Researchers at the University of Cambridge have developed a groundbreaking design for computer memory that has the potential to significantly enhance performance and reduce energy consumption in internet and communications technologies. The team created a device that emulates the functioning of synapses in the human brain using hafnium oxide, a material commonly used in the semiconductor industry, and self-assembled barriers that control the passage of electrons.
By manipulating the electrical resistance in computer memory devices and enabling information processing and memory storage to occur in the same location, this innovation could lead to the development of memory devices with higher density, improved performance, and reduced energy requirements. The findings of the study have been published in the journal Science Advances.
The escalating energy demands of our data-driven world have made it increasingly challenging to curb carbon emissions. It is projected that within the next decade, artificial intelligence, internet usage, algorithms, and other data-centric technologies will consume more than 30% of global electricity.
Dr. Markus Hellenbrand, the first author of the study from Cambridge’s Department of Materials Science and Metallurgy, explained that the surge in energy demands is largely due to limitations in current computer memory technologies. Traditional computing systems have separate memory and processing units, necessitating the back-and-forth transfer of data, which consumes both time and energy.
One potential solution to address the inefficiency of computer memory is a novel technology called resistive switching memory. Unlike conventional memory devices that operate in two states (one or zero), resistive switching memory devices can exhibit a continuous range of states. Implementing this principle in computer memory devices would enable significantly higher density and speed.
Hellenbrand and his team developed a prototype device utilizing hafnium oxide, an insulating material commonly used in the semiconductor industry. However, hafnium oxide presented a uniformity challenge for resistive switching memory applications because it lacks atomic-level structure, with hafnium and oxygen atoms randomly mixed.
To overcome this challenge, the researchers introduced barium into thin films of hafnium oxide, resulting in the formation of structured vertical barium-rich “bridges” perpendicular to the hafnium oxide plane within the composite material. These bridges allowed electrons to pass through while maintaining an unstructured hafnium oxide environment. At the junction where these bridges met the device contacts, an energy barrier was formed that electrons could traverse. By manipulating the height of this barrier, the researchers could modify the electrical resistance of the composite material.
Hellenbrand emphasized that this approach enables the material to exhibit multiple states, unlike conventional memory that is limited to two states. The hafnium oxide composites self-assemble at low temperatures, eliminating the need for expensive high-temperature manufacturing methods. The composites exhibited exceptional performance and uniformity, making them highly promising for future memory applications.
Cambridge Enterprise, the commercialization arm of the University, has filed a patent for this technology. Hellenbrand highlighted that the materials’ ability to store and process information in the same location, akin to synapses in the brain, makes them particularly exciting for the rapidly expanding fields of artificial intelligence and machine learning.
The researchers are collaborating with industry partners to conduct extensive feasibility studies on the materials, aiming to gain deeper insights into the formation of these high-performance structures. Since hafnium oxide is already employed in the semiconductor industry, integrating it into existing manufacturing processes would not pose significant challenges.
Source: University of Cambridge