Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section4.Provider.ExcessDecay.MeanControlWindowCube

MeanControlWindowCube #

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

The mean comparison, step 1: a boundary-flush cube inside the anchor's own #

window

Nothing here imports that file, and nothing here claims the anchor or any source node.

The geometric obstruction this module removes #

The mean comparison asks for a bound on the scalar

  S = |⨍_{c+□_{n+2}} (u − h)| ,    c = wellPlacedCentre x m (n+2) ,

on the boundary branch (z+□_{n+2}) ∩ ∂□_m ≠ ∅. The only extra information the boundary branch carries is the zero trace of u − h on ∂□_m, so any producer must reach ∂□_m from the cube it works on. The covering cube of the anchor's own centre x does not reach it: the branch is a statement about z, and x is only known to lie in z + □_{n+1}, so x can sit a distance 3^{n+1}/2 short of the clamp level and wellPlacedCentre x m (n+2) + □_{n+2} can then be strictly interior to □_m.

This module produces the cube that does reach it, from the anchor's own binders and with no new construction: the same wellPlacedCentre clamp, applied to z instead of x. Writing

  K := wellPlacedCentre z m (n+2) + □_{n+2} ,

three facts are proved.

The last item is what the funded coarse-grained boundary Poincaré needs and what the x-centred covering cube cannot supply: u − h has zero trace on the face, so a comparator on K that shares u's boundary data agrees with the datum h there. The n+3 ≤ m binder of the frozen statement is what makes the (n+2) clamp legitimate (n + 2 ≤ m suffices, and is implied).

1. Coordinates of a translated open cube #

theorem Algsuperdiff.Section4.Provider.ExcessDecay.mem_image_add_openCubeSet_coord_iff {d : ℕ} {j : ℤ} {z y : HCPolySupport.Vec d} :
y ∈ (fun (y' : HCPolySupport.Vec d) => z + y') '' HCPolySupport.openCubeSet (HCPolySupport.originCube d j) ↔ ∀ (i : Fin d), -(1 / 2) * 3 ^ j < y i - z i ∧ y i - z i < 1 / 2 * 3 ^ j

Membership in a translated open triadic cube, coordinatewise.

2. The negated interior gate, in clamp normal form #

The boundary branch, quantitatively.

If the parent cube z + □_j meets the frontier of □_m, then in some coordinate i and some direction σ = ±1 the centre z already exceeds the clamp level wellPlacedHalfGap m j = ½·3^m − ½·3^j of the well-placed cube of side 3^j. This is exactly the hypothesis under which the clamp saturates, so it is the bridge from the anchor's negated gate to a boundary-flush cube.

3. The clamp saturates: the cube is flush against the frontier #

theorem Algsuperdiff.Section4.Provider.ExcessDecay.wellPlacedCentre_flush {d : ℕ} {m k : ℤ} (hkm : k ≤ m) {z : HCPolySupport.Vec d} {i : Fin d} {σ : ℝ} (hσ : σ = 1 ∨ σ = -1) (hover : wellPlacedHalfGap m k < σ * z i) :

The flush face.

If the centre z exceeds the clamp level in the direction σeᵢ, then the well-placed centre of side 3^k saturates the clamp in that coordinate: σ·cᵢ = wellPlacedHalfGap m k.

theorem Algsuperdiff.Section4.Provider.ExcessDecay.wellPlacedCentre_faceLevel {d : ℕ} {m k : ℤ} (hkm : k ≤ m) {z : HCPolySupport.Vec d} {i : Fin d} {σ : ℝ} (hσ : σ = 1 ∨ σ = -1) (hover : wellPlacedHalfGap m k < σ * z i) :
σ * wellPlacedCentre z m k i + 1 / 2 * 3 ^ k = 1 / 2 * 3 ^ m

The flush face, at the frontier level. The σeᵢ face of the well-placed cube wellPlacedCentre z m k + □_k lies exactly in the frontier hyperplane {σ yᵢ = ½·3^m} of □_m.

4. The flush cube sits inside the frozen window #

The well-placed centre of z never moves z by as much as a half side.

The flush cube lies in the frozen window.

wellPlacedCentre z m (n+2) + □_{n+2} ⊆ (z+□_{n+3}) ∩ □_m: every quantity read on the flush cube transports to the frozen clause's own window at the 3^d measure ratio.