Stability studies on subtractively-fabricated CMOS-compatible superconducting transmon qubits

TL;DR

This study evaluates the temporal stability of subtractively fabricated CMOS-compatible superconducting transmon qubits for scalable quantum processors. By combining a ~95-hour single cooldown analysis of eight qubits with a long-term study over ten cooldowns spanning over a year, the work shows that TLS-induced fluctuations dominate T1/T2* dynamics while frequency drift remains modest; readout performance remains robust and aging of junctions is limited. The results demonstrate stability on par with lift-off devices, validating the CMOS approach for large-scale QPUs while highlighting TLS material improvements and stabilization strategies as key areas for further progress. Overall, the findings support the viability of CMOS-compatible qubits for fault-tolerant quantum computing and guide material/process refinements to address TLS-related limitations.

Abstract

Developing fault-tolerant quantum processors with error correction demands large arrays of physical qubits whose key performance metrics (coherence times, control fidelities) must remain within specifications over both short and long timescales. Here we investigated the temporal stability of subtractively fabricated CMOS-compatible superconducting transmon qubits. During a single cooldown and over a period of 95 hours, we monitored several parameters for 8 qubits, including coherence times $T_1$ and $T_2^*$, which exhibit fluctuations originating primarily from the interaction between two-level system (TLS) defects and the host qubit. We also demonstrate that subtractively-fabricated superconducting quantum devices align with the theoretical predictions that higher mean lifetimes $T_1$ correspond to larger fluctuations. To assess long-term stability, we tracked two representative qubits over 10 cooldown cycles spanning more than one year. We observed an average total downward shift in both qubit transition frequencies of approximately 61 MHz within the thermal cycles considered. In contrast, readout resonator frequencies decreased only marginally. Meanwhile, $T_1$ exhibits fluctuations from cycle to cycle, but maintains a stable baseline value.

Figures (17)

• Figure 1: Sketch of a single-qubits chip layout with 4 qubits structures and a test resonator coupled to a central feedline.
• Figure 2: Illustration of the cryogenic measurement setup. The qubit chips are mounted in the mixing chamber (MXC) of a Bluefors LD dilution refrigerator (DR), enclosed in a copper package and covered with Eccosorb. The MXC temperature is monitored with a $\text{RuO}_{2}$-temperature sensor. Drive and input readout signals are generated respectively by a signal generator (SHFSG) and a quantum analyzer (SHFQA), synchronized by a PQSC (all from Zurich Instruments). Both signals go through a combiner at room temperature and then are attenuated at each stage of the DR to suppress thermal noise. The output readout signal passes through filters and isolators before being amplified at the 4K stage and at room temperature.
• Figure 3: Results of 95-hour single-cooldown measurements for qubit A.2, with time traces on the left and corresponding histograms on the right in each panel: a) $T_1$, b) $T_2^*$, c)$|f_\mathrm{Ramsey}|-\Delta_f$, where $\Delta_f$ is the set detuning, d) $\Delta_\mathrm{m}$, e) readout fidelity $\mathcal{F}$, f) $T_\mathrm{eff}$, and g) $T_\mathrm{MXC}$. The red curves in histograms (a) and (b) represent fits using a Rician distribution with an offset. This choice of fit function allows to describe the observed skewness. Note that this choice is purely empirical and is not derived from an underlying physical model. The fits are used to determine the mean and standard deviation via integration. The red-shaded regions in (a) and (b) indicate where $T_1$ or $T_2^*$ fall more than one standard deviation below the mean (threshold marked by the dotted horizontal gray lines). These red-shaded regions highlight the simultaneous drops of $T_1$ and $T_2^*$.
• Figure 4: Temporal evolution of a) $T_1(t)$/$T_1(0)$, b) $T_2^*(t)$/$T_2^*(0)$, c) $|f_\mathrm{Ramsey}|-\Delta_f$, d) $\frac{\Delta_\mathrm{m}(t)-\Delta_\mathrm{m}(0)}{\Delta_\mathrm{m}(0)}$, and e) readout fidelity $\mathcal{F}$ for all qubits (except Ramsey results of A.1 and $\Delta_\mathrm{m}$ results for B.3). Panel (d) uses a logarithmic scale on the y-axis with a linear interval $\pm0.1$ around 0. The green dotted lines indicate the time limits of the largest $T_1$ drop observed in qubit B.1, and highlight its impact on the other parameters. Due to $T_2^*$ decrease, the Ramsey results were unreliable during this time interval and therefore omitted. The blue dashed lines delimit the time interval where the readout fidelity and $\Delta_\mathrm{m}$ decreased for A.1 and the Ramsey frequency could not be determined due to $T_2^*$ decrease.
• Figure 5: Standard deviation of $T_1$ with respect to its mean value, logarithmically scaled. The dotted line represents the fit following \ref{['eq:square_equation']} which gives for $a = (1.220\pm0.052)\times10^{-2}\:\mathrm{s}^{\frac{2}{3}}$.