Cross-sectional TEM images and energy-dispersive x-ray measurements

A computer memory chip modeled by the human brain

A new computer memory design has been developed by researchers that promises to greatly improve performance and reduce the energy demands of internet and communication technologies. These technologies are predicted to consume almost one-third of global electricity in the next ten years.

The researchers, led by the University of Cambridge, created hardware that processes data similarly to human brain synapses. The device uses hafnium oxide, an already existing semiconductor industry material, and tiny self-assembled barriers that can be raised or lowered to allow electrons to pass through.

Using this method to change the electrical resistance in computer memory chips could lead to the development of computer memory chips with higher performance, greater density, and lower energy consumption. The results are presented in the journal Science Advances.

Internet communication devices account for almost 1/3 of global electricity usage

Our world’s increasing demand for data has resulted in a huge rise in energy consumption, which makes it even more difficult to reduce carbon emissions. It’s expected that in the next few years, data-driven technologies such as artificial intelligence, internet usage, and machine learning algorithms will consume over 30% of global electricity.

The current computer memory technologies are largely responsible for this surge in energy demands. In conventional computing, there are two separate components: memory chips and CPU chips. Data is shuffled back and forth between these components, which takes both energy and time.

Resistive switching memory offers a range of states rather than typical on/off binary switches

One potential solution to the problem of inefficient computer memory is a new technology called resistive switching memory. Conventional memory devices can only operate in two states: either one or zero. However, a resistive switching memory device would be capable of a continuous range of states, which would make it possible to increase the density and speed of computer memory devices. For example, a typical USB stick based on the continuous range principle would be able to hold between ten and 100 times more information than today’s USB hardware.

Hafnium oxide and barium provide the solution

To create a prototype device based on resistive switching memory, Hellenbrand and his colleagues used hafnium oxide, an insulating material that is already used in the semiconductor industry. However, using this material for resistive switching memory applications can be problematic due to the uniformity problem. At the atomic level, hafnium oxide has no structure, with the hafnium and oxygen atoms randomly mixed, making it challenging to use for memory applications.

Nevertheless, researchers found a solution by adding barium to thin films of hafnium oxide. This caused some unusual structures to form perpendicular to the hafnium oxide plane in the composite material. These vertical barium-rich structures are highly organized, allowing electrons to pass through while the surrounding hafnium oxide is not.

At the point where these structures meet the device contacts, an energy barrier is created, which electrons can cross. The researchers were able to adjust the height of this barrier, which changes the electrical resistance of the composite material, allowing for multiple states to exist in the material instead of the usual two states found in conventional memory chips.

These hafnium oxide composites self-assemble at low temperatures, unlike other composite materials that require costly high-temperature manufacturing methods. The composite material showed great performance and consistency, making it a highly promising option for next-generation memory applications.

Our Sponsors

Cambridge Enterprise, the University’s commercialization arm, has filed a patent on the technology.

Dr. Markus Hellenbrand, from Cambridge’s Department of Materials Science and Metallurgy, said:

“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.

The researchers are collaborating with industry to conduct more comprehensive feasibility studies on the materials to gain a clearer understanding of how the high-performance structures are formed. Hafnium oxide is already used in the semiconductor industry, so integrating it into existing manufacturing processes should be easy, according to the researchers.