Kiessling, Altshuler and Yuzbashyan conjectured the thermodynamic transition picture for the isotropic Eliashberg spin-chain functional with vanishing Coulomb pseudopotential: the normal state is the unique minimizer above and at a critical temperature, the superconducting minimizer below it is unique up to gauge, and the transition is continuous. We establish the nonlinear part of this picture. In the variables ρn=sin2θn, the condensation energy restricted to the first quadrant is strictly convex, and folding any admissible chain into that quadrant cannot increase its energy. For every positive, non-increasing interaction with power-law decay, hidden convexity and a compactness estimate give existence and uniqueness of the minimizer up to gauge. The normal state minimizes the energy exactly when its quadratic form is non-negative, including marginal stability. Among admissible chains, a positive solution of the nonlinear Eliashberg equations exists exactly when the normal state is linearly unstable and is then unique. For temperature-dependent families with pointwise continuity and locally uniform decay bounds, the minimizing density is continuous in a weighted ℓ1 norm. Together with uniqueness of the linear-stability temperature, these results prove the thermodynamic conjectures for the γ-model, Einstein phonons and general phonon spectra; the phonon applications use separately established spectral input. Finite-dimensional algebraic and convex-analytic components are machine-checked in Lean. The proofs were obtained independently of concurrent work that reaches overlapping conclusions.
Introduction
Background
Eliashberg theory describes phonon-mediated superconductivity through a gap function Δn and a mass renormalization Zn on the fermionic Matsubara frequencies ωn=πT(2n+1)[1, 2, 3, 4]. Yuzbashyan and Altshuler [5] showed that its free energy is that of a classical Heisenberg chain in which each Matsubara frequency carries a unit spin, and that the Eliashberg equations are the conditions for the spins to be in equilibrium. Kiessling, Altshuler and Yuzbashyan (KAY) took this formulation as the starting point for a mathematical study of the theory with vanishing Coulomb pseudopotential, in three papers on the γ-model [6], on dispersive phonons [7] and on Einstein phonons [8]. In each of them the transition is described by a thermodynamic narrative, which they stated as a conjecture [6 Conj. 1], [7 Conj. 1], [8 Conj. 2]. For the γ-model it reads:
There is a critical temperature Tc(γ)>0 such that for temperatures T≥Tc, the spin chain of the normal state is the unique minimizer of H(S∣N), whereas when T<Tc a spin chain S=N for a superconducting phase minimizes H(S∣N) uniquely up to an irrelevant gauge transformation (fixing of an overall phase). Moreover, the phase transition at Tc from normal to superconductivity is continuous. [6 Conj. 1]
KAY studied the linearization of the theory about the normal state and located the temperature at which the normal state loses linear stability. In the γ-model paper they assumed that minimizers exist, “leaving a rigorous inquiry into the existence and uniqueness of minimizers to a future work” [6 Sect. 2], and in the paper on dispersive phonons they wrote that the normal state minimizes the condensation energy “only for sufficiently large T, as we will prove in a separate paper” [7 Sect. 4]. Kondratev’s proof of KAY’s bound on the stability temperature likewise states that identifying it with the thermodynamic transition is an assumption, not a theorem proved there [9]; the argument is an archived preprint, not a peer-reviewed publication. Elezaby and Abanov derived rigorous upper and lower bounds on the transition temperature of the γ-model from the Hessian of the free energy [10], again an analysis of the linearized problem. Yuzbashyan and Altshuler showed that a minimizer has all its angles in [0,π/2] (as recalled in [7 Sect. 4]) and that there is no continuous family of stationary points connected to the global minimum, and argued from symmetry that the global minimum is unique up to the overall phase [5]. For the γ-model with γ>3, in the limit T→0 of extreme retardation, Yuzbashyan, Kiessling and Altshuler reduced the problem to a universal boundary-value problem for the angle and proved that the global minimum of the free energy is non-degenerate [11]; their proof, by a monotonicity argument for θ′′=xsinθ[11 App. A], concerns that limiting local functional. In the peer-reviewed literature, existence, uniqueness and the relation between the nonlinear and the linear problem for the discrete chain at positive temperature were open; concurrent work is discussed in § 1.3. The γ-model at T=0 is known to have an infinite tower of solutions with n sign changes, of which the sign-preserving one is argued to be the global minimum and the others saddle points [12, 13]; our results concern T>0 and are consistent with this picture.
The corresponding questions for BCS theory have a long history. Existence and uniqueness of the positive solution of the BCS gap equation were proved under various hypotheses [14, 15, 16, 17, 18, 19], Hainzl, Hamza, Seiringer and Solovej proved that the normal state is the global minimizer of the BCS functional if and only if a linear operator has no negative spectrum [20], and Watanabe gave an operator-theoretical proof that the transition in the BCS–Bogoliubov model is of second order [21]. Closest to our method, Hainzl and Seiringer proved that the zero-temperature BCS minimizer is unique, for radial potentials V∈L1(R3)∩L3/2(R3) with V≤0 and V(0)<0, using the strict convexity of t↦1−1−4t in the squared pair amplitude t=∣α∣2; the unique representative is the one whose Fourier transform is non-negative [22 Sect. 5.2]. Our function g(r)=1−1−r in r=sin2θ plays the same role. The step that is new here is an exact regrouping of the two frequency-transfer sums: written in xn, cn, they become a complete square in cn−cm with a non-negative coefficient plus a negative geometric-mean term in the xn, and it is this regrouping, not frequency nonlocality as such, that exhibits the contributions that are convex in ρ. Each sum separately is not convex in the angles. Rigorous results on Eliashberg theory itself also include an obstruction to odd-frequency pairing [23], an archived preprint with the stable identifier 10.48550/arXiv.2207.01825.
Results
Write Θ=(θn)n∈N0 for the vector of spin angles, θn being the angle of the n-th spin from the normal-state direction, so that tanθn=Δn/ωn[5]. Our main observation is that the condensation energy, restricted to angles in [0,π/2] and written in the variables
ρn:=sin2θn=ωn2+Δn2Δn2,
is a strictly convex functional, and that every minimizer lies in this quadrant up to gauge. We prove the following for every pair interaction λ(k) that is positive, non-increasing in the frequency transfer k and bounded by Ck−γ for some C,γ>0 (Definition 1); this covers the γ-model for every γ>0, Einstein phonons, and every phonon spectrum with finite second moment.
Existence and uniqueness (Theorem 6). The condensation energy has a minimizer, and the minimizer is unique up to the gauge Θ↦−Θ.
Linear and nonlinear stability coincide (Theorem 7). The normal state is the minimizer if and only if it is linearly stable, i.e. if and only if KAY’s quadratic form is non-negative. At the stability boundary the normal state is still the unique minimizer.
The positive Eliashberg solution (Theorem 8). Among admissible chains, the nonlinear Eliashberg equations have a solution with all Δn>0 if and only if the normal state is linearly unstable. It is then unique, and it is the minimizer.
Continuity (Theorem 9). The minimizer depends continuously on temperature in a weighted ℓ1 norm, so the order parameter vanishes continuously as the stability boundary is approached.
The critical temperature in KAY’s conjectures is the temperature at which the normal state loses linear stability. For the γ-model the stability set is an interval [Tc,∞) by scaling, and Results 1–4 prove KAY’s Conjecture for the γ-model outright. For Einstein phonons and general spectra, the linear-stability temperature is unique for every coupling by the strict monotonicity in T of the top eigenvalue of KAY’s stability operator. This monotonicity is proved in our companion paper, Sharp bounds on the Eliashberg critical temperature: a comparison principle, convexity in the spectrum, and the Kiessling–Altshuler–Yuzbashyan conjectures (D. Rens, 2026, submitted), and, independently, in the archived preprint of Kondratev [9], an online document with the stable identifier 10.5281/zenodo.22848039 that has not been peer reviewed; KAY proved uniqueness of the crossing above a spectrum-dependent coupling threshold, using monotonicity on a restricted temperature range. With it, Results 1–4 prove KAY’s conjectures for dispersive and Einstein phonons (Corollary 10). The thermodynamic and linear-stability critical temperatures therefore coincide in this model.
Note on concurrent work
The results of this paper were obtained independently; the proofs were complete on 10 November 2026. Since then two works with overlapping results have been made public. Neither has been peer reviewed, and we have not checked their proofs; we describe them as their authors state them.
Siraj released two manuscripts with a Lean 4 formalization on 13 November 2026 [24] (we cite version 1.0.2 of 14 November). At μ∗=0 they prove KAY’s thermodynamic conjectures by the mechanism used here: the condensation energy is written as its quadratic form plus non-negative squares, and in the variables ρn=sin2θn on the first quadrant it becomes strictly convex. They go further in several directions: the global minimizer is unique among three-dimensional spin configurations, without assuming time-reversal symmetry or frequency parity, for positive, summable, non-increasing kernels; the condensation energy is bounded below by −C(Tc−T)2 as T↑Tc; and the minimization is transferred to the Yuzbashyan–Altshuler free energy of the self-energies. On 16 November 2026 Generalised Superintelligence published a white paper [25] stating that, without spin flips, the nonlinear isotropic Eliashberg equations have no nonzero real even-frequency solution at or above the linear-instability temperature, also with a symmetric Coulomb term, and that at μ∗=0 exactly one positive solution exists below it, which minimizes the condensation energy up to sign; it also excludes complex even-frequency solutions wherever the linear stability form is non-negative.
The present paper gives a self-contained proof for the planar chain, in which uniqueness, the coincidence of linear and nonlinear stability and the classification of positive solutions all follow from one rearrangement identity (Lemma 12) and strict convexity (Proposition 15). It applies to every positive, non-increasing interaction with power-law decay, summable or not, and proves continuity of the minimizer in the weighted ℓ1 norm of Theorem 9, which controls every Δn. The variational problem for the condensation energy H(S∣N)=4π2TK(Θ), the free-energy difference between a chain and the normal state, is solved over all admissible real chains, including spin-flipped (cosθn<0) and sign-changing ones. At and above Tc, every non-normal admissible chain has strictly positive condensation energy (Theorem 7(b)). So no first-order transition into a real-gap state can occur, there is no positive-gap stationary point (Theorem 8), and the thermodynamic transition temperature equals the linear-stability temperature. We do not treat non-planar spin configurations; for these see [24].
The idea of the proof
The condensation energy is a sum of a Zeeman-like term ∑n(2n+1)(1−cosθn) and of pair terms λ(∣n−m∣)(1−cos(θn−θm)) and −λ(n+m+1)(1−cos(θn+θm)), the last pair term arising from the spins at negative frequencies. With xn=sinθn and cn=cosθn, the identity 1−cncm=21(xn2+xm2)+21(cn−cm)2 rearranges it exactly (Lemma 12) into
with Λn=∑k=1nλ(k)≥0, Dnm=λ(∣n−m∣)−λ(n+m+1)≥0 and Wnm=[n=m]λ(∣n−m∣)+λ(n+m+1)>0, where [⋅] denotes the indicator of a statement. In this form the gauge reduction is immediate: replacing (xn,cn) by (∣xn∣,∣cn∣) does not increase any term. In the quadrant xn,cn≥0 write xn=ρn and cn=1−ρn. Then 1−1−ρn is strictly convex, xn2=ρn is linear, −ρnρm is convex because the geometric mean is concave, and (1−ρn−1−ρm)2 is convex for the same reason. Strict convexity gives uniqueness. The same identity shows that the functional is its quadratic part plus two non-negative remainders, which gives the stability criterion. Existence follows from a Schur-test estimate showing that the coupling term is a compact perturbation of the Zeeman term. That every positive solution of the Eliashberg equations is the minimizer is the infinite-dimensional version of the statement that a critical point of a convex function is a global minimizer.
The convexity is hidden in two senses: the functional is not convex in the angles, and it is not convex in the gap. With Coulomb repulsion, the coefficient signs used in this convexity proof are no longer automatic; we do not treat the general case μ∗>0 (§ 10).
Verification and organization
The algebraic and convex-analytic core, for every finite truncation of the chain, is formalized in Lean 4 with Mathlib (§ 9). The infinite-dimensional steps (compactness, existence, passage to the limit) are proved on paper. § 9.2 reports selected finite-truncation consistency checks for seven representative interactions, at up to 400 Matsubara frequencies, compared with fixed-point solutions of the Eliashberg equations; they are illustrations and not proofs.
§ 2 sets up the functional. § 3 states the results. § 4–§ 7 contain the proofs, § 9 the verification and § 10 the scope and open problems.
The condensation energy
Interactions
Definition 1. An interaction is a sequence λ=(λ(k))k≥1 of positive numbers that is non-increasing in k and satisfies λ(k)≤Ck−γ for all k≥1, for some constants C>0 and γ>0.
The value λ(0) (the coupling constant for phonons, infinite for the γ-model) never enters below. The three families studied by KAY are interactions:
References
[1]Eliashberg, G.M.: Interactions between electrons and lattice vibrations in a superconductor. Soviet Physics JETP 11(3), 696–702 (1960)
[2]Allen, P.B., Mitrović, B.: Theory of superconducting T_c. In: Ehrenreich, H., Seitz, F., Turnbull, D. (eds.) Solid State Physics vol. 37, pp. 1–92. Academic Press, New York (1983). https://doi.org/10.1016/S0081-1947(08)60665-7
[5]Yuzbashyan, E.A., Altshuler, B.L.: Migdal–Eliashberg theory as a classical spin chain. Physical Review B 106(1), 014512 (2022) https://doi.org/10.1103/PhysRevB.106.014512
[6]Kiessling, M.K.-H., Altshuler, B.L., Yuzbashyan, E.A.: Bounds on T_c in the Eliashberg theory of superconductivity. I: The γ-model. Journal of Statistical Physics 192(5), 69 (2025) https://doi.org/10.1007/s10955-025-03446-5
[7]Kiessling, M.K.-H., Altshuler, B.L., Yuzbashyan, E.A.: Bounds on T_c in the Eliashberg theory of superconductivity. II: Dispersive phonons. Journal of Statistical Physics 192(7), 94 (2025) https://doi.org/10.1007/s10955-025-03468-z
[8]Kiessling, M.K.-H., Altshuler, B.L., Yuzbashyan, E.A.: Bounds on T_c in the Eliashberg theory of superconductivity. III: Einstein phonons. Journal of Statistical Physics 192(7), 93 (2025) https://doi.org/10.1007/s10955-025-03469-y
[9]Kondratev, O.: 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, not peer reviewed. Version 22848039, 19 September 2026 (2026). https://doi.org/10.5281/zenodo.22848039
[10]Elezaby, A., Abanov, A.: Superconductivity near a quantum critical point: Bounds on the transition temperature in the γ-model. Physical Review B 114(2), 024512 (2026) https://doi.org/10.1103/wpn8-8j2t
[11]Yuzbashyan, E.A., Kiessling, M.K.-H., Altshuler, B.L.: Superconductivity near a quantum critical point in the extreme retardation regime. Physical Review B 106(6), 064502 (2022) https://doi.org/10.1103/PhysRevB.106.064502
[12]Abanov, A., Chubukov, A.V.: Interplay between superconductivity and non-Fermi liquid at a quantum critical point in a metal. I. The γ model and its phase diagram at T = 0: The case 0 < γ < 1. Physical Review B 102(2), 024524 (2020) https://doi.org/10.1103/PhysRevB.102.024524
[13]Zhang, S.-S., Wu, Y.-M., Abanov, A., Chubukov, A.V.: Superconductivity out of a non-Fermi liquid. Free energy analysis. Physical Review B 106(14), 144513 (2022) https://doi.org/10.1103/PhysRevB.106.144513
[14]Odeh, F.: An existence theorem for the BCS integral equation. IBM Journal of Research and Development 8(2), 187–188 (1964) https://doi.org/10.1147/rd.82.0187
[15]Billard, P., Fano, G.: An existence proof for the gap equation in the superconductivity theory. Communications in Mathematical Physics 10(4), 274–279 (1968) https://doi.org/10.1007/BF03399502
[17]Yang, Y.: On the Bardeen–Cooper–Schrieffer integral equation in the theory of superconductivity. Letters in Mathematical Physics 22(1), 27–37 (1991) https://doi.org/10.1007/BF00400375
[18]Du, Y.H.: Uniqueness results for the gap equation in the theory of superconductivity. Differential and Integral Equations 8(2), 309–324 (1995) https://doi.org/10.57262/die/1369083471
[19]McLeod, J.B., Yang, Y.: The uniqueness and approximation of a positive solution of the Bardeen–Cooper–Schrieffer gap equation. Journal of Mathematical Physics 41(9), 6007–6025 (2000) https://doi.org/10.1063/1.1286424
[20]Hainzl, C., Hamza, E., Seiringer, R., Solovej, J.P.: The BCS functional for general pair interactions. Communications in Mathematical Physics 281(2), 349–367 (2008) https://doi.org/10.1007/s00220-008-0489-2
[21]Watanabe, S.: An operator-theoretical proof for the second-order phase transition in the BCS–Bogoliubov model of superconductivity. Kyushu Journal of Mathematics 74(1), 177–196 (2020) https://doi.org/10.2206/kyushujm.74.177
[22]Hainzl, C., Seiringer, R.: Spectral properties of the BCS gap equation of superfluidity. In: Mathematical Results in Quantum Mechanics, pp. 117–136. World Scientific, Singapore (2008). https://doi.org/10.1142/9789812832382_0009 . arXiv:0802.0446
[23]Langmann, E., Hainzl, C., Seiringer, R., Balatsky, A.V.: Obstructions to odd-frequency superconductivity in Eliashberg theory. arXiv preprint 2207.01825 (2022). https://doi.org/10.48550/arXiv.2207.01825
[24]Siraj, A.: Sharp T_c bounds and uniqueness of the superconducting state in isotropic Migdal–Eliashberg theory: papers and Lean 4 formalisation (MELean). Zenodo, not peer reviewed. Version 1.0.2, 14 November 2026; version 1.0 released 13 November 2026 (2026). https://doi.org/10.5281/zenodo.24234156
[25]Generalised Superintelligence: A sharp ceiling on T_c in linearised Eliashberg theory. White paper, not peer reviewed. 16 November 2026 (2026). https://www.generalised.si/research/superconducting-tc-ceiling Accessed 2026-11-16
[26]de Moura, L., Ullrich, S.: The Lean 4 theorem prover and programming language. In: Automated Deduction – CADE 28. Lecture Notes in Computer Science, vol. 12699, pp. 625–635. Springer, Cham (2021). https://doi.org/10.1007/978-3-030-79876-5_37
[27]The mathlib Community: The Lean mathematical library. In: Proceedings of the 9th ACM SIGPLAN International Conference on Certified Programs and Proofs (CPP 2020), pp. 367–381. ACM, New York (2020). https://doi.org/10.1145/3372885.3373824
[28]Wada, Y.: Critical field and specific heat of strong coupling superconductors. Physical Review 135(6A), 1481–1497 (1964) https://doi.org/10.1103/PhysRev.135.A1481
[29]Yuzbashyan, E.A., Altshuler, B.L.: Breakdown of the Migdal–Eliashberg theory and a theory of lattice-fermionic superfluidity. Physical Review B 106(5), 054518 (2022) https://doi.org/10.1103/PhysRevB.106.054518