says it's less than three years away from introducing commercial memristor hardware, predicting it will provide far more efficient machines than flash memory and other hardware used in today's systems.
But Blaise Mouttet says the hardware giant is full of hot air.
Mouttet -- a former U.S. patent officer who specialized in nanotechnology -- has long argued that HP's technology is not really a memristor. He says that other companies -- including big names Panasonic and Samsung -- were developing similar technologies well before HP's 2008 announcement.
"All HP is doing, in my opinion, is skewing the history to make it look like they were the originators of this technology and it is really not true," Mouttet tells Wired. "To me, this is unethical."
HP didn't respond to a request for comment on Mouttet's claims. But these sorts of debates are not uncommon in tech industry -- especially when there's a place in history at stake. Engineers will always argue over who invented email, the portable music player, and even the internet itself, and the debate over the memristor is no different.
The electrical circuits that underpin today's machines use three basic components: the capacitor, the resistor, and the inductor. But in 1971, Leon Chua, a professor at the University of California at Berkeley, predicted a fourth: the memristor, short for "memory resistor." Basically, he noticed a gap in the mathematical equations describing the world's electrical circuits, and he built a new equation to plug this hole, foretelling the existence of the memristor.
Like a resistor, he said, it would create and maintain a safe flow of electrical current across a device. But unlike a resistor, it would "remember" charges even when it lost power.
HP says that it has finally built Chua's fabled memristor using two layers of the semiconductor called titanium dioxide, and it claims this contraption will serve as a more efficient breed of computer memory. It will reduce boot times, HP says, and decrease power usage.
Basically, one layer of titanium dioxide includes "oxygen vacancies" and the other doesn't. The top layer with the vacancies is conductive, and the bottom layer isn't. If you send a voltage through the device, HP says, you can move the vacancies from one layer to the other, and that lets you build a switch. It's "on" when the vacancies are in one layer, and it's "off" when they're in another.
But according to Mouttet, this isn't a memristor. Mouttet helped develop the nanotechnology patent class for the US Patent and Trademark Office. He holds eleven of his own patents related to memristors himself. His paper on the business prospects of memristive electronics was published in trade journal Nanotechnology Law and Business in the fall of 2010. And now, at least in the world of hardcore computer electrics, he has made a bit of a name for himself arguing against HP's claims.
He says that HP's creation isn't that different from the titanium dioxide resistance switching technology described by a 1967 paper that came out of the University of London, and he insists that Samsung, Panasonic, and a company called Unity Semiconductor started working on similar technology years before HP's announcement.
"The mathematical models based on Chua's memristor -- or the more generalized memristive systems -- are unable to predict most of the resistance switching behavior of metal oxides or any other materials exhibiting resistance switching effects," he says.
Chua did not respond to our request for comment. But Darrell Rinerson -- the founder of Unity Semiconductor, a company now owned by memory giant Rambus Inc. -- agrees with Mouttet's assessment. He even goes so far as to say that HP's technology is unlikely to ever reach the market. "One thing I can assure you: there is going to be no HP memristor in two years," Rinerson tells Wired. "The probability is zero."
According to a recent report, HP's Stan Williams has indicated that the technology would reach the market by the end of 2014. But he also indicated that its manufacturing partner, South Korea's Hynix, is concerned that the technology will eat into its existing flash memory business.
Dr. Kris Campbell -- who worked on memory technologies at Micron Technology before joining the faculty at Boise State University's electrical and computer engineering department -- once sided with Mouttet and Rinerson. But in recent months, she has changed her stance. She says HP has built something a little different from what Leon Chua described in 1971, but she also says that HP's papers describing the technology show that it has developed a technology that will look and act like a memristor.
HP's technology will lose some data when powered off, she says, but it does exhibit most of the properties the Cal professor predicted. "Until someone can actually prove that this device fits that role that Chua is describing -- and prove it experimentally and model it properly -- you're going to have that camp that disagrees," Campbell tells Wired. "It's pretty convincing that you could use any continuously variable or resistance variable device as a memristor. So I don't think HP is out of the ballpark."
Like most technology Who-Done-Its, says Martin Reynolds, an electrical engineering analyst with research outfit Gartner, the questions over HP's memistor ultimately boil down to semantics. Like Campbell, he says that while HP's technology HP is not precisely what Chua proposed, it does have memristor-like qualities.
"Is Stan Williams being sloppy by calling it a 'memristor'? Yeah, he is," Reynolds tells Wired. "Is Blaise Moutet being pedantic in saying it is not a 'memristor'? Yeah, he is."
In the grand scheme of things, he says, the argument is pointless. "At the end of day, it doesn't matter how it works as long as it gives us the ability to build devices with really high density storage."
But the argument will continue. That's just the way the tech world works.
