Capacitively Shunted Flux Qubit Overview

• Capacitively shunted flux qubit is a superconducting circuit that incorporates a large shunt capacitance to suppress charge noise and improve coherence.
• It features a three-junction architecture with design variants that adjust anharmonicity and flux dispersion for applications like quantum annealing and defect spectroscopy.
• Recent implementations report significantly enhanced coherence times and effective multi-qubit control through optimized circuit design and tunable coupling.

Searching arXiv for papers on capacitively shunted flux qubits and related terminology. A capacitively shunted flux qubit (CSFQ) is a superconducting flux qubit in which a three-junction Josephson loop is supplemented by a large shunt capacitance, typically across the smaller junction or across the qubit islands, so that the charging energy is reduced, the characteristic impedance is lowered, and sensitivity to charge fluctuations is suppressed (Steffen et al., 2010). In published realizations, the class includes the original low-impedance or “Zflux qubit,” tunable double-loop devices designed for coherent quantum annealing, gradiometric fully tunable variants, and high-anharmonicity multi-shunt flux circuits (Steffen et al., 2010, Trappen et al., 2023, Berlitz et al., 30 Sep 2025, Yurtalan et al., 2020). Reported coherence has progressed from $T_1,T_2^*\sim1.5\,\mu\mathrm{s}$ in the early planar low-impedance implementation to $T_1=60\text{–}90\,\mu\mathrm{s}$ and Hahn-echo coherence of about $80\,\mu\mathrm{s}$ in a 3D c-shunt flux qubit (Steffen et al., 2010, Abdurakhimov et al., 2019).

1. Circuit architecture and nomenclature

The canonical CSFQ is a three-junction superconducting loop of total loop inductance $L$, with two identical junctions of critical current $I_0$ and capacitance $C_J$, and a third smaller junction of critical current $\alpha I_0$ and capacitance $\alpha C_J$. In the original low-impedance design, the smaller junction is shunted in parallel by a large interdigitated capacitance $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$, with $\alpha\approx0.3\text{–}0.43$, and $T_1=60\text{–}90\,\mu\mathrm{s}$0 (Steffen et al., 2010). Later CSFQ realizations retained the three-junction loop but altered the biasing and shunting geometry: one hybrid CSFQ–transmon device used a split or dc-SQUID loop shunted by large interdigitated capacitors $T_1=60\text{–}90\,\mu\mathrm{s}$1 per island (Ku et al., 2020), while a tunable annealing-oriented device used one junction in the main “Z” loop and two junctions forming a DC-SQUID in the secondary “X” loop, again shunted by a large on-chip capacitor with $T_1=60\text{–}90\,\mu\mathrm{s}$2 and $T_1=60\text{–}90\,\mu\mathrm{s}$3 (Trappen et al., 2023).

A further extension is the gradiometric, fully tunable C-shunted flux qubit, in which the small $T_1=60\text{–}90\,\mu\mathrm{s}$4-junction is replaced by a tunable dc-SQUID and a single-piece shunt capacitor $T_1=60\text{–}90\,\mu\mathrm{s}$5 is added across the loop (Berlitz et al., 30 Sep 2025). In that design, the effective Josephson energy of the $T_1=60\text{–}90\,\mu\mathrm{s}$6-junction becomes

$T_1=60\text{–}90\,\mu\mathrm{s}$7

and two local flux lines provide independent control of the double-well asymmetry and the tunnel barrier (Berlitz et al., 30 Sep 2025).

The literature also contains a high-anharmonicity capacitively shunted three-junction flux circuit with three large planar shunt capacitors, one at each island of the loop (Yurtalan et al., 2020). This indicates that “capacitively shunted flux qubit” denotes a circuit family defined by the addition of large shunt capacitance to a flux-qubit topology rather than by a single unique layout.

2. Hamiltonian formulations and operating regimes

For the original low-impedance flux qubit, the phase differences across the two larger junctions are $T_1=60\text{–}90\,\mu\mathrm{s}$8, and across the smaller junction $T_1=60\text{–}90\,\mu\mathrm{s}$9. Defining

$80\,\mu\mathrm{s}$0

the two-dimensional potential is

$80\,\mu\mathrm{s}$1

with $80\,\mu\mathrm{s}$2 and $80\,\mu\mathrm{s}$3 (Steffen et al., 2010). The corresponding kinetic energy is

$80\,\mu\mathrm{s}$4

Because $80\,\mu\mathrm{s}$5, the $80\,\mu\mathrm{s}$6-mode is very massive and can be frozen out, yielding the one-dimensional Hamiltonian

$80\,\mu\mathrm{s}$7

with

$80\,\mu\mathrm{s}$8

Since $80\,\mu\mathrm{s}$9, $L$0 is reduced by roughly an order of magnitude compared with an unshunted flux qubit (Steffen et al., 2010).

In annealing-oriented and fully tunable devices, the operative description is often reduced further to an effective two-level Hamiltonian,

$L$1

where $L$2 is the tunnel splitting and $L$3 is the flux-bias detuning (Matsuzaki et al., 2020). For the tunable CSFQ studied in the context of decoherence, the same structure appears as

$L$4

with $L$5 and $L$6 set by the X-loop flux (Trappen et al., 2023).

A common oversimplification is to treat all flux qubits with shunt capacitance as conventional persistent-current double-well qubits. That is not accurate. In the original low-impedance device, the ratio of junction critical currents is chosen to make the potential have a single-well form (Steffen et al., 2010). By contrast, the tunable devices explicitly use a symmetric or biased double-well description with independently controllable $L$7 and $L$8 (Trappen et al., 2023, Berlitz et al., 30 Sep 2025). This suggests that the presence of the shunt capacitor determines the low-impedance regime, while the potential landscape remains a design choice.

3. Spectrum, sweet spots, and anharmonicity

Numerical diagonalization of the low-impedance Hamiltonian yields the lowest eigenenergies as functions of external flux. At the sweet spot $L$9, the original device had $I_0$0, while $I_0$1, corresponding to an anharmonicity $I_0$2 with $I_0$3 (Steffen et al., 2010). The positive sign is the “inverted” anharmonicity emphasized in the original report, in contrast to transmons and phase qubits, for which $I_0$4 (Steffen et al., 2010). The same work reported that the single-well shape for $I_0$5 smoothes out flux dispersion, with $I_0$6 approximately $I_0$7 smaller than in the conventional flux qubit (Steffen et al., 2010).

The hybrid CSFQ–transmon platform also operated in a positive-anharmonicity regime. At $I_0$8, the measured dressed CSFQ transition frequency was $I_0$9 and the anharmonicity was $C_J$0 (Ku et al., 2020). That sign difference relative to a transmon was central to the demonstrated elimination of static $C_J$1 interaction in the coupled pair (Ku et al., 2020).

Other CSFQ implementations occupy different spectral regimes. A 3D c-shunt flux qubit embedded in a cavity had $C_J$2 at $C_J$3 with anharmonicity $C_J$4 (Abdurakhimov et al., 2019). A high-anharmonicity capacitively shunted three-junction flux circuit, operated as a qubit or qutrit, had $C_J$5, $C_J$6, and $C_J$7 at $C_J$8 (Yurtalan et al., 2020). By contrast, a three-level quantum-battery implementation based on a capacitively shunted flux qubit reported $C_J$9, $\alpha I_0$0, and thus $\alpha I_0$1 at its working flux (Li et al., 10 Apr 2025). This suggests that the sign and magnitude of the anharmonicity are design- and bias-dependent across the broader CSFQ family.

For the fully tunable gradiometric design, diagonalization of the one-dimensional Hamiltonian yielded a qubit transition frequency that was swept over nearly $\alpha I_0$2—from a few MHz up to $\alpha I_0$3—while remaining at the flux-insensitive tilt sweet spot $\alpha I_0$4 (Berlitz et al., 30 Sep 2025). That combination of wide gap control and sweet-spot operation is specific to the gap-tunable C-shunted variant and is not a generic property of fixed-$\alpha I_0$5 CSFQs.

4. Coherence, noise protection, and decoherence channels

The large shunt capacitance lowers the qubit impedance $\alpha I_0$6, reduces the charging energy $\alpha I_0$7, suppresses sensitivity to stray capacitances and voltage fluctuations, and flattens the flux dispersion (Steffen et al., 2010). These mechanisms explain why the CSFQ was introduced as a “hybrid superconducting qubit” combining transmon-like noise immunity with flux-qubit tunability (Steffen et al., 2010).

Representative coherence metrics reported for distinct CSFQ implementations are summarized below.

Implementation	Operating regime	Reported coherence
Low-impedance flux qubit	$\alpha I_0$8	$\alpha I_0$9, $\alpha C_J$0, echo $\alpha C_J$1
Hybrid CSFQ–transmon device	Sweet spot $\alpha C_J$2	$\alpha C_J$3, Ramsey $\alpha C_J$4, echo $\alpha C_J$5
High-anharmonicity three-shunt flux circuit	$\alpha C_J$6	$\alpha C_J$7, $\alpha C_J$8, $\alpha C_J$9
3D c-shunt flux qubit	Optimal flux bias $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$0	$C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$1, Ramsey $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$2, Hahn-echo $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$3, CPMG up to $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$4
Fully tunable gradiometric C-shunt flux qubit	$C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$5	$C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$6 up to $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$7

The early low-impedance device already exhibited $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$8, and Rabi-oscillation envelope decay of $C_{\rm sh}\approx100\text{–}110\,\mathrm{fF}$9 (Steffen et al., 2010). In the 3D implementation, relaxation at the optimal point was attributed to quasiparticle tunneling; the observed temperature dependence was fit by quasiparticle-tunneling theory, with non-equilibrium quasiparticle density $\alpha\approx0.3\text{–}0.43$0 (Abdurakhimov et al., 2019). Away from the sweet spot, dephasing was fitted by Gaussian Ramsey and echo envelopes associated with flux noise, giving $\alpha\approx0.3\text{–}0.43$1 and $\alpha\approx0.3\text{–}0.43$2 (Abdurakhimov et al., 2019).

The tunable annealing-oriented CSFQ provided a more detailed decomposition of decoherence channels. At the qubit symmetry point, relaxation below $\alpha\approx0.3\text{–}0.43$3 was mainly due to intrinsic flux noise in the main qubit loop, while at higher frequencies thermal noise in the bias line made a significant contribution (Trappen et al., 2023). Dephasing was primarily due to intrinsic low-frequency flux noise in the two qubit loops, with additional contribution from low-frequency noise of control electronics used for fast annealing (Trappen et al., 2023). The same study inferred a positive noise correlation $\alpha\approx0.3\text{–}0.43$4 between the two qubit loops, possibly due to non-local sources of flux noise or junction critical-current noise (Trappen et al., 2023).

The gradiometric fully tunable device identified Purcell loss via the readout resonator and ohmic charge noise via the bias lines as dominant relaxation channels, with a combined model reproducing the observed $\alpha\approx0.3\text{–}0.43$5 (Berlitz et al., 30 Sep 2025). A common misconception is that increasing $\alpha\approx0.3\text{–}0.43$6 removes flux-noise limitations altogether. The published tunable devices instead show that large shunting strongly suppresses charge dispersion while leaving flux noise, bias-line thermal noise, Purcell decay, and TLS-induced scatter as consequential mechanisms depending on operating point and packaging (Trappen et al., 2023, Berlitz et al., 30 Sep 2025).

5. Readout, control, and multi-qubit interactions

The original low-impedance flux qubit was capacitively coupled by $\alpha\approx0.3\text{–}0.43$7 to a half-wavelength coplanar-waveguide resonator with $\alpha\approx0.3\text{–}0.43$8 and $\alpha\approx0.3\text{–}0.43$9, enabling dispersive readout in a $T_1=60\text{–}90\,\mu\mathrm{s}$00 configuration (Steffen et al., 2010). With vacuum-Rabi splitting $T_1=60\text{–}90\,\mu\mathrm{s}$01, operation in the dispersive limit $T_1=60\text{–}90\,\mu\mathrm{s}$02 produced a measured dispersive shift of $T_1=60\text{–}90\,\mu\mathrm{s}$03, read out through standard I/Q heterodyne detection (Steffen et al., 2010). The 3D implementation used the two large shunt pads as a dipole antenna coupling to the TE$T_1=60\text{–}90\,\mu\mathrm{s}$04 cavity mode, with $T_1=60\text{–}90\,\mu\mathrm{s}$05, $T_1=60\text{–}90\,\mu\mathrm{s}$06, and $T_1=60\text{–}90\,\mu\mathrm{s}$07 (Abdurakhimov et al., 2019).

Fast single-qubit control benefits from the large level spacing available in high-anharmonicity CSFQ circuits. In the three-shunt flux circuit, pulses were shaped by a $T_1=60\text{–}90\,\mu\mathrm{s}$08 AWG, with $T_1=60\text{–}90\,\mu\mathrm{s}$09 $T_1=60\text{–}90\,\mu\mathrm{s}$10-pulses, maximum Rabi rate $T_1=60\text{–}90\,\mu\mathrm{s}$11, and a measured average single-qubit randomized-benchmarking fidelity of $T_1=60\text{–}90\,\mu\mathrm{s}$12 (Yurtalan et al., 2020). The same work extracted multilevel transition and dephasing rates in the qutrit manifold, showing that the CSFQ can be used as a genuinely multilevel superconducting circuit rather than only as an effective two-level system (Yurtalan et al., 2020).

In coupled architectures, the positive anharmonicity of the CSFQ can be exploited to suppress unwanted static $T_1=60\text{–}90\,\mu\mathrm{s}$13 interaction with a transmon of negative anharmonicity. For the hybrid CSFQ–transmon system, the residual static coupling in second-order perturbation theory is

$T_1=60\text{–}90\,\mu\mathrm{s}$14

with $T_1=60\text{–}90\,\mu\mathrm{s}$15 for the CSFQ and $T_1=60\text{–}90\,\mu\mathrm{s}$16 for the transmon (Ku et al., 2020). By tuning $T_1=60\text{–}90\,\mu\mathrm{s}$17, an exact zero of $T_1=60\text{–}90\,\mu\mathrm{s}$18 was observed, yielding a $T_1=60\text{–}90\,\mu\mathrm{s}$19-free pair (Ku et al., 2020). In echoed cross-resonance randomized benchmarking, the minimum two-qubit error per gate was $T_1=60\text{–}90\,\mu\mathrm{s}$20 at $T_1=60\text{–}90\,\mu\mathrm{s}$21 when static $T_1=60\text{–}90\,\mu\mathrm{s}$22, whereas at the CSFQ sweet spot, where $T_1=60\text{–}90\,\mu\mathrm{s}$23, the error increased to $T_1=60\text{–}90\,\mu\mathrm{s}$24 (Ku et al., 2020). This directly connects the CSFQ’s spectral structure to gate-level crosstalk mitigation.

6. Tunable variants, application domains, and relation to adjacent circuits

For coherent quantum annealing, Matsuzaki et al. proposed using CSFQs because the flux qubits employed in earlier demonstrations had coherence times of tens of nanoseconds, whereas the CSFQ at its flux sweet spot achieves $T_1=60\text{–}90\,\mu\mathrm{s}$25 according to the cited experimental literature (Matsuzaki et al., 2020). Their analysis emphasized that CSFQs have persistent current $T_1=60\text{–}90\,\mu\mathrm{s}$26 tens of nA, so their direct inductive coupling is $T_1=60\text{–}90\,\mu\mathrm{s}$27 smaller than in earlier large-$T_1=60\text{–}90\,\mu\mathrm{s}$28 flux qubits, and that the bare interaction near the sweet spot contains both Ising and flip-flop terms (Matsuzaki et al., 2020). To overcome this, they proposed a spin-lock-based QA protocol in the rotating frame, which reproduces the standard annealing Hamiltonian under the rotating-wave approximation and yielded numerically high fidelity $T_1=60\text{–}90\,\mu\mathrm{s}$29 on current hardware (Matsuzaki et al., 2020).

The fully tunable gradiometric C-shunted flux qubit extends the CSFQ concept toward wideband in-situ spectroscopy. At the tilt sweet spot, it combines first-order insensitivity to flux noise with a tunability range of $T_1=60\text{–}90\,\mu\mathrm{s}$30, from a few MHz up to $T_1=60\text{–}90\,\mu\mathrm{s}$31, and relaxation times up to $T_1=60\text{–}90\,\mu\mathrm{s}$32 (Berlitz et al., 30 Sep 2025). As a model application, it was used for strain-tuned swap spectroscopy of two-level tunneling defects over nearly one octave in frequency (Berlitz et al., 30 Sep 2025). This makes the device relevant not only for gate-based processing but also for materials characterization and defect spectroscopy.

Another application direction uses the multilevel structure of the CSFQ as a three-level working medium. In the superconducting C-shunt flux quantum-battery experiment, the three lowest states $T_1=60\text{–}90\,\mu\mathrm{s}$33 were driven by STIRAP and counterdiabatic STIRAP under two global norm constraints on the driving Hamiltonian (Li et al., 10 Apr 2025). Under $T_1=60\text{–}90\,\mu\mathrm{s}$34, the optimized protocol approached the quantum speed limit and achieved charging in $T_1=60\text{–}90\,\mu\mathrm{s}$35; the same work introduced the dimensionless stability-speed metric

$T_1=60\text{–}90\,\mu\mathrm{s}$36

where $T_1=60\text{–}90\,\mu\mathrm{s}$37 is the standard deviation of the ergotropy after the charging time (Li et al., 10 Apr 2025). This illustrates that the CSFQ can serve as a platform for coherent three-level control beyond conventional qubit operation.

The CSFQ should also be distinguished from capacitively shunted fluxonium, sometimes called “heavy” fluxonium. Heavy fluxonium uses a single small-area Josephson junction in parallel with a large superinductor and a large shunt capacitance to ground, rather than a three-junction flux-qubit loop (Earnest et al., 2017, Najera-Santos et al., 2023). That circuit realizes different regimes, including metastable fluxon states with lifetimes up to $T_1=60\text{–}90\,\mu\mathrm{s}$38 and an ultra-low-frequency qubit transition of $T_1=60\text{–}90\,\mu\mathrm{s}$39 in a heavy-fluxonium charge sensor (Earnest et al., 2017, Najera-Santos et al., 2023). The distinction matters because both families exploit shunt capacitance to suppress charging effects, but they differ in topology, inductive environment, selection rules, and intended operating regime.

Taken together, the published record presents the capacitively shunted flux qubit as a family of low-impedance superconducting flux circuits in which large shunt capacitance is used to suppress charge dispersion while retaining flux control. Depending on whether the design prioritizes single-well operation, gap tunability, strong anharmonicity, multilevel control, or weak-noise annealing, the CSFQ can realize coherence from the microsecond to the tens-of-microseconds regime, positive or negative effective anharmonicity, and applications spanning dispersive readout, cross-resonance entangling gates, quantum annealing, defect spectroscopy, magnetometry, and three-level energy-transfer protocols (Steffen et al., 2010, Abdurakhimov et al., 2019, Ku et al., 2020, Matsuzaki et al., 2020, Berlitz et al., 30 Sep 2025, Li et al., 10 Apr 2025).