Five physicists in China and Brazil have proposed reading out an Andreev qubit through an oscillating supercurrent, an electric current that flows without resistance. By their calculations, the qubit produces that current on its own while it is in a superposition of its two states. Physical Review B published the theoretical work as a Letter on August 5, 2026.
The authors calculated the current and simulated the readout, and they reported no measurement on a device. In the December 2025 version of their preprint, they described the arrangement in their first figure as an envisioned experimental setup. They wrote that experimenters can use the supercurrent to probe the qubit “under currently available experimental capabilities.”
Xian-Peng Zhang, Chuanchang Zeng, and Yugui Yao of the Beijing Institute of Technology wrote the Letter with Zhen-Biao Yang of Fuzhou University and José Carlos Egues of the University of São Paulo. Zhang is also affiliated with the Hong Kong University of Science and Technology.
Andreev qubits
An Andreev qubit stores information in Andreev levels, discrete energy states that form in a weak link between two superconductors. A supercurrent can pass through such a link, which physicists call a Josephson junction.
The authors modeled a junction built around a quantum dot defined by gate electrodes in an indium arsenide nanowire. A quantum dot is a tiny island that confines electrons. On both sides, the dot connects through tunnel barriers to nanowire leads made superconducting by aluminum.
In their model, the qubit’s 0 is the junction’s lowest-energy state. Its 1 is that state with a pair of quasiparticles added to the lowest Andreev level. Quasiparticles are electron-like excitations of a superconductor.
The predicted current
In a superposition, a qubit’s state is a weighted combination of 0 and 1 with a definite relative phase between them. The authors calculated that the supercurrent through the junction then has a static part, equal to a weighted average of the currents of the 0 and 1 states, and a dynamic part that oscillates at the qubit’s frequency.
The authors attributed the oscillation to quantum interference between the two states and called it “an AC-like Josephson effect without an applied external voltage.” In the standard AC Josephson effect, an alternating current appears when a constant voltage is applied across a junction. The authors also wrote that the oscillating part vanishes at operating points where the two qubit states have the same energy.
The proposed readout
In the preprint, the authors proposed placing the junction in a large superconducting loop coupled to the small pickup coil of a SQUID magnetometer. A SQUID, or superconducting quantum interference device, is a highly sensitive detector of magnetic flux. The magnetometer would sense the flux from the current circulating in the loop.
In the scheme, the authors derived the weighting of the qubit’s state between 0 and 1 from the static current and the size of the oscillation. They derived the relative phase between the two parts of the superposition from the timing of the oscillation.
The authors wrote that nondemolition protocols “use ancilla qubits to identify the fragile quantum state of a qubit without destroying its encoded information.” An ancilla is an auxiliary qubit that takes on information from the qubit being measured and is then read out in its place. The authors wrote that their readout “avoids ancillae altogether” and that “no repetitive qubit reinitialization is needed.” A nondemolition measurement leaves intact the one property it measures and can still disturb other properties of an unknown state.
Simulated performance
In the preprint, the authors tested the scheme against imperfect measurements by adding random noise to the simulated signal, with a standard deviation of 2% of their reference current. They modeled the disturbance from the pickup coil, called back-action, with one adjustable coupling parameter.
For one test state, the authors simulated three, five, or nine measurements per oscillation period and reported readout infidelities of about 0.48%, 0.15%, and 0.19%, respectively, in the preprint. Infidelity is a measure of how far a reconstructed state is from the prepared one, where zero means a perfect match. The figures come from a simulation of the readout alone, not from an error-correction experiment.
In the preprint, the authors wrote that “these values lie well within standard quantum error-correction thresholds.” An error-correction threshold is the physical error rate below which adding physical qubits to an error-correcting code reduces the rate of logical errors.
The authors estimated the back-action in the preprint by assuming a pickup-coil current of about 0.1 microampere and a junction supercurrent of 1 to 10 microamperes, a range they took from earlier studies. With those values, they calculated that the pickup coil’s flux shifts the phase across the junction by about 0.001 of a cycle.
Preprint versions
The authors first posted the work on September 25, 2024, as an arXiv preprint under a different title. In that version, they described the readout as a “fully nondestructive measurement” of the qubit state. They simulated noise of 4% of the reference current.
In the second version, posted on December 26, 2025, the authors used the published title and described the SQUID readout and the back-action model. In the abstract of that version, they wrote that experimenters can use the supercurrent to probe the qubit “arbitrarily close to the nondestructive limit.”
Physical Review B received the manuscript on September 28, 2024, and accepted it on July 8, 2026, according to the publisher’s metadata.
Update, November 12, 2026: In an October 21 interview with Agência FAPESP, Egues said, “What doesn’t yet exist is that geometry, that setup that we proposed.”
