 Kiitos, minä olen again between you and the lunch, and I hope I will make it in time. You know, I always had talks where people talk about a lot about everything, but I wanted to learn about one thing, but now I do the thing myself. I kind of have many little things here, although I try to concentrate on the heat-based measurement of qubits and the on-demand dissipation for qubits, but I have a little bit of other stuff as well, which I think the latest stuff, which you might find interesting. For those who don't know my group, I just wanted to show the photo of the lab, that yes, we are also doing experiments. It's just an orientation for you, and of course I'd like to thank my research group. This is a group photo from last summer, and I'd like to thank the founders. We had a quantum computer project that is now ended, and then we are also funded by the ERC. By the way, thank you. There must be some people in the audience who were referees of the ERC advanced grant that was just granted to our group, and so thank you for that, and we're looking for postdocs. There's a call open, so if you know good experimental people on superconducting qubits, please let them send applications to us. We also, about three years ago, we spanned out a company from my research group, IQM, and that company is now already a significant size, so part of this work is done in collaboration with some of the nice guys working at IQM as well. But that's not all, of course. This is the quantum computer working group. That is now the late quantum computer working group, and I'd like to thank also these people, especially, of course, Jukka is the head of our center of excellence, for example, and we worked a lot with Jukka, and then I'd like to thank also Pertti, who is in the audience, and Tappi, who we worked a lot. Of course, there are a lot of international collaborators like Joachim and Fabian, and I think Jean-Louisie is online and others, so thanks for your input to this research, and we're really excited now about this Institute Q, the National Finnish Quantum Institute, where a lot of universities join and we join forces. By the way, I forgot to say that I'm a professor at Aalto University and VTT, so there are also VTT groups who have collaborated a lot with us, and especially all the work I'm going to talk about is in collaboration with VTT as well. So, we came up recently with the new superconducting qubit that we call the Uniman. I'm going to say a few words about that, and then in the ERC Advanced, the idea is to basically use the Uniman, improve it, and then also to do accurate control, readout, and reset of the Uniman circuits. And for control, we recently built up a molecular microwave source, and I'm going to say a few words about that soon. For readout, we have the world's most sensitive ballameters demonstrated, and that I'm going to talk about. We use one of our multi-channel driving schemes then in the end to do a very fast readout based on ballameters, and then for reset, we have the quantum circuit refrigerator. So, I'm going to mostly talk about the readout and the reset part. I'm going to say a few words about the Uniman and the control. So, but in the end, the idea is to combine these with the Uniman circuits, like I said, to make this sort of scalable demonstration. Demonstration of scalable qubit, very accurate qubit type in superconducting platform. So, yeah, the Uniman has been done in collaboration with IQM, so we have, this is one part, very good collaborate together. And you can see, this is the artistic image of the Uniman. In practice, we realize it by the CPW that is crowned at both ends, and we have a single social junction in the middle. So, it's a very simple structure. It doesn't have any high kinetic inductance materials. It's just the normal materials you use to build up superconduct qubits anyways. And it actually has a very high unharmonicity. It's not very high, but let's say increased unharmonicity compared to the transmon. Even if it's a 50 ohm of CPW, I think we used 100 ohm CPW characteristic impedance in our experiments. It's fully insensitive to low frequency charge noise because there's no superconducting islands. So, it's really shunted to ground with just superconducting wire. And it's also insensitive to homogeneous flux noise. It's only the flux difference through these two loops that actually bias it in a flux sense. This is the lumped element model. And with this lumped element model, we were actually able to come up with an exact mechanical analog of the Uniman, which is like this twisting beam, whereas there's a harmonic force, the twisting, restoring force of the twisting angle, and then you have a mass. And this is actually an exact mechanical analog. And the Uniman works so that the sort of the mechanical, the gravity force, the force to the gravity, the harmonic force to the gravity cancels the harmonic force of this beam that is twisting. And that's why it gives rise to this high unharmonicity of the qubit. All right. So, we made these Unimon chips. Actually the first Uniman we fabricated was the qubit C, oh, qubit B here on this plot. You can see that the unharmonicity is relatively high, about twice of the normal transmission unharmonicity. And that we get at the half flux quantum spot. That's where the sort of the mass is high up. That's that point. Zero flux quantum is when the mass is down here. And there you get low unharmonicity. And we could model this with quantum theory. And you can see that the model fits quite well at both of these spots, like zero flux quantum and half flux quantum. So we understand quite well this circuit. And we were able to do with the qubit B, I think this is the qubit B result. So this was the very first Uniman we ever measured or even tried to measure. And we were able to get three nights of gate fidelity with 13 nanosecond pulses. And actually if we could have gone even to faster pulses we could have got higher fidelity. And you can see that the fidelity is quite stable over time. So here you can see we get these three nights of fidelity over eight hours. It's such a single calibration in the beginning. So we don't have to adjust any parameters, any fluxes for eight hours. And we still get three nights of fidelity with the very first qubit we measured. So this is why we are quite excited about this. And we'll see how the next batch works out. So about microwave source. This still we have not connected with the qubits, but we measured the source itself. The source basically is a, well, we current bias it, but because of this on-chip resistor, which is about one ohm, we actually voltage bias and chosen junction. And as you know, when you voltage bias and chosen junction you get oscillating current with the frequency given by the voltage. And then when it's coupled with this LC oscillator here, then you get the Saphiro steps. And basically, when we work at the Saphiro step, we get coherent wave out, which is classical coherent wave sinusoidal voltage out. That's the idea. And this is just some images of the circuit. And we were able to measure the power and the phase noise. And I just wanted to highlight that the phase noise we get is quite low. So this is our, who is the LL? And the rate is when we turn on the RF power. So we get about minus 100 DPC per hertz at 10 kilohertz offset, which is good enough to do four nines of fidelity for single qubit gates. So that source as it is now, if we could just pulse it, could work as a nice. What about the second harmonics? So the harmonics and so on? Oh yeah, that's good. I think we didn't measure them that carefully. But we think you don't get them that much because you lock into this frequency. And actually to get this low phase noise, we did have an injection locking tone as well. So we put a weak signal on top of the bias current to lock the phase and the frequency of the oscillator. This is like 100 times less power than we get out from the device. That's okay. All right. But then this is just to a little bit compare what we achieved to do. So we got actually this tens of picowats of output power at the low phase noise. And this is what you need for this high fidelity gates. The previous millicolven sources had much less power. So this, they were not really practical, but of course we were not the first one to do this kind of source, but we were just able to do it with better parameters basically, better performance. All right. So then I move on to the readout of superconducting qubits with the balometer. So I will spend maybe about 10 minutes here or so. So as you know, nowadays qubits are read out typically by putting a microwave tone to a readout resonator, which is dispersively coupled to the qubit. So the phase you get out depends then on the state of the qubit. And then you have a quantum unit amplifier. You amplify it. And then you measure in the end the wall that's coming out by down converting it to mixers, et cetera. And this is great. This works very nice. You can get three-nines of fidelity for readout. But, and it's about, it's quite fast. You get about 100 nanoseken timescales there. But then you need, it's very bulky isolators here. And actually also the Tupa that is a quantum unit amplifier is quite large in size. So when you scale up to a very high number of qubits, you would like to do something about it. You can't build a 10,000 qubit quantum computer or 100,000 qubit quantum computer with this technology. So, OK. So this is the typical dispersive readout. In the IQ plane you get this kind of, this kind of sort of single shot experiments and average traces. So what I'm talking about here now is basically to replace this part, fully this part, it's rid of it and says pulling a 50 ohm resistor here at the end and then we measure the electron temperature of the 50 ohm resistor. So this is another way of measuring the qubits. I think nobody else has done it before. Tell me if somebody has. We will slide to the paper. So, and we use this multichannel driving here instead of the normal driving because in the multichannel driving we can basically have the qubit ground state scenario at almost zero power and then when the qubit is in excited state we can have high power coming out. So this is good for the volumetric readout because then you have low temperature here and high temperature here for the electrons. This is at least our plan to do that. You get actually quite high power out if you choose the parameters, right? There are other reasons why you would like to use a balometer like the very low pump power compared to two-pass and you know naturally cold bath or the qubits. You don't need the isolators and a small footprint. This slide, I just want to say that we built a graphene balometer recently and we read it out also using microwave reflection, but that is a half of gigahertz frequency. That frequency has almost nothing to do with the qubit frequency. We can choose it quite freely because we use this temperature to frequency conversion and that is very insensitive to the input frequency. So the input frequency of the balometer is the qubit readout pressure on the frequency and the readout frequency of the balometer itself is completely independent of that. It's about half a gigahertz anyways. But we measure very low noise equivalent powers like the lowest anybody has measured for the balometers or almost on part of the lowest numbers and then if you convert that to energy resolution you get about energy resolution of 30 gigahertz times h bar which is actually good enough for qubit readout because in the qubit readout you have a few photons at least coming out from your readout resonator so you don't only need to measure a single photon. All right, so what we did here is we have the balometer here, we have the readout resonator, we have the qubit and we can then excite the readout resonator in this experiment we did it continuously and then we can drive the qubit or the readout resonator as well. Okay, actually in here we kind of first measured the readout resonator with the balometer so this is the balometer signal that we get out after down conversion and you can see here that when we change the flux bias of the qubit the resonator frequency also changes a bit and this is basically just you would do this normally with the VMA but here we do it with the balometer. Okay, just basic characterization and then we also can measure the qubit so when the qubit excitation frequency is in resonance we see a different signal at the balometer so this is the qubit and you can see the frequency is very different about 2 gigahertz different than the resonator frequency so this is the first time I think we have seen or anybody has seen a qubit with the balometer it might be wrong it was probably somebody did earlier but we don't just know about it but we think it's the first time and then we went to measure traces after that so basically here is like a 64 average trace of the balometer signal with honor of resonance excitation for the qubit we can apply some filters then to have better signal to noise ratio and with this sort of scheme we were able to measure value oscillations of the qubit you can see quite many radio oscillations here not much decay because the qubit T1 was tens of microseconds okay, this was the first time we measured radio oscillations but then we started to do a little bit more careful analysis and we measured now like the balometer signal as function of the readout resonator power and the sort of the excitation, the pulse length to the readout resonator and it turns out that with these parameters you need a few microseconds of readout pulse I think you can get it shorter if you just optimize the parameters but this is with this sample and then we choose now 10 microseconds pulse where we have a reasonable contrast in the image and then the millivolts that I'm going to plot on the next slide is the average over this red part of the pulse so we average the wall that we get out in this part that's like what I plot on the y-axis here so this is our new radio oscillation measurement we got this just a few weeks ago no no no, I just explained that you can also drive it so that it's in the amplitude it depends on how you drive the resonator and yeah, yeah so this is the usual chevron pattern that you get you can see that you know when you're on resonance it's about here when you off resonance the fringes get a little bit faster and there's only 500 averages here so this tells you about the signal to noise the volometer actually is a rather nice detector for these you don't have any noise coming out of your amplifiers so you don't need to average the amplifier noise and then we went to single shot experiment of the qubit and you can see here the single shot distributions they're clearly offset signal to noise ratio is not very high so we could say that there is finite fidelity finite single shot fidelity but it's very low fidelity but still there is finite one but these experiments were actually done with the metallic balometer which is not as sensitive as the graphene balometer so just by moving to the graphene balometer here we would get an order of magnitude less noise I would assume just by doing that maybe even more than order of magnitude less noise that the principle works now it's about fine tuning and moving into better balometers and better setups I'm not sure I have understood that the 10 microsecond perth is just for the conversion and having the photon in the balometer or it's also for measuring the balometer signal no the balometer signal actually we kind of take only this last part here for microseconds in this case but it's basically that the balometer you need enough power to go to the balometer to have a reasonable signal and that's why we chose such a long pulse but of course you can have more power by using a resonator that has higher kappa basically broader line width for example we measure it for a long time and we only take the signal in this part basically we apply that kind of a filter that we just take this part of the signal and we do a time average there and that's what you see here basically on this scale or on this scale but that's just one way of doing it you could do it differently but we just did it now this way Mikko, do you need to wait for the balometer to reset? Yeah, the balometer also the metallic balometer has relatively long thermal time constant and if you turn off the power then the probe power then it resets I don't now remember exactly I mean it depends on the operation point also what is the thermal time constant but yes it is much higher than the microseconds might be 100 microseconds or even a millisecond scale depends on the parameters but then the graphene balometer is nice in the sense that it has only like 200 nanosecond time constant so that is kind of then a more feasible maybe for the qubit readout in terms of scaling up or using it for a good readout but this is basically what I'm now saying trying to say here is that we have tested the principle that it works and it actually we were surprised that with this metallic balometer it even works this well All right, so now about 10 minutes time left I'm going to talk about now the reset of qubits with the quantum circuit refrigerator and just to say that we did try two different approaches like a tunnable basically resonator with the resistor in it there are some experiments also on that how to reset resonators we didn't have qubits there but then now I'm going to focus on the quantum circuit refrigerator which is basically this assigned tunnel junctions that we voltage bias and when we turn on the bias you turn on the dissipation and there's also a paper where that kind of combines these two and we studied exceptional points in that and there's some latest theory papers on the QCR here in collaboration with Gianluigi and the latest experiments are in this paper I will also say a few words about those latest experiments but this is the principle like that when we turn on the so you have the SINIS junctions here so you voltage bias them and then you have a quantum device that is coupled basically to this island here in this case there are different scenarios I mean different ways you can couple but this is like now the simplest thing that we did first or maybe it's not the simplest but that's what we did first and when you voltage bias it then the photon assistant tunneling can start to take place and when your bias is less than the gap kind of the absorption of photons from your quantum device that is this one is preferred over the emission and this is kind of the phenomenon you use it's been known for a long time and as you saw the previous talk, the first talk of today was using SIS junctions and photon assistant tunneling there and there's been a lot of work on those as well but this is kind of what we did we did the SINIS junctions in these experiments so the device basically looks like this that you have a resonator here and you have the QCR at the end and we measure reflection and this is just the normal reflection data you get the phase and the amplitude and from those phases and amplitudes as function of bias voltage we can basically extract what is the dissipation rate that the QCR induces on the resonator and what is the frequency shift of the resonator due to the QCR and there's two different samples here and you can see that we can quite nicely reproduce the theoretical prediction for those with very few fitting parameters and also the frequency shift we can explain quite well with the Lamb shift given from the theory basically now we are mostly interested when you do QB3 set what is the on-off ratio and you can see that it's maybe three or two and a half orders of magnitude or so in this case so you can tune basically T1 of the QB by that order of magnitude if everything works like you want and it works we did this kind of experiment with still using a pulsed QCR on a resonator and here we can see the resonator we first excite the resonator with the coherent tone and then we let it to ring down and then during this ring down we turn on the QCR for a very short period of time we basically apply a voltage pulse here and then we can see that the signal drops very quickly and this was kind of the first demonstration that okay you can also turn it on and off in a reasonable time scale if you think about QB3 set not just in the previous experiment but applying a DC voltage on the QCR okay and now then we moved on into resetting qubits so here you can see a qubit that the qubit is now capacitively coupled to the QCR which is actually just right there it's so small you can't really see it there's an on-chip low pass filter as well and basically this is the circuit scheme so you apply the voltage here and then the current passes through this loop and the qubit island is now capacitively coupled to the QCR island as I showed you in the very beginning and what we do is that we can excite the qubit with the pie pulse in the very beginning firstly of course weight of the qubit is going to be reset by itself and then we drive it with the pie pulse of the excited state we apply a voltage pulse to the QCR with the variable length and then we measure the qubit population and as you can see here the end result the excited state probability as a function of the QCR pulse length decays with the low amplitude it decays exponentially and the exponential decay constant goes down exponential decay gets faster when we increase the amplitude of the QCR then we can see something that is a little bit funny when we go to higher amplitudes it's not purely exponential anymore it's like more to it and we think we understand at least partly why this is it's because it turns out that these junctions are actually quite resistive because there's the superconducting gap and this capacitance value is relatively large in comparison so that you get a kind of a slow RC sort of charging up of the island and I think that's why we at least could see two time constants and we're kind of in the middle of working this out in a more detailed way and I think we can get it working it's just something that we didn't think about first when we started to do these experiments that okay there's this charging up of the island that actually matters as well but anyways this is again the demonstration that the QCR can be used to reset qubits in a variable decay time and I think this opens up a nice toolbox for more experiments whether you wanna do like this environment assisted entangled state or dissipation engineering or whatever I mean you have this kind of toolbox soon that you can then apply with qubits as well I'm almost done I could say still a few words about the RFQCR that we published very recently so those previous ones we only applied like a DC voltage to bias it or pulsed voltage to bias it but in this case we actually use a second mode of the resonator and we drive that second mode of the resonator to induce this multi-photon assisted entangeling processes so when you apply power to the second mode of the resonator you can turn on and off the QCR also by that RF power and this might be nicer in some cases where you don't wanna apply fast DC pulses but instead you wanna quickly pulse RF which might be simpler from practical point of view in some scenarios so here is now again the dissipation rate as function of the DC bias voltage at different RF powers and yeah this data is yeah you can clearly see that actually even at zero bias voltage if we apply enough power we can make the dissipation rate higher and then you can see also these sort of bumps and humps and they are due to the multi-photon processes actually and we can also now plot the frequency shift of the resonator and you can see that also then has these bumps and the theory we have here we have only a few feeding parameters for the whole data so for all different powers for all different voltages we use the same sort of parameter values so you can see that even though the feed is not perfect it does explain many things that is going on in the sample alright so I think I'm pretty much done here and just to recap that what I talked about and you know if you have any further questions I'm still happy to answer them so yeah thanks for your attention thank you very much exciting stuff so we have several questions what about the back action of this ballometer readout you drop the isolators and the usual measurement chain so did you evaluate that? yeah so that's the nice thing about that basically the ballometer input is a very cold resistor so that's exactly what you want right so that's the back action that you will have back action of a very cold resistor but it's a fully destructive readout you always get zero at the end for the qubit whether it was in zero or whether it was in the one thing no no no no no no see we don't measure the qubit in this case directly we don't measure the excitation from the qubit directly but what we measure is we measure the signal coming out of the readout resonator this person will couple to the qubit so you don't convert the qubit photon into a photon in the ballometer it's just a signal coming from further resonator yeah yeah there was also a question in the back but if this is related then the discussion is not okay then I will come back to it yeah I will come back to it about your ballometer so this time resolution that you mentioned 200 nanoseconds what is it determined by? is it electron phonon coupling or some capacitor resistance do you have any idea? well first of all it's not a time resolution the time resolution is set by your electrical circuit that it is much faster than the thermal time constant of the ballometer the thermal time constant of the ballometer that was in the graphene case down to 200 nanoseconds yes that is determined by the heat capacity of the graphene electron excitations and then the thermal conduction to outside world and yeah that's a good question sometimes yeah I mean it might we have in the paper more details as far as I remember it can electron phonon can actually be there also a significant contribution because in graphene the power is not to the fifth like in the metals but it's slower power and then when you cool it down it actually can have a larger contribution relatively speaking than the photon coupling but certainly the photon photon thermal conduction also is not negligible I think in that experiment as well but yeah this is something that yeah can be also studied studied in more detail like what is exactly limiting it it's again a question about the ballometer you nicely explain the motivation as getting rid of the circulators and the hemp amplifier so you have something way more compact and scalable which is I think super cool but then on the circuit diagram you showed you were using a circulator to read out the ballometer so I guess probably dumb question there is something I missed but yeah yeah yeah that is exactly right in that experiment we did use an isolator and sometimes you even use a JPA to boost the ballometer signal you get about a factor of two by the way only with the JPA you don't get to the improvement however actually in the quite first experiment we did we didn't use an isolator we just used resistive splitter so basically lose some of the ballometer signal but the thing is that if your signal to noise is already good enough that comes out of the ballometer it doesn't matter I mean because the ballometer itself is also amplifying in a way the signal and it improved design I'm pretty sure we can do it such that then it doesn't matter then you don't have to use actually this isolator anymore I just wondered did you for the ballometer read out did you try or consider using kind of the Jains coming nonlinear read out so I mean this thing yelled many years ago just put it you know if I put in enough power into the read out resonator it bifurcates by itself right which gives you much more read out power we did not do it no I had to study I mean we went to this higher power region with this one microsecond pulse length but we were not yet smart enough to kind of employ that sort of bifurcation nonlinear to our advantage I think this is another thing that one could certainly try I would have a very quick question about your qubit sorry I forgot the name but it's a flux qubit right it's basically a flux qubit or if it's not what is the difference to a flux qubit well yeah some people say it's a transmon some people say it's a flux qubit some people say it's a fluxonium so the the charging energy is very close to that of a transmon in the one that we made but on the other hand and it has a single junction like a transmon but then inductor to ground so in a way it kind of looks like a fluxonium then because you have an inductive shunt like an inductively shunted transmon right I mean yeah or inductively shunted but the point is that that it has like it kind of what we try to do is to combine these sort of properties that you would like your qubit to have like insensitive to charge noise fully okay you put an inductive shunt it's a single junction it's very simple there's no islands but so ever that can charge up and and then you want to have high unharmonicity with the standard fabrication tools and materials that you have in your fab and when we put all that together we get the unimon it's not like one thing but it's kind of it's like the combination of everything and then it turns out that actually in this parameter range of the fluxonium circuit yeah you got one of the EL and EJ have to be equal or they are close to equal to have this cancellation effect that I talked about and it turns out that almost nobody else has gone with that parameter point I think it was just in Johannes Fink's group they actually have quite close close shot of course the parameters the circuit was completely different that's what he used but it turns out parameter point that might be very good so you always need to make compromises when you tune the parameters sometimes you know how it is so or was it actually in your group? no it was the Fink's group yeah yeah alright so I think we will have plenty of so what is the square run what are the square run stuff? I think we have one sorry let's talk about that yeah so as you saw we got three nice of single qubit gate fidelity with 13 nanosecond pulsies so you can work out what is a T2 it's about 10 microseconds in that case you know there's factor of 1000 difference between the gate length and the T2 and we don't really know why it's so short because the fabrication recipe and people and everything used to make the unimon qubit typically gives you 50 microseconds of T1 and in this case actually T1 was like 10 microseconds or even a bit lower depends on which qubit the numbers are in the archive paper and yeah I don't know whether it was just a back batch or whether there is something more to it and we have to get the design we can get around many things so there is a chance if we improve now the T1 still to 50 or 100 microseconds and the T2 as well we can get four nice we don't even need to go that far to get four nice actually but of course the real things tend to show four nice for the two qubit gates I mean that's what we want to do now three nice is the record in supercontin qubits and we want to go to fourth great thank you very much thank you all so we will resume at two o'clock