Flexible Graphene Memristors

Graphene oxide is the latest flavor of memristor, and it works on bendable plastic

Photo: Sung-Yool Choi

13 October 2010—South Korean researchers have recently made a flexible nonvolatile memory based on memristors—fundamental electronic circuit elements discovered in 2008—using thin graphene oxide films. Memristors promise a new type of dense, cheap, and low-power memory and have typically been made using metal oxide thin films. The new graphene oxide devices should be cheaper and simpler to fabricate—they could be printed on rolls of plastic sheets and used in plastic RFID tags or in the wearable electronics of the future.

"We think graphene oxide can be a good candidate for next-generation memory," says Sung-Yool Choi, who leads flexible devices research at the Electronics and Telecommunications Research Institute in Daejeon, South Korea. Choi and his colleagues reported their device last week in Nano Letters.

Memristors change their resistance depending on the direction and amount of voltage applied, and they remember this resistance when the voltage is removed. Most memory types store data as charge. But memristors would enable what’s called a resistive RAM, a nonvolatile memory that stores data as resistance instead of charge.

Although the idea for memristors was first proposed in 1971, Hewlett-Packard made the first practical device only two years ago and plans to commercialize the technology within three years. HP’s design consists of two arrays of parallel wires arranged perpendicular to each other with a titanium dioxide layer sandwiched in between. Each cross point is a memristor.

The South Korean researchers use a similar design, swapping titanium dioxide for graphene oxide. After depositing 50-micrometer-wide aluminum wires on a 6.5-square-centimeter piece of plastic, they spin a solution containing suspended graphene oxide flakes onto the surface. This forms a thin film of overlapping graphene oxide flakes over which the researchers deposit the top aluminum wire array. This results in 25 memristors, each 50 µm wide.

The graphene oxide devices are 1000 times the size of HP’s memristors, but they’re not intended as ultradense memory. Instead, cheap and flexible are the key words here. "With organic devices, it’s not really about making them really small, because they won’t then compete with silicon in terms of performance," says Max Lemme, a graphene researcher at the Royal Institute of Technology, in Kista, Sweden. "This application for graphene oxide is very nice," he adds. "The question is, ’How can we scale the process up to make a large number of devices and make them cheap?’ "

Choi and his colleagues believe that the devices switch between two resistance states because conductive filaments form when oxygen transfers between the graphene oxide and the top aluminum electrode. Flipping the voltage bias breaks the filaments, turning the device off.

The devices can cycle between the "on" and "off" states 100 000 times, a lifetime similar to flash memory. Choi says this could be increased to 1 million cycles. While the team tested only the memristors’ ability to retain their state for about 27 hours, Choi says the first few devices they made last September are still in the same state. Plus, the devices don’t degrade when bent 1000 times.

The graphene memristors show excellent characteristics for nonvolatile memory, says Curt Richter, a researcher at the National Institute of Standards and Technology (NIST) who led the work on flexible titanium dioxide memristors published last year. But, adds Richter’s colleague David Gundlach, a project leader in NIST’s semiconductor electronics division, graphene oxide devices will have to prove themselves at larger scales.

About the Author

Prachi Patel is a contributing editor to IEEE Spectrum. She writes regularly about energy, the environment, and engineering careers. In the September 2010 issue she reported on false assumptions about engineering graduates in India, China, and the United States.

Related Stories