We analyze the behavior of a dc superconducting quantum interference device phase qubit in which one junction acts as a phase qubit and the rest of the device provides isolation from dissipation and noise in the bias leads. Ignoring dissipation, we find the two-dimensional Hamiltonian of the system and use numerical methods and a cubic approximation to solve Schr\"odinger's equation for the eigenstates, energy levels, tunneling rates, and expectation value of the currents in the junctions. Using these results, we investigate how well this design provides isolation while preserving the characteristics of a phase qubit. In addition, we show that the expectation value of current flowing through the isolation junction depends on the state of the qubit and can be used for nondestructive read out of the qubit state.
We analyze the behavior of a dc superconducting quantum interference device phase qubit in which one junction acts as a phase qubit and the rest of the device provides isolation from dissipation and noise in the bias leads. Ignoring dissipation, we find the two-dimensional Hamiltonian of the system and use numerical methods and a cubic approximation to solve Schr"odinger's equation for the eigenstates, energy levels, tunneling rates, and expectation value of the currents in the junctions. Using these results, we investigate how well this design provides isolation while preserving the characteristics of a phase qubit. In addition, we show that the expectation value of current flowing through the isolation junction depends on the state of the qubit and can be used for nondestructive read out of the qubit state.
This thesis examines sources of dissipation and dephasing in a dc SQUID phase qubit. Coupling of the qubit to the bias lines and lossy dielectrics causes the qubit to lose quantum information through a process known generally as decoherence. Using knowledge of the possible sources of decoherence, a dc SQUID phase qubit is designed with parameters that should have made it resistant to dissipation and dephasing from those sources. Device PB9 was a dc SQUID with one small area 0.23 (μm)2 Josephson junction with a critical current of 130 nA, which was meant to be the qubit junction, and a larger area 5 (μm)2 junction with a critical current of 8.6 μA, which acted as part of an inductive isolation network. The qubit junction was shunted by a 1.5 pF low-loss interdigitated capacitor. The dc current bias line had an on-chip LC filter with a cutoff frequency of 180 MHz. The other control lines were also designed to minimize coupling of dissipative elements to the qubit. According to a theoretical model of the dissipation and dephasing, the qubit was expected to have an energy relaxation T1 ≤ 8.4 μs and dephasing time Tø ∼ 1 μs. Because of the relatively high Josephson inductance of the qubit junction, the device did not act perform like a conventional isolated single-junction phase qubit. Instead, the resonant modes that I observed were the normal modes of the entire SQUID. At 20 mK and a frequency of 4.047 GHz, the maximum energy relaxation time of the device was found to be 350 ± 70 ns, despite the optimized design. Through a study of T1 versus applied flux, T1 was found to depend on the strength of the coupling of the microwave drive line to the qubit. When the line was more coupled, T1 was shorter. This was evidence that the microwave line was overcoupled to the qubit, and was limiting the lifetime of the excited state T1. Through a study of the spectroscopic coherence time T2* , which measured the effects of low-frequency inhomogeneous broadening and higher frequency dephasing from noise, I discovered that device PB9 has several sweet spots. In particular, the presence of a sweet spot with respect to critical current fluctuations allowed me to identify critical current noise as a major source of broadening and dephasing in the qubit. From the spectroscopy I estimated the 1/f critical current noise power density at 1 Hz was SI01 (1Hz) ≈ 10 (nA) 2/Hz and the 1/f flux noise power spectral density at 1 Hz was SΦ (1Hz) ≈ (110 μΦ 0)2/Hz. Both of these values were quite high, possibly due to switching of the device between measurements.
We report measurements of Rabi oscillations and spectroscopic coherence times in an Al/AlOx/Al and three Nb/AlOx/Nb dc SQUID phase qubits. One junction of the SQUID acts as a phase qubit and the other junction acts as a current-controlled nonlinear isolating inductor, allowing us to change the coupling to the current bias leads in situ by an order of magnitude. We found that for the Al qubit, a spectroscopic coherence time ${T}_{2}^{\ensuremath{\ast}}$ varied from 3 to 7 ns and the decay envelope of Rabi oscillations had a time constant ${T}^{\ensuremath{'}}=25\phantom{\rule{0.3em}{0ex}}\text{ns}$ on average at 80 mK. The three Nb devices also showed ${T}_{2}^{\ensuremath{\ast}}$ in the range of 3--8 ns, but ${T}^{\ensuremath{'}}$ was 9--15 ns, just about 1/2 the value we found in the Al device. For all the devices, the time constants were roughly independent of the isolation from the bias lines, implying that noise and dissipation from the bias leads were not the principal sources of dephasing and inhomogeneous broadening.
This thesis examines the behavior of dc SQUID phase qubits in terms of their proposed use in a quantum computer. In a phase qubit, the two lowest energy states (n=0 and n=1) of a current-biased Josephson junction form the qubit states, with the gauge invariant phase difference across the junction being relatively well defined. In a dc SQUID phase qubit, the Josephson junction is isolated from the environment using an inductive isolation network and Josephson junction, which are connected across the phase qubit junction to form a dc SQUID. Five dc SQUID phase qubits were examined at temperatures down to 25 mK. Three of the devices had qubit junctions that were Nb/AlOx/Nb junctions with critical currents of roughly 30 microamps. The other two had Al/AlOx/Al junctions with critical currents of roughly 1.3 microamps. The device that had the best performance was an Al/AlOx/Al device with a relaxation time of 30 ns and a coherence time of 24 ns. The devices were characterized using microwave spectroscopy, Rabi oscillations, relaxation and Ramsey fringe measurements. I was also able to see coupling between two Nb/AlOx/Nb dc SQUID phase qubits and perform Rabi oscillations with them. The Nb/AlOx/Nb devices had a relaxation time and coherence time that were half that of the Al/AlOx/Al device. One of the goals of this work was to understand the nature of parasitic quantum systems (TLSs) that interact with the qubit. Coupling between a TLS and a qubit causes an avoided level crossing in the transition spectrum of the qubit. In the Al/AlOx/Al devices unintentional avoided level crossings were visible with sizes up to 240 MHz, although most visible splittings were of order ~20 MHz. The measured spectra were compared to a model of the avoided level crossing based on the TLSs coupling to the junction, through either the critical current or the voltage across the junction.
A dc SQUID phase qubit consists of two Josephson junctions in a loop. One junction acts as a qubit with two lowest energy levels forming the |0> and |1> status. The second junction and the loop inductance act to isolate the qubit junction from noise. In this thesis, I report on the improvement of the relaxation time and the coherence time in a dc SQUID phase qubit that used an LC filter. I also report the measurement of anomalous switching curves. In order to improve the relaxation and coherence times, I used two isolation networks, an LC isolation network and an inductive isolation network, to decouple the device from the current bias lines. This produced a very large total effective resistance of the input leads that increases the relaxation time of the qubit. In addition, I connected a low-loss SiNx shunting capacitor across the qubit junction to reduce dielectric losses. I measured two dc SQUID phase qubits. Device DS6 had a 4 (μm)2 Al/AlOx/Al qubit junction with a critical current of 0.5 μA and a 1 pF shunting capacitor. It used an LC filter made from a 10 nH inductor and a 145 pF capacitor. The capacitors contained N-H rich SiNx which produced a loss tangent of about 7×10 -4 . Device DS8 had a 2 (μm)2 Al/AlOx/Al qubit junction with a critical current of 77 nA and an LC filter similar to the first one. The shunting capacitor contained Si-H rich SiNx. Using a pulse readout technique, I measured the characteristics of the qubits, including the transition spectrum, Rabi oscillations, relaxation, Ramsey fringes and state tomography. The best relaxation time T1 for device DS6 was 32 ns and 280 ns for device DS8. The best Rabi decay time T' for DS6 was 42 ns while for device DS8 it was 120 ns. From these and other data I obtained estimates for the best coherence time T2 in device DS6 of 61 ns and 76 ns in device DS8. In DS8, I observed anomalous switching curves; i.e. switching curves which were qualitatively different from conventional switching curves. In the conventional case, the switching curve for the superposition state is the weighted sum of the |0> and |1> curves, but it was not in device DS8. Instead, the switching curve shifted along the current axis as the exited state probability increased. I present a model for understanding the behavior and use this model to extract the probability to be in the excited state.
Identifying Sources of Decoherence in a dc SQUID Phase Qubit With a Sub-μm Junction and Interdigitated Capacitor
We fabricated a dc SQUID phase qubit with a sub- μm Al/AlOx/Al qubit junction and an interdigitated shunting capacitor on a sapphire substrate. The qubit junction had a critical current of 135 nA, and the isolation junction had a critical current of 8.3 μA. The shunting capacitance was about 1.5 pF. To reduce the unwanted effects of two-level systems and increase the relaxation timeT1, we have removed unnecessary dielectrics, used a small qubit junction area (450 nm × 500 nm), isolated the qubit from the leads with an on-chip LC filter, and fabricated the device on a bare sapphire substrate. However, at a temperature of 20 mK, we foundT1≈ 300 ns and the coherence timeT2≈ 110 ns, which was much lower than one would expect from loss attributed to the leads and to dielectrics in the tunnel junction and substrate. Measurements ofT1versus applied flux (which tuned the qubit frequency) revealed a correlation between the strength of the coupling of the microwave excitation line to the qubit and the rate of energy dissipation in the qubit. This result suggests that the relaxation time was being limited by coupling to the microwave line.
Two states of a flux qubit with three Josephson junctions were shown in a single measurement with a dc-SQUID. The qubit is an aluminum superconductor loop surrounded by a dc-SQUID for readout. It has two states, which have persistent currents flowing in opposite directions. The readout data for three samples with different junction sizes suggest that the probability distribution for the double-well potential depends on the ratio of E J / E C , where E J is the Josephson energy and E C is the charging energy. The probability distribution was estimated by calculating the wavefunctions and energy levels for the measured samples. We measured the retrapping current of a dc-SQUID without the qubit as a function of the magnetic field. It was found that the relative phase relation for the magnetic field dependence of the retrapping current jumped from 0 to π at E J ≈ E C . This jump can be explained phenomenologically by taking into account the two kinds of dissipation arising from ac currents flowing through the lead and along the SQUID loop. The fact that the jump occurred at E J ≈ E C strongly suggests that the quantum nature of the electromagnetic environment needs to be taken into account in order to understand the origin of the dissipation.