 Hi, my name is Jihan Kurter, and I'm a resource scientist at IBM. Today, I will tell you a little bit about the effects of the quasi-particles on the coherence of IBM's two-dimensional transmon qubits. This work has been done with a collaboration between TJ Watson Resource Center and Almodin Resource Center, and some of my colleagues are listed here. I would like to thank them. They were involved in various stages of this work. Let's start with Josephson Junction. A Josephson junction is the main element for constructing a qubit, which is basically made of a tin layer of non-superconducting film, sandwiched between two superconductors. Here you can see a cartoon of a SIS Josephson junction. S is a superconductor, I is insulator, which is the non-superconducting part, which constitutes the barrier. Such a structure at finite temperatures can accommodate electron and hole-like excitations out of superconducting ground state, mainly we call them quasi-particles. When we apply microwave radiation to those junctions, quasi-particles directly coupled to the electromagnetic modes of the qubits and possibly limit the coherence of our devices. When a quasi-particle tunnels across the junction, a physical parameter defined as charge parity switches time, consequently we expect a small shift in the qubit transition frequency. So the charge parity can take values plus or minus one, or more traditionally odd or even, depending on the number of the quasi-particles tunneling across the junction. And this cartoon in the middle shows the various quasi-particle tunneling events, and those quasi-particle events can manifest themselves as transitions between the charge parity manifolds in the energy spectrum of a qubit. You can see it on the far right. So the question is, what type of qubit do we need for these measurements? At IBM, we focus on transmon qubits. However, at the extreme transmon regime, where EJ is significantly larger than EC, the qubit transition frequencies are insensitive to the charge fluctuations. In a different scenario, where EC is comparable to EJ, the dispersion in the energy levels are so overwhelming, so it also makes the tunneling, sorry, detection of the tunneling equally difficult for us. So what we need is a qubit operating at an intermediate regime with a moderate EJ to EC ratio. In addition to this requirement, we also need large coupling, G, and large line width kappa for fast readouts. By using the simple circuit model and associated Hamiltonian, we can calculate the eigenstates of a qubit as a function of NG, which is the offset charge, and we index those things according to plasmonic states. These calculations are based on an EJ to EC ratio of 26, and we use some of the parameters belonging to a qubit we experimentally studied. I think it was Q9. So when we focus on the first excited state of this calculation, we see it's not a flat bend. In fact, there are two charge manifolds, which are corresponding to different parity states. So we can calculate the maximum splitting from here, which is 1.36 MHz, and we can directly see the splitting from an experimentally taken Rabi spectroscopy on the far right, and these symmetrical peaks correspond to odd and even parity states, which is a direct proof of presence of those parodies. Our qubits are basically aluminum, aluminum oxide, aluminum junctions fabricated with standard Dolan-Bridge technique, and the junctions are shunted by very large niobium capacitors, which are connected to niobium resonators, and individual niobium resonators are coupled to a common transmission line. And on the right, you see the RF setup of our measurement. As you can see, the input line is heavily attenuated to be able to protect the qubit from excess radiation, and we also use the Tupa as a quantum limited amplifier to be able to improve the signal-to-noise ratio. So prior to the Ramsey experiments, sensitive to the time scale of the quasi-particle tunneling measurements, we did some standard spectroscopic measurements to be able to obtain the readout mode resonances and the transition frequencies for the qubits just roughly. And then, actually, to be able to find you in the qubits, we performed standard Ramsey experiments. Here you are seeing one example here. In this Ramsey experiment, we were able to fit the oscillations into a sum of two sinusoids fluctuating symmetrically around F01, which is the transition frequency of the qubits. And the Rabi spectroscopy I've shown you before, which was successfully showing the parity space, the peaks are separated by two delta F01. Delta F01 is the detuning from the F01, which we can get from the Ramsey experiment. So those standard Ramsey experiments are also very good to be able to determine the Ramsey wait time. It can actually determine it precisely because delta T is one port of the detuning. After determining delta T, we can do the charge parity mapping by using a Ramsey light sequence developed in the literature. So here is the pulse sequence. Considering the qubit is in the ground state, the experiment starts with an application of a pi over two pulse around the x-axis. That brings the qubit on the equator of the block sphere here. And then the qubit picks up a plus minus pi over two phase depending on the charge parity during an idle time. So that brings the different parities on the opposite side of the equator of the block sphere. Then a second pi over two pulse is applied around the y-axis to be able to map the parity to the charge basis state. So basically if the parity is odd, the qubit ends up in the excited state. If the parity is even, qubit goes back to the ground state. So to be able to observe the quasi-particle tunneling events in real time, we repeated those CP mapping in every 100 microseconds for a second. Then the power spectral density of those fluctuations were averaged over 20 traces. Finally, we fitted the power spectral density into a characteristic Lorentzian of random telegraph nodes. From there we were able to get some values really important to us like the switching time, TP, and the fidelity of the measurements. So these measurements are obtained at the IBM labs, giving us extremely long quasi-particle switching times with a very high fidelity. To be able to see the impact of the quasi-particle tunneling events on coherence, we looked at the switching times, thus the quasi-particle tunneling grades, along with T1 and T2 values, as we are changing the mixing chamber temperatures. We did it for over 20 qubits, and here you are seeing some of the representative qubits. So at low temperatures, our switching times were as long as 82 milliseconds, which corresponds to only 12 quasi-particle tunneling events per second. And the rate, actually, if you look at the rates here, so beyond 100 millikelvin, they start dramatically to increase. This is where the switching times, T1 and T2 start to drop. Okay, so we modeled the temperature dependence of quasi-particle rates for the same set of junctions. Here you can see the data. So the quasi-particle induced relaxation rate is linearly scales with the quasi-particle density, which is the exponential function of superconducting gap and temperature. We fitted the data, flooded here in the linear scale, and we have a great agreement at the high temperature regime for each qubits. The inferred superconducting gap we used as a fifth parameter turned out to be 192 microelectron volts, which is very consistent with the literature value of the aluminum tin film. We also estimated the quasi-particle density at low temperatures, which varies between 10 to minus 9 to 10 to minus 8. I think it's an order of magnitude smaller than the previously reported measurements. So if you use the same data and plot it in the log scale, you can see the disagreement in the low temperature regime between data and qubits. Although we don't fully understand the reason for this discrepancy, it might be an indicator of an interesting physics. Our qubits are a little bit different than the qubits used in similar studies, such as the qubits are controlled and measured in two-dimensional architecture. In addition to that, all our readout resonators or the transmission lines are made of niobium instead of having all circuitry in aluminum. So these are some topics to explore as a future study. To be able to complement the previous data set, we also looked at the temperature dependence of the coherence times, T1s. So you can see the data here from the same set of qubits. So at low temperatures, we have fairly high T1s, varying between 100 to 200 microseconds. However, they are significantly smaller than the Tp switching time values. And this fact doesn't change when we heated the junctions to the elevated temperatures. So beyond 100 millikelvin, despite the dramatic decrease, both in T1 and Tp values, they never become comparable, suggesting the coherence in our devices is not limited by the quasi-particle tunneling. So in conclusion, we detected extremely long switching times, thus extremely low quasi-particle tunneling rate in two-dimensional architecture. So our inferred superconducting gap values are very consistent with the literature values. The estimated Qp density is one-order magnitude smaller than the reported experiments. And the fact that Tp values are much larger than T1 at any temperature, suggesting that our qubits are not limited, the coherence of our junctions are not limited by the quasi-particle tunneling. This is it.