Gate-Controlled Andreev Molecule

• The paper demonstrates nonlocal electrostatic control of Andreev bound states, showing that gate tuning can modulate supercurrents in proximitized junctions.
• The methodology employs InAs nanowire devices and tight-binding BdG simulations to capture ABS hybridization and anomalous phase shifts.
• The findings advance scalable superconducting platforms by enabling gate-controlled tuning for topological qubit architectures and sensitive parity sensors.

Searching arXiv for papers on gate-controlled Andreev molecules and related Andreev-molecule platforms. A gate-controlled Andreev molecule is an Andreev molecule in which the coupling and spectral structure of proximitized subgap states are controlled electrostatically rather than by loop-based phase bias. In the formulation introduced in "Density of States (Gate) - Controlled Andreev Molecule and Sensor" (Shi et al., 6 Aug 2025), an Andreev molecule is formed when Andreev bound states (ABSs) from two spatially proximate Josephson weak links overlap and hybridize across a shared superconducting segment, with the condition that the separation $L$ between the junctions is shorter than, or comparable to, the superconducting coherence length $\xi$. The distinctive feature of the gate-controlled, or “Type II,” realization is that finger gates tune the local density of states (DOS), chemical potential, and effective transmission of one site, thereby producing a nonlocal modification of the ABS spectrum and critical current of a neighboring site without requiring superconducting control loops (Shi et al., 6 Aug 2025).

1. Definition and underlying physical picture

The term Andreev molecule denotes a coupled subgap system formed from hybridized ABSs rather than from isolated local ABSs. In semiconductor nanowire double quantum dots, the molecular picture was introduced for proximitized dots whose singlet, doublet, and triplet manifolds hybridize through inter-dot tunneling, providing a minimal building block for Kitaev-like chains (Su et al., 2016). In parallel InAs nanowires with a shared epitaxial Al link, hybridization via the BCS vacuum was identified as an Andreev-mediated coupling channel $J$ that produces avoided crossings, bending of subgap lines toward zero energy near charge degeneracies, and strong enhancement of the weaker dot’s signal without direct interdot single-electron tunneling (Kürtössy et al., 2021).

In the short-junction language used for coupled Josephson weak links, the essential requirement is overlap of ABS wavefunctions inside a shared superconducting segment. The multichannel scattering treatment gives the exact bound-state condition

$$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$

with the coupling controlled by the length $l$ of the central superconducting segment, the channel-resolved scattering matrices, and the superconducting phases (Pillet et al., 2020). In the decoupled limit, the spectrum reduces to the familiar single-junction ABS dispersion

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

where $\tau$ is the transmission eigenvalue (Pillet et al., 2020).

The gate-controlled Andreev molecule of (Shi et al., 6 Aug 2025) keeps the same molecular premise—overlapping ABSs across a short shared superconductor—but replaces flux as the primary control variable by local electrostatic tuning of DOS, $\mu$, and $\tau$. This change is not merely instrumental. It relocates control from a global phase-bias degree of freedom to a local materials-and-spectrum degree of freedom, so the molecule is defined operationally by nonlocal DOS transduction rather than by loop-mediated phase steering (Shi et al., 6 Aug 2025).

2. Electrostatic control through DOS, hybridization, and transmission

In the Type II architecture, gate control is achieved by tuning the local DOS in each junction via finger gates. Increasing the DOS in one site, for example in junction JJ1 by raising its local chemical potential and transmission with TG1, nonlocally modifies the ABS spectrum and supercurrent of the other site JJs through two mechanisms: wavefunction hybridization and wavefunction transmission across the short shared superconductor (Shi et al., 6 Aug 2025).

The first mechanism is wavefunction hybridization. When the DOS in JJ1 is increased, ABSs emerge and their wavefunctions extend into the common superconductor, where they hybridize with ABSs of JJs and shift the phase dependence of JJs levels. In simulation, the ABS levels of JJs versus $\phi_s$ shift when $\xi$0 is changed; the shift is negligible at low DOS in JJ1 and becomes pronounced at high DOS, signaling stronger hybridization. The associated anomalous phase sensitivity, described in the paper as a $\xi$1 shift and TRSB, is presented as the hallmark of Andreev-molecule coupling (Shi et al., 6 Aug 2025).

The second mechanism is wavefunction transmission. The effective transmission from JJs into JJ1 grows as the gap at $\xi$2 in the ABS spectrum of JJs shrinks. With JJ1 depleted, the ABS gap at $\xi$3 is large, implying suppressed transmission; with JJ1 filled, the gap diminishes, indicating enhanced transmission. Because higher transmission increases the curvature of the ABS energy-phase dispersion, it increases $\xi$4 and hence $\xi$5 (Shi et al., 6 Aug 2025).

These two processes are summarized in the paper by an electrostatic renormalization of the untuned junction:

$\xi$6

with gates raising $\xi$7 and $\xi$8, and by the nonlocal correction

$\xi$9

Within this picture, the large nonlocal increase of $J$0 in JJs follows from a gate-tuned self-energy induced by the coupled site (Shi et al., 6 Aug 2025). A particularly notable result is that the nonlocal $J$1 enhancement in JJs can exceed the local $J$2 gained in JJ1, indicating that the remote response is not a weak perturbative by-product but a dominant spectral renormalization (Shi et al., 6 Aug 2025).

The broader literature had already established that electrostatic control of ABS energies can strongly modulate supercurrent. In the ABS–QD–ABS “Andreev trimer,” varying the ABS energies using electrostatic gates altered the switching current by more than an order of magnitude, with peaks tracking the ABS minima and anticrossings in spectroscopy (Bordin et al., 2024). The Type II molecule extends that logic from local ABS control to explicitly nonlocal DOS-controlled ABS engineering (Shi et al., 6 Aug 2025).

3. Device realizations and measurement architectures

The primary gate-controlled implementation uses a single InAs nanowire of diameter $J$3–$J$4 nm with an epitaxial Al half-shell. The Al is naturally oxidized after growth to stabilize the shell. The right site, denoted JJs, is a single Josephson junction where the Al film is fully etched to expose bare InAs and is gated by SG. The left site is reconfigurable: it can be a single junction JJ1, two junctions in series JJ1 and JJ2 separated by a small Al–InAs island, or a Cooper-pair transistor when the island acquires appreciable charging energy $J$5 due to barriers at JJ1 and JJ2. JJ1 and JJ2 are tuned by TG1 and TG2, and the island by a plunger gate PG (Shi et al., 6 Aug 2025).

The common superconducting segment $J$6 between the two sites has length $J$7 nm, shorter than $J$8 at $J$9 mK, ensuring ABS overlap and molecule formation. Readout is performed by connecting the ends of the two sites into a compact SQUID loop used solely for simultaneous CPR/$$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$0 readout of multiple sites. Control is purely via gates; no loop-based phase bias is needed to form or tune the molecule. Fast counter techniques based on repeated dc current pulses measure $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$1 efficiently. Measurements are near zero field, except that CPT spectroscopy uses an in-plane $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$2 T to suppress superconductivity in thick Al contacts while keeping the thin Al island superconducting; the base temperature is approximately $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$3 mK (Shi et al., 6 Aug 2025).

This geometry should be distinguished from planar phase-controlled Andreev-molecule experiments. In a two-dimensional InAs/Al platform, two planar JJs sharing a central superconducting electrode and embedded in superconducting loops were used to reveal phase-dependent Andreev molecule states and phase-controlled induced-gap closing (Matsuo et al., 2023). In a related double-loop interferometer, independent control of the two phase differences produced a nonlocal Josephson effect and a tunable $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$4-junction, with the nonlocal phase shift strongest for inter-junction distance $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$5 nm and absent for $$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$6m (Haxell et al., 2023). Those devices are phase-engineered; the Type II nanowire architecture uses the loop only for readout and relocates control to local electrostatics (Shi et al., 6 Aug 2025).

The same nanowire platform also supports an extended three-site molecule and a sensing mode. Configuring the left site as JJ2–JJ1 or as a CPT-like island introduces a controllable charging degree of freedom while preserving the short shared superconducting segment needed for nonlocal ABS coupling (Shi et al., 6 Aug 2025).

4. Effective models and microscopic descriptions

The canonical short-junction ABS dispersion employed throughout the Andreev-molecule literature is

$$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$7

and the supercurrent derived from the ABS manifold is

$$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$8

These expressions supply the local building blocks from which molecular spectra and nonlocal CPRs are constructed (Shi et al., 6 Aug 2025).

For the two-site Type II molecule, the paper introduces a minimal ABS-level Hamiltonian

$$\det\!\left[I_{8N} - S_N(E)\,S_S(E,\phi_L,\phi_R,l)\right]=0,$$9

where $l$0 are the local ABS dispersions of sites $l$1, $l$2 are ABS quasiparticle operators, $l$3 is the hybridization across the short $l$4 segment, and $l$5 accounts for induced superconducting pairing and possible phase-offset terms. In this picture, hybridization produces level repulsion and $l$6 shifts in the coupled spectrum (Shi et al., 6 Aug 2025).

The microscopic numerics use a tight-binding BdG Hamiltonian for a one-dimensional S–N1–S–N2–S chain:

$l$7

with discretized form

$l$8

Here $l$9 are Pauli matrices in Nambu space, $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$0 in superconducting regions and zero in normal regions, and $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$1 encodes gate-induced electrostatic potentials $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$2 for JJ1 and $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$3 for JJs. The middle superconducting segment is chosen as $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$4, ensuring coupling. The CPR follows

$$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$5

with $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$6 the Fermi–Dirac distribution (Shi et al., 6 Aug 2025).

The scattering formulation generalizes this to multichannel weak links. There the exact bound-state condition

$$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$7

shows how gate-dependent channel transmissions, scattering phases, and nonlocal EC/CAR amplitudes produce avoided crossings and nonlocal Josephson coupling when $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$8 (Pillet et al., 2020). This suggests that the DOS-controlled Type II platform can be interpreted either as a localized ABS-level hybrid system or as a gate-programmable multichannel scatterer whose control variables are $$E_{\pm}(\phi)=\pm \Delta \sqrt{1-\tau \sin^2(\phi/2)},$$9, $\tau$0, and the effective nonlocal self-energy (Shi et al., 6 Aug 2025).

For practical CPR extraction in the asymmetric SQUID limit, when $\tau$1 or $\tau$2, the oscillation amplitude equals the weaker junction’s $\tau$3 while the oscillation center equals the stronger junction’s $\tau$4. In the Type II measurements, nonlocal modulation is therefore tracked through changes in the center value rather than through local weak-link amplitude alone (Shi et al., 6 Aug 2025).

5. Experimental signatures, extended molecules, and sensing modes

The two-site molecule JJ1–JJs displays two regimes. In the non-molecule regime, with low DOS in JJ1, increasing TG1 raises $\tau$5 and the SQUID oscillation amplitude grows, but the center $\tau$6 remains constant; JJ1 and JJs behave independently. In the molecule regime, further increasing the DOS with TG1 produces a large nonlocal rise of $\tau$7, sometimes exceeding $\tau$8 itself, while remaining in the asymmetric SQUID limit. Tight-binding numerics reproduce the crossover: $\tau$9 is insensitive to $\mu$0 at low DOS, then grows strongly as $\mu$1 raises DOS and ABSs appear in JJ1. The measured gate sensitivity in the two-site case is approximately $\mu$2 nA/V between TG1 $\mu$3–$\mu$4 V (Shi et al., 6 Aug 2025).

The extended molecule JJ2–JJ1–JJs introduces a third active site. When JJ2 is tuned by TG2 with $\mu$5, the remote $\mu$6 shows negligible change at small $\mu$7 and then significant enhancement when $\mu$8 is sufficiently large. Two routes are identified: direct extension of the molecular renormalization chain, and serial supercurrent limitation through effective phase biasing of JJ1 by JJ2. The paper argues that direct molecule coupling is dominant because the measured CPR becomes non-centrosymmetric around zero flux, indicative of TRSB from hybridization (Shi et al., 6 Aug 2025).

The CPT–JJs configuration turns the left island into a charge-sensitive control site. Coulomb diamonds give $\mu$9eV from diamond height $\tau$0. The island exhibits $\tau$1 periodic $\tau$2 at zero bias and $\tau$3 periodicity at higher bias. Through molecule coupling, single-Cooper-pair addition or removal on the island nonlocally modulates the ABSs and $\tau$4 of JJs. The interferometric response is strongest on Coulomb resonance and weakest in blockade, with gate sensitivity

$\tau$5

compared with approximately $\tau$6 nA/V in the two-site case, an over-$\tau$7 enhancement (Shi et al., 6 Aug 2025).

The sensing mode is explicitly noninvasive in the sense used by the paper: the sensor junction is spatially separated from the CPT island, the coupling is through the shared superconductor, and the sensor does not directly tunnel charge through the island during readout. The same logic motivates parity readout. In Andreev systems, the occupation, or parity, of ABS or Majorana states changes the set of populated $\tau$8 entering $\tau$9 via $\phi_s$0, so fixing parity can alter $\phi_s$1 and $\phi_s$2. The Type II work uses $\phi_s$3 interferometry rather than dispersive readout, but it states that the large DOS-to-$\phi_s$4 transduction suggests high susceptibility suitable for dispersive parity sensing (Shi et al., 6 Aug 2025).

A later related experiment demonstrated deterministic non-local control over the parity configuration of a proximitized quantum dot inside an Andreev molecule, with universal selection rules that depend on whether ECT or CAR is dominant and with supercurrent, directly signaled by zero-bias conductance peaks, serving as a sensor-free parity probe (Zhu et al., 27 Jan 2026). Taken together, these results place gate-controlled molecules in a dual role: they are both spectral engineering elements and readout transducers.

6. Gate control, flux control, and the route to Kitaev-chain architectures

The Type II gate-controlled molecule is explicitly contrasted with flux-controlled, loop-based “Type I” Andreev molecules. In the Type II scheme, the control knob is the set of local gates that tune DOS, $\phi_s$5, and $\phi_s$6 per site. The stated advantages are that it avoids adding superconducting loops per site, enables compact designs and scaling to many sites, allows independent fine-grained tuning, provides high sensitivity enabling sensing functions, avoids flux cross-talk and flux noise, and is compatible with complex chains and sensors. The stated trade-offs are electrostatic disorder, charge noise, and the need for careful screening and layout to minimize capacitive cross-talk, mitigated in the reported layout by grounded superconducting leads (Shi et al., 6 Aug 2025).

By contrast, flux-controlled Andreev molecules use magnetic flux in loops to set phase biases $\phi_s$7. Their advantages are direct phase engineering and rich $\phi_s$8-space manipulation; their limitations are that scaling demands multiple loops and flux lines with mutual inductance and cross-talk, increases footprint, and makes independent per-site flux control challenging (Shi et al., 6 Aug 2025). The earlier nonlocal Josephson and Josephson-diode experiments exemplify this phase-first regime: planar coupled JJs showed anomalous phase shifts up to $\phi_s$9 to $\xi$00 and anomalous supercurrents of $\xi$01 nA controlled by the phase and gate state of a neighboring JJ (Haxell et al., 2023), while nanowire-based Andreev molecules exhibited a Josephson diode effect whose efficiency changed sign as the non-local phase modulated the competition between DEC and dCAR, with a maximum efficiency of approximately $\xi$02 and a central-peak feature in gate space (Zhu et al., 19 Aug 2025).

The architectural motivation for DOS control is its direct connection to Kitaev-chain engineering. The Type II paper maps a chain of coupled ABSs onto the Kitaev Hamiltonian

$\xi$03

with the idealized topological criterion

$\xi$04

for a uniform chain with $\xi$05, yielding Majorana edge modes (Shi et al., 6 Aug 2025). In the proposed implementation, local gates set $\xi$06 and DOS on each segment, $\xi$07 is set by inter-site overlap across short superconducting segments with $\xi$08, and $\xi$09 is induced by proximity together with molecule-induced anomalous phase shifts and hybridization. Because the control is electrical rather than Zeeman-based, the approach aims at topological behavior without large magnetic fields, following the “poor man’s MZMs” strategy in minimal chains (Shi et al., 6 Aug 2025).

This program has deep antecedents. The proximitized double-dot Andreev molecule was already framed as a minimal element for emulating a one-dimensional $\xi$10-wave superconductor and for testing Majorana fusion rules (Su et al., 2016). Parallel-wire Andreev molecules established BCS-vacuum-mediated nonlocal pairing as a scalable route to artificial chains (Kürtössy et al., 2021). The ABS–QD–ABS trimer demonstrated cooperative screening and strong gate control of supercurrent in a three-site molecule (Bordin et al., 2024). A heteroatomic Andreev molecule, built from two QDs coupled by a finite superconducting island, further showed that gate control of island parity can drive a quantum phase transition from an antiferromagnetic Andreev molecular state to a heteroatomic molecule with ferromagnetically coupled QDs (Kürtössy et al., 2024). These developments collectively indicate that “gate-controlled Andreev molecule” refers not only to one device class but to a broader design principle: coupled subgap states are treated as electrically programmable artificial matter.

One recurrent interpretive caution is that phase-induced zero-bias features or gap closings in coupled Josephson structures do not by themselves establish topological zero modes. In coherently coupled planar JJs, phase-dependent gap closing was reproduced by Andreev-molecule hybridization together with Rashba SOI, and the resulting zero-bias features disappeared when either junction was pinched off, identifying them as properties of the coupled non-topological molecule rather than robust Majorana end states (Matsuo et al., 2023). In that sense, the gate-controlled molecule is best understood as a controllable ABS platform whose relevance to topological qubits depends on subsequent chain formation, parity control, and edge-mode stabilization rather than on the mere presence of nonlocal hybridization.

The immediate implication is that scaling to long Kitaev chains requires explicit treatment of nonlocal renormalizations. The Type II work states this directly: fine gate tuning of DOS and couplings must account for nonlocal DOS effects when building multi-site chains and seeking sweet spots for robust edge modes (Shi et al., 6 Aug 2025). A plausible implication is that, in future large arrays, electrostatic calibration will need to optimize molecular couplings globally rather than site by site, because each gate can renormalize not only local $\xi$11 and $\xi$12 but also neighboring $\xi$13 and effective pairing channels.