How does a twisted superconducting quantum diode enable high fidelity anharmonic qubits?

Moderate rectification beats maximal rectification

The paper asks whether a superconducting diode should be pushed toward perfect one-way conduction before it is useful in a quantum circuit. Its answer is no. In twisted NbSe2 the authors measure a diode efficiency of 27.6%, far below the 100% ideal, and their simulations show that this intermediate value is the one that best preserves qubit anharmonicity and stabilizes a two-level system.

Two quantities are being traded against each other. Diode efficiency, written η, is the difference between the forward and reverse critical currents divided by their sum, expressed as a percentage; it measures how strongly current prefers one direction. Anharmonicity is the spacing that separates a qubit's two working levels from higher levels; without it, drive pulses can push the qubit into states it should never occupy. The paper's central claim is that raising η improves directionality but tilts the potential energy landscape in a way that shrinks the effective well and suppresses that desirable anharmonicity.

A one-degree twist creates the asymmetry

The device is a twisted stack of NbSe2, a layered superconductor that can be exfoliated into thin flakes. Two layers are assembled with a twist angle of 1°, and the resulting heterostructure is placed on prepatterned Cr/Au electrodes and encapsulated with h-BN to limit oxidation. The authors fabricated devices with overall thicknesses of 16, 30 and 60 nm, corresponding to 8, 15 and 30 nm per NbSe2 layer.

Transport was measured in a four-probe geometry with current–voltage scans inside a Physical Property Measurement System spanning 1.6 K to 400 K and magnetic fields up to ±9 T. Before each diode measurement the superconducting magnet was degaussed, and the sample–field alignment was calibrated with full 360° rotation-dependent resistance measurements at 10 K. That calibration matters because the paper later argues that field misalignment cannot explain the in-plane response.

The reference temperature is a superconducting transition at 6.87 K, extracted from resistance–temperature data. Critical currents were extracted from the I–V curves, with error bars from bootstrap resampling: 100 resamplings per critical current, drawn from a window of up to 11 data points around the detected critical point.

In-plane and out-of-plane fields peak differently

The 27.6% figure is not a single global property of the material; it is the maximum efficiency of the 16 nm device under an out-of-plane field, reached at −13 Oe. Under an in-plane field applied parallel or antiparallel to the current, the same 16 nm device reaches a maximum η of about 24.0% at −100 Oe. These are separate field orientations with separate optimal values, and the paper treats them as distinct responses.

The two orientations are attributed to different mechanisms. The in-plane asymmetry is linked to broken inversion symmetry in the superconducting state, with interfacial spin–orbit coupling, possibly including a radial Rashba component, predicted to generate a superconducting diode effect. The out-of-plane response is attributed to the vortex diode effect: twisting the two layers makes the edges inequivalent, which changes the heights of the Bean–Livingston surface barriers and therefore the onset of resistance depending on current polarity.

Thickness changes the picture. The 16 nm device shows a pronounced field-dependent efficiency, while the 30 nm and 60 nm devices show much weaker nonreciprocity. The paper explains this through the vortex diode effect: in thicker samples the Meissner screening current density is reduced and the effective Pearl penetration depth becomes smaller, which strengthens the asymmetry of the vortex surface barriers and binds vortices more firmly to the edges, suppressing the transition into the resistive state. The 60 nm device also shows hysteresis in its I–V response, which the authors relate to weak damping and a Stewart–McCumber parameter above 1.

A residual zero-field response and its caveat

At fields below 10⁻⁴ T the authors measure η ≈ −2.1%, whereas NbSe2-based diodes are reported to show vanishing rectification at B = 0. They interpret this finite intrinsic anisotropy as a possible signature of an emergent correlated phase with a spontaneous twist-induced magnetization, arising from strong electronic interactions, the moiré landscape and non-trivial band topology combining to produce orbital magnetization and spontaneously broken time-reversal symmetry.

The authors also state a limitation directly: a residual magnetic field of several Oe from trapped flux in the superconducting magnet cannot be completely excluded and may contribute to the observed field offset. The zero-field claim should therefore be read as a possible intrinsic effect rather than a fully isolated one.

Efficiency tilts the well that defines the qubit

The quantum part of the paper is a simulation, not a measurement on a fabricated qubit. The authors model a transmon qubit chain with quantum diodes using QuTiP, representing quantum states, density matrices, Hamiltonians and time evolution, with a Josephson junction supplying the diode efficiency. They then compute the potential-energy landscape and the energies of the qubit levels inside the wells for representative efficiencies.

The mechanism is a competition inside a single parameter. A larger η does two things at once: it enhances the asymmetry of the current–phase relation, improving directionality and strengthening nonreciprocal transport, but it also tilts the potential U(φ), reducing the effective potential well and suppressing the anharmonicity that keeps higher excited states out of reach. Because both effects grow with η, fidelity has an optimum rather than a maximum at η → 100%.

The paper reports that a perfect two-level system can be achieved for η close to 27.6%, the value measured in the twisted device. This is the step that connects the transport measurement to the qubit claim: the experimentally attained efficiency is presented as sitting near the simulated optimum, not as a compromise awaiting improvement.

Directional fidelity survives realistic noise

Fidelity here is defined as the overlap between the intended and achieved quantum states, F = |⟨ψ|φ⟩|², and forward and reverse fidelity measure how accurately an excitation is transferred in each direction. In noiseless simulations the forward fidelity approaches unity while the reverse fidelity falls near zero, producing high-contrast regions of ideal directional selectivity.

The noisy simulations add parameter fluctuations of ±0.5 in the Josephson energy, charge noise of amplitude 0.03 representing Cooper pair number fluctuations, phase noise of amplitude 0.02 from flux variations, energy dissipation at rate 0.01 modeling environmental coupling, and measurement uncertainty of ±0.002. Noise broadens the features and slightly lowers peak fidelity but preserves the asymmetry. The maximum forward–reverse difference in the two-level case occurs around η ≈ 22.9–44.8% and 67.1–77.4% for evolution times of 1–10 ns, and the signal-to-noise ratio exceeds 10:1 in the optimal parameter regime.

The range matters for reading the headline number. The measured 27.6% falls inside the first simulated window, but the simulations also identify a second high-asymmetry window at much higher efficiency, so the paper's claim is about an optimum region rather than a single sharp threshold.

Design for fidelity, not for the on/off ratio

The paper's stated departure from prior work is a change of objective. Much of the field pursues higher η and higher on/off ratios, with reported efficiencies around 70% in sputtered materials and about 50% maintained at higher temperatures in some van der Waals systems based on high-temperature superconductors. This study argues that in a quantum circuit the relevant target is fidelity and anharmonicity under noise and design constraints, and that an intermediate efficiency already achieved experimentally can serve that target better.

The practical reading is a priority ordering for twisted-diode design: the 16 nm thickness, the out-of-plane field near −13 Oe and an efficiency near the simulated optimum are the conditions the paper identifies as favorable, while thicker devices weaken the nonreciprocity. The authors are explicit that integration into practical quantum circuits and experimental testing of fabricated quantum diodes remain future work, so the qubit benefit is a simulation-based design principle rather than a demonstrated device result.

Sources used in this answer