Synapse

Researchers from the University of Cambridge, Linköping University, Purdue University, University College London, the University at Buffalo, and the Los Alamos National Laboratory have come up with a new approach to computer memory based on hafnium oxide — and capable of dramatically improving both performance and efficiency.

"To a large extent, this explosion in energy demands is due to shortcomings of current computer memory technologies," claims first author Markus Hellenbrand, PhD, of predictions that computers driving the internet and other communications network will grow to consume nearly a third of the world's energy supplies in the next decade. "In conventional computing, there's memory on one side and processing on the other, and data is shuffled back between the two, which takes both energy and time."

To solve that, the researchers say, will require a rethink of how memory works — and to highlight one possible approach, they set about building a prototype memory device based on hafnium oxide doped with barium. Unlike traditional memory in which data is represented as clearly-delineated zeroes and ones, the team's prototype is a form of resistive memory which stores information as a continuous range of values — boosting the density, speed, and efficiency. "A typical USB stick based on continuous range," Hellenbrand claims, "would be able to hold between ten and 100 times more information, for example."

The device works thanks to the creation of vertical bridges standing up from the hafnium oxide plane, creating a highly-structured highway through which electrons can flow — whereas the unbridged hafnium oxide sits in an unstructured state and blocks the flow of electrons. By controlling the height of the energy barrier at which the bridges meet the device contacts, the researchers' prototype is able to store continuous values.

"What's really exciting about these materials is they can work like a synapse in the brain," Hellenbrand adds. "They can store and process information in the same place, like our brains can, making them highly promising for the rapidly growing AI and machine learning fields."

The team's work has been published in the journal Science Advancesunder open-access terms; A patent on the technology has been filed by the commercialization arm of the University of Cambridge.