Method obstructions to linear-in-frequency pairing ceilings in electron-phonon models: integrated cumulants, positive determinants, and the vanishing effective attraction
This preprint asks whether the pair-ordering temperature of a Holstein or bond-SSH electron-phonon model can be bounded at a temperature scale proportional to the phonon frequency, uniformly in that frequency and at fixed hopping and coupling. That linear step remains open, and the note does not close it. It records three exact obstructions that identify which proof methods cannot reach it, together with the regimes in which a frequency-uniform ceiling does survive. First, for the Holstein model with arbitrary real static site fields, the exact first coefficient of the local dynamic cumulant, computed with the dynamic Gaussian already integrated out, is at least U Omega2/(1200 tau3) on an open three-site path for 0 < Omega <= tau <= t/4, and at least U Omega2/(2400 tau3) on every square torus of side at least four with a finite static barrier. At tau = C Omega this diverges as Omega goes to 0, so no expansion whose uniform local activity must dominate it can stay small. Second, a strictly positive single-impurity determinant does not bound the impurity propagator: on a two-slice history the fermion determinant exceeds one, all monodromy eigenvalues are positive, and the conditional Green function grows as B2/(2v2); an interval-certified fixture at B = 6, v = 1/2 gives det(1+U) > 15 and |G(1,0)00| > 52.89. Third, for the Hubbard-Holstein-SSH model at every effective on-site interaction Ueff, the high-temperature regime closes with identical constants: for beta|t| + beta Gb <= 2-10 the pair correlator is at most (341/103)(32/341)d, uniformly in Ueff, phonon frequencies, chemical potential and density, so kBTpair <= 1024(|t|+Gb), while three further proof routes remain obstructed at Ueff = 0. Combining this uniform ceiling with strong-coupling bounds gives explicit minimum ceilings for Ueff < 0 at all densities and for Ueff > 0 at half filling, whereas exact dimer calculations show that neither the pair susceptibility nor the pair correlator |<Deltax†Deltay>| is monotone non-increasing in repulsive U0 >= 0. Version 3 adds the annealed static field. On square tori of side at least four, with 0 < kappa <= U/4, 0 <= t <= kappa/64 and sqrt(252) Omega <= tau <= U/64, averaging the first dynamic cumulant over the exact interacting static-field marginal gives E[ax(W)] < (62/675 U/kappa + 12) Omega2/tau2, which at kappa = U/4 is 8348/675 Omega2/tau2 < (25/2) Omega2/tau2; along a ray tau = C Omega with C >= sqrt(252) it is below 8348/(675 C2), with limsup at most 248/(675 C2) at kappa = U/4. Annealing therefore removes the tau-3 resonance divergence of the fixed-field witness at the level of the first cumulant (PROVED, constants CERTIFIED). What blocks the route is everything above that level: the two-site dimer's direct reflection kernel has a negative 2 x 2 minor, and the four-site zero-flux ring's single-spin and two-spin particle-hole kernels have negative principal minors while the dimer particle-hole control of the same size is positive (PROVED computer-assisted, CERTIFIED by finite minors). The gapped corridor itself is arithmetic and is PROVED: 227t2/U <= 219t and min(210t, 227t2/U) <= 210t. The separating-contour and corridor syntheses are labelled HEURISTIC and carry no theorem status. Version 3 supersedes version 2 (10.5281/zenodo.24409383) and leaves every statement of version 2 intact. It adds the annealed static-field section, records its two syntheses as HEURISTIC, lists the annealed route among the open problems and the explicit non-claims, and corrects the page range of one cited article. Every statement concerns model quantities of the precisely specified Hamiltonians; nothing here is a bound on a measured transition temperature, and no materials claim is made. The statements are labelled PROVED (the analytic argument in the source record's proof notes), CERTIFIED (exact-rational or outward-rounded interval arithmetic, replayed in the supplement together with a separately written independent implementation and deliberately corrupted runs that must be rejected), NUMERICAL (finite-matrix and floating-point sanity checks that are not part of any proof) or HEURISTIC (a recorded synthesis that carries no theorem status). The replays do not formalize the analytic arguments. AI agents generated the proofs, code and drafts under D. Rens' direction, and separate AI agents performed the independent reviews. This preprint has not been peer reviewed.
Authors
- Davis Rens
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-21
- DOI
- https://doi.org/10.5281/zenodo.24408870
- Citations
- 4
- Primary Topic
- Physics of Superconductivity and Magnetism
- Type
- preprint