Skip to main content

Part 0· SCC Hero · T11

T11 — Sharp-Interface Γ-Convergence

Hero · Phase + Stability group · Cat A. Source: C-0011 / P-0011. Verification: E-0014. Canonical version: CV-1.0. Full proof: Canonical Spec — Part 5 · §13.

Statement

Let ε=α/β\varepsilon = \alpha/\beta be the smoothness-to-double-well ratio. As ε0\varepsilon \to 0, the rescaled boundary-morphology energy

Ebdε(u)=εx,yNt(x,y)(u(x)u(y))2+1εxW(u(x))\mathcal{E}^\varepsilon_{\mathrm{bd}}(u) = \varepsilon \sum_{x,y} \mathbf{N}_t(x,y)(u(x) - u(y))^2 + \frac{1}{\varepsilon} \sum_x W(u(x))

Γ-converges to a perimeter functional

F0(χE)=cWPerG(E)\mathcal{F}_0(\chi_E) = c_W \cdot \mathrm{Per}_G(E)

defined on characteristic functions χE\chi_E of subsets EXE \subseteq X, where cW=012W(s)dsc_W = \int_0^1 \sqrt{2 W(s)}\,ds is the surface-tension constant and PerG(E)\mathrm{Per}_G(E) is the graph perimeter (number of edges crossing E\partial E).

Minimizers of Ebdε\mathcal{E}^\varepsilon_{\mathrm{bd}} converge (up to subsequence) to characteristic functions of sets minimizing the graph perimeter subject to the volume constraint E=m|E| = m. Self-referential correction terms (closure, separation) modify the effective surface tension at higher order in ε\varepsilon.

As ε = α/β → 0, soft cohesion fields u^ε Γ-converge to characteristic functions χ_E of minimal-perimeter sets. The soft-to-crisp bridge — object-like crisp entities are recovered as the asymptotic limit of the soft framework.

Proof idea

Standard Modica-Mortola structure. The Allen-Cahn energy on a graph has the form ε(Dirichlet)+ε1(double-well)\varepsilon \cdot (\text{Dirichlet}) + \varepsilon^{-1} \cdot (\text{double-well}). As ε0\varepsilon \to 0, configurations are forced to either side of the double-well wells (i.e., u0u \approx 0 or u1u \approx 1 everywhere except on a thin transition layer); the energy concentrates on the transition layer with surface-tension density cWc_W.

Lim sup (recovery sequence). For any EXE \subseteq X, construct uεu^\varepsilon that approximates χE\chi_E with smooth transition of width O(ε)O(\varepsilon) across E\partial E; verify Ebdε(uε)cWPerG(E)\mathcal{E}^\varepsilon_{\mathrm{bd}}(u^\varepsilon) \to c_W \cdot \mathrm{Per}_G(E).

Lim inf (compactness + lower bound). Sequences with bounded energy are pre-compact in L1L^1 via the standard equicoercivity argument; any limit is a characteristic function (the double-well term forces this); the lim inf inequality follows from the slicing lemma.

Self-referential corrections. The closure (Ecl\mathcal{E}_{\mathrm{cl}}) and separation (Esep\mathcal{E}_{\mathrm{sep}}) terms are perturbations of the Allen-Cahn substrate. Their effect on Γ-convergence is via modified surface tension cWcW+O(λcl+λsep)c_W \to c_W + O(\lambda_{\mathrm{cl}} + \lambda_{\mathrm{sep}}) — handled by perturbation analysis. \square

Why this is a hero

T11 is the soft-to-crisp bridge. SCC commits to the primacy of the soft cohesion field u:X[0,1]u : X \to [0,1], but it must explain how object-like crisp entities are recovered. T11 provides the mathematical mechanism: in the sharp-interface limit, soft minimizers automatically approach characteristic functions of crisp sets minimizing perimeter.

This connects SCC to:

  • Geometric measure theory — perimeter minimization is the classical problem; SCC inherits its tools (slicing, BV functions, isoperimetric inequalities).
  • Allen-Cahn / Cahn-Hilliard theory — the variational substrate is shared, but SCC adds self-referential corrections that modify the effective surface tension.
  • Crisp recovery protocol (open problem) — the asymptotic limit is rigorous, but the threshold rule for finite ε\varepsilon remains under-specified. T11 establishes that some crisp recovery is mathematically inevitable; the protocol for any specific ε\varepsilon is downstream work.

The substantive content of T11 for SCC specifically is that the four-term energy is consistent with a sharp-interface limit. The deliberate non-Allen-Cahn structure of SCC (closure self-reference, distinction asymmetry) does not destroy the limit — it perturbs the surface tension. This is non-trivial: a more aggressive non-locality could have broken Γ-convergence entirely.

Logical dependencies

  • Builds on: T1 (minimizer existence), T8-Core (non-trivial minimizers exist for β/α\beta/\alpha above threshold), discrete isoperimetric structure of the graph.
  • Builds into: Deep Core Dominance 2b (uses isoperimetric inequality directly), T-Merge (b) (perimeter ordering on connected graphs is the limit of Γ-convergence), formation birth analysis at βcrit\beta_{\mathrm{crit}}.

See also