From synapses to semiconductors: Cambridge team unveils brain-inspired memory breakthrough
Researchers from the University of Cambridge have made a groundbreaking discovery in computer memory design that could revolutionize internet and communications technologies’ performance and energy efficiency. As these technologies are predicted to consume almost a third of global electricity within the next decade, this development is crucial for reducing energy demands and mitigating carbon emissions.
The research team’s innovative device operates similarly to the synapses in the human brain, processing data and storing information in a single location. By utilizing hafnium oxide, a material commonly used in the semiconductor industry, and incorporating self-assembled barriers, the researchers can alter the electrical resistance of computer memory devices. This breakthrough could create memory devices with significantly enhanced density, performance, and energy efficiency. The findings were published in the journal Science Advances.
In today’s data-driven world, the escalating energy demands pose a challenge for carbon reduction efforts. It is estimated that artificial intelligence, internet usage, algorithms, and other data-intensive technologies will consume over 30% of global electricity in the coming years.
Markus Hellenbrand, the study’s first author from Cambridge’s Department of Materials Science and Metallurgy, highlights the shortcomings of current computer memory technologies as a major contributor to this energy explosion. The traditional separation of memory and processing in conventional computing requires energy-intensive data shuffling between the two components, resulting in increased energy consumption and time delays.
Researchers have explored a promising solution, resistive switching memory, to address these inefficiencies.
“This allows multiple states to exist in the material, unlike conventional memory which has only two states,” said Hellenbrand. “A typical USB stick based on continuous range would be able to hold between ten and 100 times more information, for example.”
The team at Cambridge tackled the challenge of utilizing hafnium oxide for resistive switching memory applications, known as the uniformity problem. Hafnium oxide lacks atomic-level structure, with hafnium and oxygen atoms randomly distributed, making it unsuitable for memory applications. However, by introducing barium to thin hafnium oxide films, the researchers observed the formation of unique perpendicular structures within the composite material.
These structured vertical barium-rich “bridges” allow the passage of electrons while maintaining an unstructured hafnium oxide surrounding. An energy barrier forms at the interface where the bridges meet the device contacts, which electrons can traverse. By controlling the height of this barrier, the electrical resistance of the composite material can be modified, enabling multiple states to exist.
Importantly, these hafnium oxide composites self-assemble at low temperatures, avoiding the need for expensive high-temperature manufacturing processes. The composites exhibited excellent performance and uniformity, making them highly promising for future memory applications.
“What’s really exciting about these materials is they can work like a synapse in the brain: 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,” Hellenbrand said.
This characteristic makes them particularly well-suited for the rapidly expanding fields of artificial intelligence and machine learning.
The researchers are now collaborating with industry partners to conduct larger feasibility studies on the materials, aiming to gain a deeper understanding of the formation of these high-performance structures. Since hafnium oxide is already used in the semiconductor industry, integrating this technology into existing manufacturing processes would be relatively straightforward.
With the potential to greatly enhance computer memory’s performance and energy efficiency, this discovery has significant implications for the future of Internet and communications technologies, offering a pathway to address the rising energy demands of our data-centric world.
