PORTLAND, Ore.—Electrical engineers who expressed skepticism that Hewlett Packard Co.'s memristors could switch as fast as DRAM and yet retain their memories millions of times longer than flash can now rest easy, according to their inventor, senior HP Fellow Stanley Williams.
"What we have discovered is that an electric field and a current act together to enable a memory device that can both be switched very rapidly and hold its state indefinitely," said Williams. "Not only does an applied voltage drive the migration of oxygen vacancies in the device, but at the same time there is a current that heats it up to about 300 degrees Celsius—just enough to turn the amorphous film into a crystalline film."
Memristors are touted as the future "universal memory" device because they are as fast as DRAM, as small as flash, and as durable as read-only-memories, according to HP. As the fourth fundamental passive circuit element—after resistors, capacitors and inductors—memristors retain either a high- or low-resistance state by virtue of introducing or removing oxygen vacancies in oxide thin films.
Synchrotron x-rays probed the memristor in a 100 nanometer region with concentrated oxygen vacancies (right, shown in blue) where the memristive switching occurs. Surrounding this region a newly developed structural phase (red) was also found to act like a thermometer revealing how hot the device becomes when read or written.
Using their favorite formulation—titanium oxide—HP recently used high-energy synchrotron x-rays to correlate the device's electrical characteristics with its atomic structure, chemistry, and temperature in three dimensions. The until now unforeseen conclusion was that a hot spot near the bottom electrode heats enough during switching to induce a crystallization of the oxide. After driving out vacancies (for a 1) or introducing them (for a 0) in one-to-two nanometers thick region, the film cools in an annealing-like like process which leaves the film in a fixed crystalline state that should remain that way indefinitely.
"In testing, we have switched these devices over 30 billion times and counting, with no degradaton in their ability to retain information," said Williams.
If this technology is dependent on hot-spots for phase transitions, I would be surprised if reliability were not an issue. They're claim of cyclability is a little suspect. How many devices are switching for 3x10^10 cycles?
Colin""""Not only does an applied voltage drive the migration of oxygen vacancies in the device, but at the same time there is a current that heats it up to about 300 degrees Celsiusójust enough to turn the amorphous film into a crystalline film."""
So it now appears rather than just a device that uses voltage driven migration of vacancies to change the condition of a thin region, the memristor is a phase change is required, this is now another PCM a phase change memory. It needs a temperature of 300 C to crystallize, what temperature is required to return it to the amorphous state? Or how does it return to the amorphous state and is the formation of the crystalline state a one time initial forming process into which vacancies are forced or removed to create the memory states?
It will be interesting to see if the memory patent portfolio of ENER and or Ovonyx cover phase changes in materials other than chalcogenides and or materials in general, I suspect they might. If so HP might find they have more than a technology manufacturing problems to deal with.
Colin, that is interesting to note, thanks. But for these analog operations, it is even more important to first confirm the distributions are tight. Otherwise relying on the sensitivities will lead to lots of errors.
Charges jump in and out all the time even at zero current. Should we expect natural memristance fluctuations?
HP is also planning to tune these sensitivities for analog operation, both to add levels for multi-bit per memristor storage, as well as for brain-like connections that become more conductivity the more you use them.
Join our online Radio Show on Friday 11th July starting at 2:00pm Eastern, when EETimes editor of all things fun and interesting, Max Maxfield, and embedded systems expert, Jack Ganssle, will debate as to just what is, and is not, and embedded system.