Memristors are vanguard of next cyber revolution
HOUGHTON – For decades, the story of computing has been about getting smaller and faster, devices shrinking in size and increasing in power until today’s cell phone has more computing power than yesteryear’s mainframe. There has been constant incremental improvement but just one true computing revolution – the introduction of the microchip in 1959.
Now, the world could be poised for a second revolution, and Michigan Tech could be at the leading edge of the next generation of computing. Memristors, said Tech materials science and engineering professor Yun Hang Hun, could do away with microchips, exponentially increasing computing power and even leading to artificial intelligence to replicate the thinking of the human brain.
Recently, said Hun, he and two colleagues developed the most successful memristor to date, shaving molybdenum sulfide (MoS2) down to nano-thickness and coating it with electric-conducting silver resisters to create chips that can record a broad spectrum of electric signals.
The MoS2 is resistant to changes in voltage in the resisters, Hu explained. “The memristor holds the wattage change in its memory.
“But a microchip can record only zeros and ones,” he said. “This can use any number in between.”
In other words, the same space that can record a single yes or no bit of data can now hold a theoretically infinite number of signals for the computer.
Tech did not invent the memristor, Hu noted. The concept was first proposed in 1996, and HP, Inc. (formerly Hewlett Packard), published a paper on an early design in 2008, announcing a plan to build and commercialize a chip by 2020. But Tech’s memristor ups the ante for current developers by improving the symmetry of the electric current in the chip to increase its storage capabilities.
The secrets, Hu said, were exfoliating the MoS2 down to a next-to-nothing thin nanolayer, then sandwiching two layers of MoS2 between three layers of electric resistors.
“We got two layers into a single memristor so the interface on both sides is exactly the same,” Hu said, adding that finding a way to go even thinner with the MoS2 should further increase performance.
“We’d like to get it to the atomic layer, but we’re not there yet,” he said.
Memristors won’t change computing on their own, Hu said. They’ll also need computers designed to take advantage of all that memory.
There’s still work to do, he said, but the potential appears endless.
“To go to unlimited memory, I think, is feasible,” Hu said.