Josephson current in clean ballistic superconductor-normal-superconductor junctions: the continuum (scattering-state) contribution beyond bound-state dominance and its cancellation structure

The Josephson current of a clean superconductor-normal-superconductor (SNS) sandwich whose normal-layer thickness $2d$ greatly exceeds the superconducting coherence length was derived by Bardeen and Johnson (Phys. Rev. B 5, 72 (1972)) from the discrete Andreev bound states trapped in the normal layer, giving at sufficiently low temperature a sinusoidal current-phase relation with amplitude suppressed as $\exp(-4\pi d^* k_B T/\hbar v_F)$. They explicitly left open whether the scattering states above the superconducting gap add a larger near-gap term of the type proposed by Kulik, how to evaluate the scattering-state contribution in the crossover regime where bound-state dominance fails, and how impurities, imperfect interfaces, the magnetic field generated by the current, and self-consistency of the pair potential modify the result. Kulik's larger term is now documented to have been incorrect; a unified Matsubara-frequency expression containing both the bound-state and continuum contributions exists in one dimension; and the leading terms of a high-temperature expansion in inverse normal-layer thickness have been evaluated, confirming the Bardeen-Johnson exponential as the leading sinusoidal term. A precise open core survives: the closed-form analytic derivation of the leading continuum contributions beyond order $1/L^{3/2}$, of the temperature-independent term that survives past the established $1/L^{3/2}$ cancellation, and of the current in the regime above the crossover temperature $T^*$ where power-law corrections dominate the exponential; the closely related construction of a fully charge-conserving, self-consistent theory is the subject of a still-unresolved controversy running from 2021 onward.

Background

A Josephson junction is a weak electrical link joining two superconductors, through which an equilibrium supercurrent flows as a periodic function of the phase difference $\phi$ between the superconducting order parameters of the two electrodes (the order parameter is a complex field whose magnitude $\Delta_0$ sets the superconducting energy gap). The system studied here is an SNS sandwich: a slab of normal metal of thickness $2d$ between two superconducting leads. "Clean" or "ballistic" means the electronic mean free path is large compared with $2d$, so impurity scattering is negligible, and the long-junction regime of interest has $2d$ large compared with the Bardeen-Cooper-Schrieffer coherence length $\xi_0 = \hbar v_F/\pi\Delta_0$ (with $v_F$ the Fermi velocity), so that superconducting correlations induced in the normal metal by its contact with the superconducting banks (the proximity effect) are negligible.

The elementary scattering process at a normal-superconducting interface is Andreev reflection, introduced by Andreev in his analysis of the intermediate state of superconductors: an electron incident from the normal side with energy below the gap $\Delta_0$ cannot propagate into the superconductor as a quasiparticle and is instead reflected as a hole, transferring charge $2e$ to the condensate as a Cooper pair. Two such interfaces facing each other trap quasiparticles, since the electron and its retroreflected hole repeatedly convert into one another; the normal layer therefore acts as an effective potential well of depth $\Delta_0$. Bardeen and Johnson showed that the well has an effective half-thickness $d^* = d + \hbar^2 k_{zF}/(2m\Delta_0)$, where $m$ is the electron mass and $k_{zF}$ the Fermi momentum component along the junction normal, the increment being the penetration depth of the evanescent quasiparticle wave into the superconductors. The resulting discrete subgap levels (Andreev bound states) have energies $E_n = \hbar^2 k_{zF}\,\pi(n+\frac{1}{2})/(2md^*)$ for integer $n$, and each level's energy depends on the phase difference $\phi$.

The equilibrium supercurrent follows from the phase dispersion of the occupied levels. At zero temperature Bardeen and Johnson, by a Galilean-invariance argument (a phase difference is equivalent to a uniform velocity of the condensate, in whose frame the bound-particle calculation becomes elementary), recovered Ishii's result: a sawtooth (piecewise-linear) current-phase relation $J(\phi) = n_e e\hbar\phi/(4md)$ for $|\phi|<\pi$ (with $n_e$ the electron density), periodic in $\phi$ with jumps of sign at odd multiples of $\pi$. At finite but still low temperature the current takes the sinusoidal Josephson form $J = J_1\sin\phi$ with $J_1 = (6 n_e \hbar e/2 m d^*)\exp(-4\pi d^* k_B T/\hbar v_F)$: the level quantization in the normal region makes the current decrease extremely rapidly with temperature (a sinusoidal form of this type had been assumed without derivation in a contemporary review by Waldram, Pippard, and Clarke). This exponential is valid only when $k_B T \lesssim \hbar^2 k_{zF}\pi/(2md^*)$, the Andreev-level spacing; outside this window the bound-state approximation fails.

Above the bound levels lies the continuum of scattering states with energies $E \geq \Delta_0$. These also carry a phase-dependent current, and their resonance structure — the scattering-state density is peaked when the phase accumulation in the normal layer is a multiple of $2\pi$, with peaks sharpening as $E \to \Delta_0$ and broadening into a uniform continuum at higher energy — made Bardeen and Johnson warn that "the net current is a very small difference between large terms" and that obtaining the current when the bound-state formula is no longer valid requires a more careful joint analysis of bound and scattering states. Their idealized model additionally neglected the magnetic field produced by the current, treated impurity and interface scattering only at the level of a survival probability, and took the pair potential to be a prescribed step function rather than a self-consistent solution of the gap equation (the equation determining the pair potential from the quasiparticle states it creates). The tool that makes a joint treatment of bound and continuum states tractable is the Matsubara-frequency method: the current is evaluated at the discrete imaginary energies $\epsilon_p = i\,\pi k_B T\,(2p+1)$ with integer $p$ (the Matsubara frequencies), where the resonant structure of the continuum is smoothed out and the sums over states converge rapidly; in one dimension the entire current, bound-state and continuum parts together, collapses into a single such expression.

The modern reformulation splits the total current into a condensate current $J_s$ (from the phase gradient in the superconducting leads), a vacuum current $J_v$ (from the filled negative-energy levels in the normal layer), and an excitation current $J_q$ (from the thermally occupied levels). Charge conservation requires the latter two to cancel appropriately, and in the high-temperature long-junction limit the exponentially small supercurrent emerges from a mutual cancellation of large temperature-independent terms, established through order $1/L^{3/2}$ (with $L = 2d$ the normal-layer thickness). Because the step-function pair potential breaks gauge invariance, the model admits mathematically correct solutions that violate charge conservation and must be filtered out, or else the pair potential must be determined self-consistently. Quantitative questions about the relative roles of bound and continuum states therefore remain tied to the conceptual question of which solutions of the model are physical. Impurity scattering, imperfect interface transmission, the self-generated magnetic field, and a self-consistent pair potential are separate generalizations of this clean step-potential model rather than parts of the continuum question itself.

Problem Statement

For a clean ballistic superconductor-normal-superconductor junction modeled by a step-function (square-well) pair potential, with normal-layer thickness $2d$ (written $L = 2d$ where $L$ appears below) large compared with the superconducting coherence length: obtain the controlled, quantitative theory of the equilibrium Josephson current in the regime where the bound-state-dominated approximation fails, that is, when the thermal energy exceeds the Andreev-level spacing $\hbar^2 k_{zF}\pi/(2md^*)$. This requires, jointly, (i) the leading continuum (scattering-state, $E \geq \Delta_0$) contribution beyond the established $1/L^{3/2}$ expansion, that is, the first order past it, together with its phase dependence and its interplay with the discrete Andreev bound states — Kulik's proposed larger near-gap term, suppressed only as $\exp(-2\pi d^* k_B T/\hbar v_F)$, is now documented to be incorrect; (ii) a closed-form, analytic determination, independent of (i), of the temperature-independent residual that survives past the cancellation of the vacuum and excitation currents established through order $1/L^{3/2}$, including the mechanism by which that cancellation fails and the phase dependence of the residual; and (iii) the closed-form analytic derivation of the current in the temperature regime above $T^*$, with its phase dependence, and whether its leading power of $L$ is universal across the three dimensionalities and persists at orders beyond the first surviving correction; here $T^*$ is the temperature at which the temperature-independent power-law term equals the exponentially small current. The numerical finding that the temperature-independent current decays as a power law rather than exponentially in this regime, together with the cancellation of the leading temperature-independent terms established through order $1/L^{3/2}$, are premises of this problem rather than its targets. The one-dimensional forms of the established leading results (the unified Matsubara expression and the Bardeen-Johnson exponential) are known, but the open sub-questions (i)-(iii) must each be resolved in all three dimensionalities (one, two, and three dimensions): because the transverse momentum integration changes the relative weight of the bound and continuum contributions, results obtained in one dimension do not transfer trivially to the others. Impurity scattering, imperfect interface transmission, the magnetic field generated by the current, and self-consistency of the pair potential are separate generalizations of the clean step-potential model studied here; in particular, whether a charge-conserving treatment must include phase gradients in the superconducting leads is itself an open controversy (see Refs. 20-22) and is not presupposed here. The current-phase relation obtained should reduce, in the appropriate limits, to the established results: the zero-temperature sawtooth, the Bardeen-Johnson exponential, and the one-dimensional Matsubara-sum expression (the unified current obtained by summing over the discrete imaginary energies $\epsilon_p = i\,\pi k_B T\,(2p+1)$).

References

1. A. F. Andreev, 'The thermal conductivity of the intermediate state in superconductors', Zh. Eksp. Teor. Fiz. 46, 1823 (1964) [Sov. Phys. JETP 19, 1228 (1964)] (no DOI registered), https://scholar.google.com/scholar?q=Andreev+%22thermal+conductivity+of+the+intermediate+state+in+superconductors%22+1964
2. I. O. Kulik, 'Macroscopic quantization and proximity effect in S-N-S junctions', Zh. Eksp. Teor. Fiz. 57, 1745 (1969) [Sov. Phys. JETP 30, 944 (1970)] (no DOI registered), https://scholar.google.com/scholar?q=Kulik+%22Macroscopic+quantization+and+proximity+effect+in+S-N-S+junctions%22
3. J. R. Waldram, A. B. Pippard, J. Clarke, 'Theory of the current-voltage characteristics of SNS junctions and other superconducting weak links', Philosophical Transactions of the Royal Society of London Series A 268, 265 (1970), doi:10.1098/rsta.1970.0075, https://doi.org/10.1098/rsta.1970.0075
4. C. Ishii, 'Josephson currents through junctions with normal metal barriers', Progress of Theoretical Physics 44, 1525 (1970), doi:10.1143/PTP.44.1525, https://doi.org/10.1143/PTP.44.1525
5. J. Bardeen, J. L. Johnson, 'Josephson current flow in pure superconducting-normal-superconducting junctions', Physical Review B 5, 72 (1972), doi:10.1103/PhysRevB.5.72, https://doi.org/10.1103/PhysRevB.5.72
6. A. V. Svidzinskii, T. N. Antsygina, E. N. Bratus', 'Superconducting current in wide S-N-S junctions', Zh. Eksp. Teor. Fiz. 61, 1612 (1971) [Sov. Phys. JETP 34, 860 (1972)] (no DOI registered), https://scholar.google.com/scholar?q=Svidzinskii+Antsygina+Bratus+%22Superconducting+current+in+wide+S-N-S+junctions%22
7. C. Ishii, 'Thermodynamical properties of Josephson junction with a normal metal barrier', Progress of Theoretical Physics 47, 1464 (1972), doi:10.1143/PTP.47.1464, https://doi.org/10.1143/PTP.47.1464
8. A. V. Svidzinsky, T. N. Antsygina, E. N. Bratus', 'Concerning the theory of the Josephson effect in pure SNS junctions', Journal of Low Temperature Physics 10, 131 (1973), doi:10.1007/BF00655245, https://doi.org/10.1007/BF00655245
9. J. Niemeyer, G. von Minnigerode, 'Effects of nonmagnetic and magnetic impurities in the normal metal layer of SNS junctions', Zeitschrift für Physik B Condensed Matter 36, 57 (1979), doi:10.1007/BF01333954, https://doi.org/10.1007/BF01333954
10. M. Yu. Kupriyanov, Fiz. Nizk. Temp. 7, 700 (1981) [Sov. J. Low Temp. Phys. 7, 342 (1981)] (no DOI registered), https://scholar.google.com/scholar?q=Kupriyanov+1981+SNS+junction+current+arbitrary+thickness
11. H. van Houten, C. W. J. Beenakker, 'Andreev reflection and the Josephson effect in a quantum point contact', Physica B Condensed Matter 175, 187 (1991), doi:10.1016/0921-4526(91)90712-N, https://doi.org/10.1016/0921-4526(91)90712-N
12. C. W. J. Beenakker, H. van Houten, 'Josephson current through a superconducting quantum point contact shorter than the coherence length', Physical Review Letters 66, 3056 (1991), doi:10.1103/PhysRevLett.66.3056, https://doi.org/10.1103/PhysRevLett.66.3056
13. P. F. Bagwell, 'Suppression of the Josephson current through a narrow, mesoscopic, semiconductor channel by a single impurity', Physical Review B 46, 12573 (1992), doi:10.1103/PhysRevB.46.12573, https://doi.org/10.1103/PhysRevB.46.12573
14. U. Gunsenheimer, U. Schüssler, R. Kümmel, 'Symmetry breaking, off-diagonal scattering, and Josephson currents in mesoscopic weak links', Physical Review B 49, 6111 (1994), doi:10.1103/PhysRevB.49.6111, https://doi.org/10.1103/PhysRevB.49.6111
15. R. A. Riedel, L.-F. Chang, P. F. Bagwell, 'Critical current and self-consistent order parameter of a superconductor-normal-metal-superconductor junction', Physical Review B 54, 16082 (1996), doi:10.1103/PhysRevB.54.16082, https://doi.org/10.1103/PhysRevB.54.16082
16. H. X. Tang, Z. D. Wang, Y. Zhang, 'Normal reflection effect on the critical current in a clean-limit superconductor-normal-metal-superconductor junction', Zeitschrift für Physik B Condensed Matter 101, 359 (1997), doi:10.1007/s002570050220, https://doi.org/10.1007/s002570050220
17. Svidzinsky and Sakhnyuk, 'Current states in SNS junction for arbitrary concentrations of nonmagnetic impurities', Condensed Matter Physics 3, 683 (2000), doi:10.5488/CMP.3.3.683, https://doi.org/10.5488/CMP.3.3.683
18. Sakhnyuk and Svidzinsky, 'Current states in an SNS junction for arbitrary thickness of the normal layer at temperatures close to critical', Condensed Matter Physics 6, 159 (2003), doi:10.5488/CMP.6.1.159, https://doi.org/10.5488/CMP.6.1.159
19. A. A. Golubov, M. Yu. Kupriyanov, E. Il'ichev, 'The current-phase relation in Josephson junctions', Reviews of Modern Physics 76, 411 (2004), doi:10.1103/RevModPhys.76.411, https://doi.org/10.1103/RevModPhys.76.411
20. E. B. Sonin, 'Ballistic SNS sandwich as a Josephson junction', Physical Review B 104, 094517 (2021), doi:10.1103/PhysRevB.104.094517, https://doi.org/10.1103/PhysRevB.104.094517
21. E. Thuneberg, 'Comment on Ballistic SNS sandwich as a Josephson junction', Physical Review B 108, 176501 (2023), doi:10.1103/PhysRevB.108.176501, https://doi.org/10.1103/PhysRevB.108.176501
22. E. B. Sonin, 'Reply to Comment on Ballistic SNS sandwich as a Josephson junction', Physical Review B 108, 176502 (2023), doi:10.1103/PhysRevB.108.176502, https://doi.org/10.1103/PhysRevB.108.176502
23. E. B. Sonin, 'Andreev reflection, Andreev states, and long ballistic SNS junction', Journal of Low Temperature Physics 217, 24 (2024), doi:10.1007/s10909-024-03137-7, https://doi.org/10.1007/s10909-024-03137-7
24. E. B. Sonin, 'Erratum: Ballistic SNS sandwich as a Josephson junction', Physical Review B 110, 139902 (2024), doi:10.1103/PhysRevB.110.139902, https://doi.org/10.1103/PhysRevB.110.139902
25. E. Thuneberg, 'Square-well model for superconducting pair-potential', Journal of Low Temperature Physics 217, 12 (2024), arXiv:2405.07659, doi:10.1007/s10909-024-03169-z, https://doi.org/10.1007/s10909-024-03169-z
26. E. B. Sonin, 'Comment on Square-well model for superconducting pair-potential', arXiv:2410.12580 (2024), https://arxiv.org/abs/2410.12580
27. S. Krekels, V. Levajac, K. Moors, G. Simion, B. Sorée, 'Current conservation in the self-consistent theory of Josephson junctions', Physical Review B 112, 144508 (2025), arXiv:2505.09288, doi:10.1103/1xbv-bvch, https://doi.org/10.1103/1xbv-bvch
28. E. B. Sonin, 'The theory of planar ballistic SNS junctions at T=0', Physical Review B 113, L060502 (2026), arXiv:2503.21583, https://arxiv.org/abs/2503.21583
29. E. Sonin, 'Theory of ballistic SNS junctions revisited', Journal of Low Temperature Physics 222, 29 (2026), arXiv:2511.02967, https://arxiv.org/abs/2511.02967