From Bogoliubov-de Gennes to Ginzburg-Landau: Critical Points Near $T_{\rm c}$ in the Non-Magnetic Case

Abstract: We study the relation between the Bogoliubov-de Gennes equation and the Ginzburg-Landau equation for a BCS model without external fields. While previous rigorous derivations of Ginzburg-Landau theory from BCS theory have focused on energies and minimizers, here we consider arbitrary critical points in the relevant energy regime. For temperatures close to the critical temperature, we prove that every sufficiently small solution of the BdG equation admits an asymptotic factorization into a microscopic Cooper-pair profile and a macroscopic order parameter. The latter satisfies the Ginzburg-Landau equation up to an error that vanishes in the scaling limit. Our analysis relies on a Birman-Schwinger reformulation of the BdG equation, a Lyapunov-Schmidt type reduction, and semiclassical estimates at low regularity.

• The paper proves that every sufficiently small Bogoliubov–de Gennes critical point near the critical temperature factorizes into a universal Cooper-pair profile and a macroscopic order parameter satisfying the Ginzburg–Landau equation approximately.
• Using Birman–Schwinger reformulation, Lyapunov–Schmidt reduction, and semiclassical estimates, the authors establish a remainder bound of order h^{7/6}, an H² bound for the order parameter, and an L² Ginzburg–Landau residual of order h^{1/6}.
• The result extends the rigorous BCS-to-Ginzburg–Landau correspondence beyond minimizers in dimensions 1–3 without external fields, while leaving field-dependent critical points and the existence of non-minimal small solutions open.

Context and motivation

The paper by Frank, Hainzl, and Ou Yang addresses the rigorous relationship between the microscopic Bardeen–Cooper–Schrieffer (BCS) theory of superconductivity and its macroscopic Ginzburg–Landau (GL) description. Prior rigorous derivations of GL theory from BCS theory, beginning with Frank–Hainzl–Seiringer–Solovej [FHSS-12] and extended to external magnetic fields in subsequent works, established that minimizers of the BCS functional converge to minimizers of the GL functional in a suitable scaling limit. The present work extends this correspondence to arbitrary critical points: every sufficiently small solution of the Bogoliubov–de Gennes (BdG) equation — not merely an energy minimizer — is shown to factorize asymptotically into a universal microscopic Cooper-pair profile and a macroscopic order parameter satisfying the GL equation up to a vanishing error.

The setting is a translation-invariant BCS model with no external fields, for dimensions $1 \le d \le 3$, with reflection-symmetric interaction $V \in L^1 \cap L^\infty$ satisfying $|x|^2 V \in L^\infty$, chemical potential $\mu \in \mathbb{R}$, and inverse temperature $\beta = \beta_c(1 + D h^2)$ with $D > 0$ fixed and $h \to 0$. The critical temperature $T_c$ is characterized via the linear operator $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$ on $L^2_{\rm symm}(\mathbb{R}^d)$, which is assumed to have a simple zero eigenvalue with eigenfunction $V \in L^1 \cap L^\infty$0 (Assumptions (A1)–(A5)).

Main result

The central theorem states that if $V \in L^1 \cap L^\infty$1 solves the BdG equation

$V \in L^1 \cap L^\infty$2

and satisfies the smallness/regularity condition $V \in L^1 \cap L^\infty$3 — where $V \in L^1 \cap L^\infty$4 is a Sobolev-type norm measuring regularity only in the center-of-mass variable $V \in L^1 \cap L^\infty$5 — then $V \in L^1 \cap L^\infty$6 admits the decomposition

$V \in L^1 \cap L^\infty$7

with a $V \in L^1 \cap L^\infty$8-periodic order parameter $V \in L^1 \cap L^\infty$9 obeying

$|x|^2 V \in L^\infty$0

and solving the GL equation

$|x|^2 V \in L^\infty$1

in the sense that the residual is bounded by $|x|^2 V \in L^\infty$2 in $|x|^2 V \in L^\infty$3. The coefficients $|x|^2 V \in L^\infty$4 (positive definite matrix), $|x|^2 V \in L^\infty$5, and $|x|^2 V \in L^\infty$6 are given by explicit momentum integrals involving $|x|^2 V \in L^\infty$7, the Fourier transform of $|x|^2 V \in L^\infty$8, and the functions $|x|^2 V \in L^\infty$9; they coincide with those obtained in the earlier minimizer-based derivations. Notably, the theorem delivers bounds on $\mu \in \mathbb{R}$0 in $\mu \in \mathbb{R}$1 and an $\mu \in \mathbb{R}$2-control of the GL residual, going beyond the natural energy space ($\mu \in \mathbb{R}$3 / $\mu \in \mathbb{R}$4). A further consequence: since no lower bound on $\mu \in \mathbb{R}$5 is provided, degeneration $\mu \in \mathbb{R}$6 could only occur if $\mu \in \mathbb{R}$7 were an eigenvalue of $\mu \in \mathbb{R}$8 on $\mu \in \mathbb{R}$9, so for all but a discrete set of $\beta = \beta_c(1 + D h^2)$0 the extracted order parameter does not vanish.

Two remarks place the result in context. First, when $\beta = \beta_c(1 + D h^2)$1 is a BCS minimizer, the smallness assumption holds automatically, so the theorem recovers part of [FHSS-12] — though the latter covers external fields and slightly weaker assumptions on $\beta = \beta_c(1 + D h^2)$2. Second, while the absence of external fields makes the BdG equation translation invariant, the authors seek general, non-translation-invariant solutions, so the analysis retains most of the difficulties of the field-coupled case.

Proof strategy

The argument proceeds through several stages:

Birman–Schwinger reformulation. Using Matsubara-frequency representations of $\beta = \beta_c(1 + D h^2)$3, the nonlinear BdG equation is rewritten as

$\beta = \beta_c(1 + D h^2)$4

with a linear operator $\beta = \beta_c(1 + D h^2)$5 and a cubic-plus nonlinear map $\beta = \beta_c(1 + D h^2)$6, both acting on periodic operators. This reformulation, due to prior work of the authors and collaborators, converts unbounded into bounded operator problems; the novelty here is that the final statement is expressed in terms of the original Cooper-pair wave function.

Linear analysis. The operator $\beta = \beta_c(1 + D h^2)$7 has a one-dimensional kernel spanned by $\beta = \beta_c(1 + D h^2)$8 (normalized), with a spectral gap $\beta = \beta_c(1 + D h^2)$9 on the orthogonal complement. A Lyapunov–Schmidt scheme is set up using the projection $D > 0$0, where $D > 0$1 projects onto $D > 0$2 and $D > 0$3 cuts off center-of-mass momenta at $D > 0$4; crucially, the "relevant" subspace has dimension growing like $D > 0$5 as $D > 0$6, finite only because of the cutoff. Invertibility of $D > 0$7 with quantitative bounds, and its persistence for $D > 0$8 via $D > 0$9, are established. The choice $h \to 0$0 balances the resulting error terms.

Nonlinear analysis. Mapping properties of $h \to 0$1 from $h \to 0$2 to $h \to 0$3 ($h \to 0$4) are proved via resolvent estimates and Schatten-class/Hölder inequalities on local trace ideals, together with a decomposition $h \to 0$5 isolating the leading cubic term. These estimates operate at low regularity — smoothness in $h \to 0$6 is not available a priori, which distinguishes this semiclassical analysis from standard settings.

Extraction of the GL equation. Writing $h \to 0$7, the projected equation yields, to leading order,

$h \to 0$8

with errors $h \to 0$9 and $T_c$0 respectively. The kinetic coefficient arises from a Taylor expansion of the symbol difference $T_c$1, and the mass term from the expansion of $T_c$2 under $T_c$3. After rescaling $T_c$4, elliptic bootstrapping of the approximate GL equation — dropping the positive quartic term and absorbing gradient terms whose powers exceed the threshold — upgrades the a priori $T_c$5 bound to $T_c$6 and converts the $T_c$7 control of the residual into $T_c$8 control. Finally, the remainder $T_c$9 is bounded in $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$0 by $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$1, confirming that the factorized term dominates.

Limitations and open questions

The authors identify several restrictions explicitly. The exclusion of external fields is described as their most serious assumption, which they hope to remove in future work; the results of [FHSS-12] for minimizers do cover fields, so extending the critical-point analysis to that setting is the natural next step. The smallness condition $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$2 encodes both amplitude smallness and slow variation on the scale $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$3; whether non-minimal solutions of the BdG equation exist within this regime (or at all) is stated as an open problem, with no known examples cited. The exponent $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$4 on the remainder is acknowledged to be technical and improvable, though any value strictly above $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$5 suffices for the factorization. The theorem also provides no lower bound on $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$6, leaving open the possibility of degenerate sequences at exceptional values of $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$7.

Conclusion

This work establishes, for the first time at full mathematical rigor, that all sufficiently small critical points of the BCS functional near $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$8 — not only minimizers — are approximated by solutions of the Ginzburg–Landau equation, with quantitative rates: an $\mathfrak{h}/\tanh(\mathfrak{h}/2T_c) - V$9 bound and $L^2_{\rm symm}(\mathbb{R}^d)$0 residual control of order $L^2_{\rm symm}(\mathbb{R}^d)$1 for the order parameter, and an $L^2_{\rm symm}(\mathbb{R}^d)$2 remainder in the factorization. The combination of Birman–Schwinger techniques, a Lyapunov–Schmidt reduction over an infinite-dimensionally-growing relevant subspace, and semiclassical estimates at low regularity constitutes the technical core. The result confirms Gor'kov's physical picture at the level of general critical points in the field-free case, and leaves the extension to external fields and the existence of non-minimal BdG solutions as clearly posed open problems.