Flux-noise-resilient transmon qubit via a doubly-connected gradiometric design

TL;DR

This work introduces the 8-mon, a doubly-connected gradiometric transmon with a cross-bridge (nano-airbridge) that links two loops to suppress low-frequency flux noise while preserving full tunability. Experiments show that the 8-mon achieves $T_{1}$-level coherence similar to standard X-mons and, at small flux bias, substantially extends Ramsey coherence times ($T_{2}^{*}$) by a factor of 2–3 without requiring dynamical decoupling, along with remarkable long-term frequency stability (drift < $0.1$ MHz over 10 h, and minimal drift even without shielding). A spatially correlated flux-noise model demonstrates that the gradiometric geometry filters long-wavelength noise by interfering currents in the two loops, predicting regimes where suppression scales as $(\xi/d)^2$ and matching observed Ramsey dephasing trends. The results provide a practical route to more coherent, stable tunable superconducting qubits, compatible with existing X‑mon control/readout and lacking additional measurement overhead. The study further highlights the coexistence of short- and long-correlation-length magnetic noise in the chip environment, informing future materials and geometry optimizations for scalable quantum processors.

Abstract

Frequency-tunable superconducting transmon qubits are a cornerstone of scalable quantum processors, yet their performance is often degraded by sensitivity to low-frequency flux noise. Here we present a doubly-connected gradiometric transmon (the ``8-mon") that incorporates a nano-airbridge to link its two loops. This design preserves full electrical tunability and remains fully compatible with standard X-mon control and readout, requiring no additional measurement overhead. The airbridge interconnect eliminates dielectric loss, which enables the 8-mon to achieve both energy relaxation times $T_{\rm 1}$ comparable to reference X-mons and, in the small flux-bias regime, a nearly threefold enhancement in Ramsey coherence time $T_{\rm 2}^*$. This improved $T_{\rm 2}^*$ reaches the same order as $T_{\rm 1}$ without employing echo decoupling. The device also exhibits superior long-term frequency stability even without any magnetic field shielding. We develop a spatially correlated flux-noise model whose simulations quantitatively reproduce the experimental coherence trends, revealing the coexistence of short- and long-correlation-length magnetic noise in the superconducting chip environment. By unifying high tunability with intrinsic flux-noise suppression through a robust geometric design, the 8-mon provides a practical pathway toward more coherent and stable superconducting quantum processors.

Figures (10)

• Figure 1: Device architectures of frequency-tunable superconducting qubits.a-c, Schematic illustrations of a conventional SQUID (a), a triply-connected gradiometric SQUID (b), and a doubly-connected gradiometric SQUID (c). In b, dashed lines indicate circuit segments that may vary across different types of superconducting qubits. In c, a cross-bridge connects the two loops, forming the gradiometric structure. d, Pseudo-colored scanning electron microscopy (SEM) image of a reference X-mon qubit, identical to the devices measured in this work. e, SEM image of an 8-mon qubit implemented with an embedded SiO$_2$ dielectric layer to bridge the two superconducting loops. f, An alternative 8-mon design utilizing a nano-airbridge for inter-loop connection (outlined by the red dashed area). g, A zoom-in SEM image of the nano-airbridge highlighted in f.
• Figure 2: Basic characterization of an 8-mon qubit.a, Qubit transition frequency as a function of the bias voltage applied to the flux line, demonstrating that the 8-mon qubit remains broadly tunable, akin to a conventional transmon. b, Rabi oscillations, demonstrating high-fidelity coherent control of the qubit state. c, Energy relaxation and echo decoherence measurements, revealing good coherence times with exponential fits yielding $T_{\rm 1}=41.3$$\mu$s and $T_{\rm 2}^{\rm E}=44.8$$\mu$s. d, Ramsey oscillation fringe, from which a Ramsey coherence time $T_{\rm 2}^*\approx 33$$\mu$s is extracted. All data in panels b-d were acquired at the flux-insensitive sweet spot (optimal point). Measurements were performed on device $\rm Q_{\rm a3}^*$.
• Figure 3: Coherence time characterization.a, Energy relaxation time $T_{\rm 1}$ for three 8-mon qubits and two reference X-mon qubits from the same chip, measured near the flux-insensitive sweet spot. The $T_{\rm 1}$ values, ranging from 20 $\mu$s to 45 $\mu$s, indicate that the 8-mon design preserves a similar relaxation time to the conventional transmon. b, Ramsey coherence time $T_{\rm 2}^*$ under small applied flux bias. The 8-mon qubits exhibit a substantial enhancement, with $T_{\rm 2}^*$ values nearly 2–3 times longer than those of the X-mons near the optimal point. This improvement diminishes as the flux bias increases. c, Echo decoherence time $T_{\rm 2}^{\rm E}$ within the same bias regime. The $T_{\rm 2}^{\rm E}$ values for both 8-mon and X-mon qubits are of the same order of magnitude. d, Flux-noise variance $\langle\Phi^2\rangle$ as a function of the spatial magnetization correlation length $\xi$ for X-mon and 8-mon circuits, calculated using the continuous magnetization model. e, Noise-suppression factor $S(\xi)=\langle\Phi_X^2\rangle/\langle\Phi_8^2\rangle$ as a function of $\xi$. f, Experimental measurements of the Ramsey dephasing time $T_2^*$ as a function of flux bias $\Phi_{\rm bias}$ for two representative devices (Q$^*_{a3}$ and Q$_{a5}$). The solid and dashed lines show fits obtained using the same dephasing model.
• Figure 4: Long-term frequency stability of 8-mon and X-mon qubits.a-b, Representative Ramsey fringes measured over 10 hours for a reference X-mon (a) and an 8-mon (b) qubit, both biased at 0.1 $\Phi_0$. c, Qubit frequency drift extracted from the Ramsey measurements in a and b. The 8-mon qubit exhibits superior stability with a drift $<0.1$ MHz, markedly lower than the $\approx 0.4$ MHz drift of the X-mon. d-e, Direct spectroscopic monitoring of the qubit transition frequency $f_{\rm 01}$ over 10 hours for an X-mon (d) and an 8-mon (e) in the absence of a cryogenic magnetic shield. f, Frequency drift extracted from Gaussian fits to the spectroscopic data in d and e. Without shielding, the X-mon experiences a drastic frequency drift exceeding 20 MHz, whereas the 8-mon maintains remarkable stability with no observable drift.
• Figure S1: Wiring schematic of measurement setup.