Lieven Vandersypen | Building Silicon Spin Quantum Processors

Daniel Rodan Legrain: Hello everyone and hello Leven.
Lieven Vandersypen: Hello Daniel.
Daniel Rodan Legrain: ⁓ I'm Daniel Rudan Legrein. I'm the Quantum Science and Engineering Lead for North America at QBlocks, the sponsor of today's And this Fireside Chat basically brings together voices shaping the quantum landscape from fundamental physics to scalable hardware and global ecosystems. We've previously hosted Will Oliver, Berta Young, Peter Shore, Steve Girvan, Gretchen Campbell, Black, David Ashallon, Karzhenyak, among others. And today it's a real pleasure to welcome one of the pioneers of semiconductor spin qubits and one of the key figures behind the quantum Delft ⁓ ecosystem, Professor ⁓ Lieber van der Seypen.
Lieven Vandersypen: Thanks.
Daniel Rodan Legrain: Leven is a professor at EU Delft and the director of quantum research at Q-Tech. His work, as many of you may know, spans some of the earliest demonstration of quantum algorithms during his PhD at Stanford and IBM Almaden, ⁓ all the way to pioneering the control of think electron spins and building scalable semiconductor quantum processors today. We're also speaking at a particularly exciting moment for the field, I would say, at the APS Global Physics Summit. ⁓ Last month, the SpinQubit community generated a remarkable level of momentum, including Leaven's own results. And we'll hear firsthand about that from him today. Leaven, it's a real pleasure to have you with us today.
Lieven Vandersypen: Glad to be here and thanks for all these kind words, Daniel.
Daniel Rodan Legrain: For everyone who is watching live, please feel free to post your questions in the chat and I will bring a few towards the end. So, Livan, why don't we get started with your journey, let's say from the early experiments to building one of the world's leading quantum hardware ⁓ programs. So your work in particular, ⁓ your PhD work, I would say, demonstrated one of the first implementations of quantum algorithms. You factor 15 using nuclear spins. ⁓ What drew you from those early experiments into pursuing semiconductor spin qubits as a long-pick platform?
Lieven Vandersypen: ⁓ Yeah, so indeed my PhD on NMR quantum computing. This was very early days, right? I started in 1997. We did some of the first algorithms. It was really fun actually to experiment with these molecules in liquid solution, many copies in a liquid solution, because we could very early on already test the functioning of quantum algorithms on real hardware. hardware also had very well understood limitations. It was very quick to get going, but also ⁓ difficult to get very far ⁓ in terms of scaling. Imagine how are you going to design a molecule with a thousand nuclear spins that are all addressable and interactive and coherent. And in addition, even worse, perhaps there were well-known ⁓ limitations in the state preparation. ⁓ Yeah, the state preparation methods that we used basically weren't scalable. And so, yeah, at that time around 2000, 2001, I was looking around. I was very enthusiastic about quantum computing. I stumbled upon it at the start of my PhD. I was immediately blown away by the possibilities. I literally couldn't sleep from excitement when I first heard about it. And I definitely wanted to continue working quantum computing. And so I started looking at the literature. early work on trapped ions, on superconducting qubits. And I also had read ⁓ theoretical proposals from Daniel Los and David DiVincenzo, as well as the paper by Bruce Kane. And those works really interested me a lot, especially the Los DiVincenzo proposal, because I thought that if it were possible to build high quality quantum bits based on semiconductor, based on the same technology that's driven the information revolution, of past decades, that's a very powerful starting point. So that's one thing I liked about it. What also drew me to it is that the spin of an electron is the canonical quantum bit. It's a two-level system. ⁓ It's the prototype pictured of a quantum bit. We always use it in textbooks as an example of a two-level system. So that part I liked. And the third aspect, I think, that drew me to semiconductor qubits, this is more like at a personal level. What I liked is that the first qubit had still to be built. Nothing had been done yet. There was a basis, were quantum dots. I had been reading also experimental papers where people demonstrated that electrons can be confined in these small semiconductor nanoscale devices that we call quantum dots, that these can be stable, that you can apply high-frequency electrical drives to them and so on. So there was some foundation, but no qubit work had been done. And that also appealed to me to because if this field was going to go somewhere, which at the time I really had no idea, I would be there from the start. And that also drew me to this approach.
Daniel Rodan Legrain: That's a great answer, Liven. And as you very well said, many qubit plasforms were being ⁓ explored. What made you personally believe that electron spins in quantum dots could actually be a ⁓ scalable path moving forward?
Lieven Vandersypen: Yeah, so I mean, of course, on the one hand, it's the fact that you create them by lithography. we know how to use lithography to create enormously complex structures where many transistors all work together. And similarly, I sort of envisioned, not the only one in the world, that the same could be true for semiconductor qubits. At the same time, when I started working on this back in 2001, I I really had no idea where it was going to go. For all I knew, it could also die within a few years and nobody would talk about it anymore. So I can't claim that I could anticipate where things would go. Although at the same time, it's a bit, I mean, it's sort of a multi-sided story. Also, if I look back at the one, on the one hand, I wasn't sure where it was going to go. On the other hand, I was optimistic. And I definitely thought we were going to go fast and much faster than we actually did as a field. ⁓ But yeah, I think the bottom line is it was early days and it was hard to predict what would come out.
Daniel Rodan Legrain: So if you look back at those times, Liven, what were the key insights or the key moments that made, let's say, spin qubits feel viable beyond proof of principle experiments?
Lieven Vandersypen: Well, ⁓ I don't think we can look back at those early days to say that because then it was really proof of principle experiments. But in more recent years, I really feel like the momentum has picked up. There are now many groups around the world pursuing this and gaining ⁓ control over more and more quantum bits with better and better precision. There are large industry teams that with very focused, well-funded efforts move very fast. ⁓ and it's impressive to see and to help push that. you know, I think that, well, looking back at the history, there are certainly a couple of points where we can identify that there was like a jump, you know, where our field took a jump and, or made a step up and the prospects started to look better. I think one clear one is the transition from the three, five quantum dots to group four quantum dots. So what this means is in the beginning, we all used quantum dots made in gallium arsenide. So these were gallium arsenide, aluminum gallium arsenide, heterostructures, a very well controlled, clean system that had been the workhorse of much of physics at the time, around 2001. And that was a natural way to build the first quantum dots, and we did. But the difficulty is that every single arsenic or gallium atom in the universe carries a nuclear spin. And even though we operate our devices at low temperature and in a large magnetic field, the probability of finding the nuclear spin up or down is still about 50-50. They're pointing in random directions, basically. And so now if your electron wave function is in contact with, let's say, a million nuclear spins from the substrate, on average, equally many point up or down. But there is always some statistical fluctuation around the average. And that led to very fast dephasing of the electron spin. Basically within a few tens of nanoseconds, if you put the spin in a superposition, it would rotate in the block sphere to a position that you no longer know. And that's of course the essence of decoherence. Now, it was known from the beginning that if you could make quantum dots in silicon, in germanium, in carbon, that there would be very few nuclear spins and that things should get a lot better. But for many years, it's been a painful road to fabricate devices that were sufficiently stable to be able to perform qubit measurements. We also did some measurements together in collaboration with Mark Erickson from the University of Wisconsin. And the first generations of devices, it was just frustration because by the time that you had to tune up all the gate voltages just right to start experiments, something in the background would shift and you had to start over pretty much. And so that changed because of improvements in the materials from Mark, also from the team at HRL, where we understood how to engineer these heterostructures ⁓ in a way that the devices became stable. There had been parallel developments at UNSW in Sydney where people had worked hard on silicon-silicon oxide quantum dots and started to get results. And so, yeah, about 10 years into, let's say around early 2010s, as a community, we started to see first results in silicon and immediately ⁓ everything looked a lot better. If you would look at the Rabi oscillation in Gallia Mars night, we were super proud to see the first time a single spin be rotated and we could see several periods of oscillation before it would damp and we were very excited. But it was nowhere near good enough to be the basis for the technology. And in silicon, if you look at the early Rabi oscillations, from what is it, 2013 or around there, 14, from Mendel Welthorst when he was a postdoc with Andrew Zurak in Sydney. The oscillations just keep going. There's basically barely any damping. And this was enabled not only by silicon, but even a special form of silicon, where the remaining 5 % or so nuclear spins silicon-29 atoms in the silicon substrate that is naturally present. where that 5 % was largely eliminated. So this is a process that we call isotopic enrichment or purification, where you keep the ⁓ non-magnetic spinless atoms and eliminate all of the atoms that carry, or most of the atoms that carry nuclear spin. And so in this way, Rabi oscillations were very long lived. T2 star went up by four orders of magnitude, going from...
Daniel Rodan Legrain: Wow.
Lieven Vandersypen: From gallium arsenide to silicon was a factor of 100. From silicon to isotopically purified silicon, another factor of 100. And so that was really one big boost in our confidence. I mean, now so many years after the facts, maybe I can reveal that there was a time where I started thinking, what shall I do after I stop working on spin qubits because it no longer looks attractive? And that was before Silicon showed, or before as a community, we showed that Silicon qubits can work. And so seeing them work really ⁓ boosted my motivation and confidence and made me go all in and really wanting to make this happen. And so that's what I did. So that was definitely one moment. ⁓ And now, you know, what... What I mentioned earlier is that one of the promises of this technology is that you can use industrial manufacturing, right? And we can talk about it in more detail later. But basically that promise in the last 10 years has also started to materialize, notably by the team at Intel, also IMEC, ⁓ HRL. I mean, if you really put industrial methods behind the manufacturing of these devices, you can get very far very quickly. quickly. As in, you need to build up the momentum and the team and the expertise and so on, and then things start to move really fast. And so that is a second transition that we're in the middle of, where the number of qubits, I think, is certainly number of quantum dots is rapidly exploding. ⁓ instead of being limited by device fabrication, ⁓ we're now entering a new phase where we're pulled forward, ⁓ propelled. by advances in device fabrication.
Daniel Rodan Legrain: that's a fascinating story, Livan, and I'm happy that the field got these inflection moments and that your interest only kept increasing during time. So let's now move from proof of concept, early experiments, individual qubits to systems. And let's talk about scaling and what makes spins qubit today. You've already hinted at it. Spins qubits are often highlighted for their small footprint and their compatibility with the semiconductor fabrication. ⁓ What do you see are the strongest advantages compared to other platforms?
Lieven Vandersypen: Yeah, these are definitely two strong ones, right? So to reiterate, they are very compact quantum bits. ⁓ In terms of lithographic dimensions, I think about 10,000 or so spin qubits would fit in the size of a single superconducting qubit. ⁓ Then, yeah, can really very directly leverage all of the ⁓ infrastructure from the semiconductor technology in the semiconductor world to bring in signals and extract signals, so the wiring. ⁓ These quantum bits are also quite robust. Spins are natural two-level quantum systems. ⁓ They are well protected from their environment in general. So spin lifetimes can be very long. And in our field, we distinguish two time scales. Well, many more, but simplifying two time scales. One is you put the spin in the superposition. How long does it take before you lose track of the phase in the superposition? That is a long time scale, but even much longer for these spins is the T1. That is to say, you put it in the excited state. How long does it take to relax to the ground state by giving energy to its environment? And that time scale is really long. Let's call it a second, easily many milliseconds. Let's call it a second. So that's good. But there are other aspects. There are also, for instance, quite robust to temperature. So we typically operate in dilution of the refrigerators. ⁓ base temperature of 1020 millikelvin, but that's not needed. We use it because it's a technology that exists and the colder things are, the quieter the environment should be. But we don't need that. Actually, we've seen as a community that even at, as we call it, very hot temperatures, as warm as a Kelvin, the qubits can be very well behaved. is a little bit of degradation in the performance compared to 10 millikelvin, but not too much. And that actually then brings a lot of advantages. It means that you can use cooling technology where you can have a lot more cooling power. A lot means several orders of magnitude more cooling power. That means that you could imagine some local electronics for routing signals, maybe even for generating or extracting signals very close to the qubit chip. So yeah, it's a technology. It's a type of qubit that has a lot going for it. And that's, of course, why we're working hard and motivated to make it happen. Maybe for completeness, it's important, of course. ⁓ I'm a scientist to acknowledge that it's not only advantages. Every technology has its strengths and also its difficulties. And I also want to be ⁓ straightforward about some of the challenges. ⁓ The footprint, the compact footprint, like I said, it's a major strength. ⁓ But in a way, the footprint is so small that this actually in itself presents a challenge. We call it the wiring challenge. So every qubit needs to have at least one wire going to it for the moment to control it and to read out its state. You could say it needs to be two or three, but it's of the order of one. ⁓ to first order. Okay, so with such a compact footprint, how do you get all the wires in? That's not easy. And so we also have schemes where we imagine that maybe we can separate the qubits a bit from each other, create some breathing room, and we can come back to what it takes to do that. I will already throw in the word shuttling. Let's come back to that later. ⁓ But so that's one challenge. And the second that I would note is... And it has to do again with the compact size in a way. These pink qubits are each, well, each and every one is a little bit different because of microscopic variations between them. We often say qubits have personalities and personalities are wonderful in people, know, everybody's a bit different. That's fantastic. That makes us rich as people. In qubits, it's of course helpful if every qubit if all qubits are really exactly alike. And that's not the case in the semiconductor qubits. They're all slightly different. And we have to learn how to navigate that. Can we improve the homogeneity just well enough? Or do we need to account for the differences by speaking to each in a slightly different way by adjusted individually calibrated signals? So these are some of the open questions that we're gonna have to tackle further.
Daniel Rodan Legrain: Let's talk little bit more about these challenges and in particular about scaling challenges. mean control cross talk calibration. Where do you see the main bottlenecks for large scale spin qubit processors?
Lieven Vandersypen: So far, I would say I don't particularly see it in crosstalk. I think much of the crosstalk is manageable in the sense that there is instance capacitive crosstalk. What does that mean? It means you have two quantum dots, two electrons, and then you have gate electrodes above them. Ideally, you want this gate electrode to talk to that quantum dot, but it unavoidably also talks to the neighbor just because it's close by. The reason I don't worry about it too much is there's several reasons actually. One, it's now pretty routine in our field to compensate for that crosstalk. The crosstalk is local. We construct a crosstalk matrix. It's a very sparse matrix. We invert it and we account for it. So anytime we send a control signal or a gate voltage to one of the quantum dots that hosts the quantum bit, In reality, we sent a linear combination of several voltages applied to several gate electrodes so that the net effect is what we want. Namely, this signal has a net effect here and any effect on the neighbor is automatically accounted for. So that's one reason. Another, as we are seeing in recent experiments also in our group, it can actually be helpful for a variety of reasons to separate the quantum bits a little bit from each other. And then crosstalk is much reduced. We are now doing experiments where we have basically no capacitive crosstalk between the qubits. This being said, there can still be other types of crosstalk, microwave crosstalk. Maybe we could get into it, but let me move on. I'm more worried. In the end, I think I'm the most worried about the wiring challenge. Let's say if I had a magic wand and I could solve the wiring challenge once and for all, really for good, scaling to arbitrary numbers of qubits. That's the... That's what I would pick using my magic wand and having one shot at fixing something. Because then if you can do that, then I think also the qubit personalities can be more easily dealt with and so on. I think that I see as the most serious. There are ideas for overcoming it. And as a community, we have to demonstrate that these ideas really work and come together in a way that we can scale all the way to the many millions that we anticipate will be needed.
Daniel Rodan Legrain: Well, I mean, this brings us directly to the momentum we saw at APS this year, even, at the APS Global Physics Summit, where we met in person for the first time. ⁓ As an attendee, I saw a lot of excitement around the spin-cubit session, know, from HRL, Intel, Dirac, Sego Tarucha, and of course, your own presentation. So maybe for those of ⁓ the audience that were not... ⁓ present there. What would you say made this moment feel different and what did it seem that spin qubits really took a meaningful step forward?
Lieven Vandersypen: Yeah, OK, so I'll answer it in two ways. ⁓ One is I think it was meaningful because there was real exciting progress from several teams. And the other somehow, and I think I want to credit the HRL team for this study as well as colleagues, they did a fantastic job at generating a bus. And as a result, in this massive room where we both were, ⁓ it was packed. And that by itself already creates excitement. But of course, you can draw lots of people and you still have to have something to show, right? Otherwise people walk away disillusioned and like, ⁓ what was this about? Why am I here? And I don't think that was the case. think there was real good progress. ⁓ the progress I think was, I would say there are three flavors that I really liked and which to me maybe summarize also a bit why I feel that it's really good momentum. in our community at the moment and why things are really moving. On the one hand, I think what we saw, especially from the industry teams, is the scaling in the qubit chips. I said earlier, or maybe a bit more explicitly, until a few years ago, five years ago maybe, ⁓ I think many groups' progress was slowed down by challenges in nano fabrication. These quantum dots are compact. They're made with electron beam lithography traditionally. They're not easy to make. takes, yeah, the yield often was low. Every device was a little bit different. But over the past 10 years, ⁓ large industrial teams, Intel, IMEC, HRL have really invested in applying all the tools of the semiconductor world. Also the team at Cialeti have invested all the tools of the semiconductor world to realize quantum chips. And after a fairly long period of building the capabilities of getting the processes right, getting the design right for the basic building blocks, you reach a point where going from 10 qubits to 100 or 1000 or 10,000 at the level of the qubit chip is, let's say, well, at the level of the qubit plane is straightforward because it's just repeating patterns which we can do with lithography. So that's one thing. And we see this happening. We saw this in the talks. Now the second is if you have what people call the front end, right? So this lowest layer where the gates are that control the quantum dots, that form the quantum dots, you could lay out even more quantum dots up to a million, but then we wouldn't know at the moment how to wire them up. But this is where I think there was a second advance from some of the teams in real impressive advances in wiring up these 100, 1000, 10,000, well, 10,000 wasn't yet done, there was more in the planning, qubit chip devices. And this is really, again, leveraging the semiconductor industry technology to have multi-layer interconnects to take signals out. But even then you're not done because then you're sort of at your cupid's chip and you still need to take your signals ⁓ to typically room temperature instruments. And so here, one of the visions in our field is that it may be possible to bring some of that electronics to the cold, into cryogenic conditions. And ⁓ also in our group and with my engineering colleagues, Fabio Sebastiano and so on, we've worked on this already for more than a decade. to explore what it takes to bring some classical electronics close to the chips, make it work in the cold and make it interact with the qubits. And I think here also we've seen real progress in very sophisticated cryo CMOS chips and also in the interface, the wiring that connects a cryogenic chip to the qubit chip, you know, because you can have two individual chips with each capable of either hosting. or controlling a thousand quantum bits, but you still have to connect them. so also there we've seen impressive advances in super compact, low cross-talk superconducting ⁓ cables. ⁓ So that's the second. ⁓ So the first was the manufacturing, the second is on the wiring and the addressing in the cold. And the third ⁓ is that I think once again, we've seen how... creative novel ideas are making us rethink how to scale up. And this is an area where my group has been very active in the last couple of years and that I also spent quite a bit of time speaking about at ⁓ APS meeting. You know, one example is that ⁓ in our field until a few years ago, everybody operated on electrons. They were trapped in static locations that we call quantum dots, you know, and these are remarkable. You can isolate an electron, one single electron in a quantum dot, you can isolate one single electron in the next quantum dot and in the next one over and so on. That in itself I think is very remarkable, right, that this can be done. One single electron if you think about it. ⁓ But that's, and that's a very stable well-known ⁓ condition already. For 25 years we can do this. ⁓ And all this time we've operated on the spins of these electrons while they were trapped in their quantum dots. Now Maybe we can get, this technology all the way doing that. But imagine that additionally, you could take an electron and literally displace it some distance over the chip. You know, what would that bring? And so this is where I think, ⁓ yeah, we've really seen novel ideas being that we tested in our group and that we ⁓ have been experimenting with. And we actually discovered that this really brings many new possibilities. The first motivation was that we could imagine that there are several registers of qubits on the chip kind of laid out in a tile, right? So there is a group of qubits here, maybe 20, maybe a thousand, I don't know, some number, and then another group here, another group here, and in this way you would build this whole tile of little qubit arrays. And then shuttling could be used to connect them to each other in two directions. And in this way you can build up a structure where all of these registers work together as one large processor. Okay, so that was good. And that's still a good application of shuttling. But what we've shown now is that you could also take a register and shuffle electrons within it. And that actually brings another advantage. It means that you have essentially all to all connectivity. Every qubit talks not just to its neighbors, but you can bring it next to any other qubits in the register, or if you want, even in another register. And so bring any two of them together. You could imagine taking a whole array and overlaying it with another array and separating them again, like people are doing with Rydberg atoms. ⁓ So this creates new possibilities. And then even beyond that, what we've discovered is that ⁓ you can also compute on the qubits while they are moving. And not only while they are moving, but you could operate on them controlled by the motion itself. So depending on how fast or how far you displace an electron, ⁓ you could perform one type of rotation on the qubit or another. You could shuttle two qubits towards each other, make them interact and separate them again ⁓ while they remain trapped in these moving quantum dots. So it's literally a quantum dot. in which an electron is confined, that is moving, and it carries the electron along. so, long story short, you know, I think this is another area where I feel there is real momentum in our field. There are other examples, for instance, Menno Viltors, my colleague at Q-Tech here at Theodelt, has shown a novel method of controlling spins by making them hop between two adjacent quantum dots. And in properly engineered structures, the quantization axis in these two quantum dots can be different from each other. So now imagine you start with a spin in the up direction and you make it hop over to its neighboring position. It's still in the up direction, but the rotation axis is this one. So now it's going to perform this rotation and you can carefully time it. And after some time you come back and then maybe you've performed, ⁓ you know, maybe you've performed a 90 degree rotation about the Y axis or maybe you timed it so that you got a 180 degree rotation about the X axis or directly a Hadamard. operation for the experts and so on. So this is just another example and I could give you more. yeah, think these are the three types of advances that I think we've been seeing and that make me optimistic and give me encouragement ⁓ that we're on a good path as a community.
Daniel Rodan Legrain: And those were indeed really impressive results, including of course your own that you fantastically well summarized. Thank you. Since you've mentioned some of your colleagues at Delft, let's move beyond the hardware itself and talk about the Netherlands and Delft in particular. You guys have built something quite unique at an ecosystem level. So from the United States, I can even say it's one of the most prominent quantum ecosystems globally. What makes the Delft model work so well in practice?
Lieven Vandersypen: Yeah, it's a good question. ⁓ What to say? So I've been in Delft for 25 years. And one ⁓ of the aspects that I've really noticed several times is that here in this environment of nanoscience, quantum science, and so on, we have managed to continuously reinvent ourselves. So we don't end up falling asleep. ⁓
Daniel Rodan Legrain: Wow.
Lieven Vandersypen: Let me make that concrete. When I arrived here as a postdoc, working with Leo Karunov in 2001, ⁓ we were part of the department of nanoscience. Then I think within two or three years afterwards, I don't remember the exact timing, very soon afterwards, out of that department, Kees Dekker, Nike Dekker, Sergio LeMay founded a completely new department ⁓ focused on single molecule biophysics and they called it bio nanoscience. Okay. And they were strongly supported by the Kavli Foundation, by the university board and they could build a completely new department. So those of us who were in the nano science department from which that other department had spun off then scratched their heads. Okay. So we are nano science, they are bio nano science. How does this work? You know, or who are we really? And after quite a bit of discussion, we settled on, we are quantum nanoscience. And I think this was a great choice. ⁓ I think quantum nanoscience is now worldwide a very active field. It was, in that sense, a bit visionary. It didn't literally include all the activities that were happening in our department at the time. But I think that's fine. You just need to pick something and go with it. And it gives you a phase, an image, something that you can work from. And so that was one example. Then another 10 years or so later ⁓ came the Q-Tech initiative, right? So Leo Kauwenhoeven here had the vision that ⁓ we could build an institute expanding on the couple of PIs that were working in Delft on quantum technology and bring in new disciplines, bring in computer science, bring in computer architecture, bring in circuit design, bring in material science, bring in mathematics and so on, error correction, many aspects that we were not covering. and envisioned what it could be if we did that. And even more, what would it bring if we could build something that would, in a way, transcend our own science, build around us economic activity companies. People called it an ecosystem. And to be perfectly honest, back in 2012, 13, something like that, when somebody said ecosystem, it meant nothing to me. I could not, to me it felt like a buzzword which was empty and hollow and with no meaning. But now I've experienced it, you know, in a way sort of inadvertently I helped, well, not really inadvertently, but without knowing that this is the word, this is what I also helped build. And it's all these companies that have spun out of QTEC. QTEC by itself has grown a lot, right? So from... maybe 60 or so at the start, we're now at over 300. So we've grown considerably, but the Delft Quantum startup companies have surpassed the size of QTek and FTE already, what is it, two years ago or something with QBlocks that you're working for, of course, as one of the biggest and most successful companies in terms of number of employees. don't know, are you at 150 or 200? I don't know, something like that, right?
Daniel Rodan Legrain: Around that number indeed,
Lieven Vandersypen: And so the total is perhaps 500, 600. I don't know, I lost count. Others are keeping count, but I forget what the number is at the moment. And these companies are growing much more rapidly than Qtek itself. Qtek is growing fast, and then these companies came up and grow much faster. so in this way, our environment has constantly been in flux, in motion, and changing. And I think that has been one of the strong points. that we aren't just happy with what's there, but that we also can imagine what can be and where it's going to go. And ⁓ along with that, I think comes a certain combination of ambition and a desire to do, for instance, research at the world level, at the world leading level, not to be satisfied too quickly. But also an open culture where for instance with these quantum startup companies, there is lots of collaboration. But also within Q-Tech, there is lots of collaboration. Also within the earlier department of quantum nanoscience. think when I... I mean, I'm still part of it. About 10 years ago, I made for myself a little of the inventory. I think we were maybe 18 PIs and I had joint papers with about 12 of them, something like that. I could be off by one or two, but it's something like that.
Daniel Rodan Legrain: That's amazing.
Lieven Vandersypen: And so it shows, I think, the willingness to work together to make the sum of one and one greater than two. And in Q-Tech, we've done that very consciously. We've also stimulated that. was also a scientific director of Q-Tech for four and a half years, and this was a very conscious effort from the leadership to encourage, promote, facilitate collaboration so that we really could take advantage of the opportunities that having this collection of experts here bring that go much beyond what each can do by themselves.
Daniel Rodan Legrain: Let's follow up a little bit on the collaboration part. mean, you've been deeply involved also with collaborations with industry, including ⁓ Intel. So how important would you say is this, know, co-design between academia and industry, especially thinking about scaling quantum hardware?
Lieven Vandersypen: No, I think it's super important. ⁓ The mindset in industry and academia is very different, but at the stage of the field we're at, I think we can really help each other a lot. we're academics primarily, it's not black and white, but primarily bring in novel ideas, test them out, things that nobody's ever thought of before, that nobody's ever tried before. And we're industry teams with a lot of focus, march down a path. And by this... can also go very fast, build up real momentum. And so I think this combination is very strong. And ⁓ in the last couple of years, the collaboration with Intel has not been very active, basically, or stopped, more or less. But for the first six years or so, it was very active. We were talking to each other every week, visiting each other several times a year, back and forth. It was super nice. And I think we learned a lot from working with Intel. And Intel learned a lot from working with us, I think. have to ask them, but I think so. And ⁓ also now, what you see now is a new phase where Intel is not working exclusively with us, but they're sharing devices with many groups around the world. And then these groups all perform a variety of measurements, and that provides ⁓ feedback to Intel. This HRL is doing the same thing. IMEQ is starting to do the same thing. So I think these are really great models where
Daniel Rodan Legrain: I'm sure.
Lieven Vandersypen: There are only a few facilities in the world that can manufacture these devices. It's expensive and it takes a lot of effort to put it together. And then to involve others in the feedback and in the design and in the next generations, I think is very worthwhile. Sometimes IP gets in the way. ⁓ It's not even always the quantum IP, but it can also be classical processes that get very close to, or let's say the quantum technology can, well, and that's of course what we want. The classical technology can get very close to the, sorry, the quantum technology can get very close to the classical technology, to the classical transistor technology. It can leverage that. But if that's the case, then of course, ⁓ know, trade secrets and IP become even more sensitive because now you're talking about an active multi-multi-billion euro industry. So yeah, but so yes, in short, I think that these two worlds stand to gain a lot from each other. And that's what we've been seeing in our field and still continues today. And I understand that the industry teams cannot share the secret sauce of certain things, but overall they are still fairly open about what they're doing. And I think that's ⁓ a healthy sign of both self-confidence and also of ⁓ an understanding that as a community, we can really go faster if we're sharing certain information. And this helps everybody.
Daniel Rodan Legrain: Absolutely. ⁓ Livan, let me now turn to the people that are behind all of this ecosystem that we're talking about. ⁓ And in fact about the whole quantum momentum itself. mean, you've trained many students and researchers that are now leading groups worldwide. ⁓ What do you think are the most important skills for the next generation of quantum scientists and engineers that are maybe thinking about going into quantum?
Lieven Vandersypen: Yeah, okay. Well, so yeah, actually it's true. So many group members have moved, have continued to stay in the field and build their own groups or joined the industry teams. And I can tell you, it gives me really a very, actually a very warm feeling to go to speak with conferences these days because I see so many former group members back and doing so well in, you know, all around the world. ⁓ So, okay, now. coming to your question. ⁓ So I think one of the skills that in this field is definitely super valuable is the ability to really work well together and be willing to learn other languages. Languages I mean not ⁓ French or German or whatever, but I mean ⁓ speak the language of ⁓ fabrication experts, speak the language of software engineers, speak the language of circuit designers and so on and vice versa. building a quantum computer, It's extremely hard. The quantum computer at scale is an extremely complex product system. And in this system, many, many aspects need to come together. And so that makes it a team effort automatically. ⁓ And so the ability to work well together with others, I think is a very important skill. Together with that, or what comes with that is also the ability to communicate well. And yeah, in terms of hard skills, obviously, It helps to have some understanding of ⁓ basic notions of quantum physics, quantum information science, and then pick something that you want to be really good at. And what do you want to really be good at? I would say first and foremost, what you really enjoy. This is what I always advise my own group members. Don't think too strategically like... Is this skill going to be needed five years from now or 10 years from now? I think that there will be room for good people regardless of what specific skills they have acquired. following your passion, your interest and be really good at something, become really good at something, become the world expert at something, I think that is a very valuable process and a very good learning experience that gives you confidence that you can do it again. ⁓ in a different setting. So yeah, that's what I would say. And also trust that your instincts are right, that if you're really interested in something, if you think something is the coolest thing in the world, that you will not be the only person excited about it, that others will also be excited and that somebody will be interested in your skills also.
Daniel Rodan Legrain: And I definitely think your story really illustrates that last principle, Leaven. Let me actually twist the question a little bit. And as we see the field becoming more engineering-driven from an educator point of view, how should education evolve to adapt for the current needs of the market and the science?
Lieven Vandersypen: Well, the first is, and you see it already for several years, that there are now many dedicated programs on quantum science and technology. ⁓ companies, I noticed it also here, companies are really looking for experts. And there are some areas where I think it's nowadays very difficult as a company to attract experts. And I noticed it from quantum error correction to... ⁓ measurement engineers of spin qubits. These are ⁓ professions where it helps to have a certain background. And yeah, let's see now I lost my train of thought. ⁓
Daniel Rodan Legrain: From an education point of view, from TU Delft for example, what are the plans that you guys are thinking to train the next generation?
Lieven Vandersypen: So, right, yeah, so. Yeah, we've, right, so we have started a masters, it's called QIST, Quantum Information Science and Technology. I think we are now doing it for the third time already. So the third cohort has started or is active this year. And yeah, industry is crying for people with this background, but I think they are also crying for more electronics experts or for more mathematicians or material scientists. know, there is a lot of technical expertise. that needs to be brought together in building this system. There is a need in systems engineers. And so, yeah, I think on the one hand, what we need to do is to continue to educate people across all of the traditional disciplines because they will be needed. And in addition, there is this new field of quantum information science and technology and many universities, including TU Delft is now offering programs. I can tell you, if I were a student now, I wouldn't hesitate for a second. I would definitely study that. In fact, what is it 30 years ago when I started my master's at Stanford and then PhD. If you look back at the program I put together for myself by combining a number of electives and so on. It was already a bit this program, of Havanna-Letre. Yeah, I took courses in information science and in computer science and of course in quantum physics, but sort of at the intersection of all these disciplines. And that is also what we offer. I think it's a super interesting program, but of course that's just my personal interest and others may have different interests and they should, I would say, once again, pursue their interests and do what they're really excited about and trust that things will work out.
Daniel Rodan Legrain: That's fantastic advice, Leven. ⁓ So let me actually now try to zoom out a little bit and let's look ahead and let's try to connect all the pieces that we've been discussing today, the physics, the systems, the ecosystem, and let's discuss what ⁓ the future of spin qubits looks like from your point of view. So if you look, for example, five to 10 years ahead. What do you believe would be a successful spin-qubit quantum processor and how would it look like? Both from an architect's point of view and from a system's point of view.
Lieven Vandersypen: Yeah, okay, so maybe we can start with what they will be able to do, right? And I definitely think on that time scale, you'll see many ⁓ companies offering commercial spin-based quantum processors, and I hope they will be out in the field and users will be using them. They may not yet be solving the world's challenges, computational challenges, but they will be out in the field at scale and be used and tested and show themselves under the hood. ⁓ If you ask me, I'm convinced that shuttling is going to be part of the future architecture. I think it's extremely powerful. It relaxes constraints or demands for error correction. ⁓ It relaxes hardware constraints. I think it's very powerful. I'm involved in a startup company. It's called Groove Quantum. It is built around germanium quantum dots and these, I think, have a lot going for them. often still a silicon substrate, but then there's a thin layer of germanium where electrons, in particular in this case holes, so missing electrons, represents the qubits and this has a lot going for it as well. So maybe that will be the workhorse for the future. ⁓ What else can I say? ⁓ Yeah, it's going to be a system. so this is a shift that we're going through as a community that, and that shift hasn't started last year. It started 10 years ago or so, you know, where we began, well, we're together with, for instance, the foundation of QTIC or the collaboration with Intel and so on, right? Where we're realizing that a future quantum computer is not going to be built by a collection of physics, PhD students led by a PI. You need really to think of this quantum computer as a system where everything needs to come together. so this is where I think in the coming years, there will be more and more emphasis in our field to take these systems and generating point of view and really think through how everything can come together. And that's a super hard thing to do because everything is connected with everything, you know, from the choice in the materials and the substrate. to the way that you encode the quantum bit, to the way that you generate the control signals, what are the requirements? Is it microwaves? Is it baseband? What are the timing precisions required? ⁓ To the error correction scheme that you run on top, which depend on the connectivity enabled by shuttling, ⁓ to the higher levels of the stack, the operating system, so to speak. ⁓ And then the application, well, the... the applications on top of that, it all comes together. Even the application, in a way, it comes together from the lowest to the highest level. ⁓ The choice on the materials can have an impact on the applications ⁓ in the end, ⁓ in a certain sense. There is, of course, the picture of the universal quantum computer, and then any application can be run on it. But you could also have intermediate steps where you build a system that can deliver intermediate applications, not yet the full thing, fully programmable. And then maybe that can take advantage of specific choices in, well, anywhere lower than in the stack. So yeah, I think that's the big transition thinking at the systems level.
Daniel Rodan Legrain: That might be a loaded question, but if you had to pick one, what do you see as the most critical breakthrough that is still missing for us to get there?
Lieven Vandersypen: Yeah, I think despite all of the recent advances, including the APS March meeting, I still think that we're not, we have not yet gathered or certainly not yet demonstrated all of the necessary ideas for wiring up a million qubits, you know, all the way to 10,000. I see it's happening on a single chip that is already quite a bit beyond where the superconducting where the superconducting qubit world is going. There were projections in the IBM roadmap where they had a plan to go from, I don't know, let's say 50 to 100 to a few hundred to a thousand. Their latest roadmaps are all on the same chip. The latest roadmaps have, I think, something like of the order of between one and 200, I think, on a chip. And beyond that, it's connecting chips together. And that can maybe work, we'll see. They're certainly working hard to make it work. But the links between the chips could become the weak links. ⁓ Of course, that industry will try to prove me wrong and we'll see what that brings. But for the semiconductors, all the way to 10,000, it can be on a single chip. That's already pretty far. ⁓ But I think we can also go to a million on a single chip, not yet. Well, we have not yet demonstrated all the pieces that are needed to make that happen. And so that's, think, fundamentally what I see. still the most open questions to be answered. There are of course challenges at other levels, know, from uniformity in the devices all the way to how to run error correction and how to get the classical electronics at scale at an affordable cost and so on. And it's hard all around, but if I had to pick one, that would be my answer on the wiring.
Daniel Rodan Legrain: I mean, we've been hearing about wiring ⁓ throughout the APS talks and today's conversation, and I can see it's at the very much center of many scientific minds nowadays and in the future. Liven, thank you very much. Let's maybe take one or two questions from the audience ⁓ watching live. ⁓ So here goes one. You've mentioned that the spin community is transitioning from good concept ⁓ academic research to quantum industry with large scale fabrication goals. Are there any different research needs and approaches necessary now to support this industry in the pre-competitive stage?
Lieven Vandersypen: Let's see, so are there research needs?
Daniel Rodan Legrain: How can we, as I understand the question, how can we as academics support the industry in this pre-competitive stage?
Lieven Vandersypen: Well, I think in two ways, or in several ways. One is to provide feedback on devices that are fabricated in industry. But I'm also convinced that there is still plenty of room for good ideas. you know, I've shuttling, I've mentioned performing gate operations by making electrons hop, so-called hopping gates. There are other examples where you could say they really come out of left field. Even Germanium, know, like 10 years ago there was very little effort on Germanium. Now it's a big effort throughout our field. And these have come from academic groups like ourselves. And ⁓ I think this is going to continue to feed the industry. And I'm convinced that no single industry player will come up with all the creative good ideas that are going to be part of a future large scale. spin-based quantum processor. So coming up with these new ideas, testing them out, ⁓ I think will really continue to be a good role for academics ⁓ for many years to come.
Daniel Rodan Legrain: Another question here from the audience is, what are your thoughts on possibilities for setting up a program similar to what exists in the US, which is the Squid Foundry at Lincoln Labs, that basically allows groups at universities to send, it's for superconducting qubits, apply it for spin qubits to send spin qubit designs and then get them fabricated to be able to measure them.
Lieven Vandersypen: Yeah, now I expect this is going to be a major trend in the coming years and a very important one. ⁓ So when I said earlier that devices are shared by industry until now, I think with few exceptions, it's mostly that industry creates a design in isolation, performs measurements on it, and at some point starts sharing devices with others in the community. But I think where there is a real opportunity is to allow the community to submit their own designs. And with IMEC and ⁓ CAA Letty and partners around Europe, we've built a European consortium precisely with this goal. So it's called SPINZ. It's a European, well, it's funded half by the European commission and half by the member states. It was recently launched. And in that program, the goal is to create PDK, so process development kit. It's basically a way, well, it's basically ⁓ a process where you know how to make high quality quantum bits. And following that, there will be a phase where academics can submit a design to these ⁓ institutions, to these facilities and get a device back. And I think this is going to be a very powerful model. ⁓ It's going to change the dynamics in the field, you know, the The approach used to be that everybody fabricated their own device. There was a time where every PhD student fabricated their own device. In my group, we've stopped doing that already 10 years ago. ⁓ Nowadays, most devices in my group are fabricated by professionals who are embedded in the group, not by PhD students and postdocs, but permanent staff. ⁓ And I think eventually the next phase will be this interaction with the large scale manufacturing facilities. At the same time, interaction is, the turnaround is typically slower, especially if you need mask making, is the standard in industry. It's a slower turnaround. so for sure in Delft, we're not going to give up our capability of fabricating devices locally. It keeps us flexible and we can implement ideas, well, within two weeks, basically. ⁓ But at the same time, yeah, I think this is... This is a very good model and also in Europe this is starting to be the case and I think it's a very good strategy.
Daniel Rodan Legrain: We're getting short on time, leaving, and I see more and questions populated. So let me pick one more, maybe in 30 seconds, although it's a pretty loaded question again. And it's about your own results on wait for parity checks in silicon and the spin-chatteling architecture. And the question is, why does that matter and how will this influence quantum error correction in spin qubit platforms?
Lieven Vandersypen: So in that experiment what we did is we have four quantum bits stored in well-separated locations and then a fifth one comes by, interacts with the first, the second, the third and the fourth. So we use this shuttling connectivity to make them interact with each of these four data qubits. And in this way you can collect the parity of these four quantum bits. What that means is you distinguish whether there is an even number of zeros or an odd number of zeros. And that actually is the cornerstone of quantum error correction. In quantum error correction, you want to avoid measuring your qubit states. If you measure the qubit states, your qubit state has collapsed, it's gone. What you want to do is compare pairs of qubits or groups of four qubits ⁓ and see if they have the same state or opposite states without measuring it. So you don't measure the state, but you just compare them. And so that's what we've done. And I think the power of that experiment is maybe even more than in the fact that we did this for... wait for parity check in how we've done it, using the shuttling to create arbitrary connectivity. And I think we're going to be ⁓ benefiting from this possibility and the power that it brings a lot in the coming years.
Daniel Rodan Legrain: Thank you so much for sharing your time and insights with us today.
Lieven Vandersypen: Yeah, you're very welcome.
Daniel Rodan Legrain: It's been really impressive to see how your career has connected some of the early experiments in quantum computing to some of the most exciting advances in ⁓ scalable spin-based processors today. At QBlocks, our motto, as you might know, is building quantum together. And I think Reed's conversation is really a perfect example of how progress emerges from the collaboration across physics, engineering, and ecosystems. ⁓ As a small token of appreciation, Leven, we've sent you one of our QBlocks mugs with our motto and we hope it sees many more breakthroughs from the Vanderseipel lab. ⁓ For everyone watching, ⁓ the talk today will be available shortly on YouTube, Apple Podcasts Spotify under Quantum Builders. Please stay tuned. More exciting speakers ahead. Leven, thank you again. Thank you for everyone watching us.
Lieven Vandersypen: Thanks, Daniel. Thanks for having me. fun.