Characterization of Gap-Engineered Josephson Junctions and Gate Fidelities for a Superconducting Qubit

Abstract

Quantum computing promises applications in physics, cryptography, material science, pharmaceuticals, and a wide range of other science. Superconducting qubits offer a possible platform for developing a quantum computer. To perform useful quantum computations, the coherence and control of present day superconducting qubits must be greatly improved. In this dissertation, I present two main results to improve the performance of transmon qubits.

For the first project, I fabricated and characterized the coherence of transmon devices with asymmetric superconducting gaps. Previous models suggested that devices with asymmetric superconducting gaps on either side of the Josephson junction can be designed to be less subject to loss from quasiparticle tunneling. To gap-engineer the Josephson junctions, I used Ti metal to proximitize and lower the superconducting gap of the Al counter-electrode. Unfortunately, the energy relaxation time constant for an Al/AlOx/Al/Ti 3D transmon I fabricated and tested was T1 = 1 us, over two orders of magnitude shorter than the measured T1 = 134 us of an Al/AlOx/Al 3D transmon with Al capacitor pads and the measured T1 = 143 us of an Al/AlOx/Al 3D transmon with Ta capacitor pads. DC IV measurements of proximitized Josephson junctions showed a reduced superconducting gap, demonstrating that the gap-engineering in the Al/Ti layer was successful. However, these same IV measurements showed greatly increased excess current for voltage biases below the superconducting gap compared to my Al/AlOx/Al junctions. This suggests the addition of Ti caused the junction quality to worsen, potentially being a source of tunneling loss in the transmon devices. Intentionally adding oxygen disorder between the Al and Ti layers reduced the proximity effect and subgap current in DC measurements while increasing the relaxation time of a 3D transmon to T1 = 32 us.

Additionally, I designed an Al/AlOx/Al SQUID device to perform DC IV measurements of junctions with tunable total critical current. In a single junction, subgap tunneling features can be due to the critical current interacting with the environment, subgap quasiparticle processes, or other sources. Reducing the critical current allows these features to be differentiated and more accurately measure the effects from quasiparticle tunneling alone. Characterizing this device showed subgap tunneling features consistent with inelastic Cooper pair tunneling and quasiparticle transport via multiple Andreev reflection in a low transparency junction. This measurement technique could be used to further study gap-engineered junctions.

For the second project, I characterized an Al/AlOx/Al 2D transmon device with Ta features and performed high-fidelity single qubit gates. First, I used error amplifying pulse sequences to fine-tune the qubit gate pulses. I evaluated the gate error with randomized benchmarking. I characterized gates with Gaussian and cosine shaped pulses at a variety of pulse lengths. Analyzing the pulse envelopes in the frequency domain and directly measuring leakage to the transmon's second excited state revealed that leakage from driving higher qubit transitions was a major source of gate error. Next, I characterized gates using a pulse shape designed by a physics informed neural network designed by Güngördü and Kestner and found improved gate error for 16~ns pulses achieving an average error per gate of (3.36 +/- 0.03) x 10^-4. This outperformed errors of (5.54 +/- 0.24) x10^-4 for a cosine shaped pulse and (3.93 +/- 0.12) x10^-4 for a Gaussian shaped pulse of the same length. Further optimization of the pulse using predistortion or leakage reduction strategies may yield even greater performance.