Low-frequency noise caused by critical-current fluctuations in Josephson junctions can lead to substantial, measurement-induced dephasing in superconducting qubits. The purpose of this work is to measure this noise as a function of temperature and to simulate its effect on the measurement dynamics associated with superconducting qubits. In Nb trilayer junctions we measure the noise power at 1 Hz down to 10 mK and verify a T2 dependence for T > 1 K, below which we observe a flattening of the noise magnitude versus temperature. In Al trilayer junctions we measure the critical-current noise power at 1 Hz from 10 mK to 1.4 K and find no dependence on temperature over this range. Possible reasons for the deviation from a T2 dependence are explored, including heating and sample quality. Simulations of the effect of 1/f noise are performed to examine the consequence of using different qubit sampling methods on expected dephasing times. Using the two sampling methods, we can probe the spectral nature and possible source of the dominant noise in a qubit.
For over three decades, the presence of magnetic flux noise with a power spectral density scaling roughly as
SPhi(f)propto1/falpha, where
alphalesssim1, has been known to limit the low-frequency performance of dc superconducting quantum interference devices (SQUIDs). In recent years, experiments indicate that this same noise persists to frequencies up to 1~GHz and is a dominant source of dephasing in flux-sensitive superconducting quantum bits (qubits). Thus, the reduction of flux noise presents a major hurdle towards the successful realization of scalable quantum computers that are based on flux-based qubits. In this thesis, we present experimental measurements, theoretical analyses, and numerical simulations that support a more detailed understanding of both the microscopic and macroscopic properties of flux noise.Our experimental work begins with flux noise measurements of a large number of SQUIDs in the temperature range from 0.1~K to 4~K. We report on measurements of ten SQUIDs with systematically varied geometries and show that
alpha increases as the temperature is lowered; in so doing, each spectrum pivots about a nearly constant frequency. The mean square flux noise, inferred by integrating the power spectra, grows rapidly with temperature and at a given temperature is approximately independent of the outer dimension of a given SQUID washer. We show that these results are incompatible with a model based on the random reversal of independent, spins that are located at the surface of the SQUID washer.In the course of our flux noise measurements, we became aware of a spurious contribution to low-frequency critical current noise in Josephson junctions---normally attributed to charge trapping in the barrier---arising from temperature instabilities inherent in cryogenic systems. These temperature fluctuations modify the critical current via its temperature dependence. By computing cross-correlations between measured temperature and critical current noise in Al-AlOx-Al junctions, we show that, despite excellent temperature stability, temperature fluctuations induce observable critical current fluctuations. Particularly, because
1/f critical current noise has decreased with improved fabrication techniques in recent years, it is important to understand and eliminate this additional noise source.Next, we introduce a numerical method of calculating the mean square flux noise
Phisq from independently fluctuating spins on the surface of thin-film loops of arbitrary geometry. By reciprocity,
Phisq is proportional to
langleBbf(rbf)2rangle, where
Bbf(rbf) is the magnetic field generated by a circulating current around the loop and
rbf varies over the loop surface. By discretizing the loop nonuniformly, we efficiently and accurately compute the current distribution and resulting magnetic field, which may vary rapidly across the loop. We use this method to compute
Phisq in a number of scenarios in which we systematically vary physical parameters of the loop. We compare our simulations to an earlier analytic result predicting that
PhisqproptoR/W in the limit where the loop radius
R is much greater than the linewidth
W. We further show that the previously neglected contribution of edge spins to
Phisq is significant---even dominant---in narrow-linewidth loops.To calculate theoretical dephasing rates in qubits, we consider flux noise with a spectral density
SPhi(f)=A2/(f/1 hboxHz)alpha, where
A is of the order of 1~
muPhi0,hboxHz−1/2 and
0.6leqalphaleq1.2;
Phi0 is the flux quantum. For a qubit with an energy level splitting linearly coupled to the applied flux, our calculations of the dependence of the pure dephasing time
tauphi of Ramsey and echo pulse sequences on
alpha for fixed
A show that
tauphi decreases rapidly as
alpha is reduced. We find that
tauphi is relatively insensitive to the noise bandwidth,
f1leqfleqf2, for all
alpha provided the ultraviolet cutoff frequency
f2>1/tauphi. We calculate the ratio
tauphi,E/tauphi,R of the echo (
E) and Ramsey (
R) sequences, and the dependence of the decay function on
alpha and
f2. We investigate the case in which
SPhi(f0) is fixed at the ``pivot frequency''
f0neq1~Hz while
alpha is varied, and find that the choice of
f0 can greatly influence the sensitivity of
tauphi,E and
tauphi,R to the value of
alpha.Finally, we conclude with a brief review of our principal results and conclusions. We also comment on promising avenues of future research.
The microscopic description of 1/f magnetic flux noise in superconducting circuits has remained an open question for several decades despite extensive experimental and theoretical investigation. Recent progress in superconducting devices for quantum information has highlighted the need to mitigate sources of qubit decoherence, driving a renewed interest in understanding the underlying noise mechanism(s). Though a consensus has emerged attributing flux noise to surface spins, their identity and interaction mechanisms remain unclear, prompting further study. Here, we apply weak in-plane magnetic fields to a capacitively shunted flux qubit (where the Zeeman splitting of surface spins lies below the device temperature) and study the flux-noise-limited qubit dephasing, revealing previously unexplored trends that may shed light on the dynamics behind the emergent 1/f noise. Notably, we observe an enhancement (suppression) of the spin-echo (Ramsey) pure-dephasing time in fields up to B=100 G. With direct noise spectroscopy, we further observe a transition from a 1/f to approximately Lorentzian frequency dependence below 10 Hz and a reduction of the noise above 1 MHz with increasing magnetic field. We suggest that these trends are qualitatively consistent with an increase of spin cluster sizes with magnetic field. These results should help to inform a complete microscopic theory of 1/f flux noise in superconducting circuits.
For many types of superconducting qubits, magnetic flux noise is a source of pure dephasing. Measurements on a representative dc superconducting quantum interference device (SQUID) over a range of temperatures show that ${S}_{\ensuremath{\Phi}}(f)={A}^{2}/{(f/1\phantom{\rule{0.28em}{0ex}}\text{Hz})}^{\ensuremath{\alpha}}$, where ${S}_{\ensuremath{\Phi}}$ is the flux noise spectral density,
A is of the order of 1 $\ensuremath{\mu}{\ensuremath{\Phi}}_{0}\phantom{\rule{0.28em}{0ex}}{\text{Hz}}^{\ensuremath{-}1/2}$, $0.61\ensuremath{\leqslant}\ensuremath{\alpha}\ensuremath{\leqslant}0.95$, and ${\ensuremath{\Phi}}_{0}$ is the flux quantum. For a qubit with an energy level splitting linearly coupled to the applied flux, calculations of the dependence of the pure dephasing time ${\ensuremath{\tau}}_{\ensuremath{\varphi}}$ of Ramsey and echo pulse sequences on $\ensuremath{\alpha}$ for fixed
A show that ${\ensuremath{\tau}}_{\ensuremath{\varphi}}$ decreases rapidly as $\ensuremath{\alpha}$ is reduced. We find that ${\ensuremath{\tau}}_{\ensuremath{\varphi}}$ is relatively insensitive to the noise bandwidth, ${f}_{1}\ensuremath{\leqslant}f\ensuremath{\leqslant}{f}_{2}$, for all $\ensuremath{\alpha}$ provided the ultraviolet cutoff frequency ${f}_{2}>1/{\ensuremath{\tau}}_{\ensuremath{\varphi}}$. We calculate the ratio ${\ensuremath{\tau}}_{\ensuremath{\varphi},E}/{\ensuremath{\tau}}_{\ensuremath{\varphi},R}$ of the echo (
E) and Ramsey (
R) sequences and the dependence of the decay function on $\ensuremath{\alpha}$ and
f2. We investigate the case in which ${S}_{\ensuremath{\Phi}}({f}_{0})$ is fixed at the ``pivot frequency'' ${f}_{0}\ensuremath{\ne}1$ Hz while $\ensuremath{\alpha}$ is varied and find that the choice of
f0 can greatly influence the sensitivity of ${\ensuremath{\tau}}_{\ensuremath{\varphi},E}$ and ${\ensuremath{\tau}}_{\ensuremath{\varphi},R}$ to the value of $\ensuremath{\alpha}$. Finally, we present calculated values of ${\ensuremath{\tau}}_{\ensuremath{\varphi}}$ in a qubit corresponding to the values of
A and $\ensuremath{\alpha}$ measured in our SQUID.
We present measurements of 1/f frequency noise in both linear and Josephson-junction-embedded superconducting aluminum resonators in the low power, low temperature regime—typical operating conditions for superconducting qubits. The addition of the Josephson junction does not result in additional frequency noise, thereby placing an upper limit for fractional critical current fluctuations of 1×10−8 (1/Hz) at 1 Hz for sub-micron, shadow evaporated junctions. These values imply a minimum dephasing time for a superconducting qubit due to critical current noise of 40–1400 μs depending on qubit architecture. Occasionally, at temperatures above 50 mK, we observe the activation of individual fluctuators which increase the level of noise significantly and exhibit Lorentzian spectra.
We measure the dependence of qubit phase coherence and surface spin induced flux noise on inductor loop geometry. While wider inductor traces change neither the flux noise power spectrum nor the qubit dephasing time, increased inductance leads to a simultaneous increase in both. Using our protocol for measuring low frequency flux noise, we make a direct comparison between the flux noise spectrum and qubit phase decay, finding agreement within 10% of theory. The dependence of the measured flux noise on inductor geometry is consistent with a noise source correlation length between 6 and 400 um.