A novel computer memory design has been created by researchers, offering significant enhancements in performance while simultaneously addressing the pressing need to minimize energy consumption in internet and communications technologies. Projections suggest that these technologies could account for nearly one-third of the world’s electricity consumption within the next decade.
A team of researchers has devised 1 a device that emulates the data processing capabilities of synapses in the human brain. This innovation utilizes hafnium oxide, a material already employed in the semiconductor industry, along with minute self-assembled barriers that can be raised or lowered to facilitate the passage of electrons.
The approach of modulating the electrical resistance in computer memory devices, enabling coexistence of information processing and memory, holds significant potential for the creation of memory devices with substantially enhanced density, superior performance, and reduced energy consumption.
The exponential growth of data-driven technologies, including artificial intelligence, internet usage, algorithms, and other data-driven applications, has resulted in an unprecedented surge in energy requirements. This escalating demand poses a significant challenge in reducing carbon emissions. Projections indicate that these technologies are anticipated to account for over 30% of global electricity consumption in the coming years.
Dr. Markus Hellenbrand, the primary author from the Department of Materials Science and Metallurgy at the University of Cambridge, highlighted the significant role played by existing computer memory technologies in driving the surge in energy demands. He explained, “To a large extent, this upsurge is attributed to the limitations of current computer memory technologies. Traditional computing relies on a separation between memory and processing, necessitating the transfer of data back and forth between the two, which consumes both energy and time.”
A possible remedy for the inefficiency of computer memory lies in a novel technology called resistive switching memory. Unlike conventional memory devices that operate in binary states of one or zero, resistive switching memory devices have the potential to exhibit a continuous range of states. Exploiting this principle, computer memory devices based on resistive switching can achieve significantly higher density and speed.
According to Hellenbrand, employing a continuous range in a standard USB stick would enable it to store between ten and 100 times more information compared to current technology.
Hellenbrand and the research team successfully built a prototype device utilizing hafnium oxide, an insulating material already employed in the semiconductor industry. However, they encountered a hurdle known as the uniformity problem when employing this material for resistive switching memory purposes. At the atomic level, hafnium oxide lacks a defined structure, as the hafnium and oxygen atoms are randomly mixed. This randomness poses a significant challenge in utilizing hafnium oxide for memory applications.
Nevertheless, the researchers made an intriguing discovery during their study. By introducing barium into thin layers of hafnium oxide, they observed the emergence of peculiar structures within the composite material, perpendicular to the plane of hafnium oxide.
The vertical “bridges” enriched with barium exhibit a distinct and organized structure, facilitating the passage of electrons. Meanwhile, the surrounding hafnium oxide retains its unstructured nature. At the junction where these bridges intersect with the device contacts, an energy barrier forms, enabling electron transmission. Notably, the researchers successfully manipulated the height of this barrier, consequently altering the electrical resistance of the composite material.
As Hellenbrand explained, this breakthrough enables the material to sustain multiple states, setting it apart from conventional memory technologies that can only accommodate two states.
In contrast to other composite materials that necessitate costly high-temperature manufacturing techniques, these hafnium oxide composites possess the unique ability to self-assemble at low temperatures. Moreover, these composites demonstrated exceptional performance and uniformity, rendering them exceptionally promising for future memory applications.
Cambridge Enterprise, the commercialization division of the University, has filed a patent for this technology, securing its intellectual property rights.
Hellenbrand expressed his enthusiasm, stating, “The truly exciting aspect of these materials is their capacity to function akin to synapses in the human brain. They possess the ability to store and process information in a unified manner, similar to our brain’s functionality. This remarkable attribute holds immense promise for the rapidly expanding fields of artificial intelligence (AI) and machine learning.”
Presently, the researchers are collaborating with industry partners to conduct comprehensive feasibility studies on the materials at a larger scale. The objective is to gain deeper insights into the formation of these high-performance structures. As hafnium oxide is already employed in the semiconductor industry, the researchers emphasize that integrating it into existing manufacturing processes would pose no significant challenges.