Dutch discovery brings energy-efficient computers a step closer

As the world struggles with the amount of energy used for computers, datacentres and other electronics, researchers are looking for alternatives to provide the power to fulfil those requirements.  

Christian Nijhuis, professor of Hybrid Materials for OptoElectronics at the University of Twente’s science and technology faculty, is one such expert.

“Wind farms are now literally being built next to the large datacentres of Microsoft and Google,” he says. 

About half of all energy generated is used to transfer and process digital information. “The system squeaks and creaks,” says Nijhuis – the reason for him, and his department, to look into more energy-efficient ways of transferring and processing data.

“Nature has always fascinated me infinitely,” he says. “The way our bodies process and store energy, for example, and how plants convert energy from sunlight into nutrients via photosynthesis.”

The human brain is a significant inspiration for Nijhuis’s research. “Our brain is an incredibly smart and energy-efficient supercomputer,” he says. “Of course, there are some things a digital system can do better than us, such as processing huge amounts of information quickly, but in simply recognising a cat in a picture, our brain can still do much better than a computer. Also, when you are in a jam-packed stadium, and someone calls your name, you will hear it, but your neighbour won’t.”

Nijhuis says this is because our brains are neural networks. Although the technology sector has been working for years to develop artificial neural networks, it’s proving extremely difficult to mimic our brains digitally. “Moreover, computers are huge energy guzzlers, whereas our brains are very energy-efficient,” he adds.

Molecules mimicking synapses 

Nijhuis and his team set out to find molecules that mimic our brains’ synapses. “An important feature is that synapses can be on or off, and everything in between,” he says. “Moreover, the extent to which they turn on or off depends on how often and quickly they have already done so.”

The researchers looked for that learning ability in molecules. The actual discovery is a nice anecdote, says Nijhuis. “Before returning to Twente, I spent 10 years researching this field in Singapore. The student who made the final discovery did not dare tell me that his research was not going as expected because he was afraid of losing face.”

The student researcher spent months putting a molecule under tension and seeing if it turned on or off like a light bulb, but this molecule behaved strangely and didn’t do what the student expected. It didn’t clearly switch on or off but seemed to differ in how much it switched on or off. “When I heard that, I immediately knew: this is exactly what we are looking for! The molecule was learning from previous behaviour,” he says.

Having succeeded in mimicking one synapse, the next step is an entire network. To develop that, Nijhuis returned to Twente for further research. He received a Vici grant from the Dutch Research Council to support this research. A Vici grant is one of the Netherlands’ most significant personal scientific grants, and is aimed at advanced researchers who have demonstrated their ability to develop their own innovative line of research successfully. The Dutch Research Council is one of the leading science funders in the Netherlands, and ensures quality and innovation in science.  

One of the reasons why it was attractive for Nijhuis to return to the University of Twente is the presence of a nano lab, the largest cleanroom in Europe.  

Moreover, the university is also a very strong player in microtechnology. “And we have recently set up a molecular centre,” he says. “This combination of optical, nanotechnology and microelectronics makes it possible to create an optical-electronic neurocomputer in the future. Everything just comes together at the University of Twente.”

More energy efficient 

Follow-up research by Nijhuis and his team focuses on scaling up to millions of molecular switches in a complete artificial neural network to prove it can be done energy-efficiently.

“Outside our bodies, we will probably never reach the energy efficiency of our brain, but if we manage to make devices 10 to 100 times more energy-efficient, it would have a major impact,” he says.

He dreams of making computers that are much more energy-efficient in a decade or so. “That would solve the energy problems in the datacentre sector,” says Nijhuis.  

Another big wish is to apply his invention to soft robots and healthcare. “Since we work with molecular materials, it is attractive to incorporate that in soft robots or applications in the body,” he says.

Nijhuis is also associated with the EBRAIN-NL consortium, which aims to develop implants to stimulate the visual cortex.

“The Brain Institute in Amsterdam, for example, has already implanted a silicon disc in a patient that stimulates nerve cells and allows digital information to be seen by a blind person,” he says. “As you can imagine, silicon is quite invasive to a body, so we prefer to use soft bio-compatible materials. For such applications, molecular materials are ideally suited.”

Innovation at your doorstep 

When it’s possible to process information 10 to 100 times more energy-efficiently, self-driving cars suddenly become more within reach, too, says Nijhuis. The computer and battery in an electric vehicle would provide 30% less range if such a car were self-driving.

“Can you imagine what we can achieve when such a computer system consumes much less energy,” he says.

The number of possible applications is abundant, according to the professor. “It has added value for any situation where battery life is important,” says Nijhuis.

But there are also possible applications if sensors can process collected data locally to some extent, and data does not always have to be sent back and forth over a network.

“Our brain not only stores information but also processes it at the same time,” he says. “Whereas in computers, the processor and memory are still separate. Our invention will benefit any application in which data storage and processing plays a role.”

His ultimate dream is an application in intelligent soft robots that can actively respond to their environment and provide feedback. “If I can help make a device that can give someone a better quality of life, it would mean a lot,” says Nijhuis.