Hybrid computers made of nerves and fibre optics

A long-cherished dream in medical technology is to control prostheses using only the power of thought – as is common in science fiction films –. A bionic arm, for example, surgically fitted after an accident, should feel like a natural part of one’s own body. And function just as well as the old arm. Or even better. Star Wars hero Luke Skywalker sends his regards.

Initial attempts at this already exist, but signal translation is still often problematic and movements remain limited. It is very difficult to replicate the complex natural exchange of information between the brain and muscles on a one-to-one basis. The crucial component for this is the so-called brain-computer interface – a microchip connected to the nervous system. It translates brain impulses into electronic control signals for the prosthesis and, conversely, the prosthesis’s tactile signals into sensations.

Out of the box

Wolfram Pernice, Professor of Experimental Physics at Heidelberg University, is working on such chips, which could be used not only in medicine but also in computer technology more generally.

Pernice's approach is entirely new, which is why the Volkswagen Foundation is supporting him through its Momentum funding initiative: this supports researchers in projects that show great promise but take an unconventional approach. They are venturing into uncharted scientific territory with an uncertain outcome – and therefore receive less attention from other funding organisations.

That, says Pernice, is precisely what makes basic research and a programme like Momentum so exciting: "Here, I am allowed and able to really think ‘outside the box’ and try out outlandish ideas." In fact, he adds, it is usually the case that nothing works at first. "But if you stick with it, you sometimes find a completely new solution to an important problem."

Light instead of electrons

The physicist is researching microchips modelled on the brain’s neural networks, but which use optical signals instead of electrical ones. Fibre optics instead of copper cables, so to speak. As with the internet, this would also offer various advantages for brain-computer interfaces: "Fibre optics," says Pernice, "can transmit data much faster; the material is inexpensive, flexible, easy to install, durable and – which is particularly important for implants – biocompatible."

Together with researchers at the University of Oxford in England, where Pernice once completed his PhD, his team developed a thumbnail-sized microchip years ago that simulates four neurons and 60 synapses. Such chips are known as ‘neuromorphic photonic systems’. The light circulates within a special waveguide geometry made of semiconductor materials such as silicon nitride or silicon. These function like optical fibre, but allow for even smaller structures on the chip. In this way, the light triggers switching processes in phase-change materials, similar to those on DVDs: they switch between transparent and opaque using a laser.

Progress in this field is urgently needed. This is because digitalisation and the AI boom are driving up our electricity demand. Data centres already consume 1.5 per cent of global electricity generation and, according to estimates, will double their demand by 2030. "Without greater energy efficiency, we will reach the limits of our capacity expansion, particularly due to the rapid spread of artificial intelligence," warns Pernice. Optical chips could help here: initial laboratory tests show power consumption per computational operation that is ten to a thousand times lower.

Computing with real nerve cells

But that is not yet enough for the interaction between brain and machine. An optical chip can hardly mimic the three-dimensional, highly interconnected neural tissue of our brain. The chips are constructed on one or more levels. However, a cubic millimetre of brain tissue contains around 100,000 nerve cells, which criss-cross the space and are each connected via synapses to hundreds or even thousands of other nerve cells – an unmanageable tangle of connections that nevertheless processes data very efficiently. "I don’t see how we could replicate something like that in the foreseeable future to capture brain signals one-to-one," says Pernice. "Our somewhat bold idea was therefore: could we instead grow real neurons on our chips and stimulate them with light signals? Then their natural 3D architecture could be integrated into our artificial chips."

Pernice knows Klingauf, as he himself conducted research at the University of Münster from 2015 to 2023. He was then appointed to the Kirchhoff Institute for Physics at the University of Heidelberg. The project, which began in 2019, is now continuing at both locations: "The research is mainly carried out in Heidelberg. We build the chips in Münster."

The vision is to develop hybrid computing systems that combine the advantages of both worlds – the biological and the technical: the highly interconnected 3D architecture of brain cells with quantum-photonics data processing at the speed of light.

Such systems could not only make prosthetics more flexible and AI more energy-efficient. They could also help quantum computers achieve a breakthrough and facilitate data encryption. It may even be possible to use them to bypass damaged nerve tissue, to help people with spinal cord injuries get back on their feet.

Pernice stresses, however, that there is still a very long way to go. Nor can it be ruled out that chips containing living neurons might not ultimately work as well as hoped. But even then, the effort would not have been in vain: "We are learning so many valuable things from this project that will help us in other areas too, so it is definitely worth it." That, too, is part of the nature of basic research: "Even when straying from the path, you may stumble upon unknown effects that could lead to entirely new technologies."