At Eindhoven University of Technology, researchers are working on the next generation of computer chips that enable the future digital revolution. Their solution to handling the ever increasing amount of data lies in combining electronics and photonics.
From your morning coffee to your binge-watching series at night: our current society is built on computer chips. Over the past few decades, the electronics in these chips have become increasingly cheaper, faster, and smaller. 'But that development is reaching its limits,' states Peter Baltus , Professor of Electrical Engineering in TU/e's Integrated Circuits group. 'And yet, new applications still require further evolution. We will no longer get the required improvements from electronics alone. Photonics, which uses light instead of electrons to process information, thus enabling high bandwidths and fast processing, is a young and promising field, but it is unlikely that technology alone will be able to solve everything. That is why I expect that the coming decades we will witness applications that combine both technologies.'
Electronic and photonic factories of today vastly differ from each other. One simple aspect is that the wafers have a different size and are made of different materials. How do you bridge that difference?
Prof. Peter Baltus (TU/e)
Organic move
Over the past years, photonics has already slowly but surely been moving closer to electronics, says Martijn Heck , full professor at TU/e's Photonic Integration group. 'Look for instance at electronic chips that are surrounded by optical transceivers, used in applications like datacenter switches. On the other hand, state-of-the-art optical modulators and detectors increasingly need to be closely interfaced with fast driver electronics and digital signal processing.' The problem is to connect the two in such a way that the ever-increasing amount of data can be generated, processed and transmitted at the right speeds, and with an energy consumption pattern that is sustainable in the long term.
What's more, it is not clear yet what business models would be viable for such a combined technology, Baltus emphasizes. 'Electronic and photonic factories of today vastly differ from each other. One simple aspect is that the wafers have a different size and are made of different materials. How do you bridge that difference? Will a single company harbor both an electronic and a photonic production line, or will two different companies each provide one part of the chip? And in the latter case, who will take care of the integration?'
Collaboration is key
Baltus, who has a thorough background in high frequency electronics himself, is thrilled to be expanding new horizons, he says. 'In electronics, we have entered a reasonably predictable phase, where there is still plenty to improve and research, but the major developments are history already. This hybrid technology is terra incognita again, which is a very exciting topic to work on.'
TU/e is rather uniquely positioned to be working on the challenges of combining electronics and photonics. There are several reasons for that. First of all, TU/e belongs to the international top in the field of semiconductor research and is well-known for its expertise in chip design. And for decades now, the Eindhoven university has been leading in indium phosphide photonics, also integrating the photonics material on silicon wafers. 'For me, what stands out is the fact that we are exceptionally strong in collaboration - both with industry and with each other, 'Baltus states. 'When you work on the integration of multiple technologies, you need to be able to look beyond borders, both in terms of disciplines and institutions.'
When designing such a heterogeneously integrated system, the key is to not put the best possible electronics solution together with the optimal photonics design, but to optimize the combined technology as a whole.
Prof. Martijn Heck (TU/e)
Tackling the basics together
Heck agrees and adds: 'When designing such a heterogeneously integrated system, the key is to not put the best possible electronics solution together with the optimal photonics design, but to optimize the combined technology as a whole. To this end, you need to put people from both disciplines together at the drawing board, and start from the basic questions: what is your device supposed to do, and which design can you come up with that maximizes bandwidth, minimizes use of energy, is compatible with mature semiconductor technology nodes, and minimizes interference, crosstalk and electronic parasitic effects?'
And the design process goes beyond the chip itself alone, Baltus emphasizes. 'We also need to come up with new methods and procedures. How do you design the architecture of such a complex system? How can you produce the system in such a way that the process steps required to make photonic components do not ruin the performance of the electronic parts and vice versa? How can you check a photonic and an electronic function at the same time?'
Variety of projects
At the moment, TU/e researchers are working on a number of projects to tackle these and other open questions. Heck: 'One example is the HiCONNECTS project, in which the TU/e groups of Photonic Integration, Electro-Optical Communication and Integrated Circuits collaborate to realize highly efficient, ultra-fast optical transmitters on an indium phosphide membrane platform, that are suitable to be heterogeneously integrated with driving electronics.'
Baltus also names a few examples. 'We are trying to make transistors in indium phosphide, the same material that our photonics circuits are based on. The goal is to make transistors that are both faster and easier to integrate with photonics than the mainstream silicon devices. In another project, we are developing ways of transferring photonic chips on top of electronic chips, looking into applications like sensing and neuromorphic computing.'
I am convinced that we are on the cusp of a major transition, like when transistors replaced radio tubes or when LEDs overtook light bulbs. For scientists like us, these are golden times.
Prof. Peter Baltus (TU/e)
Clock is ticking
Even though there is no market pull yet, both men agree that now is the time to work on the heterogenous integration of electronics and photonics. Heck: 'These types of disruptive technological innovations usually take at least ten years, and the clock is ticking if we want to keep meeting the ever growing need for data without consuming all of the world's available energy along the way.' Baltus concludes: 'We need to create realistic expectations here. I think that in ten to fifteen years from now, there will be multiple components on the market that combine both technologies for a variety of applications. But it will be a gradual process, and photonics might never completely replace electronics. There will always be things that electronics are better at. Having said that though: I am convinced that we are on the cusp of a major transition, like when transistors replaced radio tubes or when LEDs overtook light bulbs. For scientists like us, these are golden times.'
