Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section4.Provider.Schauder.CubeSchauderFreezing

CubeSchauderFreezing #

Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.

Cube Schauder: the freezing step (harmonic approximation with divergence forcing) #

If u solves -Δ u = ∇·G on a window U and V ⊆ U is an interior sub-domain, then for any constant vector c the same u solves -Δ u = ∇·(G - c) on V, because a constant field is weakly divergence free. Solving the zero-trace problem -Δ w = ∇·(G - c) on V and setting h := u - w produces a weakly harmonic comparison function on V together with the sharp energy bound

  ∫_V |∇w|² ≤ ∫_V |G - c|² .

Taking c = G(x₀) for a point x₀ of the sub-cube and using the C^{0,1/2} bound on G gives the freezing gain: the harmonic approximation error is controlled by the oscillation of the forcing over the sub-cube, hence by (side)^{1/2} · [G]_{C^{0,1/2}} rather than by ‖G‖.

This is Armstrong--Kuusi ellipticregularity.tex, the display e.harmapprox.Schauder (~2660) specialized to a = I_d (where the ‖a - I_d‖_{L^∞} leg vanishes identically), in the development's own carriers.

Main results #

References #

1. The identity background #

The identity coefficient field drops out of the flux pairing.

-Δ u = ∇·g in the development's carrier, unfolded at the identity background.

2. A constant field is weakly divergence free #

A constant field pairs to zero against zero-trace gradients. This is the weak statement ∇ · c = 0 for a constant vector c, and it is the algebraic heart of the freezing step.

Freezing invariance of the equation. Subtracting a constant from the forcing does not change the divergence-form weak equation.

3. Zero-trace solvability and the sharp energy bound #

The zero-datum comparator exists on any open bounded convex domain. For F ∈ L²(U) there is w ∈ H¹₀(U) with -Δ w = ∇·F weakly.

The sharp zero-trace energy bound. A zero-trace weak solution of -Δ w = ∇·F has Dirichlet energy at most the L² energy of the forcing.

The proof tests the equation with w itself and applies Young's inequality |ξ·η| ≤ |ξ|²/2 + |η|²/2 pointwise; the balanced split reproduces the sharp constant 1.

4. The freezing step #

The freezing step (harmonic approximation with divergence forcing).

Let u solve -Δ u = ∇·G weakly on an open bounded convex domain U and let c be an arbitrary constant vector. Then there is w ∈ H¹₀(U) with

  • -Δ w = ∇·(G - c) weakly on U,
  • u - w weakly harmonic on U, and
  • ∫_U |∇w|² ≤ ∫_U |G - c|².

With c := G(x₀) the right side is the squared L² oscillation of the forcing, which is the freezing gain of the Schauder argument.

5. The freezing gain on a triadic sub-cube #

The Euclidean square length is at most d times the squared sup norm.