ºÚÁϸ£ÀûÍø

Integrating two of ºÚÁϸ£ÀûÍø’s unique research strengths to build ultrafast, energy-efficient memory chips. That is the dream Bert Koopmans from the Department of Applied Physics and Science Education is realizing together with his colleagues at the Department of Electrical Engineering. A story on how decades of developments in integrated photonics and switching thin films of magnetic materials are now culminating in energy-efficient solutions for fast, broadband data storage.

‘When I arrived here in the late nineties, Eindhoven was worldwide recognized for its expertise in integrated photonics and in engineering thin magnetic layers. I added light manipulation to that mix,’ Koopmans starts telling the tale of how a shared vision is now turned into reality. ‘After decades of innovations in both fields, we are now at a stage where we can integrate what is called spintronics and photonics on a single chip. The aim is to build an ultrafast, energy-efficient, integrated magneto-photonic memory, in which data is directly copied back and forth between the photonic and magnetic domain without the need for any intermediate electronic steps. These devices can be operated at higher frequencies and speeds than their electronic counterparts.’

Even though photonics offers interesting possibilities for fast and energy-efficient data transmission, it is not well-suited for building memories. Magnetism, on the other hand, is an established technology to base non-volatile, long term data storage on. Where traditional magnetic hard drives are rather slow due to their moving mechanical parts, spintronics – which uses spin, the natural magnetic property of electrons, to store and process data – is significantly faster.

Starting from fundamental discoveries 

The road toward spintronic-photonic integration began around 2012, when several groups worldwide had demonstrated the possibility to quickly and efficiently switch electron spins with extremely short flashes of light. ‘However, this was done in alloys that were less relevant for spintronic memory chips. In Eindhoven we had a strong expertise in ultrafast laser-induced magnetization dynamics in layered materials, which are better suited for spintronic applications. We started to explore whether these materials might also host the optical switching process, Koopmans says. 

Around that time, the Eindhoven departments of Applied Physics and of Electrical Engineering came together to discuss possible collaborations. ‘We proposed the wild idea to use optical switching to add a memory functionality to photonics. Back then, it was hard to convince people of the feasibility of this idea,’ Koopmans remembers with a smile. ‘Together, we had managed to obtain a large-scale, ten-year NWO Gravitation grant to build a Center for Nanophotonics. This grant enabled us to explore our idea with two successive generations of PhD students. Where the first generation of students was mainly looking for suitable switchable materials, the second generation could move toward the challenge of doing this on-chip.’

By 2017, the researchers achieved a series of important breakthroughs when studying so called racetrack memories, in which the electron spins are pushed along a track by an electrical current. Koopmans: ‘We demonstrated that there are specific layered materials that display the ultrafast optical switching, in which this optically-written train of tiny magnets can reach speeds of several kilometers per second, leading to a racetrack that can reach data transfer rates in the order of 50 GHz.’

Crossing lanes

Ultimately, the team came up with a chip design consisting of a photonic chip covered by a sandwich of ultrathin magnetic layers. It looks like a highway interchange, where in the overpass electron spins cross light pulses racing along underneath.  

‘Worldwide, we are the first to integrate these technologies on a single chip in an attempt to make a memory application,’ Koopmans says with pride. The first results look promising. ‘Over the past ten years, we have delivered three crucial proofs of concept. We can optically switch layered magnetic structures, we can create a fast-propelling magnetic racetrack, and we recently demonstrated that we can switch a magnetic bit on top of a waveguide, using polarized light. Not only have we demonstrated the three crucial functions that are required to realize our vision, we also did so with promising parameters in terms of speed, storage capacity, and energy consumption.’ 

No delays 

Integrating different technologies on a single chip can significantly advance future computing. If you can directly store data on a photonics-based chip, you don’t need  separate units for memory and for processing, and thus no delaying data communication between the two, Koopmans explains. This is interesting for buffering large, fast data streams in an energy-efficient way. ‘I don’t think this type of technology will end up in applications for large-scale storage, but rather for fast buffering of incoming broadband data in a data center, or for other applications where speed and energy-efficiency are the decisive factors.’ 

This development is also relevant for new applications such as neuromorphic computing, he adds. ‘The storage process in our chips closely resembles how firing synapses work in the brain: you need a certain amount of pulses arriving simultaneously to write a bit.’ 

Upscaling

Now that the team has demonstrated its ability to read and write one bit on one waveguide, the next step is to scale up. ‘Currently, two PhD students are working on this topic, reading and writing multiple bits on multiple waveguides. Even though this technology is still at rather low Technology Readiness Levels, companies like Lionix and are very interested in it. ’ The momentum in this field has also created opportunities beyond academia, he says. ‘One of our former postdocs has co-founded , a startup developing non-volatile working memory hardware for AI and next-generation computing systems.’  

If anything, the proposed photonic memory is an excellent example of how long-term multidisciplinary collaborations can lead to radical innovations, Koopmans states. ‘This is a great collaboration between different research capabilities in Eindhoven. These were brought together in the Gravitation program, which formed the basis for the later Institute for Photonic Integration. Ultimately, that merged into the Casimir Institute, through which we can now further expand this initiative toward wider ecosystems. Our work shows how something that started as a wild idea can develop toward actual applications with societal impact, while raising new fundamental research questions at every step along the road.’