Electronic chips are hitting efficiency limits due to heat generation. Photonics offers a more efficient and faster alternative. The Dutch ecosystem has a strong starting position to break its dependence on established players.
End-to-end optical data processing still lacks a good optical memory.
The current memory for photons is very volatile; DUAL3M is investigating the use of new materials for stable, compact, non-volatile, and rewritable photonic storage.
TU Eindhoven and Motion Imager are joining forces to research this new photonic memory and bring it closer to applications. This could lead to a breakthrough in end-to-end optical data processing.
Energy consumption and heat generation in the current generation of computer systems are increasingly constituting a societal problem. For example, a data centre consumes as much electricity as a medium- to large-sized city. Some hyperscalers also require up to 20 million litres of water per day to dissipate the heat generated by server chips. That development cannot continue like this for much longer.
‘Breakthroughs come when people become desperate’, is the rule of thumb of Peter Baltus, professor of Integrated Circuits at TU Eindhoven. He has a long career at companies including Philips and NXP. "When existing technology no longer provides a solution to a problem, and the costs and risks of further stretching the existing capabilities become unacceptable, then there is room for alternatives," he observes. It is clear that the growth curve of conventional transistor-based computer technology is finite. "We're not hitting a wall, but rather a soft trampoline." You have to push harder and harder to make progress.
"Breakthroughs come when people become desperate."
Peter Baltus, Professor Integrated Circuits, TU Eindhoven
Continuously improving computer technology is a must
Information processing and communication using photons do not suffer as much from the high energy consumption and significant heat generation that occur with electronic data processing. Photons, unlike electrons, have no mass and do not cause electrical resistance. Information transfer can therefore be up to a thousand times faster than with electrons. Everyone who has connected their computer to a fibre optic network knows those advantages. Fibre-optic communication is already commonplace, with more than half of households in the Netherlands having an active connection. Communication over long distances has used fibre-optic cables for decades.
Even in data communication over ultra-short distances – that is, between components in a server – photonics already offers advantages by reducing energy consumption and heat generation. Even performing calculations with photons (optical computing) is on the rise. When end-to-end optical data processing becomes possible, information no longer needs to be converted between photons and electrons, resulting in increased speed and reduced energy consumption. The Dutch ecosystem around photonics, where knowledge organisations and companies collaborate closely, also offers opportunities to develop the next generation of computer technology in Europe and reduce dependence on non-European providers.
Apu Saha, CEO and founder of Motion Imager, wants to leverage the benefits of end-to-end optical data processing. The photonic memory could complement the capabilities of the company's new unconventional computer technology. In current computers, there are always noticeable delays and accuracy issues that can have disastrous consequences. Saha gives an example from the manufacturing industry in which a surface is machined with a milling cutter operating at 30,000 revolutions per minute. In that machine, real-time decisions must be made about slowing down, stopping, or changing the milling head. "A minimal deviation of, for example, 0.001 mm can significantly increase the risk of breakage," Saha explains. For real-time analysis in this situation, significant computing power is required. End-to-end optical data processing can ultimately deliver the required computing speed, yielding digital models that more accurately represent reality, while consuming less energy and generating minimal heat.
Saha mentions other examples of processes that require fast, energy-efficient, and accurate emulation. For example, automated welding systems in industrial production processes must respond to many variables changing simultaneously. "You cannot properly simulate the physical variables of a process without real-time verification and validation, which is why many errors still occur in the initial phase."
Need for fast, energy-efficient computing power
In classical electronics, bits are stored as charge in capacitors or as transistor states. That is a reliable and stable way to store information. The downside is that reading and writing information consume a lot of energy and generate heat. "It's not just the interface between photons and electrons: electronic storage is simply slow," says Baltus, who is also a member of the advisory board of Motion Imager, Saha's company. "You want essential functions, such as memory, to be as fast as the other components of the computing process." Otherwise, you have, so to speak, a beautiful Formula 1 race car with the engine of a city car."
Optical storage is much more complicated than electronic storage because you cannot 'stop' photons like the electrons in a capacitor or memory cell. It is indeed possible to make photons circulate in an optical resonator. That is simply put, a space where light bounces back and forth between mirrors or other reflective materials for some time. A special variant of this is a resonator in the form of a closed ring, where photons circulate until a process occurs at specific wavelengths. Such a ring resonator can even be integrated onto a chip.
Optical storage is still too volatile
A major disadvantage of this form of optical storage is its volatile nature. A ring resonator can hold information for at most a few nanoseconds because there is loss when the photons circulate. Researchers are therefore exploring ways to store optical information for longer periods. In the DUAL3M project, among other things, the development of new materials is being examined in which a state change can be induced with photons and then reverted. Saha: "This is really not just classic CMOS silicon memory technology."
TU Eindhoven is a leader in research on optical technology. Weiming Yao, assistant professor at the Photonic Integration research group at TU Eindhoven, has been working for years on generating, detecting, and manipulating optical signals on a chip. During that time, the foundation for photonic chip technology was laid, leading to new ideas for developing optical memory. "We want to store information in material states that we can manipulate with photons. We hope to create a system that can reliably perform billions of write/read actions.”
Saha has extensive knowledge and expertise at his company in the development and construction of devices and chips. "We are a full-stack company and create our own hardware. A memory module like this is on our list of intellectual property that we would like to develop in-house," he explains as the reason for his enthusiasm for this DUAL3M project. Saha and his team also have extensive experience in materials development.
"We want to store information in material states that we can manipulate with photons."
Weiming Yao, assistant professor at the Photonic Integration research group, TU Eindhoven
Project is in the initial phase
In the coming period, it will be about smartly combining photonic and electronic computer technology. Baltus: "Photonic computer technology is developing rapidly, but there is a significant gap in maturity between photonics and electronics. We still have a few steps to take before you have a photonic alternative to a ten-euro SD card that can hold 120 GB." Baltus compares the situation to the transition in the automotive industry from combustion engines to electric motors. "Cars don't really become more fuel-efficient at a certain point. Then the development of the electric car follows, and another step can be taken. There are initially disadvantages to that, so there are now hybrid cars as an interim solution."
Research like this is important because it serves as a link between materials research and research into new computer technology, Yao adds. "We aim to develop a demonstrator that can be integrated into a computer platform." This way, the gap between research and application can be quickly bridged. This creates advantages for companies that are the first to seize this opportunity and opportunities for the ecosystem in this region.
Societal and economic impact can be significant
Disruptive transitions typically do not originate from established players. They are more focused on refining the technology that has enabled their success, so that it remains relevant for as long as possible. That offers opportunities for start-ups or spin-offs to specialise in a new approach without the baggage of the past.
Often, various developments emerge in parallel. Which technology will emerge as the winner is not predictable in the early stages. In the Netherlands, ASML became a major player after the company spun off from Philips. A spin-off from TU Delft later developed a promising alternative lithography approach, but failed to scale it up and thus achieve a competitive position. This project contributes to the realisation of a new computing paradigm capable of handling new types of data processing that current computer technology struggles with. The joint team of TU Eindhoven and Motion Imager has access to high-quality expertise, advanced tools, production facilities, and the photonic ecosystem for research, creating a very favourable starting position. Saha: "But caution in predicting the unknown is important."
Baltus believes that the Netherlands is already doing a lot to bring promising initiatives to fruition, for example, with PPP project financing from Holland High Tech. "If you invest in several promising developments, the chance that a few will be successful is quite high. Success does depend on many variables over which you have no control, and which you sometimes don't even know." In the Netherlands, the startup phase is going well, but many companies get stuck in the growth phase. "The government could consider whether it is possible to better support companies in that phase," Baltus thinks.
Such innovative companies are essential for the development of new ecosystems and technological autonomy and independence. "If we manage to make our plan a success, it will be of enormous strategic importance for the Netherlands and the entire continent," says Saha. "With new proprietary intellectual property in Europe and a good plan to commercialise it, new Key Enabling Technologies will emerge here. As a result, various industries can emerge here, generating spin-offs and laying the foundation for new research and development projects.” He points out that, in the Netherlands, this way a whole supply chain can emerge around processing of new critical raw materials, which can be seen as an IPCEI (Important Project of Common European Interest).
“If we manage to make our plan a success, it is of enormous strategic importance.”
Apu Saha, CEO and founder, Motion Imager
