Hybrid Josephson Rhombus: Tunable Interferometer

• Hybrid Josephson rhombus is a superconducting interferometric element composed of four gate-tunable Josephson junctions that enable controlled interference between 2e and 4e transport processes.
• It achieves tunable current–phase relations and nonreciprocal superconducting diode effects by balancing magnetic frustration with electrostatic gate control.
• Arrays of hybrid rhombi offer protected manifolds and collective spin behavior, with design parameters optimized to balance noise channels and qubit transition frequencies.

The hybrid Josephson rhombus is a superconducting interferometric element formed by four Josephson junctions arranged in a loop, with the junction couplings controlled by magnetic frustration and, in the hybrid implementation, by local gate voltages on semiconductor–superconductor weak links. In its balanced and fully frustrated limit, the device suppresses the first Josephson harmonic and realizes a $\pi$-periodic $\cos(2\phi)$ potential associated with coherent charge-$4e$ transport; away from that limit, the same architecture supports tailored current–phase relations, a superconducting diode effect, and qubit implementations ranging from parity-protected designs to a deliberately softened, biased-noise qubit with directly addressable low-energy transitions (Banszerus et al., 2024). In array form, rhombi shunted by a large capacitance admit an effective spin description with a protected manifold and a flux-driven quantum phase transition, while in the multi-mode single-rhombus realization a large shunt capacitor and a weakened junction produce localized phase-valley states with strongly asymmetric relaxation and dephasing characteristics across flux bias regimes (Schrade et al., 2021).

1. Device topology and control variables

The hybrid Josephson rhombus comprises four gate-tunable semiconductor–superconductor Josephson junctions, $J_1,\dots,J_4$, arranged as two parallel arms in a superconducting loop. Two junctions in series form each arm, and each junction’s Josephson energy $E_{J,i}(n_{g,i})$ is controlled electrostatically by a local gate voltage $V_i$. Magnetic frustration is defined by

$f=\Phi_R/\Phi_0,$

where $\Phi_R$ is the flux through the rhombus loop and $\Phi_0=h/2e$, while a phase bias $\phi$ is placed between the two superconducting leads (Banszerus et al., 2024).

A more explicit multi-mode description treats the rhombus as a four-junction loop with two “small” islands (nodes 2 and 4) and two “large” islands (nodes 1 and 3), together with a large shunt capacitor $\cos(2\phi)$0 connecting nodes 1 and 3. After choosing a spanning tree, one finds three independent degrees of freedom labeled by superconducting phases $\cos(2\phi)$1 and conjugate Cooper-pair numbers $\cos(2\phi)$2,

$\cos(2\phi)$3

with one convenient choice

$\cos(2\phi)$4

and the constraint $\cos(2\phi)$5 encoding the external flux. In this description the physical offset charges on the large islands appear as parameters $\cos(2\phi)$6, $\cos(2\phi)$7 (Sanchez et al., 7 May 2026).

These two descriptions emphasize complementary aspects of the same element. The gate-defined hybrid viewpoint foregrounds in situ tunability of Josephson harmonics and arm asymmetry, whereas the multi-mode circuit viewpoint foregrounds charging anisotropy, phase-space localization, and the role of the shunting capacitor in suppressing charge motion on the small islands.

2. Energy functionals and current–phase relation

In the classical limit $\cos(2\phi)$8, the total potential energy of the hybrid rhombus is

$\cos(2\phi)$9

subject to the fluxoid constraints

$4e$0

Minimizing $4e$1 over the internal phases $4e$2 gives the ground-state energy $4e$3, from which the current–phase relation follows:

$4e$4

For the full rhombus, interference between the two arms yields the Fourier expansion

$4e$5

equivalently

$4e$6

with $4e$7 (Banszerus et al., 2024).

At the single-arm level, two SIS-type junctions in series with energies $4e$8 form an effective single-mode junction characterized by

$4e$9

and current–phase relation

$J_1,\dots,J_4$0

For $J_1,\dots,J_4$1 or $J_1,\dots,J_4$2 this reduces to a nearly sinusoidal form, whereas for $J_1,\dots,J_4$3 it becomes strongly nonsinusoidal with substantial higher harmonics (Banszerus et al., 2024).

In the multi-mode soft-rhombus circuit, orthodox circuit quantization gives

$J_1,\dots,J_4$4

Here the capacitance matrix $J_1,\dots,J_4$5 encodes self- and cross-capacitances among the three modes, and the external flux $J_1,\dots,J_4$6 enters through the last cosine term (Sanchez et al., 7 May 2026).

The coexistence of Fourier-CPR and multi-mode Hamiltonian descriptions is central to the subject. The former makes harmonic engineering explicit; the latter resolves the internal phase valleys and the charging structure that determine whether the element functions as an interferometric nonlinear circuit component or as a qubit.

3. Destructive interference, $J_1,\dots,J_4$7 behavior, and nonreciprocal transport

For balanced Josephson couplings at full frustration, the rhombus displays the canonical interference effect that motivated protected-rhombus proposals. If $J_1,\dots,J_4$8 and $J_1,\dots,J_4$9, odd harmonics interfere destructively so that $E_{J,i}(n_{g,i})$0, and the dominant term is the second harmonic,

$E_{J,i}(n_{g,i})$1

The first harmonic corresponds to single Cooper-pair ($E_{J,i}(n_{g,i})$2) tunneling, whereas the surviving $E_{J,i}(n_{g,i})$3 contribution implies an elementary tunneling event carrying charge $E_{J,i}(n_{g,i})$4. In an RF-driven experiment this maps, through $E_{J,i}(n_{g,i})$5, to steps of $E_{J,i}(n_{g,i})$6 (Banszerus et al., 2024).

The same cancellation can be expressed in the harmonic language used for gate-tunable interferometers. If the upper and lower arms have balanced first-harmonic amplitudes,

$E_{J,i}(n_{g,i})$7

then at $E_{J,i}(n_{g,i})$8 the two $E_{J,i}(n_{g,i})$9 paths acquire Aharonov–Bohm phases $V_i$0 and interfere destructively,

$V_i$1

while double–Cooper-pair processes acquire $V_i$2 and add constructively, leaving

$V_i$3

when only up to second harmonic is retained (Schrade et al., 2021).

The hybrid rhombus is not confined to this balanced limit. By keeping each arm internally balanced but deliberately mismatching the two arms’ total $V_i$4, inversion symmetry is broken. Off a time-reversal-symmetric frustration, $V_i$5, the current–phase relation becomes asymmetric and the superconducting diode effect appears,

$V_i$6

with diode efficiency

$V_i$7

The measured efficiency reaches $V_i$8 in the optimized regime, while vanishing at $V_i$9 and at the arm-balance point (Banszerus et al., 2024).

A recurring misconception is that the rhombus is intrinsically a pure $f=\Phi_R/\Phi_0,$0 element. The published results show instead that $f=\Phi_R/\Phi_0,$1 behavior requires balanced couplings and half-flux frustration; gate imbalance or flux detuning reintroduce odd harmonics or asymmetric transport. This suggests that the hybrid Josephson rhombus is best viewed as an interferometric platform for controlled harmonic selection, of which the parity-protecting $f=\Phi_R/\Phi_0,$2 regime is a special case.

4. Soft rhombus and biased-noise qubit realization

The soft-rhombus approach intentionally departs from the perfectly balanced protected element. In the realized circuit, one junction is weakened according to

$f=\Phi_R/\Phi_0,$3

so that at $f=\Phi_R/\Phi_0,$4 the destructive-interference structure is softened rather than made exact. The stated purpose is to investigate the “soft version of the rhombus qubit,” directly probe the qubit transitions over several GHz, and reduce the potential drawbacks of interferometer-based protection (Sanchez et al., 7 May 2026).

Ignoring charges, the potential-energy landscape at frustration is

$f=\Phi_R/\Phi_0,$5

For the symmetric rhombus, $f=\Phi_R/\Phi_0,$6, the minima occur at

$f=\Phi_R/\Phi_0,$7

forming a double-well structure in three dimensions. More generally, the two minima move continuously with $f=\Phi_R/\Phi_0,$8, and away from $f=\Phi_R/\Phi_0,$9 the potential becomes tilt-biased. In the soft-rhombus case, bringing the wells closer in phase space lifts the degeneracy to $\Phi_R$0 at $\Phi_R$1, while away from half flux the tilt localizes the two lowest states in different minima (Sanchez et al., 7 May 2026).

Near frustration, projecting onto the two lowest eigenstates $\Phi_R$2 yields

$\Phi_R$3

with $\Phi_R$4. More generally,

$\Phi_R$5

where

$\Phi_R$6

so that

$\Phi_R$7

The transition frequency increases with flux detuning, and the resulting operating mode is described as a biased-noise qubit (Sanchez et al., 7 May 2026).

The measured coherence metrics are strongly regime-dependent. In the biased-noise regime, for large $\Phi_R$8, the reported averages are

$\Phi_R$9

At the frustration sweet spot, the averages are

$\Phi_0=h/2e$0

with Gaussian Ramsey decays and echo extending $\Phi_0=h/2e$1 by approximately $\Phi_0=h/2e$2 (Sanchez et al., 7 May 2026). The contrast between long relaxation away from frustration and longer Ramsey dephasing at frustration is one of the defining empirical signatures of the soft-rhombus implementation.

5. Arrays, protected manifolds, and collective-spin description

A single balanced rhombus with charging energy can be written as

$\Phi_0=h/2e$3

and in the balanced half-flux case this reduces to

$\Phi_0=h/2e$4

The logical structure in this formulation is associated with even and odd Cooper-pair parity sectors, which motivated the original protected-rhombus qubit concept (Schrade et al., 2021).

Protection becomes more explicit in an array of $\Phi_0=h/2e$5 identical rhombi in series, producing $\Phi_0=h/2e$6 superconducting islands. Neighboring islands are coupled by small capacitance $\Phi_0=h/2e$7, while the two end islands are shunted by a large bus capacitance $\Phi_0=h/2e$8. After rotating away the offset charges, the charge–phase Hamiltonian is

$\Phi_0=h/2e$9

In the tight-binding limit $\phi$0, each rhombus contributes two nearly degenerate Bloch bands, or valleys at $\phi$1, and the low-energy theory becomes

$\phi$2

Here $\phi$3 is a single-rhombus tunneling splitting, $\phi$4 is a diagonal hybridization proportional to $\phi$5, and $\phi$6 encodes flux detuning (Schrade et al., 2021).

For $\phi$7, one finds $\phi$8, and the two ferromagnetic configurations

$\phi$9

become the two lowest eigenstates, split only by processes exponentially small in $\cos(2\phi)$00. These states carry opposite total Cooper-pair parity and are robust against any single-site $\cos(2\phi)$01 or $\cos(2\phi)$02 noise. For uniform detuning, the Hamiltonian can be rewritten as the Lipkin–Meshkov–Glick form

$\cos(2\phi)$03

with a protected phase for $\cos(2\phi)$04 and an unprotected phase for $\cos(2\phi)$05. The critical point occurs at

$\cos(2\phi)$06

Numerically, the protection window in flux broadens with $\cos(2\phi)$07 under balanced conditions (Schrade et al., 2021).

The same analysis also identifies a stringent condition: any imbalance

$\cos(2\phi)$08

reintroduces a residual $\cos(2\phi)$09 term and destroys flux protection, even for large $\cos(2\phi)$10, because parity is no longer an exact symmetry (Schrade et al., 2021). The array architecture therefore sharpens, rather than relaxes, the requirement of in situ balancing.

6. Noise channels, operating windows, and research directions

In the multi-mode soft-rhombus qubit, the reported relaxation channels at low frequency are flux noise and quasiparticle tunneling. Flux noise couples through

$\cos(2\phi)$11

with relaxation rate

$\cos(2\phi)$12

using $\cos(2\phi)$13. For quasiparticles tunneling across junction $\cos(2\phi)$14,

$\cos(2\phi)$15

with $\cos(2\phi)$16 and $\cos(2\phi)$17. Dielectric loss and Purcell loss affect high-frequency plasmon modes but are negligible for the low-$\cos(2\phi)$18 fluxon states (Sanchez et al., 7 May 2026).

The loss analysis identifies a nonmonotonic operating window. At low $\cos(2\phi)$19, below $\cos(2\phi)$20, flux noise and quasiparticles dominate with approximately $\cos(2\phi)$21 scaling; above $\cos(2\phi)$22, dielectric loss and Purcell loss dominate with approximately $\cos(2\phi)$23 dependence. Their crossing predicts an optimal $\cos(2\phi)$24–$\cos(2\phi)$25 for the longest $\cos(2\phi)$26, using $\cos(2\phi)$27, $\cos(2\phi)$28, $\cos(2\phi)$29, $\cos(2\phi)$30, and $\cos(2\phi)$31 (Sanchez et al., 7 May 2026).

For the hybrid-gated rhombus as a CPR-engineering element, the recommended regimes are explicit. Pure $\cos(2\phi)$32 transport requires $\cos(2\phi)$33 matched within approximately $\cos(2\phi)$34 and $\cos(2\phi)$35 pinned to $\cos(2\phi)$36. Maximum diode efficiency requires arms balanced internally, $\cos(2\phi)$37, but with $\cos(2\phi)$38–$\cos(2\phi)$39, and with $\cos(2\phi)$40 midway between integer and half-integer frustration. Large loops or reference junctions with higher critical current are recommended for clean phase biasing when extracting the CPR directly (Banszerus et al., 2024).

Across the literature, protection and tunability are therefore complementary rather than identical objectives. The protected-array proposals emphasize exact balance, half-flux frustration, and strong collective coupling; the soft-rhombus qubit trades exact charge-parity protection for a moderate $\cos(2\phi)$41; and the hybrid CPR study emphasizes deterministic control of harmonic content and arm asymmetry (Schrade et al., 2021). Possible extensions that have been explicitly proposed include concatenating multiple soft rhombi for longitudinal coupling in Ising arrays, using higher-harmonic junctions such as semiconductor weak links to restore full $\cos(2\phi)$42 protection, engineering cat-codes to exploit the strong $\cos(2\phi)$43 bias, and optimizing $\cos(2\phi)$44 to tune $\cos(2\phi)$45 and the curvature for a targeted noise sweet spot (Sanchez et al., 7 May 2026).

Taken together, these results establish the hybrid Josephson rhombus as a tunable interferometric building block whose defining feature is controlled interference between Josephson harmonics. In one limit it approximates a pure $\cos(2\phi)$46 element and supports parity-based protection; in another it functions as a biased-noise qubit with localized phase-valley states and long relaxation; and in still another it provides in situ synthesis of asymmetric or higher-harmonic current–phase relations for nonreciprocal transport. The unifying constraint is that the most distinctive regimes arise only under carefully engineered balance, frustration, and capacitive environment.