In linearised Migdal–Eliashberg theory without Coulomb repulsion, the instability temperature obeys for every electron-phonon spectrum, with coupling , mean square phonon frequency and , and no spectrum attains it. Allen and Dynes 6, Leavens 9 and Moussa and Cohen 7 stated the bound or its asymptote, and Kiessling, Altshuler and Yuzbashyan (KAY) conjectured it sharp 3. Kondratev posted the first complete proof we know of on 19 September 2026 1. Beyond 1 we prove a continuity theorem, the exclusion of complex gaps, and sharp versions for several bands, anisotropy and, in the symmetric convention at stated parameters, a static Coulomb term; much of it overlaps Siraj's postings of 13 and 14 November 2026 2. None of it says that room-temperature superconductivity is impossible.
The question
Eliashberg theory describes phonon-mediated superconductivity. Does the temperature at which the normal state becomes unstable to pairing have a ceiling in terms of , where is the electron-phonon spectral function and .
In units the model is isotropic, with constant density of states, infinite bandwidth, particle-hole symmetry and the kernel between Matsubara frequencies steps apart, without vertex corrections. The normal state is stable at if the quadratic form of the linearised gap equation is non-negative, and is the supremum of the temperatures where it is not 3, 4. A Coulomb term of dimensionless strength 5 acts on Matsubara indices, on both indices of the pairing kernel (symmetric convention) or on the summed index only (standard convention).
Leavens 9, as reported in 2, and Moussa and Cohen 7 stated the bound. KAY 3 proved a bound about 2.034 times larger at large coupling, conjectured the sharp bound, and left the nonlinear transition as a conjecture.
What we found
Here unless stated.
Theorem 1 (Sharp bound). Let be nonzero with . Every is stable, so , where and is the top eigenvalue of the stability operator of the kernel . The unstable set is exactly , and increases with . The constant is sharp: Einstein spectra with at fixed approach it from below.
Theorem 2 (Einstein ceiling). For one Einstein mode of frequency and any , exists and is unique; increases strictly and increases to . For every spectrum with coupling and , , with equality only for one Einstein mode. For spectra supported in a fixed interval , , under constraints with continuous , one of them fixing the coupling , where , is attained by a spectrum of at most atoms.
Moussa and Cohen 7 state as known from 6; Kondratev 1, Section 12, proves it.
| 1 | ||
| 2 | ||
| 4 | ||
| 10 | ||
| 100 | ||
| 1000 |
Theorem 3 (Continuous transition). In the nonlinear equations for gap and mass renormalisation with the positive root (no spin flips), take a real gap even in Matsubara frequency. No nonzero solution exists at ; the same holds with a symmetric-convention Coulomb term or any real symmetric pairing-channel matrix, then being the instability temperature of the form that includes it. For exactly one solution has all gap values positive; it is the global minimiser of the condensation energy up to sign and vanishes as , so the transition is continuous in this sector.
Theorem 4 (No complex gap). Let solve the equations at without spin flips, complex and even in frequency, with arbitrary phases between frequencies and bands and a Coulomb term as in Theorem 3. With , for the linear stability form at , Coulomb term included. So no such solution exists wherever , in particular at and above the instability temperature of that form. An imaginary part odd in frequency is stable at every temperature.
Theorems 3 and 4 need non-increasing and at most , , true for every spectrum of finite , and for several bands the coupling of Theorem 7. A certified unphysical signed kernel breaks them.
Theorem 5 (Coulomb pseudopotential). In the symmetric convention, for the eight pairs below, over all is the soft-phonon limit, that of single Einstein modes with at fixed ; it is enclosed in the symmetric column, no spectrum attains it, and every spectrum is stable above it. The same holds for every at and at , and, at and , for the absolute cutoffs with the constants of .
| Symmetric convention | Standard convention, truncated to 200 sites | ||
|---|---|---|---|
| 0.10 | 5 | ||
| 0.10 | 10 | ||
| 0.10 | 20 | ||
| 0.10 | 40 | ||
| 0.13 | 5 | ||
| 0.13 | 10 | ||
| 0.13 | 20 | ||
| 0.13 | 40 |
Proposition 6 (Standard convention, truncated). Truncate the standard-convention problem to Matsubara sites, keeping the principal block of the stability matrix with the normal self-energy summed over all frequencies. In 58 cases, namely with at , the eight pairs of the table at , and at , every spectrum is stable above the soft-phonon threshold, and is bracketed to a relative width of about .
Theorem 7 (Several bands). Take bands of weights and pair masses with , finite for and an arbitrary spectral shape for each pair. Let be the largest eigenvalue of . Every is stable, so , sharp when all row sums are equal.
Theorem 8 (Exact multiband supremum). Let over nonempty sets of bands. At fixed and , over all spectral shapes, with , so with . Also . For two bands with pure interband coupling and , .
Theorem 9 (Anisotropy). Let be a Fermi surface with its density of states, atoms allowed, and positive spectral measures (detailed balance). Let and the static coupling have bounded row integrals, and be the norm of on . Then , strictly for finitely many patches. At fixed , lies between and , , and can exceed the ceiling of the Fermi-surface average by any factor.
Theorem 10 (Screening). Let a dimensionless Coulomb kernel act in the gap equation, with window matrix on the Coulomb indices. If it is positive semidefinite at every temperature above the phonon-only , it cannot raise ; this covers the symmetric term and positive-definite kernels. For several bands or anisotropy the same holds for a static positive semidefinite matrix, with Theorems 7 and 9. For an indefinite static pairing-channel matrix on the bands or patches, if , with and the smallest number with for every in ; for bands of weights summing to one it is the larger of and minus the lowest eigenvalue of .
Weaker conditions fail (certified examples). The non-negative non-increasing kernel for in a window of 20, far above physical strengths, destabilises the soft phonon kernel , , at , where phonons alone are stable. An idealised two-band model, with equal weights, soft Einstein intraband spectra, no interband phonon coupling and a sign-changing gap under interband repulsion of strength 0.05 to 0.2, gives in five certificates with to that to times .
Why it is true
The form is in , with and lag forms . Each is non-negative on non-negative non-increasing sequences, and the critical mode of a non-increasing kernel is such a sequence. So the top eigenvalue increases strictly with , and every , the kernel: Theorem 1. Jensen's inequality gives , and convexity of the stable set the atoms.
For Theorem 3, turns the equations into critical points of a spin-chain energy , with the exact identity and an explicit sum of non-negative terms, so a nonzero solution has . Existence comes from minimising , uniqueness from comparing positive solutions, Theorem 4 from the identity for and . Theorem 5 reduces to Einstein atoms and covers by boxes; Theorems 7 and 8 split the form into pair forms, one certified by a Z-matrix supersolution; Theorem 9 compresses the Fermi surface to finite partitions.
How it was checked
Certificates are checked by short standalone programs in interval or exact arithmetic. A second party re-derived each result and recomputed it with separately written code that imported none of ours. It reproduced the exact identities, re-certified key lemmas, and found no violation in adversarial searches. The estimates behind Theorem 3 were re-derived, not formalised, and the checker for Theorem 4 was run, not audited. A separately designed certifier re-certified all 58 cases of Proposition 6, but for Theorem 5 only eight brackets and two of 151 sweep intervals were re-certified. For Theorem 8 the separate code certifies and , our checker .
The separately written code's own interval enclosure of reached only , so the 13 digits of and the enclosures of rest on one method, implemented twice. Theorems 1 to 4 and the first statement of Theorem 10 are proofs, the Einstein table is a certified computation, and Theorem 5, Proposition 6 and Theorems 7 to 9 are computer-assisted proofs. Theorems 5 and 7 to 10 take Theorem 1 and the enclosure of as input. The proofs have not been peer reviewed.
What it does not show
- Scope. The spin-flip solutions of Yuzbashyan and Altshuler 11 and a first-order transition into a state outside Theorems 3 and 4 are not excluded, apart from the global minimiser at treated in 2.
- Regime. The ceiling is approached only by soft phonons at very large , where Migdal–Eliashberg theory is not expected to hold 17. For published data, see the companion note.
- Coulomb term. Proposition 6 is for truncations: its constants fall by between 200 and 400 sites at , and a code that sums the self-energy over the retained frequencies only is a different matrix, with soft threshold of 0.9810024 against 0.9810778 at . At the truncated soft limit stops being the supremum near . At , , outside the physical range, a hard phonon atom stabilises a certified unstable temperature: the form is not monotone. No multiband bound with is proved.
- Hypotheses. Three bands with repulsive interband Coulomb coupling have a certified time-reversal-breaking solution of a truncated functional, as in BCS models 8, not excluded by Theorem 4 where . Theorem 8's pure-interband extreme is not tied to physical matrix elements.
- Dynamic screening. In the gap equation alone, the kernel of bare strength 1 with and a sharp symmetric cutoff at gives a soft-phonon certified in , where the sharp window admits one more Matsubara frequency. A smooth 1 to 10 per cent edge gives 1.80 to 1.71, and the kernel also in the mass renormalisation (both measured).
Earlier and parallel work
Kondratev 1 proves the bound in the scalar linearised model at by comparing kernels coefficientwise, as we do. It is unrefereed, not formalised, and says that AI systems produced it. We found 1 after our proof was written.
Siraj 2 posts two papers and a Lean 4 formalisation, under review, AI-generated by its own record and Lean-checked only in the clauses so tagged. We read the two main papers and the Liouville statements of Supplement A, and checked no proof. It overlaps Theorems 1 to 5, 7 and 9 and the static part of Theorem 10, and we claim no priority over 2 there. We add a tighter enclosure of , the Einstein table, Theorem 8, the norm for anisotropic kernels, constants enclosed to with the supremum shown equal to the soft limit, Proposition 6, complex gaps with Coulomb terms and several bands, and the counterexamples of Theorem 10.
| Source | Year | What it states |
|---|---|---|
| Allen and Dynes 6 | 1975 | The strong-coupling asymptote , as reported in 4; KAY note that it is presumably an upper bound at every coupling 3. |
| Leavens 9 | 1977 | According to 2, , which is the bound of Theorem 1 in non-strict form to the three digits printed (), by a functional-derivative argument. We read only the abstract, which claims rigorous upper bounds. |
| Moussa and Cohen 7 | 2006 | , citing 6 and 9; the multiband version with the largest eigenvalue of , without derivation; and that an Einstein spectrum maximises at fixed and , as already known from 6, without proof. They say the upper bound corresponds to of an Einstein model; by Theorem 1 it is approached there and never attained. |
| Hainzl and others 13 | 2008 | The BCS analogue of the nonexistence in Theorem 3, without frequency-dependent mass renormalisation. |
| Yuzbashyan and Altshuler 11 | 2022 | The spin-chain form of the equations, its spin-flip solutions, and the argument that the free-energy minimiser has a non-negative gap. |
| Langmann and others 12 | 2022 | No pure odd-frequency gap in one band. |
| Pimenov and Chubukov 14 | 2022 | The zero-temperature quantum phase transition under repulsive dynamical interaction; sign-changing gaps under dynamically screened repulsion were known before, see its introduction. |
| Semenok and others 15 | 2024 | as an upper bound, without proof. |
| KAY 3, 4, 10 | 2025 | The sharp bound as Conjecture 2 of 3, and as Conjecture 1 of 3 the thermodynamic picture that Theorem 3 proves for real gaps without spin flips; Jensen's inequality for one lag; part I 10 defers the nonlinear equations. |
| Sadovskii 16 | 2025 | A review of the strong-coupling limit with coefficient 0.182. |
| Kondratev 1 | 2026 | The unique onset at every coupling, the bound and its optimality (Theorem 1); his Propositions 1 and 2 and Section 10 imply the strict monotonicity of in the spectrum and of , with its limit; the proof of , with equality only for one Einstein mode, by Jensen's inequality (Section 12, not claimed as new there); the constant to six digits; the nonlinear transition left open. |
| Siraj 2 | 2026 | The bound for every in both conventions with a Lean-checked ; Einstein maximality with certified ceilings at and at on ; constants for within 0.051 per cent above the soft limit for the symmetric cutoff at and , and for other the constant of the largest of these values not above (proof sketch); the spectral-radius multiband bound, Lean-checked for a common spectrum in the even sector and sketched otherwise; an anisotropic bound; at global uniqueness without time-reversal symmetry and Liouville-type exclusion of other no-flip and of half-plane complex solutions. |
| Das Sarma and others 18 | 2026 | Doubts about Eliashberg results with a screened Coulomb coupling in simple metals. |
Apart from these overlaps we found no earlier statement of Theorems 5, 8 and 10, of Proposition 6, of Theorem 4 with a Coulomb term or several bands, or of a bound with a frequency-dependent Coulomb kernel in the sources searched (arXiv, Zenodo, Crossref and web search, 5 to 15 November 2026; GitHub through web search only). This is absence of evidence, not a priority claim.
References
- 1.O. Kondratev, "A proof of Conjecture 2 of Kiessling, Altshuler and Yuzbashyan: the sharp spectral-moment bound on the linear-stability onset temperature in Eliashberg theory", Zenodo preprint, v3 (2026). doi.org/10.5281/zenodo.22848039
- 2.A. Siraj, "Sharp T_c bounds and uniqueness of the superconducting state in isotropic Migdal-Eliashberg theory: papers and Lean 4 formalisation (MELean)", Zenodo, v1.0.2 (2026). doi.org/10.5281/zenodo.24234156
- 3.M. K.-H. Kiessling, B. L. Altshuler and E. A. Yuzbashyan, "Bounds on T_c in the Eliashberg theory of superconductivity. II: Dispersive phonons", J. Stat. Phys. 192, 94 (2025). arxiv.org/abs/2409.00532
- 4.M. K.-H. Kiessling, B. L. Altshuler and E. A. Yuzbashyan, "Bounds on T_c in the Eliashberg theory of superconductivity. III: Einstein phonons", J. Stat. Phys. 192, 93 (2025). arxiv.org/abs/2409.02121
- 5.P. Morel and P. W. Anderson, "Calculation of the superconducting state parameters with retarded electron-phonon interaction", Phys. Rev. 125, 1263 (1962). doi.org/10.1103/PhysRev.125.1263
- 6.P. B. Allen and R. C. Dynes, "Transition temperature of strong-coupled superconductors reanalyzed", Phys. Rev. B 12, 905 (1975). doi.org/10.1103/PhysRevB.12.905
- 7.J. E. Moussa and M. L. Cohen, "Two bounds on the maximum phonon-mediated superconducting transition temperature", Phys. Rev. B 74, 094520 (2006). arxiv.org/abs/cond-mat/0607832
- 8.V. Stanev and Z. Tešanović, "Three-band superconductivity and the order parameter that breaks time-reversal symmetry", Phys. Rev. B 81, 134522 (2010). arxiv.org/abs/0912.5214
- 9.C. R. Leavens, "Important parameters in strong-coupling superconductivity", J. Phys. F 7, 1911 (1977). doi.org/10.1088/0305-4608/7/9/034
- 10.M. K.-H. Kiessling, B. L. Altshuler and E. A. Yuzbashyan, "Bounds on T_c in the Eliashberg theory of superconductivity. I: The gamma-model", J. Stat. Phys. 192, 69 (2025). arxiv.org/abs/2409.00533
- 11.E. A. Yuzbashyan and B. L. Altshuler, "Migdal-Eliashberg theory as a classical spin chain", Phys. Rev. B 106, 014512 (2022). arxiv.org/abs/2205.06442
- 12.E. Langmann, C. Hainzl, R. Seiringer and A. V. Balatsky, "Obstructions to odd-frequency superconductivity in Eliashberg theory", arXiv:2207.01825 (2022). arxiv.org/abs/2207.01825
- 13.C. Hainzl, E. Hamza, R. Seiringer and J. P. Solovej, "The BCS functional for general pair interactions", Commun. Math. Phys. 281, 349 (2008). doi.org/10.1007/s00220-008-0489-2
- 14.D. Pimenov and A. V. Chubukov, "Quantum phase transition in a clean superconductor with repulsive dynamical interaction", npj Quantum Mater. 7, 45 (2022). arxiv.org/abs/2112.06273
- 15.D. V. Semenok, B. L. Altshuler and E. A. Yuzbashyan, "Fundamental limits on the electron-phonon coupling and superconducting T_c", arXiv:2407.12922 (2024). arxiv.org/abs/2407.12922
- 16.M. V. Sadovskii, "Upper limit for superconducting transition temperature in electron-phonon superconductors: very strong coupling", J. Supercond. Nov. Magn. 38, 234 (2025). arxiv.org/abs/2506.19326
- 17.A. Chubukov, I. Esterlis, A. Abanov and N. Prokof'ev, "Breakdown of the Migdal-Eliashberg theory for electron-phonon systems. Role of polarons/bi-polarons", Phys. Rev. B 114, 185127 (2026). arxiv.org/abs/2604.14294
- 18.S. Das Sarma, J. D. Sau, Y.-T. Tu and S. Wang, "Conventional and practical metallic superconductivity arising from repulsive Coulomb coupling", Phys. Rev. B 114, 055202 (2026). arxiv.org/abs/2511.00625