A New Twist on Memristance
NIST researchers create flexible memory circuits that act like memristors
10 June 2009—Researchers at the U.S. National Institute of Standards and Technology (NIST), in Gaithersburg, Md., report in the July 2009 issue of IEEE Electron Device Letters that they have created a low-power, inexpensive flexible memory that has the properties of a memristor. Memristors can be used to make brainlike circuits and nanoelectronic memories, because they ”remember” the amount of current that has flowed through them, and that memory is reflected in the device’s resistance.
Though these devices were first theorized in 1971, no one was able to make a practical memristor until Hewlett-Packard figured out how to do it in 2008. Producing a flexible form of memristor could make a fourth fundamental circuit element usable in implantable medical electronics where stiff, brittle silicon wouldn’t work, says Curt A. Richter, head of the Nanoelectronic Device Metrology Project at NIST’s Semiconductor Electronics Division.
”Interestingly, we didn’t originally set out to make a memristor,” says Nadine Gergel-Hackett, one of the researchers on the project. The NIST group’s original work, begun two years ago, was part of an existing program on printed and flexible electronics.
But when Hewlett-Packard reported its work on memristors last year, the NIST team immediately recognized the similarities between its own work and what the HP researchers had done. Though the NIST researchers weren’t—and still aren’t—certain of all the advantages of having flexible memristors, they were sure it couldn’t hurt. ”We changed the geometry of the devices we were making, and sure enough, we noticed memristor-like properties,” says Richter.
From the outset, Richter and Gergel-Hackett envisioned flexible memory being used in medical applications. But flexible memristors could also be used for advanced analog as well as digital memory. ”You can picture using the analog properties of the memristor to remember over time what was going on, while using other memristors to store the state at a given instant,” says Richter.
The key to the NIST technology is a method for spinning a precursor solution onto a surface and letting it react with the water in the air to form a solid gel of titanium dioxide. Current applied to the gel changes its state to produce the memristor-like effect.
The major technical hurdle that still must be overcome is researchers’ limited understanding of what causes the state change in the titanium dioxide. ”The TiO 2 we use is not perfect, stoichiometrically speaking,” says Richter. He notes that although TiO 2 itself is a very good electrical insulator, the HP researchers theorized that the material ”is TiO 2- x, where x is a small number.” He says that because the titanium dioxide’s crystal lattice is a bit deficient in oxygen atoms, the arrangement of the existing oxygen can be manipulated to leave vacant spots on the lattice in positions where the effect on free charge carriers is most dramatic.
Spreading the vacancies evenly across a sample of titanium dioxide creates a lower-resistance state—or at least that’s what scientists think is the case. Conversely, they think that moving the vacancies all to one side creates a high-resistance, low-current state. ”But to actually probe that on an atomic level is very challenging, and that’s one of the things we’re working on now,” Richter says.
Finding out what’s going on electrochemically would help improve the device’s reliability. ”Now, each individual device seems to have a nice, large on-off ratio. But when you compare the high and low states of two or more devices, the variability is more than we would like to see,” says Gergel-Hackett.
Another improvement the NIST team is working on is switching from using metal evaporation to deposit the electrodes to printing them with a machine akin to an inkjet printer. Once they can do that, ”we would literally put a transparency similar to the ones used for overhead projectors in the machine, print on a contact, spin on the gel, print on the second contact, and have a completely [automated] processing procedure,” says Gergel-Hackett.
”In that paradigm, you could imagine getting to make your own circuits at home,” says Richter.