27. Shuttling Spins: The Road to Scalable Quantum Computing with Lars Schreiber, Mats Volmer, Max Beer

00:00:05: Hi, I

00:00:05: am Chandana Rao and I'm Meera.

00:00:08: We are associated members of ML-FourQ And you're listening to ML Four Q&E a show where members from

00:00:17: the matter & light for Quantum

00:00:19: Computing Cluster talk about their careers Their research and future

00:00:23: in quantum.

00:00:24: Today we're excited.

00:00:26: welcome one of ml four q's member Dr.

00:00:29: Lars Schreiber a lecturer at Arviteha Aachen University and a senior researcher in the Quantum Technology Group together with his students, Max Farmer & Max Beer.

00:00:41: Dr.

00:00:42: Schreiber leads effort on silicon based spin-cubit devices.

00:00:47: is also coordinator of European consortium SI Cubus which aims to realize quantum bus A key building block for true scalability of Silicon Quantum Computer architectures.

00:01:01: It's question many

00:01:02: ask

00:01:03: If semiconductor technology is already so advanced, why don't we yet have a million qubit quantum computer based on semiconductor spins?

00:01:13: In this episode.

00:01:14: We explore exactly that.

00:01:16: We discuss the major challenges in scaling semiconductor spin-cupid architectures today and hear directly

00:01:23: from

00:01:23: Macs as they dive into technical details of their impressive work on SpinCupid Shuttling & T-junctions, which is an approach that's becoming increasingly important across different quantum computing platforms.

00:01:41: Welcome to all three guests!

00:01:43: We have Dr.

00:01:44: Lars Treiber leading the effort on silicon-based Qubit devices with his two PhD students, Matt Spalmer and Max Beer.

00:01:54: I'm particularly excited for this episode because i've worked in the theory side of spin qubits And also know Matt's and Max for some time now start with the career journeys.

00:02:09: For large,

00:02:10: your diploma thesis which was done at Ayurveda was on spin coherence and spin defacing of hot electron spins in three-five semiconductors.

00:02:22: And then you're master thesis

00:02:23: Was

00:02:25: Done At The University Of Cambridge

00:02:27: On the growth

00:02:28: of gallium arsenide Aluminum Gallium Arsenide Heterostructures Using Molecular Beam Epitaxy.

00:02:36: Unfortunately, I couldn't find them online.

00:02:38: But I see that they are already going in the direction of SpinCubit architecture.

00:02:43: This was also after The Seminal Paper by Daniel Loss and David DuVincenzo on spin cubits.

00:02:49: So did that influence your choice of topics?

00:02:52: In any way?

00:02:53: No at that time wasn't aware of this so it was mainly interested in semiconductors from lectures for diploma lectures in Aachen.

00:03:04: And that made then my choice to go to Cambridge and there, look into growth which was basically a random decision.

00:03:11: So I learned about molecular beam epitaxy and high-quality growth of three five semiconductors... ...and also an important step to know how tricky it is to grow these heterostructures and learn the language that I can now use to communicate with the growers.

00:03:34: We cooperate with many excellent groups, they grow for us the heterostructures and then we can communicate about these issues And all this.

00:03:42: The quality of the hetero structures is extremely important For the quantum computing know so this helps a lot.

00:03:49: and Then in your PhD thesis you were already looking into information processing using spins.

00:03:54: So where are you at?

00:03:57: This time introduced to quantum computing already,

00:04:03: and

00:04:04: these were also the three five semiconductors.

00:04:07: Would you say that the methods

00:04:09: used back

00:04:09: then are similar to the method your students use today?

00:04:14: How has your PhD shaped your research

00:04:16: today?".

00:04:18: That was very different.

00:04:19: it was a femtosecond spectroscopy working so with optics lasers quite complex experiment.

00:04:29: These optical experiments are very delicate and hard to handle.

00:04:35: I was then later, very happy that I could go to transport experiments which were more stable.

00:04:40: so...I remember many days there where i tuned the mirrors over-the day.

00:04:50: hard work.

00:04:52: But what the topic was about in these semiconductors, I kept this topic about other material that i've learned on Cambridge and then added this spin aspect to it.

00:05:03: so Spintronics.

00:05:04: And there It Was About The Transport.

00:05:08: And That Influenced Also My The Main Topic That We Are Doing Now Transporting Spin Cubits.

00:05:14: but at that Time There Was Not A Single Spin many, many spins and it was not really quantum.

00:05:21: Everything was defined by magnetization classical magnetization vectors very interesting topic And at the moment's phytonics also leads to something.

00:05:33: but At that time I was a bit unsatisfactory because i wanted To do something more quantum.

00:05:40: then I came across colloquium talk by Lieben van der Sijpen in Aachen, and he showed that one can control single electrons.

00:05:50: That was then the point.

00:05:52: I wanted to go to Thio Delft.

00:05:55: So now

00:05:56: we got to know how Lars' journey is as a PhD student.

00:06:00: In this podcast you also have his PhD students, Max & Max.

00:06:06: Can you tell us about your PhD?

00:06:09: And

00:06:09: what

00:06:10: attracted you to work on these topics?

00:06:12: Yeah,

00:06:14: so I actually started my bachelor's thesis in somewhat different topic theory of topological insulators.

00:06:25: And then over the time just got to know that... ...I'm a person and need do stuff with my own hands.. ..I need see physics directly myself!

00:06:35: So i was naturally drawn towards experiment more and more....and still wanted stay field that is quantum computing adjacent.

00:06:47: And there's the back then and still today this very strong spin qubit group here in Aachen, I joined for my master thesis.

00:06:58: So

00:06:59: Max,

00:06:59: how was your

00:07:00: journey?

00:07:00: For me it was even less exciting.

00:07:05: I've been here at Adderbeteer H for the entire time with the same smarts but also doing my bachelor thesis in the same group as I'm doing my PhD now and master's thesis.

00:07:15: I started out my bachelor thesis working on cryo amplifiers, so heterojunction bipolar transistors, coolant down in liquid helium.

00:07:24: A lot of dipstick testing, dipping the samples into liquid helium and just checking the characteristics.

00:07:30: And then i kind-of moved from that to my masters where also did quantum information specialization.

00:07:39: That was quite new a second year and they really wanted to go into semiconductor fabrication.

00:07:48: So I elected my courses properly, to do the theoretical basis here at the Electrical Engineering Institute And then also did a lab course.

00:07:57: Actually i did two lab courses.

00:07:59: one of that was voluntary That was the Fabrication Lab Course in the Electical Engineering Department where you could actually going through clean room together with Professor in small group.

00:08:09: That was quite fun.

00:08:10: The devices obviously didn't work as they usually do for the first fabrication that you do, so yeah then I moved on to my master thesis where i then fabricated the Cuba samples set by my then supervisor and I measured also and fortunately got a publication out of afterwards which uh... So it's somewhat productive!

00:08:31: And then essentially just continued onwards in my PhD, fabricating my T-junction devices much larger.

00:08:38: And yeah, I've been working on the measurement and publication of this data now.

00:08:42: How's it doing a PhD as an experimentalist?

00:08:46: Do you have to wait very long for results?

00:09:05: There can be times where you, in principle just wait for good data to pile-in.

00:09:11: It all depends on the sample at least... ...at a moment when still somewhat more academic style of fabrication which is very flexible.

00:09:24: but some samples work better than others.

00:09:27: and sometimes you measure a sample and notice that a gate might not work warm it up again and start over.

00:09:38: And I guess that's just part of the process, but then when you get one sample that works... You can make a lot of progress in very short amount of time.

00:09:50: Back at Davos was actually quite an intense time where i measured quite alot of data.

00:09:58: Actually most majority of my data so far have recorded during my PhD within the total duration of three months.

00:10:07: It was quite important to keep the experiment running and alive, so I was working somewhat around-the-clock during these three months.

00:10:16: but yeah... Quite happy that i did that!

00:10:20: And looking forward to next sample which hopefully will work soon.

00:10:25: Yeah remember you said that you were running this experiments overnight or something?

00:10:31: In principle they can like a certain Experiment can run for quite a while alone by itself.

00:10:38: Sometimes they drift out of the sensitive, so... For instance your charge sensor Can drift out off its sensitive region or you spin to judge conversion for Spin readout that can also drift Out-of-the-sensitive region.

00:10:54: So you have to regularly check whether everything's still going on fine and Also sometimes You just want to change experiment parameters because we are happy with Your earlier measurements.

00:11:05: There needs to be some interaction, but sometimes you can also just let it run overnight.

00:11:09: And well actually regularly we just run... Let it run over night.

00:11:13: just see whether something occurred

00:11:15: during the night.

00:11:17: Yeah So then after your PhD Lars You did a postdoc at Delft in The Group of Van Der Sippen Another popular name In the SpinCupid community.

00:11:30: And then you came back to Aachen, where have been teaching since twenty eighteen I think.

00:11:37: and i had the opportunity attend your course two years ago on theory behind spin-cupid experiments.

00:11:44: it really helped me develop a solid understanding of device how things work in practice.

00:11:50: for me its served as valuable bridge between theory & real world.

00:11:55: How challenging was is design such a course?

00:11:59: So first of all, it's great that its helped you.

00:12:02: Actually I started lecturing right away so i was announced in a lecture after a lecture in two thousand eighteen.

00:12:12: but during my whole career I helped with lectures exercises and also when I came here to Aachen which is already some time ago at the end of twenty eleven I started designing.

00:12:26: this one was not so difficult because it is my specialized knowledge.

00:12:33: So designing the fundamental lectures, this more tricky.

00:12:39: It's of course then a lot fun because it has my main interest and we could brought many experiments in them in this course And you can learn about other state-of-the art experiment.

00:12:52: I'm also teaching a Python programming course since two thousand and fifteen every year.

00:12:59: And that was an interesting challenge because it's something we just noticed.

00:13:08: after our students have very different background knowledge in programming, then this is as optional course at the beginning of evening with some gamification.

00:13:19: people had fun to come through this course, and then later it was included into the main curriculum.

00:13:26: And now of course all the bachelor students in physics have to attend since three years ago.

00:13:34: so that is also a course I designed over here as well.

00:13:39: there's stuff like lab courses and seminars.

00:13:44: Yeah i think programming is quite important for a physicist, whether you do experimental physics or theory.

00:13:51: Also, Do You think this lecture that you did on spin qubits can also be taught at bachelor's level?

00:13:59: Can we introduce to students earlier.

00:14:02: Well it needs to be digested in the way.

00:14:05: there are other lectures That We have at our TH Aachen.

00:14:09: so we have master specialization branch in quantum technology And that starts then with a platform course.

00:14:17: That gives an overview of the different qubit platforms and probably this one is more suitable to start with, um...and people if they like spinqubits can go deeper into their topic.

00:14:31: Of course on spinquits teach to bachelor students important points because it's not all about deep quantum technology.

00:14:46: There are so many things that go together, electronics programming and it probably is nice to have students from the beginning see how many bricks enter into such a field.

00:15:02: Yes Let's now talk about your research.

00:15:08: You're extensively working on the silicon-based spin qubits.

00:15:12: So to set this stage for our listeners, could you explain in simple terms what a spinqubit is?

00:15:19: Yes yes so actually I started with gallium arsenide and spinqubits in DERFT.

00:15:25: that was the established system at their time And then one of first ones who moved into Silicon in two thousand and ten.

00:15:36: But first of all, what is it about?

00:15:38: So for qubits we need a true-level quantum system And the electron spin also holds spin but let's say electrons been because this is what I'm mainly dealing with.

00:15:51: Is there a natural system?

00:15:52: because It has only spin up and spin down state?

00:15:58: The energy of these states are different.

00:16:00: when you apply a magnetic field Therefore, we get already a nice qubit.

00:16:05: And now we need to capture this electron so that you do not lose the qubit.

00:16:11: and it can be done nicely in such a semiconductor like silicon or a silicon germanium hetero structure which provides one confinement direction.

00:16:25: Then there are various options for other confinement directions And our preferred way is to have a couple of gates on top off the device that then sets the barriers in other directions.

00:16:39: In this way we form a quantum board where you can capture single electron.

00:16:44: There are options, one can use wires, quantum wires or one could use donors some crystal defects that can bound the electron hole and at end it's scalability, what systems can be scaled the easiest.

00:17:07: So first of all we now retrap our electron in this quantum dot and then use its spin as a qubit.

00:17:15: And there are ways to manipulate it by electrons spin resonance or some variance of that.

00:17:23: One can read out with certain scheme the spin and transfer into for example a current that can be measured, there are ways to couple two spins together.

00:17:35: So one has then the full set of qubit operations defined by the differential criteria.

00:17:45: But what motivated you to shift to silicon?

00:17:49: Yes so to be honest it was at that time an idea a project that Dean van der Sijpen had.

00:18:00: So first of all, I started with gallium arsenide and also the gallium arsonite.

00:18:04: is my whole history that I worked within indian gallium-arsenite systems or was very familiar to three five system.

00:18:10: i know it's a very good heterostructure because the lattice constant of gallium as night and aluminum gallium are the same.

00:18:18: one can nicely grow them on top each other.

00:18:22: but then we also realized in Well, the nuclear spins are something annoying.

00:18:29: So we did for example manipulation experiments where you have a certain frequency and this frequency has to match the energy difference of up-and down state that is given by magnetic fields.

00:18:45: but then one also some random magnetic field nuclear spin bar that gives them so-called overhose of fuel, which is not fully controlled.

00:18:56: And in gallium arsenide this is an unavoidable thing because all the arsenic and gallium atoms always have a nuclear spin.

00:19:08: That's related to the fact they are three and five material.

00:19:12: So there will be odd number of protons.

00:19:16: means then that they have a nuclear spin, cannot be compensated by the number of neutrons in the nucleus.

00:19:22: And so one has to kind of live with it and there are... Then at that time there were different ways.

00:19:29: for example person called Hendrik Blum tried to stabilize on the nuclear-spin baths and other people considered for example two holes which couple less onto the nuclear spins.

00:19:46: Then it might be a good idea to go into the material system, which has no nuclear spins.

00:19:52: And this in principle can be silicon at that time and was not isotopically purified but had some silicon-twenty nine nuclei, which have then small hyperfine interaction of small overhose fields that acts on the nuclear spins But gives us immediately better coherence times.

00:20:10: It's easier to manipulate.

00:20:12: But in the beginning, it was very painful because it also meant that the heterostructure is not so perfect.

00:20:18: And then one has more charge noise and to just trap the electrons where it was really hard.

00:20:24: Also It had a large effective electron mass which means that the fabrication of these devices Is much more difficult compared to Gallimau's night.

00:20:36: The gate structures we put on top have to be done.

00:20:40: The feature size is smaller, and that was just at the edge of what can be academically be fabricated.

00:20:48: And it took then some time for results also from other groups to come in – there were later a complete transition between all these groups towards silicon systems either Silicon or Silicon Germanium.

00:21:01: So we see already that spin qubits are not really confined to one type of material, you can use different materials and have these trapped electrons.

00:21:14: Are there also different types of qubits?

00:21:18: Yes so they are various materials indeed silicon, galamazinite we're always working on zinc selenite a two-six system And there's also germanium, silicon germanium but their also indeed different types.

00:21:30: So the ones that is obvious as to use one electron which has a spin up and spin down.

00:21:37: One can also use in the material, a hole.

00:21:40: so it's simply not a charge that is in the conduction band of the material but in the valence band Which has then simply other properties In terms for example properties like spin-orbit interaction or Has an also different effective mass influences on the device.

00:21:59: And then there are also encodings.

00:22:01: one can not just one electron, but for example two electrons to define a qubit.

00:22:07: First of all then one has four states on... But one can pick them two states out off the silver space and get some advantages For example To be robust against some variation global magnetic fields.

00:22:23: This is for example the single triplet qubit.

00:22:26: another one Is The exchange only qubit.

00:22:30: that then consists three electrons that have to be set in.

00:22:34: Three quantum dots, there are tonal coupled next to each other.

00:22:38: the advantage is then one can use other methods for manipulation not just ESR and EDSR but um...just exchange coupling between these qubits from manipulation.

00:22:51: That comes with a cost that it's more complex system.

00:22:57: So many of this encodings materials work very well when you look at a single device with one or two qubits.

00:23:06: The big question now is, which of these approaches material and encoding works best for scaling up the system?

00:23:15: And the goal has to go through millions of qubits.

00:23:18: this what we need them for having really quantum advantage with a quantum computer.

00:23:22: Now that you mentioned it do want focus towards the quantum advantage and toward scaling off these quantum computers.

00:23:30: So

00:23:31: qubits cannot be confined to just one place and that doesn't help us in scaling these quantum computers.

00:23:38: How are you

00:23:39: kind of addressing this questions, transporting these qubits across the chip let's say?

00:23:48: Okay I start a little bit before to understand why we want to transport.

00:23:53: so first thing is really nice about our qubits.

00:23:58: So they are the smallest among all of the other qubit platforms.

00:24:03: Thirty nanometer in diameter, for example is a little bit larger than in Garimah's night and then silicon.

00:24:09: but it doesn't matter because at the end when we fabricate by industry or chip It will have a size of ten by ten millimeters or twenty by twenty millimeters.

00:24:19: And if you just calculate how much space that is We can put billions on this chip.

00:24:28: So why are we not there yet?

00:24:30: The problem is that our qubit lives in a material.

00:24:33: Not like in vacuum, like ion traps or neutral atoms and they're influenced by material imperfections.

00:24:40: And these imperfections have to be balanced out By tuning voltages.

00:24:46: Every qubit basically needs its own little trap But tuned up special voltages for this qubit For this place.

00:24:54: In the end you need then For all these qubits, one can calculate that two defined voltages.

00:25:02: One for its chemical potential and one to four it's barrier towards neighbor as an example.

00:25:07: So in the end we need millions of voltages fine tuned.

00:25:12: if you would just follow scaling up single-two-qubit demonstrator chips that have now And this is simply not possible.

00:25:25: So a chip cannot have millions of terminals.

00:25:27: We don't want to have billions of wires going up to room temperature electronics, so the key point now scale this up is make it compromise not having these little dots close each other but making them sparser and extra space that we get then could be sufficient to implement cryo-electronics that generates the signals we need.

00:25:57: So, and... We need very specific signals then for the points where you do manipulation because this is a delicate thing.

00:26:05: so we need a manipulation zone or readout zone there we have to tune up the voltages very precisely.

00:26:10: For that we need then this on-chip electronics an individual tuning of different regions.

00:26:15: The electronics has been called as Tiled up with the quantum layer.

00:26:21: So we have a certain tile in the quantum player on top of a tile Of electronics and then the footprint of these two tiles match.

00:26:29: Then one can tie it up And I will ever large chip, and the balance that you see is now to have A tile which has a size of ten by ten micrometer.

00:26:40: so It's not a ten nanometer qubit anymore but ten point ten by Ten micrometre and they're in such a tile.

00:26:46: We can put into the electronics a couple of qubits, so then we can scale this up.

00:26:53: But you also now have to make it possible that the qubits communicate with each other because they only do two qubit gates when they are very close together and need to have tunnel coupling For example transport them from one to another in order for us.

00:27:17: that only can work if one finds a transport mechanism, which is from the signals.

00:27:26: That easy so it could be also implemented on this cryoelectronic tiles.

00:27:31: So we were seeking for very simple method to transport the qubits and there we had to make compromises.

00:27:37: We couldn't do individual tuning anymore but we could use just a few signals.

00:27:43: In best case required for transporting should be independent of the transport length.

00:27:51: And that was then the idea for our transport method, we generate with only four simple signals a kind of wave that transports electrons within this way from one side to another.

00:28:07: it's rough.

00:28:08: The qubits have to survive that, the qubit information has to survive it.

00:28:12: It's rough because all the potential disordered that we have or those stones you just go over there with this wave and the way fast be large enough That is possible And then bring into next manipulation zone.

00:28:26: in this manipulation zone We are very careful tuning up the voltages again.

00:28:32: This was key challenge not only transporting but also easy in terms of the signals that we need and therefore be implemented or implementable in a cryoelectronic tiles.

00:28:49: And then this propagating wave, which you mentioned is also known as the conveyor belt or conveyer mode shuttling.

00:28:58: I think There's also bucket brigade shuttlings.

00:29:04: Can you tell us about it?

00:29:08: Why do you prefer doing the conveyer mode shuttling?

00:29:14: OK.

00:29:18: So first of all, people have a series of quantum dots in their device which are tongue coupled and then the first idea one could have is to transport an electron by letting it hop from one dot to the next And this called bucket brigade.

00:29:35: so bucket here is this quantum dot And then we calculated that this is only possible if the tunnel couplings are fine-tuned.

00:29:48: That requires for each tunnel coupling a tuned up voltage and when you want to go at distance of ten micrometer, which now it's required for implementing the cryoelectronics Then one need to hop about two hundred or let's say, one hundred quantum dots.

00:30:12: And that needs then to transport.

00:30:15: in this way it won't require a lot of fine-tuned voltages and is exactly what we want to avoid.

00:30:21: There were also efforts on that.

00:30:23: I had very nice project, a quantaire project started in the year two thousand seventeen.

00:30:28: together with CODIRFT they tried to bucket regate put a little effort into fine tuning these dots and could also transport them to spin.

00:30:37: But then the big question was, is it possible to do this than with only a few voltages?

00:30:42: And the answer was no way.

00:30:43: that's not possible.

00:30:45: on the other hand without devices we could do this transport but only if you are.

00:30:50: voltages and therefore I'm just conveyor belt method is the better one.

00:30:56: an is now also implemented by other groups.

00:31:01: Maybe you heard there's even a conference that is the International Conference on Spin Shuffling.

00:31:11: That first took place in Aachen, was now in Los Angeles this next year it should be in South Korea.

00:31:18: so people really pick up a complete change in architecture and gives many, many more possibilities even beyond just transporting.

00:31:28: One can find new manipulation schemes now when the qubits can be shuttled this way.

00:31:34: When transporting a qubit across a silicon device what factors begin to dominate its behavior?

00:31:42: So you start transporting a cubital across a device via shuttling.

00:31:50: Well, some things that are more important now is a material parameter.

00:31:54: That's called valley splitting.

00:31:57: You can imagine it basically as two additional states to your spin subspace That behave pretty much very very similar to the spin subspaces just a tad bit different in other categories.

00:32:13: so The main difference between these two sets of States The excited valley spin states have a slightly different g-factor.

00:32:25: And one thing that can happen if you shuttle around an area where the valley splitting gets rather small, so two sets of states align pretty much in energy space is you get to hop into other two states and this happens.

00:32:44: What will happen to the qubit is it'll experience this slightly different g-factor of other valley.

00:32:53: And if the hop is probabilistic and not deterministic, what might happen is you get a probabilistic phase that your qubit sees via this slightly changed G factor.

00:33:05: That could lead to decoherence.

00:33:12: main thing that we at the moment are worrying about and which is why We Are analyzing material now more and more using the methods that we develop.

00:33:23: And The other thing That could happen Is if you're Zeeman splitting Alliance with your so the spin splitting alliance With Your valley Splitting?

00:33:32: And You are in like In a superposition or in the excited state Like Spin upstate You get an additional decoherence channel.

00:33:42: So basically via this valley your spin can relax quite a lot faster.

00:33:46: so these are the two factors that begin to dominate its behavior.

00:33:50: one thing That is at actually advantageous when moving spins around or having mobile spin qubits Is you have to worry a bit less about nuclear spins because?

00:34:00: Uh, you get in effect called motion narrowing where if your wave function sees more and more spins or basically overlap of the wave function, your electron with your alloy integrates over more and more nuclear spins or more configurations of nuclear spins.

00:34:18: Let's say it affects coherence less?

00:34:25: So

00:34:26: you

00:34:26: said that the coherence is affected mainly because of disorder in this system.

00:34:32: right And what are then doing to address?

00:34:41: Yeah, so the main thing what we are worrying about is The fact that these material parameters can change and do change Across the alloy.

00:34:51: And as you say due to alloys order because in your Silicon Germanium barrier You have yeah, seventy percent silicon thirty percent germanium and they are ordered not In a specified manner but I can be ordered in a random way.

00:35:12: and then depending on what your overlap with your barrier is, the two valley states do have quite different overlaps because they belong to quite opposing K vectors.

00:35:23: And so there are... They oscillate pretty much out of phase.

00:35:29: What happens is material parameters or like the energy scales off the valley splitting?

00:35:39: across the whole wafer.

00:35:42: And what we are trying to do about this is there's some approaches at a moment that have been grown and fabricated, so my ideas of how design a waiver specifically to mitigate these issues.

00:35:56: for instance something called a wiggle.

00:35:57: well where you don't just fabricate or Quantum Well, but you actually include some germanium atoms into the quantum well in an oscillating manner that aligns with your oscillations of a block wave.

00:36:14: And if we do that is proposed to enhance the value splitting.

00:36:19: This something has been now research quite heavily on growth and I think there's been recent publication from a collaborator And we now have these devices in our labs, and we want to measure them as quickly as possible.

00:36:45: So essentially looking for ways to enhance the valley splitting so that ground states don't escape into excited state?

00:36:56: I can add to this the bigger wireless bond approach.

00:36:59: We also have from other collaborators, for example Tivo Delft.

00:37:03: you get devices where they think that already solved us by a better heterostructure.

00:37:11: and on these heterostructures there are very exciting results on shuttling.

00:37:16: but no interesting is it really now solve very splitting problem?

00:37:23: Do we do not understand the theory completely?

00:37:25: You also have other projects with collaborators, for example to put strain on devices that can help.

00:37:34: To split the value splittings.

00:37:35: there will be stressors One of my PhD students is doing membrane device and tries to put some strain in these devices.

00:37:48: And then we also want to evaluate the value splitting.

00:37:51: Because what is now really nice, that this transport method... ...is not only good for transporting and transforming one side of each other,... it's a perfect way to measure or map properties of the heterostructure.

00:38:04: For example the value-splitting or charge disorder.

00:38:07: So experimentally

00:38:08: how are you

00:38:09: mapping the valley splitting across the entire terrain?

00:38:16: For this, what we are doing is we're basically shuttling a spin coherently or firstly initialize the spin singlet.

00:38:25: Basically... ...a bell pair of two spins.

00:38:29: We leave one static and other one.

00:38:32: we shuttle as fast as can to spot inside the channel.

00:38:38: so.. ..we have this Cubus type device design that Lars already had mentioned And with that, we move it to a distance d. We then let it precess and basically interact with the material parameters which are interested in like the valley splitting or what you can also measure as g-factor For both of us have slightly different pulses for the valley spitting.

00:39:07: What do is wait for constant time Then shuttle the electron back and recombine.

00:39:17: We get a spin singlet or spin triplet via polyspin blockade.

00:39:23: And actually what we are interested about when we measure the value splitting is not really whether it's a spin-singlet, but rather that there still entanglement and how we probe this.

00:39:38: once What we see is in the vicinity of the valley splitting, so when the valley spitting aligns with a Zeeman splitting.

00:39:52: The phase... We don't see a phase anymore.

00:39:55: basically the state population drops to a fifty percent between spin singlet and spin triplet because it becomes a mixed-state.

00:40:04: And now When we do this for one distance It's not yet super clear.

00:40:10: where are the value splitting?

00:40:11: Is but if we do an entire patch inside our channel, what we get is a coherent background of some phases that vary due to local g-factors and wait times.

00:40:29: So within this coherent background with quite high contrast the thin line background considerations.

00:40:41: And when we correlate both the magnetic field and distance, it becomes quite clear to sketch out this trace of valley splitting.

00:40:49: in This way We can record a one-dimensional trace of value splitting something that so far I think only has been done over A couple of nanometers.

00:40:59: twenty nanometers ish?

00:41:01: We have now Been able To do Over a Distance Of four Hundred Nanometers.

00:41:06: In principle There's not that much stopping us from going further rather than just getting a device.

00:41:14: That can shuttle further coherently.

00:41:19: and so having this done now for one dimension, we actually have on these shuttling devices.

00:41:26: We have additional gates with which we can shift the position of this one-dimensional electron channel.

00:41:33: So there's Cubus channel where we move the electron along not just in the distance coordinate, but also laterally and why coordinate.

00:41:43: We can basically displace the channel up-and down.

00:41:46: if you think about shuttling horizontally You can move it vertically And then you just repeat rinse and repeat the measurement until you get a full two dimensional map of your entire canvas where we can shuttle.

00:42:03: your electron can be used to record a valley splitting map, and actually pretty similarly we could also record the g-factor.

00:42:14: There instead of varying the magnetic field what you do is... We vary the weight time then record oscillations that result from this weight time variation And these local variations at the local spot.

00:42:29: From frequency.

00:42:30: of these we can extract the g factor difference or the g factors relative our static electron of the spin singlet.

00:42:40: And yeah, these two material parameters already give us quite a good insight into what's happening in our wafer and now... In future I think we aim to look at all this improved heterostructures that Lars and I have been talking about.

00:42:59: with exactly these methods

00:43:06: These measurements, you conduct them only at the end of a shuttle cycle.

00:43:12: Right?

00:43:15: Yeah so basically we measure that the spin state is only in the end for the full shuttling experiment.

00:43:22: So... We initialize and then we shuttle into the channel.

00:43:29: let it evolve there.

00:43:31: This where when its on the channel.

00:43:33: this is important part Where actual physics the physics we are interested in happen, physics happens all of a time.

00:43:41: We shut it back and when we have put into a double quantum dot can make use of poly spin blockade

00:43:51: where

00:43:52: the spin triplet is blocked from tunneling.

00:43:57: The singlet isn't if you tune your system .

00:44:00: If via this spin selective tunneling, one could say you are able to measure the spin state.

00:44:13: So the measurement actually only happens really at that end

00:44:17: as I said and then you get a g-factor directly from the valley splitting?

00:44:21: No so the value splitting in the G factor they're different like the pulse is slightly different.

00:44:28: there's different experiments.

00:44:32: okay

00:44:34: You basically for the value splitting you have a constant way time and For the g-factor, we are very in waitime.

00:44:40: And for the values bidding instead of varying your weight I'm you vary your magnetic field.

00:44:45: Those are just two different parameters.

00:44:48: Then from this you can reconstruct.

00:44:50: So when do they shuttling?

00:44:52: so your electron literally

00:44:53: moves?

00:44:54: So could you define something like velocity or speed?

00:45:00: Yes There's actually quite some theory behind it, because I've earlier spoken about these excitations to the higher energy valley state.

00:45:13: They happen depending on called a Landauzina transition.

00:45:22: They did the theory on this quite awhile ago already and it's basically when you approach an anti-crossing, energy splitting between two states gets small.

00:45:31: they are coupled with some sort of tunnel coupling.

00:45:35: then depending upon the variation or like the energy variations in principle your excitation probability changes.

00:45:51: we have thought about what would be an optimal shuttling speed to both fast enough so that the spin does not decoher during shuttling, but also not too fast for us get a lot of excitations.

00:46:08: That seems around ten meters per second.

00:46:11: and yeah these shuttlings speeds has been achieved basically also far beyond that, is in principle possible electrostatically.

00:46:25: Or last do you agree or is it an accurate representation?

00:46:29: That's very accurate.

00:46:30: so the point of all this shuttling as we move the electron into this quantum dot and there it is nicely preserved and confined And we have to take care clock rate of the quantum computer in the end doesn't, it's not slowed down by the shuttling.

00:46:56: And um... It is actually a nice coincidence that this around ten meters per second sometimes about fifty or forty meters per seconds are working quite well and they're good for the clock speed as sufficient in the speed.

00:47:13: so there were also impressive results from other groups.

00:47:19: But the problem at this moment is that most people shuttle forth and back a couple of times, and then accumulate their distance.

00:47:28: Not shuttling over large distances.

00:47:30: in one way We did it with charge on our device which was ten micrometers long but what you can see?

00:47:39: there are some crystal defects coming into the way, misfit dislocations maybe other effects or simple effects like the gates not working some other things in the setup is not working.

00:47:52: So what we see now, it's what happens when we scale up?

00:47:56: This is know-what because you have the largest devices of community and if they see whats happening its not only about valley but also that this location makes electrons stop moving there a too large barrier.

00:48:14: These are not things that we have to take care of, and it's again back something they need to be optimised in the material.

00:48:23: So you also partly touched upon what was going to my next question... You said there is already a sort-of limit on shuttling speeds.

00:48:35: so does that create a limit at which you can shuttle?

00:48:41: Because I think In the bigger picture, you need to really take into account the shuttling times and then the gate time in order do something useful out of computation.

00:48:54: At this moment we have all operations taken into account.

00:48:59: The slowest operation besides shuttlng is readout.

00:49:06: This is probably at that moment a weaker spot for spin qubits of all the platforms.

00:49:14: And what we try is with a shuttling to be well faster than the readout, and the read out has to not done at the very end of an operation because if you want to do quantum error correction, readouts have been done couple times within your circuit algorithm.

00:49:29: At the moment one can see that when take this as limit it can shuttle quite far sufficient still for this idea to implement cryo-electronics, so go from one tile to the next and maybe even go beyond.

00:49:49: What is in their way?

00:49:51: Not only shuttle the spin but also make sure that a charge can follow.

00:49:55: then these kinds of new things.

00:49:57: Misfit dislocation we start seeing there... And perfection on the device or all gates have to work get at the limit of our academic devices, and we have now since a couple years established an industry road off devices.

00:50:17: We also measure these and hopefully they are more reliable than in end.

00:50:23: Could you give us some time scales?

00:50:26: You know like T-one times, T two star times And overall gate times and shuttling times that typically has.

00:50:34: Yeah, so the T two star time in isotopically purified silicon you can get think up to fifty microseconds for unfortunately sometimes it's a bit shorter around ten microsecond even less.

00:50:48: The one-time actually don't know its much longer.

00:50:52: I dunno last whether you have that number?

00:50:55: It depends on spinoid injection and the silicone.

00:50:58: we do not have much spinode injection there is.

00:51:00: second depends also on the size of magnetic field.

00:51:02: It's so long that we do not really care about that anymore, but if one simulates the depolarization channel versus defacing channel it plays a crucial role and is something you can benefit from in future when T-one time is superlong compared to t-two.

00:51:22: And maybe regarding the shuttling distance that's possible, I think in recent Delft-Gurion shuttleneck paper there was an interesting calculation.

00:51:31: The fastest shuttlings speed they can shuttle more than two millimeters which is obviously a significant fraction of the chip and probably more.

00:51:39: then you would require if we think about... You have a two dimensional grid for tjunction but implement surface code unit sets that you need logical qubits and obviously much less footprint than these millimeters.

00:51:57: Right, it's basically already at the size of a chip that we have in the end.

00:52:01: so its point is to go beyond something like twenty millimeters because this will be the chip size.

00:52:08: So... That looks very promising

00:52:14: Max!

00:52:15: You've mentioned T-junction.

00:52:19: coming into that I think When you have multiple shuttered regions, especially regions where these shuttling lanes are perpendicular to each other then would think of something like the T-junction and this is also what we've been working on.

00:52:37: Can tell us more about Where they are used?

00:52:41: why?

00:52:41: They're used.

00:52:42: Yeah, so in principle that there kind of the natural extension.

00:52:45: if you have a linear shuttler Just our standard cubers conveyor more shuttle You can obviously populate them with qubits who could move from around but he won't get one cubits to jump over The other.

00:52:56: it's not really possible.

00:52:57: He could do some some gate-based swap With some error rate and then you can compute in a linear chain.

00:53:04: But obviously I watch it was two dimensional would be very nice for us too.

00:53:07: also extend architecture in that sense.

00:53:10: Also, you know if we have a two-dimensional architecture it naturally implements the topology for example of surface code like quantum error correction scheme and this is exactly where this T junction comes from.

00:53:21: so here's an example to conveyors might be independent on each other uh... You can join them at ninety degree angle And I did actually do some device.

00:53:33: my recent advice stop measuring.

00:53:35: now writing under publication There, I have a ten micrometre long cubars and then in the very center put another five micrometer-long cubars perpendicular to that.

00:53:46: And they actually joined.

00:53:47: so it is possible for electrons.

00:53:50: my experiment only shut at the charge because this device is so large and its difficult technologically They are.

00:54:00: there's possibly two route than these electron charged if you might initialize into horizontal shuttle lane into the vertical shuttle lane.

00:54:09: And obviously there are some difficulties with that, you have quite dense gating in the center of such a device which we're able to fabricate even academically because they offer very good technicians.

00:54:21: at H&F Nano facility and Research Center Ulish They've got an e-beam set up so electron beam lithography setups.

00:54:30: That allows us essentially bring these perpendicular up to a distance, down to the distance of about twenty nanometers only.

00:54:40: Which is then sufficient.

00:54:41: yeah as I said the quantum dot c electrons have an electron wave function.

00:54:45: in the end it's about thirty nanometers forty nanometers in size and if you get these gates close enough that just coaxes your electron from the horizontal to vertical shuttling.

00:54:55: That works quite well.

00:54:56: Does this

00:54:56: affect

00:54:58: spin when changing its direction?

00:55:04: So in my experiment I only shuttled the charge.

00:55:09: The change and shuttling direction, that sounds a bit like you have an electron, shoot it down on linear direction then apply force or Newton to steer the electron around.

00:55:24: This is common misconception.

00:55:26: people get about this shuttling.

00:55:28: it is essentially from a perspective of the kinetics, what's going on?

00:55:32: It's an stationary process.

00:55:35: We are shuttling with a few millimeters per second but the kinetic energy in electron basically negligible.

00:55:42: so you don't have to steer the electron just give a confinement potential that you deform and then t-junction the quantum dots that they form, they merge and then you break them apart again such that the electron follows on...the direction that we want to.

00:56:04: But the electron just follows a potential minimum if it has enough space for the ray function not yet to be very confined.

00:56:11: And so this doesn't break up.

00:56:12: because yeah For example potential minima next which other with a barrier in between?

00:56:16: but If you ensure that the potential is well defined The electron just follows that.

00:56:20: And actually, these experiments maybe you wonder why we care about electrons alone and not about the spin?

00:56:26: There's always two stages of this experiment... ...and they're both very challenging!

00:56:30: Just moving the electron sounds may be easy but also great to work with from a theory side.

00:56:39: But it is already big challenge because of this restriction which gives us only use few signals to control that.

00:56:50: In fact, in this experiment that Max did we had more than fifty quantum dots involved and the filling of all these quantum dots were controlled by just eight signals.

00:57:04: This is quite remarkable because you could set different patterns move them around.

00:57:08: one can even think of classical computing with these electrons as a digital unit.

00:57:20: And the storage works also remarkably well, they do not escape us.

00:57:25: and it gives completely new ways of thinking that when you have this kind of shuttling... ...and we even some collaboration between people who try to use shuttlings for the current, defining ampere.

00:57:48: So a metrology project... Metrology Project!

00:57:51: But is there any use if you can't maintain this spin coherence and just

00:57:56: move the electrons?

00:57:57: Well so in principle that's what I wanted to say..so If we could move the charge in principle This linear shuttling should not be too different from each other.

00:58:08: We know that we can shuttle coherently in a Linear Shuttling.

00:58:11: Of course, you have some details in there.

00:58:14: The confinement of the wave function changes a bit.

00:58:19: Quantum dot-the-wave function changes its shape and that will probably have some amount of influence.

00:58:25: but in principle we know from coherent linear shuttling apply proper confinement.

00:58:31: voltage drive amplitude needs to be sufficiently large In this sense would expect us get spin around the corner principle, nothing goes against it.

00:58:42: That obviously has to be demonstrated on.

00:58:45: my focus was more in this.

00:58:46: very large devices set fifty four quantum dots into these two shutters lanes so you have quite a lot of distances involved.

00:58:54: where I'm shutting through coherently still remains bit off the challenge.

00:58:58: but principal should be possible

00:58:59: and maybe add too this.

00:59:01: So there's also some projects with PTB on standardizing like yeah basically current right.

00:59:11: maybe last can actually tell that story a little better, but also is just relying on the electron being shuttered and not necessarily this spin.

00:59:22: And I think the shuttler has strong contender there in trying to get back into good standard.

00:59:33: What were challenges you faced when realizing these T-junctions?

00:59:40: unexpected results or surprises in your observation?

00:59:45: Yeah, so the one challenge I already stated that's difficulty to fabricate this from a gate distance perspective.

00:59:52: But then again we have quite well under control twenty nanometers.

00:59:56: but still there is device now it's ten micrometers by five micrometres has over two hundred individual gates why they're in principle connected together and these cooms on these clever chargates such that you only have to apply voltages for four gates per shuttling, so total eight.

01:00:15: All of these fingers on this comb still has to be fabricated and For that You really need a process quite well under control.

01:00:25: These are tiny nanostructures.

01:00:27: They're about sixty nanometers in width.

01:00:30: There's maybe fifteen

01:00:32: or

01:00:32: forty nanometers height And they what is it roughly, I think its about one micrometer long in the end.

01:00:41: The finger itself maybe a bit shorter but yeah.

01:00:44: so this has to be fabricated and the samples need to be screened.

01:00:47: i've fabricated them also myself.

01:00:50: So you have quite careful there.

01:00:52: You have check every step.

01:00:54: don't check too much.

01:00:55: There are some concerns regarding SEM electron bombardment because that might degrade the heterostructure interfaces.

01:01:05: Yeah, but once you have that or all of it under control I could fabricate the sample and then in principle.

01:01:10: The measurement itself wants to have this setup up on running is quite simple conceptually.

01:01:18: But obviously we have these these details.

01:01:21: You have to be sure that your shuttling actually works linearly.

01:01:24: so i can't immediately verify this junction transfer where I go around the corner.

01:01:29: First, you have to verify that they even get there.

01:01:32: Tuning these in and at the beginning not much of an idea.

01:01:34: what these voltages are that we apply.

01:01:37: You're unsure about failed now?

01:01:39: Did linear shuttling fail or did the transfer fail ?

01:01:41: Did the transfer backfail only ,and not the transfer into other shuttlings but only back?

01:01:47: All of those problems had to be figured out and then carefully analyzed.

01:01:54: But once you have that working, as Loz said here are these eight pulses.

01:02:01: Once calibrated, typically they stay quite consistent even through thermal cycles.

01:02:07: So if you warm the sample up it's measured in a dilution refrigerator and cool back down we're actually surprised how stable this was.

01:02:13: And then when doing an experiment there are few more things that were quite interesting.

01:02:18: For example we observed You can easily shuttle more than one electron into quantum dots.

01:02:24: These drive amplitudes as you apply Can be quite strong Even within the limited thermal budget.

01:02:31: In that case, you can stack multiple electrons in the same quantum dot and perfectly fine shuttle them linearly around a corner.

01:02:40: You could even... I tried this by playing around at the end of my measurement ...you can load these electrons into separate arms where independent control moves towards center then stacks there.

01:02:54: so something like is quite interesting.

01:02:57: I guess this probably won't work coherently with the electron spin.

01:03:00: Maybe we are surprised there, but for charges you can play around a lot.

01:03:05: and i think that other final aspect of what's interesting from me from the T-junction is it allows to do essentially alluded earlier on the swap gate by just swapping electrons.

01:03:17: So normally if you have a linear chain, it's quite difficult to do a swap.

01:03:20: You'll have these actual quantum gates in order to transfer the quantum state or swap the quantum states of two qubits and this is error prone principle.

01:03:30: while t-junction swapping just take two electrons shuttle one ahead and one following shuttle through the junction go up with your one goes straight for other then swap them around.

01:03:45: This works quite well.

01:03:46: Obviously, I didn't test it for spin but from the charge perspective It worked as well As we do is the shuttling does and then it was very much better than what you would expect.

01:03:57: So we won't expect that.

01:03:58: the chart shutting is a limiting factor For this been shuttled And You can see them when you get this swap operation essentially for free.

01:04:08: Could also imagine doing other two qubit gate operations in these shuttling channels, as you already mentioned that you could have multiple electrons coming together?

01:04:20: Yes certainly.

01:04:20: So for that the T-junction is obviously not really required.

01:04:24: but if we have a linear shuttle device and for example here's small magnet on this side which might use other aspects of two qubit gates then when you shuttle away from it magnetic field gradient that is changing, and then depending on how fast you shuttle through it.

01:04:42: You can cause essentially rotations in the block sphere And I think this has been now demonstrated...I don't know whether this was published by now but yeah This has been demonstrated in a device.

01:04:59: so this definitely possible.

01:05:01: Yeah, at the moment.

01:05:03: The fantasy is triggered by this shuttling and one can have completely new schemes.

01:05:08: usually One tries to avoid too many qubits together because it's difficult To control them.

01:05:14: so one big point Big plus of those shuttling approaches that That only two qubits meet each other in a manipulation zone And all the other cubits are far away.

01:05:25: So some crosstalk Can be avoided.

01:05:31: Yes.

01:05:32: So now that we have discussed so many fascinating results from your experiments, what's really interesting about research is all the resources you've got are not just sitting in a lab but then taken these insights and dead into creation

01:05:46: of start-up called

01:05:47: ARC systems.

01:05:49: Lars can tell me actually what motivated you and co-founders to turn this academic research result?

01:06:01: So first of all there are a couple of startups in the quantum technology, quantum computing business.

01:06:08: But off course some requirements and one thing is that we have this idea vision how we want to have millions of qubits working together.

01:06:20: it's all around this shutting approach And we filed on twenty patents.

01:06:26: That is also good base for startup And now we can try to realize this vision.

01:06:35: and there's one definite problem in this is that We have still kind of small demonstrators.

01:06:41: We are can operate fifty electrons, but what do you aim at his millions?

01:06:48: and What we then need is a high degree off professional professionality In many ways software fabrication.

01:07:00: So we look at what stage do have to transition from more academic devices, two industry devices.

01:07:07: And this is already what are doing in projects where we have collaborations strong collaboration with Infineon and there's some other Leibniz Institutes, Fraunhofer Institutes involved.

01:07:19: it's a great opportunity Germany that these many institutes when they're working together can really something remarkable.

01:07:27: And now we have to also bring this other things together, one thing for example is that when you want to reach there it has to do mass characterization a high degree of professionality in all the steps.

01:07:42: I can give you an example where at moment limitations increase the number of qubits and a number of wiring.

01:07:57: We have to take more care about setups, reliability of bonding wires.

01:08:04: The PCBs are getting more complex smaller capacitors and resistors on it.

01:08:13: that is more difficult to handle And this all has to be professionalized.

01:08:17: one has specific groups that working hand-in-hand and fabricate these devices.

01:08:23: And it's still working also quite well in conjunction with the academic projects, for example we have now a master thesis very simple on bonding if you communicate.

01:08:35: but our institutes here in Aachen who are working on modules good experience in this high degree of reliability that you need for such a project.

01:08:54: And all these groups have to work together

01:09:00: and we've seen a lot of spin-offs from academic institutions like researchers, becoming entrepreneurs especially the quantum industry.

01:09:09: Do think it helps come form an academic background as founder?

01:09:14: And how do the founders expertise and roles complement?

01:09:20: with one another?

01:09:22: So I'm CTO together with Hendrik Bloom.

01:09:25: He's also CTO.

01:09:26: it definitely helps that we have this scientific background.

01:09:30: It gives us deep knowledge, but i think its not sufficient for a startup company.

01:09:37: so you're all very happy That We Have co-founders who have A Very Different Background.

01:09:44: The other co-founder is Dr.

01:09:46: Markus Beckers, he's an engineer side.

01:09:51: He's a very agile networker Knows the patent business very well worked as a patent engineer and there also Doctor

01:10:03: Wolfgang

01:10:04: Meisner And he has a large experience in background In large projects that work with billions of dollars, fusion reactors.

01:10:18: He also was at McKinsey as a consultant.

01:10:22: so he has this background on how the economic part such startup works and these are very essential to do such project.

01:10:34: it's really complicated endeavor many, many factors.

01:10:39: So I think with these co-founders and you asked about them they're back on to be have a good start.

01:10:45: we have the patents And what we now also need are Good employees students.

01:10:52: in the end We hope that The excellent students that we have at WTH Aachen and our group That they like to stay In the company and then can attract very good employees from other companies.

01:11:10: At the moment, the status was in last years that we had kind of a brain drain to other company.

01:11:18: you have this student couple students working now at Intel Microsoft and so on And okay these are very attractive positions.

01:11:28: We hope with ArcSystems can also give some opportunities for the students here.

01:11:34: Also,

01:11:35: did you feel personally?

01:11:36: You need to change your way of thinking in order to meet the industry requirements like.

01:11:41: The very approach research or they will plan.

01:11:44: so Did you feel that he had to change their way because not to meet startup culture?

01:11:49: on the industrial requirements

01:11:51: yeah there's some different Way off thinking.

01:11:56: this is something that we implemented already in the past and institute one has to do a much stricter project management.

01:12:06: These specialized groups working together has also a couple of consequences.

01:12:11: I can here give one example and on our academic side, Max Baer fabricated these T-junctions in his own.

01:12:20: he precaracterized them tested then did some simulations put into fridge repaired the fridge measure device.

01:12:28: it worked at end.

01:12:31: very fascinating result then also big effort, because you covered all this production chain.

01:12:37: It took a couple of years but it's possible.

01:12:41: But now when we want to professionalize that and have special groups working together And what we know witnessed within the year is there are some interface weaknesses.

01:12:55: So first fabricate your device Then give it somebody else who has to bond them We need to have very good workflows, too.

01:13:07: Always qualify whether the previous step was working and this is something that I haven't learned during my studies.

01:13:17: And now it's the hard way of working and we cannot do just trial-and-error anymore.

01:13:21: So from academia some thinking about how all these works probably will also work next time qualifies on things and really check how high is the yield.

01:13:34: Is the yield sufficient?

01:13:36: For example, for bonding we tested some bonding with two hundred bonds.

01:13:41: that gives us a certain yield when everything works.

01:13:43: but if it's really sufficient or you want to have large device that consists of one thousand bond wires then you need much larger sample size in these kinds of things.

01:13:53: now enter And make protocols.

01:14:00: We have to work on these workflows.

01:14:02: And that also involves now some kind of different thinking.

01:14:12: Kudos to Max!

01:14:16: Yeah, yeah... That's actually then where it comes from – I know the sample is dated across thermal cycles and had quite a few of those.

01:14:23: It happens if you fix your own fridge

01:14:25: so….

01:14:27: So now A lot of experiments happen at your research labs.

01:14:33: How do you decide which all these experiments are?

01:14:36: will be the industry related requirements and how do you shift it from a research lab to start-up company?

01:14:47: So what I think is that in academia for paper, its sometimes sufficient have one star chip networks operates And in similar way than also on benchmarking is done.

01:15:05: So we have, of course certain benchmark schemes that are common among quantum computing how to measure the fidelity of a single gate or two-qubit gate and so on... And this also something that transfers to the company.

01:15:24: but then probably what's no different it has to be done continuously.

01:15:30: It not only one lucky shot by a PhD student within his project.

01:15:36: It has been done all the time reliably, for that one needs to have software and equipment for mass characterization.

01:15:47: we get from industry wafers of chips on such a wafer.

01:15:55: there are several hundred thousand chips also to give feedback through the industry, which of their fabrication steps worked and which have not worked.

01:16:07: So this is completely different than fabricating ten devices in a clean room in academia And for that reason we also within the company make our own equipment.

01:16:20: We have an example device called M-Pact a career start with electronics that we can use for mass characterization.

01:16:31: It's also something, they had to be sell to other companies and this is generally something I think it was good at.

01:16:38: these companies there are lot of them come from academia in their work together project.

01:16:43: They have some useful exchange.

01:16:47: now on the startup side And you know what?

01:16:51: In the end We still need this common effort to bring quantum computing forward specifically for semiconductors because it's this fabrication, and so complex.

01:17:04: Yeah before we move into the end I just want to ask you for the listeners if you have open positions at your startup in your group

01:17:17: You

01:17:17: can let them know.

01:17:21: Yes We now a couple of new projects.

01:17:27: Yes, I think this is now confirmed and therefore we have many open positions.

01:17:33: And these concern many different fields.

01:17:37: you need a good programmer if there are possibilities opportunities for theory but We also need technicians working For example on bonding testing devices.

01:17:57: We have here an open position at the moment in Jülich, there will be also more positions announced for ARK soon.

01:18:07: so this is because we just got a lot of projects and yes it's definitely worth to check our webpages And as I said, it covers a lot of different topics.

01:18:27: Electro-engineers technicians theory and experiments.

01:18:33: programming is the joint effort also

01:18:37: like

01:18:37: in the near term.

01:18:38: what are the technological milestones?

01:18:40: Are you aiming from our

01:18:43: system's point of view an arc systems?

01:18:45: we're at the moment tuning up five qubit chip that Is set up in Yulich.

01:18:53: So this is the first coin, so we aim at selling quantum computers.

01:18:58: The first quantum computer will have five qubits in silicon and it has implemented our shuttle technique.

01:19:06: then next project would be around a two hundred qubit chip And that is then the stage... This probably also comes from industry.

01:19:21: And then we have at some point the transition step, the AMET that around three years in the future where you have to implement seriously the cryo electronics.

01:19:31: There are some first steps for we have a cryo electronic chip next to our quantum chip but they have to be packaged together.

01:19:38: so really make this this tie-up structure and I think than will be time when we have scale up boosts can bring them together high yield from The industry chips can bring them together with cryo-chips, have a thermal management.

01:19:58: And when this works we will see where are the limits and to what extent we know it can then scale up?

01:20:11: Also like... Are there any technical breakthroughs you hope to achieve so that you can significantly scale these qubits?

01:20:20: So the main technical breakthroughs that we need is first to get this last stopping point for shuttling out of the way.

01:20:31: This is the valley splitting, and we put at the moment a lot effort into it from all different perspectives We try to approach.

01:20:41: there's one way of improving material but they are plan A B & C to just invest more into complexity of the signals and get around the points, so do a kind of slalom shuttle.

01:20:57: When we know this two-D map as much described later, I believe you can also do these kinds things.

01:21:03: there is also a plan maybe go through another material system that doesn't have this valley issue but then may be other issues.

01:21:11: It's Germanium silicon, Germanium is a candidate That suffers from large spinobit interaction.

01:21:19: So this is one side and the other side, which I think it's a big challenge to implement the cryo-electronics And there you need not just cryo electronics working but also on small footprint and thermal management.

01:21:36: when we bring together the cryoelectronics and quantum chip We are not allowed heat up the quantum chips that much.

01:21:44: The called Quantum Chip or quantum layer, is more robust than superconducting cubes.

01:21:49: So the next big challenge was really to have the cryogenics working and for that we need them not just working as a side chip but what we need is to have right footprint of their electronics.

01:22:05: We need to have thermal management when everything is packaged together And this will be quite exciting.

01:22:12: And this is not the core knowledge of physicists.

01:22:17: It also involves circuit designers, electrical engineers and it's a big cooperation effort.

01:22:25: Thank you Lars for your insights.

01:22:27: So before

01:22:27: we wrap

01:22:28: up today I'd love to hear some personal reflections.

01:22:31: so looking back what have

01:22:33: been most

01:22:34: important lessons learned in translating academic research into

01:22:38: functioning startup?

01:22:40: Oh that's very good.

01:22:46: So first of all, it is really an endeavor and looks like a big challenge.

01:22:53: And one needs what I said this high degree of professionality talking to investors in different ways than talking to physicists.

01:23:03: This definitely won't work but its similar towards researchers are always doing.

01:23:09: we have of robustness against frustration.

01:23:15: And yeah, it's maybe a continuation off all the frustration that we had also while we tried to get into this shuttle.

01:23:26: so as an example I told you we have this Quanterra project for the shuttling in two thousand seventeen and then fortunately ML-FORQ are supporting there too which was where they really started running before.

01:23:38: I had two proposals that didn't work and they also have the shot thing in it.

01:23:43: And there was eminuta, Sonderforschungsbereich... These kind of things then continue.

01:23:51: when you go to a company.

01:23:55: these issues need to be solved.

01:23:58: Also for other researchers who are listening Who might consider launching spin-off companies from their works What advice would give them?

01:24:08: I think you need a good team.

01:24:11: It's not sufficient to have just knowledge about the technology, You also need to know about the economic part of start-up.

01:24:21: One has to be brave to start with this.

01:24:23: it is time invest and one can try.

01:24:28: maybe in Germany people are hesitant.

01:24:34: with new technologies and other companies like us seems to be easier, but in the end one needs to try.

01:24:44: Okay so that I think we come to an end.

01:24:48: thank you Lars, Matz & Max for sharing your insights experience and the exciting work you are doing.

01:24:54: this topic was particularly very interesting.

01:25:02: It's been a pleasure having

01:25:03: you on the podcast and I'm

01:25:05: sure our listeners have gained a lot from

01:25:07: hearing

01:25:08: about SpinCubits, The Science & Story also behind

01:25:12: ARK

01:25:12: systems.

01:25:14: Yeah thankyou very much for inviting us!

01:25:16: Thank You.