Open quantitative core of the Cooper–Stölan strongly coupled shell model: exact cycle-weight function f(α) with its analytic continuation f(−1), and the magnitude of the omitted non-BCS graphs

In 1971 Cooper and Stölan proposed a diagrammatic derivation of BCS (Bardeen–Cooper–Schrieffer) superconductivity from a full strongly coupled two-body Hamiltonian: confining fermions to a thin shell about the Fermi surface, replacing the kinetic energy by its shell average, and extracting the 'generalized BCS' (GBCS) graph subset, they showed that a BCS-like ground state and gap emerge with corrections ordered by the small parameter Q/N ~ (δ/k_F)^2. The analysis left its key quantitative links uncontrolled: the weight function f(α) was modeled but not determined (its analytic continuation f(−1), which fixes the ground-state energy, was only bounded), the random-walk phase-space estimate Q^(s) ≈ Q/s^(3/2) was unrefined, the contribution of the omitted non-BCS graphs — including possible gapless collective modes — was not estimated, the shell-averaging of the kinetic energy had no controlled limit, and the numerical size of Q/N for real superconductors was not assessed. No work to date determines f(α) or f(−1), bounds the omitted non-BCS graphs, or validates the Q/N ordering for realistic materials, so the quantitative core of the Cooper–Stölan program remains open.

Background

Bardeen–Cooper–Schrieffer (BCS) theory explains superconductivity from a 'reduced' Hamiltonian in which, out of the full two-body interaction, only scattering of time-reversed pairs — an electron of momentum k and spin up with one of momentum −k and spin down — is retained. Why this drastic reduction captures the physics of superconductors has been a recurring conceptual question since 1957: the retained pair-scattering processes are a small minority of all two-body processes, and no naive counting makes the discarded ones small. Leon Cooper, in earlier work on strongly coupled many-fermion systems (Cooper 1961; Stölan and Cooper 1966), developed a diagrammatic answer to the parallel question of when a subset of the linked-cluster (Goldstone) perturbation series — the expansion of the vacuum-to-vacuum amplitude in powers of the interaction — can legitimately approximate the full series. The key notion is the exclusion-principle-violating (EPV) process: individual diagrams pass through intermediate states that violate the Pauli principle; these unphysical contributions cancel only between different diagrams, so an approximation that keeps some diagrams but not their cancellation partners is internally inconsistent and, at high order, can even violate exact bounds on the amplitude. A consistent graph subset must therefore be 'properly completed', i.e. closed under EPV cancellations, as the expansion of a genuine Hamiltonian would be.

Cooper and Stölan's 1971 paper 'Origin of the Pairing Interaction in the Theory of Superconductivity' applies this machinery to a deliberately simplified model of a superconductor: all single-particle states lie in a thin spherical shell of thickness δ about the Fermi momentum $k_F$ (the Fermi surface is the boundary between occupied and empty states at zero temperature), the kinetic-energy variation across the shell is replaced by its shell average (then shifted to zero), and the two-body interaction is a constant attraction −V between shell states. With N one-spin states in the shell, the interaction Hamiltonian is $H_1=-V\sum c^{\dagger}c^{\dagger}cc$ over all shell momenta. By a phase-space argument — the weight of pair-scattering processes peaks at small pair total momentum q — the authors extract the graph set of a 'generalized BCS' (GBCS) Hamiltonian: the diagrams built from the four BCS-type two-body vertices (particle–particle, hole–hole, and the two pair-conversion vertices) among the sixteen vertex topologies of $H_1$, but, unlike the reduced BCS Hamiltonian, not restricted to exactly opposite momenta. Counting uses a cycle decomposition introduced by Gaudin in nuclear physics: splitting each two-body vertex into creation and annihilation parts decomposes every diagram into directed cycles, and the order-n connected-graph sum organizes as $S_n=N^n\sum_{\alpha}S_n^{(\alpha)}(-1/N)^{\alpha}$, where $S_n^{(\alpha)}$ counts graphs with $n-\alpha$ cycles and $S_n^{(0)}=(n-1)!$ is the ladder count; ladder graphs dominate only up to order $n\lesssim(3N)^{1/2}$, so ladder-only resummations are inconsistent. Phase-space sums over pair momenta with $|q|\le q_0\sim\delta$ are estimated by a random-walk argument, $\sum N_s(q_1\ldots q_s)=NQ^{(s)}$ with $Q^{(s)}\approx Q/s^{3/2}$, where Q is the number of allowed pair momenta; the intrinsic small parameter of the whole scheme is $Q/N\sim(\delta/k_F)^2$. In the GBCS graph sums, the phase-space-reduced weight of an $n$th-order connected graph with $n-\alpha$ cycles takes the form $QN^n(Q/N)^{\alpha}f[g]$ (Eq. 3.25 of the paper), where $[g]$ is a number attached to the graph's reduced polygons — the same number for large classes of graphs, e.g. 1 for the ladder diagrams — and, in the limit $n\to\infty$ at fixed $\alpha$, $f[g]$ becomes a function of $\alpha$ alone: the cycle-weight function $f(\alpha)$. In the resulting '$f(\alpha)$ model' ($f[g]=f(\alpha)$), the total $n$th-order contribution of graphs with $n-\alpha$ cycles is $QN^nS_n^{(\alpha)}(-1)^{\alpha}(Q/N)^{\alpha}f(\alpha)$ (Eqs. 5.1–5.2); $f$ is thereby defined on non-negative integer values of $\alpha$ by these graph sums, normalized by $f(0)=1$ from the ladder graphs, and represented as a Laplace-type integral $f(\alpha)=\int_0^{p_{\max}}g(p)e^{-\alpha p}\,dp$ with $g(p)=0$ for $p>p_{\max}$ (Eqs. 5.5, 5.64), through which its analytic continuation into the left half-plane is taken. Resumming the graph series in this model yields a ground-state energy $E_0=-VN^2[f(-1)+Q/N]$ (Eqs. 5.16–5.17), where $f(-1)$ is the analytic continuation of $f(\alpha)$ to α = −1; a hard single-particle gap $2\Delta=2NV$ identical to strongly coupled BCS; and excited levels grouped in bands of width of order $VQ$ near integer multiples of $2NV$. The authors argue that the GBCS set is the proper EPV-complete extension of the ladder subset, so deviations from BCS phenomenology are of order $Q/N$. For comparison, the pairing Hamiltonian restricted to exact (−k, k) pairs is exactly solvable by Richardson's approach (Richardson 1966), while no exact solution is known for the GBCS Hamiltonian.

The authors themselves delimit what remains unestablished: the shell-averaging of the kinetic energy has no controlled limiting justification; the interaction was taken momentum-independent over the shell (assumption (iii) of the paper's conclusion, which the authors regard as relatively trivial and made primarily for simplification); the magnitude of the omitted non-BCS graphs — and whether they generate gapless collective modes — was not estimated; $f(\alpha)$ was modeled, not determined, with only heuristic bounds on $f(-1)$; the polygon weights rest on the unrefined random-walk estimate; and the numerical smallness of $Q/N$ for physical superconductors was not assessed. The qualitative claim — BCS structure emerges with corrections ordered in $Q/N$ — is therefore not yet a controlled quantitative statement about the stated model.

Problem Statement

For the strongly coupled thin-shell model of Cooper and Stölan — fermions in a shell of thickness δ about the Fermi momentum $k_F$, kinetic energy replaced by its shell average and shifted to zero, constant attraction −V between shell states, $N$ one-spin states in the shell, $Q$ allowed pair momenta with $|q|\le q_0\sim\delta$, small parameter $Q/N\sim(\delta/k_F)^2$ — establish quantitatively the claim that the GBCS graph subset extracted from the full shell Hamiltonian $H_1$ approximates the full problem with corrections ordered in $Q/N$. Three components jointly serve this single objective:

(1) Determine exactly, or with controlled error, the cycle-weight function $f(\alpha)$ governing the GBCS graph sums — including a refinement (or error bound) of the random-walk phase-space estimate $Q^{(s)}\approx Q/s^{3/2}$ for cycles with $s$ vertices — together with its analytic continuation to α = −1, so that the ground-state energy $E_0=-VN^2[f(-1)+Q/N]$ is fixed beyond the existing heuristic bounds of the form $1+\frac{Q}{4N\ln 2}\ln\frac{\pi N}{Q}\lesssim f(-1)\lesssim 2+\frac{Q}{4N\ln 2}\ln\frac{\pi N}{Q}$.

(2) Provide a quantitative estimate of the summed contribution of the omitted non-BCS graphs — those generated by the two-body vertex topologies discarded when passing from $H_1$ to the GBCS Hamiltonian — to the ground-state energy and to the low-energy spectrum; in particular, determine whether these graphs generate gapless collective modes within or below the gap $2\Delta=2NV$, or only renormalize the order-$Q/N$ corrections.

(3) Justify the replacement of the kinetic-energy operator by its shell average with a controlled error statement in a limit such as $\delta/k_F\to0$ (in which the strong interaction competes with the kinetic-energy variation), and assess the numerical magnitude of $Q/N\sim(\delta/k_F)^2$ for realistic superconductors, thereby validating — or refuting — the perturbative ordering on which the whole GBCS reduction rests.

References

1. L. N. Cooper and B. Stölan, 'Origin of the Pairing Interaction in the Theory of Superconductivity', Physical Review B 4, 863–893 (1971). DOI: 10.1103/PhysRevB.4.863, https://doi.org/10.1103/PhysRevB.4.863 (source paper)
2. B. Stölan and L. N. Cooper, 'Cycle analysis of diagrams generated by Hamiltonians with the BCS topology', Il Nuovo Cimento B 10, 291–324 (1972). DOI: 10.1007/BF02911427, https://doi.org/10.1007/BF02911427 (direct authors' follow-up)
3. L. N. Cooper, 'Theory of Strongly Coupled Many-Fermion Systems. I. Convergence of Linked-Cluster Expansions', Physical Review 122, 1021–1028 (1961). DOI: 10.1103/PhysRev.122.1021, https://doi.org/10.1103/PhysRev.122.1021 (predecessor)
4. B. Stölan and L. N. Cooper, 'Approximation Methods in Many-Body Perturbation Theory', Physical Review 143, 209–214 (1966). DOI: 10.1103/PhysRev.143.209, https://doi.org/10.1103/PhysRev.143.209 (predecessor)
5. M. Gaudin, 'Sur le développement de la grande fonction de partition pour des systèmes de particules identiques', Nuclear Physics 20, 513–532 (1960). DOI: 10.1016/0029-5582(60)90193-0, https://doi.org/10.1016/0029-5582(60)90193-0 (cycle-decomposition method used by the source paper)
6. R. W. Richardson, 'Numerical Study of the 8-32-Particle Eigenstates of the Pairing Hamiltonian', Physical Review 141, 949–956 (1966). DOI: 10.1103/PhysRev.141.949, https://doi.org/10.1103/PhysRev.141.949 (evidence that the (−k,k)-restricted pairing Hamiltonian is exactly solvable)
7. I. I. Fal'ko and V. L. Fal'ko, 'Density of states of superconductors with overlapping bands', Solid State Communications 10, 409–411 (1972). DOI: 10.1016/0038-1098(72)90908-8, https://doi.org/10.1016/0038-1098(72)90908-8
8. I. I. Fal'ko and V. L. Fal'ko, 'Energy spectrum of transition metals in the superconducting state', Soviet Physics Journal 16, 754–757 (1973). DOI: 10.1007/BF00895682, https://doi.org/10.1007/BF00895682
9. J. Bostock, 'A Summary of the Course', in 'The Science and Technology of Superconductivity', pp. 735–753 (1973). DOI: 10.1007/978-1-4615-8978-5_14, https://doi.org/10.1007/978-1-4615-8978-5_14
10. S. Ø. Aks, J. Sienicki, and B. B. Varga, 'Quasisecular Renormalization of the φ⁴ Model of Quantum Field Theory', Physical Review D 6, 520–528 (1972). DOI: 10.1103/PhysRevD.6.520, https://doi.org/10.1103/PhysRevD.6.520
11. W. W. Schmidt, R. Block, and L. Jansen, 'Superexchange and superconductivity: A possible correlation', Physical Review B 26, 3656–3672 (1982). DOI: 10.1103/PhysRevB.26.3656, https://doi.org/10.1103/PhysRevB.26.3656
12. R. Block and L. Jansen, 'Theoretical analysis of equilibrium geometries and barriers of rotation in molecules H2X2, with X=O, S, Se, and Te', The Journal of Chemical Physics 82, 3322–3328 (1985). DOI: 10.1063/1.448231, https://doi.org/10.1063/1.448231
13. R. Shankar, 'Renormalization-group approach to interacting fermions', Reviews of Modern Physics 66, 129–192 (1994). DOI: 10.1103/RevModPhys.66.129, https://doi.org/10.1103/RevModPhys.66.129 (adjacent modern program)
14. R. Frank, C. Hainzl, R. Seiringer, and J. Solovej, 'Microscopic derivation of Ginzburg-Landau theory', Journal of the American Mathematical Society 25, 667–713 (2012). DOI: 10.1090/s0894-0347-2012-00735-8, https://doi.org/10.1090/s0894-0347-2012-00735-8 (adjacent modern program)