Andreev Qubits Overview

• Andreev qubits are superconducting qubits encoded in subgap Andreev bound states that arise from coherent Andreev reflection in weak links.
• They support versatile implementations—including even-parity pair and odd-parity spin encodings—in platforms like atomic contacts, nanowire and quantum-dot junctions.
• Recent experiments demonstrate coherent control, photon-mediated long-range coupling, and cavity-QED integration, highlighting their scalability and potential in quantum computing.

Andreev qubits are superconducting qubits encoded in subgap Andreev bound states (ABS) of superconducting weak links, including short Josephson junctions, quantum-dot Josephson junctions, semiconductor nanowire junctions, and related hybrid structures. In the simplest short-junction limit, a single high-transmission channel supports ABS with energies $$E_A(\varphi)=\pm \Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$, and these microscopic states can be promoted to qubit degrees of freedom by selecting either an even-parity pair excitation or an odd-parity spin doublet (Janvier et al., 2015). Subsequent work established coherent manipulation and single-shot readout of such microscopic quasiparticle states in superconducting atomic contacts (Janvier et al., 2015), photon-mediated long-range coupling between two Andreev level qubits over $\approx 6$ mm (Cheung et al., 2023), and architectures spanning ultrastrong circuit-QED coupling (Shvetsov et al., 13 Feb 2025), quantum-dot implementations (Zhang et al., 2024), multi-terminal junctions (Piasotski et al., 2024), and topological-edge realizations (Latini et al., 29 Jan 2026). The term therefore denotes a family of qubit modalities rather than a single hardware instantiation.

1. Physical basis in Andreev bound states

Andreev bound states are discrete subgap eigenstates localized at weak links between superconductors. In a short superconducting junction, coherent Andreev reflection at the two superconducting interfaces produces a pair of subgap levels whose phase dependence is set by the superconducting phase difference $\phi$ or $\varphi$ across the junction. For a single conduction channel of transmission $\tau$, the canonical short-junction dispersion is

$$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$

or equivalently $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$, depending on notation (Janvier et al., 2015). The associated supercurrent is determined by the phase derivative of the ABS energy, $I(\phi)=(2e/\hbar)\partial E/\partial \phi$, which is the central mechanism by which Andreev qubits couple to microwave circuits (Cheung et al., 2023).

In superconducting weak links formed from semiconductors, the ABS are not merely orbital states. Spin–orbit coupling, Zeeman fields, and Coulomb interaction reshape the ABS manifold, producing spin-split or interaction-renormalized subgap states. In quantum-dot Josephson junctions, the superconducting Anderson impurity model

$$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$

captures the even- and odd-parity subgap sectors, with $U$, $\approx 6$0, $\approx 6$1, spin–orbit coupling strength $\approx 6$2, and background tunneling $\approx 6$3 controlling the crossover between ABS- and Yu–Shiba–Rusinov-like regimes (Iličin et al., 28 Dec 2025). This places Andreev qubits at the intersection of mesoscopic superconductivity, impurity physics, and circuit QED.

A common misconception is that all Andreev qubits are equivalent to short-junction, noninteracting ABS pair qubits. The literature instead distinguishes several encodings. Even in nominally similar Josephson junctions, the relevant qubit can be an even-parity “pair” excitation, an odd-parity spin excitation, or a more strongly correlated state shaped by charging energy and multi-orbital physics (Iličin et al., 28 Dec 2025). Another recurring source of confusion is the relation to Majorana devices: trivial, semilocal ABS can mimic local Majorana signatures, but they do not inherit the nonlocal noise protection of a true topological Majorana qubit (Mishmash et al., 2019).

2. Encodings and microscopic qubit manifolds

The most established encoding is the Andreev pair qubit, sometimes also called an Andreev level qubit. In a short, highly transmissive weak link, the two even-parity many-body states $\approx 6$4 and $\approx 6$5 formed from the lowest ABS define a qubit with transition frequency $\approx 6$6 (Janvier et al., 2015). In the nanowire implementation of “Photon-mediated long range coupling of two Andreev level qubits,” the qubit is encoded in the lowest pair of ABS of a short, highly transmissive superconducting weak link, with transition frequency

$\approx 6$7

and a sweet spot at $\approx 6$8 where the frequency is first-order insensitive to phase noise (Cheung et al., 2023).

A second major encoding is the Andreev spin qubit, in which information is stored in a singly occupied, odd-parity ABS doublet. In nanowire and quantum-dot Josephson junctions, spin–orbit coupling and phase bias can produce a spin-dependent supercurrent and a finite spin splitting without requiring the same control resources as conventional semiconductor spin qubits (Pita-Vidal et al., 2022). In the strongly interacting quantum-dot realization, the relevant low-energy Hamiltonian is

$\approx 6$9

with $\phi$0 MHz and $\phi$1 MHz extracted at the operating point in the reported device (Pita-Vidal et al., 2022). In that implementation the qubit is encoded in the spin-split doublet ground state of an odd-occupied quantum-dot Josephson junction, which suppresses leakage because higher excitations are separated by at least $\phi$2 GHz (Pita-Vidal et al., 2022).

The distinction between even-parity and odd-parity encodings becomes sharper once interactions are included. In the even-parity “Andreev pair qubit” of an interacting quantum-dot Josephson junction, the qubit splitting is

$\phi$3

and interactions admix Yu–Shiba–Rusinov components into the ABS-like singlets. This enhances spin transitions in the presence of spin–orbit coupling and introduces sensitivity to local magnetic field fluctuations, especially near the ABS–YSR crossover $\phi$4 (Iličin et al., 28 Dec 2025). This suggests that the same device family can interpolate continuously between charge-dominated, inductive, and spin-sensitive operating regimes rather than occupying a single universal “Andreev qubit” limit.

The literature also contains more specialized encodings. A topological-edge Andreev spin qubit has been proposed in a magnetically doped quantum-spin-Hall Josephson junction, where the qubit basis is the pair of spin-resolved ABS wavefunctions $\phi$5 and microwave electric fields drive transitions through dipole matrix elements $\phi$6 activated by transverse magnetic exchange (Latini et al., 29 Jan 2026). A solitonic Andreev spin qubit has likewise been proposed in a Corbino-disk Josephson junction, where a fluxoid mismatch creates Jackiw–Rebbi-like spin-degenerate ABS localized at a controllable angular position and holonomic gates arise from phase-driven transport of the soliton state (San-Jose et al., 18 Jun 2025). These variants preserve the core Andreev-qubit logic—encoding in discrete superconducting subgap states—while changing the geometry, symmetry class, and control manifold.

3. Control, cavity coupling, and readout

Andreev qubits couple naturally to circuit-QED hardware because the ABS energy depends on superconducting phase and therefore on zero-point flux fluctuations. For a pair qubit coupled to a resonator, the interaction can be written in Jaynes–Cummings form,

$\phi$7

with dispersive readout in the detuned regime through a qubit-state-dependent shift $\phi$8 (Souto et al., 2024). This same basic logic underlies atomic-contact experiments, nanowire ABS qubits, and quantum-dot spin-supercurrent devices.

The earliest clear demonstration of coherent control of microscopic ABS as a qubit was performed in superconducting atomic contacts. In that work, a single high-transmission channel with $\phi$9 was embedded in a circuit-QED architecture with a niobium quarter-wave resonator at $\varphi$0 GHz, and the measured qubit–resonator coupling was $\varphi$1 MHz (Janvier et al., 2015). The device exhibited single-shot readout, Rabi oscillations, $\varphi$2s, $\varphi$3 ns, Hahn-echo $\varphi$4 ns, and parity switching at $\varphi$5 kHz (Janvier et al., 2015). This established Andreev qubits as experimentally coherent microscopic superconducting degrees of freedom rather than merely spectroscopic features.

Subsequent nanowire work pushed coupling far beyond the standard strong-coupling regime by maximizing resonator impedance. In a lumped-element NbTiN resonator with differential impedance above $\varphi$6 at $\varphi$7–$\varphi$8 GHz, the maximum measured coupling to an Andreev pair qubit reached $\varphi$9 GHz, corresponding to $\tau$0 for one device, while coupling to an Andreev spin transition reached $\tau$1 MHz (Shvetsov et al., 13 Feb 2025). The reported system therefore approached the ultrastrong-coupling regime and exhibited spectroscopic features consistent with short-junction and finite-length spin–orbit models (Shvetsov et al., 13 Feb 2025).

Readout modalities are no longer limited to standard dispersive spectroscopy. In quantum-dot Josephson junctions, the oscillatory supercurrent generated by coherent superpositions of many-body Andreev states provides a non-collapsing electric readout channel. For a prepared state $\tau$2, the current expectation value contains a coherent interference term,

$\tau$3

or, in the form used in the paper,

$\tau$4

with $\tau$5 (Zhang et al., 2024). This readout is explicitly non-collapsing in the sense that it monitors the evolving expectation value of the current rather than projecting onto the qubit eigenbasis.

For Andreev spin qubits, control selection rules can be highly geometry-dependent. In a clean phase-driven nanowire ASQ, direct spin-flip matrix elements of the current operator vanish, motivating Raman protocols. However, time-reversal-preserving impurities generically create spin-flip transmission and a nonzero current matrix element, enabling direct single-tone phase driving with Rabi frequency

$\tau$6

(Fauvel et al., 2023). In topological-edge proposals, electric-dipole transitions are instead activated by transverse magnetic exchange in the weak link, with realistic gate pulses at $\tau$7–$\tau$8 V/m and $\tau$9 ps sufficient for $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$0 and $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$1 rotations in the simulated HgTe/Nb platform (Latini et al., 29 Jan 2026).

4. Coupled Andreev qubits and scalable architectures

A central development has been the move from single-qubit spectroscopy to coherent inter-qubit coupling. The most direct demonstration is the photon-mediated long-range coupling of two Andreev level qubits in a two-mode superconducting cavity coupler (Cheung et al., 2023). In that architecture, two InAs full-shell nanowire weak links were placed $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$2 mm apart and coupled through an engineered pair of hybridized coplanar resonator modes. The symmetric mode at $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$3 GHz, with $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$4 MHz, enabled fast dispersive readout, while the anti-symmetric mode at $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$5 GHz had strongly suppressed coupling to the readout port and functioned as a low-loss quantum bus (Cheung et al., 2023).

The single-qubit couplings to the anti-symmetric mode were $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$6 MHz and $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$7 MHz, with qubit decay rates $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$8 MHz and $$E_A(\varphi,\tau)=\Delta \sqrt{1-\tau \sin^2(\varphi/2)}$$9 MHz (Cheung et al., 2023). When both qubits were tuned into resonance with the bus mode, the resulting spectra exhibited the bright–dark structure of the Tavis–Cummings Hamiltonian,

$E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$0

including a characteristic photon-free dark state

$E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$1

that becomes a maximally entangled two-qubit state for $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$2 (Cheung et al., 2023). The work was explicitly spectroscopic rather than tomographic, but it established long-range coherent coupling and entangled eigenstates for spatially separated Andreev pair qubits (Cheung et al., 2023).

Scalability has also been addressed at the architectural level for spin qubits. A blueprint for all-to-all connected superconducting spin qubits proposes using spin-dependent supercurrents and a common coupling junction to implement selective, distance-independent longitudinal couplings

$E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$3

with

$E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$4

(Pita-Vidal et al., 2024). In the proposed operating regime, putting qubits at $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$5 or $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$6 turns their spin-dependent supercurrent “ON,” while $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$7 turns it “OFF,” permitting selective pairwise interactions while all other qubits remain effectively uncoupled (Pita-Vidal et al., 2024). This is a superconducting-spin analog of flux-programmable Ising connectivity rather than a cavity-bus architecture.

Three-terminal junctions add another route to large, phase-dependent spin splittings. In the idealized three-terminal Andreev spin qubit, non-Abelian spin–orbit holonomy around the loop produces large pseudo-spin splittings at zero magnetic field, but pseudo-spin conservation blocks direct transitions between the lowest pseudo-spin-split pair (Piasotski et al., 2024). The paper therefore concludes that operating the system as a practical spin qubit requires explicit pseudo-spin mixing through external magnetic field or magnetic impurities (Piasotski et al., 2024). This clarifies that large equilibrium spin splitting alone is not sufficient; accessible matrix elements must coexist with the desired low-energy structure.

5. Decoherence, parity dynamics, and material constraints

Parity switching is a defining issue for Andreev qubits because the encoded states often coexist with odd-parity or even-parity manifolds that are separated only by quasiparticle poisoning. In the atomic-contact Andreev qubit, continuous measurement directly revealed both $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$8 quantum jumps and parity switches into and out of the odd manifold, with a measured parity-switching rate of $E_A(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)}$9 kHz at 30 mK (Janvier et al., 2015). In nanowire ALQ experiments, odd-parity ABS states were currentless in the short-junction limit and therefore decoupled from the cavity, appearing spectroscopically as bare cavity lines and as parity-resolved overlays in Tavis–Cummings spectra (Cheung et al., 2023).

Decoherence mechanisms differ strongly between pair and spin encodings. For Andreev pair qubits, phase and transmission noise are natural dephasing channels because the qubit frequency is directly tied to the ABS energy–phase relation. In the atomic-contact implementation, the linewidth minimum at $I(\phi)=(2e/\hbar)\partial E/\partial \phi$0 identified that point as first-order insensitive to flux noise, while the remaining dephasing was modeled using 1/f transmission noise together with white and 1/f flux noise (Janvier et al., 2015). In interacting quantum-dot pair qubits, increasing $I(\phi)=(2e/\hbar)\partial E/\partial \phi$1 admixes Yu–Shiba–Rusinov components into the even-parity states and can make the qubit magnetically sensitive through spin–orbit-enabled spin matrix elements, especially in the crossover region $I(\phi)=(2e/\hbar)\partial E/\partial \phi$2 (Iličin et al., 28 Dec 2025).

For Andreev spin qubits, the central question is whether decoherence is dominated by charge-like fluctuations, magnetic fluctuations, or parity dynamics. A dedicated theoretical study concluded that the dephasing rate follows the local spin difference for magnetic noise and the local charge difference for electric noise, and that comparison to experiment points to magnetic noise as the dominant source in present nanowire ASQs (Hoffman et al., 2024). In the realistic parameter set matched to experiment, tuning the semiconductor filling from $I(\phi)=(2e/\hbar)\partial E/\partial \phi$3 meV to $I(\phi)=(2e/\hbar)\partial E/\partial \phi$4 meV increased the simulated $I(\phi)=(2e/\hbar)\partial E/\partial \phi$5 from $I(\phi)=(2e/\hbar)\partial E/\partial \phi$6 ns to $I(\phi)=(2e/\hbar)\partial E/\partial \phi$7 ns, reflecting a magnetic-noise sweet spot near a subband anticrossing (Hoffman et al., 2024).

This interpretation is consistent with broader materials trends. InAs-based Andreev spin qubits have demonstrated direct all-electric control and strong coupling to transmons, but their coherence is limited by short $I(\phi)=(2e/\hbar)\partial E/\partial \phi$8 and echo times. In the strongly coupled QD-JJ implementation, $I(\phi)=(2e/\hbar)\partial E/\partial \phi$9s at $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$0 mT and $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$1 GHz, while $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$2 ns, Hahn-echo $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$3 ns, and Carr–Purcell dynamical decoupling yielded $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$4 ns for $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$5 (Pita-Vidal et al., 2022). The paper explicitly attributes the short dephasing times primarily to nuclear spin dynamics in InAs, with flux noise deemed subdominant (Pita-Vidal et al., 2022).

This motivates alternative materials, particularly germanium. Theoretical work on Ge-based Josephson junctions argues that the absence of resolved Andreev spin qubits in existing Ge devices can result from geometry-induced qubit splittings below the thermal scale $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$6, rather than from the absence of spin splitting itself (Hoffman et al., 16 Jun 2025). The same study gives concrete design guidance: for $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$7 and $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$8, the optimal width is $$H = H_{\mathrm{dot}} + H_{\mathrm{leads}} + H_t + H_{\mathrm{SOC}} + H_{\mathrm{bg}} + H_Z$$9 nm, while for thinner $U$0-confinement $U$1 nm and the maximum qubit frequency can be more than doubled (Hoffman et al., 16 Jun 2025). Since Ge can be isotopically purified and hole hyperfine coupling is reduced relative to InAs, this suggests a plausible route to ASQs with substantially improved coherence if geometry and filling are chosen to push $U$2 above the dilution-refrigerator thermal scale (Hoffman et al., 16 Jun 2025).

A distinct protection strategy uses Franck–Condon blockade. In a transmon-embedded Andreev spin qubit with spin-dependent Josephson potentials,

$U$3

the two spin states shift the equilibrium phase by $U$4, with $U$5 (Kurilovich et al., 10 Jun 2025). The overlap of the corresponding plasmon wavefunctions yields a Franck–Condon parameter $U$6, and the zero-temperature relaxation rate becomes

$U$7

(Kurilovich et al., 10 Jun 2025). For realistic parameters $U$8 GHz and $U$9–$\approx 6$00 MHz, the paper estimates $\approx 6$01, corresponding to a $\approx 6$02 enhancement by a factor $\approx 6$03 at sufficiently low temperature (Kurilovich et al., 10 Jun 2025). This does not improve pure dephasing, but it rebalances the noise budget in favor of phase errors rather than relaxation.

6. Emerging directions, variants, and broader significance

One active direction is to adapt Andreev qubit concepts to new symmetry classes and geometries. Magnetically doped quantum-spin-Hall junctions yield spin-resolved ABS whose dipole transitions are activated by exchange-induced spin mixing, with simulated NOT and Hadamard gates using $\approx 6$04 ps Gaussian pulses and transition frequencies in the $\approx 6$05–$\approx 6$06 GHz range for HgTe/Nb parameters (Latini et al., 29 Jan 2026). Josephson-vortex-bound Andreev spin qubits in planar junctions use a weak out-of-plane field to create Josephson vortices that bind isolated spinful ABS; in the reported parameter set, the qubit splitting is $\approx 6$07eV, corresponding to $\approx 6$08 GHz, with flux-driven Rabi frequencies $\approx 6$09 MHz for $\approx 6$10 and cQED readout shifts $\approx 6$11 MHz for $\approx 6$12 (Laubscher et al., 11 Dec 2025). These proposals aim to reduce hardware overhead relative to gate-defined nanowire spin qubits while keeping supercurrent-based readout and long-range interactions.

Another direction is heterogeneous integration. Fast readout of semiconductor dot spin qubits via electrically tunable coupling to Andreev spin qubits has been proposed for germanium devices, with MHz-scale dispersive or longitudinal shifts and calculated readout fidelities above $\approx 6$13 in sub-$\approx 6$14s times (Jakob et al., 24 Jun 2025). Here the ASQ acts as a fast superconducting-spin ancilla, while the quantum-dot spin qubit retains its longer coherence during idle periods because the DSQ–ASQ coupling can be turned off electrically (Jakob et al., 24 Jun 2025). A plausible implication is that Andreev qubits may become infrastructural elements for readout and networking even in processors whose logical qubits are not themselves ABS-based.

At the conceptual boundary with topological quantum information, ABS remain double-edged. Precise engineering of subgap states in quantum-dot Josephson structures has produced both Andreev qubits and minimal Kitaev chains, showing that trivial subgap states can be valuable quantum resources in their own right rather than mere nuisances on the path to Majorana devices (Souto et al., 2024). At the same time, noisy Majorana-qubit analyses emphasize that semilocal ABS-based qubits remain fundamentally distinct from nonlocal Majorana encodings because their dephasing and splitting do not exhibit the exponential protection with device length that characterizes topological qubits (Mishmash et al., 2019). This distinction is critical in interpreting zero-bias and parity signatures.

Taken together, the current literature supports a technical view of Andreev qubits as a broad platform class defined by controllable superconducting subgap states rather than by any single microscopic realization. The field now spans microscopic pair qubits in atomic contacts (Janvier et al., 2015), long-range-coupled Andreev level qubits in nanowire weak links (Cheung et al., 2023), directly driven and transmon-coupled Andreev spin qubits in quantum-dot Josephson junctions (Pita-Vidal et al., 2022), ultrastrong-coupling implementations using high-impedance resonators (Shvetsov et al., 13 Feb 2025), and multiple proposals for topological-edge, multi-terminal, vortex-bound, and holonomic variants (Latini et al., 29 Jan 2026). The main unresolved issues are no longer merely whether ABS can act as qubits—they can—but which encodings, materials, and architectures best suppress parity switching, dephasing, and control overhead while preserving the unusually strong and tunable coupling of ABS to superconducting circuits.