TU/e plans ‘hybrid chips’ to make the most advanced square centimeter even more efficient

Even though our smartphones have more speed and memory with the release of each new model, we don't have to charge the battery more often. The energy consumption of those compact computers in our pockets remains limited, despite the increasing computing power of electronic chips. Yet the limit is coming into sight, especially with the gigantic amounts of data we increasingly send and receive in society. Photonics – which involves sending signals using light – may be the solution. But how can photonic chips be combined with today's electronic chips? At TU/e, we are pioneering so-called 'hybrid chips' because we have both the right ecosystem and areas of expertise to facilitate this key future innovation.

Photonic chips, which operate using light instead of electricity, already exist in various technologies. One is based on silicon, which is the material that forms the basis of today's semiconductor industry, so theoretically, with existing capabilities, these photonic chips could be mass-produced soon.

"Silicon is a kind of like baked sand," simplifies full Professor Peter Baltus. "That’s convenient, because the beaches are full of it. And it is an easily manipulated and controlled material with special properties. It can carry or conduct a little bit of current through it, but it can also stop current. So, silicon can act as an insulator at the same time. Half conduct and half insulate, hence the term semiconductors."

New technology in chip production

However, an essential aspect of photonic chips is that instead of current (electricity), its light that passes through them. "So, then you are less dependent on semiconductors, but rather mainly on glass fibers. Many people know about glass fibers when it comes to our Internet connections, but in chip production it is still an emerging technology. And we cannot yet replace all the electronic components on current chips to allow light to pass through them," Baltus says as he outlines the problem.

Silicon, however, still leaves something to be desired regarding its use in photonic chips. For example, you cannot make lasers or light sources in conjunction with silicon.

Indium phosphide, on the other hand, is a material well-suited for use in photonic chips with lasers. However, it is still incompatible with other essential components in chip production. "The possibilities for connection between electronics and photonics are still too limited," Baltus summarizes simply. "The ideal is to eventually make everything on one chip that has electronic and photonic components on it. But for now, we mainly need to see what the options are to combine different technologies and figure out where we put our money.”

In the coming years, TU/e will expand its key position in Dutch academia in the field of semiconductors with its Future Chips flagship. With the initiative, the university is giving a boost to the development of chip technology to accelerate the creation of solutions to the major societal, economic and geopolitical challenges in this field. 

Gigantic amount of data

In the chip industry, then, photonics cannot do without electronics to enable the jump in scale that the rapid digitization of our modern lives demands. But why do electronics need light on chips? Baltus' scientific colleague Martijn Heck phrases the problem in another way: "It has to do with the gigantic amount of data that we are constantly sending back and forth, from devices through data centers to other devices. Especially now that the development of AI is growing exponentially."

"All those devices and data centers work with chips, and on them, data is transmitted electronically through metal pins. But nowadays there is so much data that it hardly fits through the pins anymore. At some point, you must replace those pins with light channels. And that means those chips also must contain photonic components. Not so much to do complex calculations, but to carry the data to and from the chip."

Another challenge is that the ever-smaller chip components, such as transistors, are now approaching the size of an atom. "One current electronic chip hypothetically contains as many as 160 million transistor radios," Baltus says. The manufacturing methods for making such components on an increasingly smaller scale are the subject of research at TU/e. Nevertheless, this limitation plays into the need for a new type of chip that can keep up with the growth of our data traffic demands.

In short, there is no one-size-fits-all solution for all the major digital challenges of the coming decades. The worlds of chip design, processes and materials, and devices (equipment) must converge.

"This is the era of heterogeneous integration," Heck articulates. Here, heterogenous integration is about the process of including computing components on one chip when the components have been made using different methods and operate in different ways. But how can this be done?

"That's the open research question,” says Heck. “The chip is already the most advanced technology ever, and now we need to move to a heterogeneous version, which is even more complex. On that square centimeter, the whole world is currently working together."

Printing technology for heterogeneous chips

Within that worldwide collaboration, TU/e has a special leading position. At least with the INSPIRE project, a European project under Heck's coordination in which a scalable, robust, and cost-efficient printing technique is being developed for such a heterogeneous chip.

"You can combine chips in different ways. The easiest is to put two different chips side by side and literally bridge them with small wires. Or you can put them on top of each other with small solder contacts, so you get vertical electrical connections. In INSPIRE, we also put chips on top of each other, but to get vertical optical connections. For that, you need very precise equipment."

Heck emphasizes that such heterogeneous photonic-electronic chips are not yet intended for the applications and needs that we face in everyday life. "Think mainly about communication in data centers, for example, where all those servers are connected to each other with optical fiber optics. Or sensors that measure the integrity of parts in airplane wings, bridges, or roads so we know if something is wrong there. That also goes through fiber optics, so it's important to have the right chips for that."

Tech push and not application pull

So, the huge array of sensors for all kinds of applications, such as for the automotive industry, the Internet of Things, and for wearables, is leading to a great need to combine different technologies on chips.

"By the way, this demand is not always from the market," Heck argues. "This is one of the few technologies where we may talk about tech push rather than an application pull. History shows that we will need this development in the future. And it takes 10 to 15 years to develop, so that's where the urgency is. Applications will come naturally. We don't have to worry about that."

Pioneering in all respects is being done and needed then. This also gives TU/e an edge as part of the high-tech hotspot Brainport, an ecosystem with major parties in the semiconductor industry such as ASML, ASM and NXP. Or, as Baltus describes it, "a nice environment where everyone is open to those crazy ideas of ours."

“The university is a great place to try out risky possibilities,” Baltus adds. Students and PhD researchers just do it; they are not concerned with 'why not.' And the companies around us, in turn, are willing to work with those solutions."

"What we generate here in terms of innovations thanks to this intensive cooperation between companies and institutes is quite unique. We grant each other something; and more than that, we enjoy innovating together. That has certainly contributed to our success in chip development. The excellent cooperation between different research groups within TU/e is also not a given. It's evidently different in other scientific environments. We walk in and out of cleanrooms with each other, so to speak."

Speaking of cleanrooms; that's another area where we as a university aim high, says Heck. "We have a world-leading cleanroom for producing indium phosphide chips – the NanoLabTUe, which is academically in the international top three."

In addition, TU/e has a strong track record of out-of-university startups such as ProDrive, Smart Photonics, Effect Photonics, MicroAlign, MantiSpectra, and ANTENNEX. All those forces combined make this ecosystem the ideal place to accelerate the production of heterogeneous computer chips.

This article was first published 27 May 2024 by TU/e.