Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section3.Provider.Whitney.BadSetDefinitions

BadSetDefinitions #

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

d.whitney.bad-set.definitions: the Whitney partition and its bad set #

This module defines local implementations of the definitions in ABK26's sss.whitney.partition:

The recursion, unrolled #

The manuscript's construction is: divide □_m into its 3^d children; a child that touches ∂□_m is carried to the next layer and re-divided; a child that does not touch ∂□_m is a layer-n cube and is subdivided h_n further generations. The cubes so produced are the elements of 𝒲(□_m).

A cube reaches layer n exactly when every one of its ancestors of depth 1, …, n-1 touches ∂□_m, and by touchesBoundary_of_mem_descendantsAtDepth that is the same as its depth-(n-1) ancestor touching ∂□_m. So the layer is the closed-form family

𝒲(□_m, n) = ⋃ { descendantsAtDepth C (h n) :
                  C a non-touching child of a touching depth-(n-1) descendant } ,

which is the definition taken here (whitneyLayer). 𝒲(□_m) is the union of all layers, without any fuel truncation. whitneyLayer_eq_filter_scale proves that this agrees with the manuscript's own e.W.n.def, which defines 𝒲(□_m, n) as the cubes of 𝒲(□_m) of size 3^{m-n-h_n}.

Carriers #

Cubes are Homogenization.TriadicCube d; □_m is originCube d m; triadic subdivision is childCubes / descendantsAtDepth. "□ touches ∂□_m" is the extreme-index predicate TouchesBoundary: a depth-t descendant of □_m is the box (index ± ½)·3^{m-t}, and □_m is (0 ± ½)·3^m, so the closed cube meets ∂□_m in the coordinate i exactly when index i = ±(3^t - 1)/2 = ±E_t. The bad event 𝓑(□) is the proved BadEvents.bad.

Reading notes (recorded) #

References #

The scale separation sequence e.hn.def #

noncomputable def Algsuperdiff.Section3.Provider.Whitney.whitneyScaleSeq (b : ℝ) (hs k₀ n : ℕ) :

ABK26 e.hn.def: the scale separation sequence h_n = ⌈b (1-b)⁻¹ n⌉ + h_hat_sep + k₀, as a function of the value hs of the minimal scale separation h_hat_sep and of the parameter k₀.

Equations
Instances For

    h_n ≥ h_hat_sep + k₀: the bottom of the sequence, used to check the h ≥ h_hat_sep hypothesis of the deterministic percolation bounds.

    theorem Algsuperdiff.Section3.Provider.Whitney.whitneyScaleSeq_slope_le_one {b : ℝ} (hb0 : 0 < b) (hb : b ≤ 1 / 8) :
    b * (1 - b)⁻¹ ≤ 1

    The slope of e.hn.def is at most 1 when b ≤ 1/8 (the true bound is 1/7; ≤ 1 is what the gap proof consumes).

    theorem Algsuperdiff.Section3.Provider.Whitney.whitneyScaleSeq_succ_le {b : ℝ} (hb0 : 0 < b) (hb : b ≤ 1 / 8) (hs k₀ n : ℕ) :
    whitneyScaleSeq b hs k₀ (n + 1) ≤ whitneyScaleSeq b hs k₀ n + 1

    ABK26 e.hn.gap: for b ≤ 1/8 consecutive scale separations differ by at most one, so the Whitney scales grow by at most one layer at a time.

    theorem Algsuperdiff.Section3.Provider.Whitney.whitneyScaleSeq_mono {b : ℝ} (hb0 : 0 ≤ b) (hb1 : b < 1) (hs k₀ : ℕ) :
    theorem Algsuperdiff.Section3.Provider.Whitney.le_whitneyScaleSeq_cast (b : ℝ) (hs k₀ n : ℕ) :
    b * (1 - b)⁻¹ * ↑n + ↑hs + ↑k₀ ≤ ↑(whitneyScaleSeq b hs k₀ n)

    The real lower bound h_n ≥ b(1-b)⁻¹ n + h_hat_sep + k₀ used in the cluster diameter estimate.

    noncomputable def Algsuperdiff.Section3.Provider.Whitney.whitneyScale {d : ℕ} (M : ABKModel d) (m : ℤ) (E b : ℝ) (k₀ : ℕ) (omega : Cutoff.CutoffSample d) :
    ℕ → ℕ

    The Whitney depth sequence anchored at the sample's separation height.

    Equations
    • One or more equations did not get rendered due to their size.
    Instances For
      theorem Algsuperdiff.Section3.Provider.Whitney.hsep_le_whitneyScale {d : ℕ} (M : ABKModel d) (m : ℤ) (E b : ℝ) (k₀ : ℕ) (omega : Cutoff.CutoffSample d) (n : ℕ) :
      Percolation.hsep M m E b omega ≤ whitneyScale M m E b k₀ omega n

      Boundary-index arithmetic #

      The extreme index coordinate of a scale-s triadic descendant of □_m.

      Equations
      Instances For

        □ touches ∂□_m: some index coordinate of □ is extreme. For a depth-t descendant of □_m, whose closed realization is the box (index ± ½)·3^{m-t} inside □_m = (0 ± ½)·3^m, this is exactly the condition that the closed cube meets the boundary ∂□_m.

        Equations
        • One or more equations did not get rendered due to their size.
        Instances For

          Triadic subdivision #

          The index of a depth-s descendant stays within E_s of the rescaled index of its ancestor.

          Boundary touching propagates from a child to its parent.

          Boundary touching propagates to ancestors. Contrapositive: no descendant of a cube in the interior of □_m ever touches ∂□_m.

          The Whitney partition #

          ABK26 𝒲(□_m, n): the layer-n cubes of the Whitney partition of □_m at the scale profile hn. A layer-n cube is a depth-h_n triadic descendant of a child C of a boundary-touching depth-(n-1) descendant A of □_m, where C itself does not touch ∂□_m. Layer 0 is empty: the recursion starts at layer 1.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For

            ABK26 𝒲(□_m): the Whitney partition, the union of all its layers. No truncation: the family is the literal infinite one.

            Equations
            Instances For
              theorem Algsuperdiff.Section3.Provider.Whitney.scale_eq_of_mem_whitneyLayer {d : ℕ} {m : ℤ} {hn : ℕ → ℕ} {n : ℕ} {Q : HCPolySupport.TriadicCube d} (hQ : Q ∈ whitneyLayer m hn n) :
              Q.scale = m - ↑n - ↑(hn n)

              e.W.n.def, the size clause: every layer-n cube has size 3^{m-n-h_n}.

              A layer-n cube is a depth-(n + h_n) triadic descendant of □_m.

              The termination clause: no element of 𝒲(□_m) touches ∂□_m.

              theorem Algsuperdiff.Section3.Provider.Whitney.whitneyLayer_disjoint {d : ℕ} {m : ℤ} {hn : ℕ → ℕ} (hmono : Monotone hn) {n n' : ℕ} (hne : n ≠ n') :

              Distinct layers are disjoint.

              The bad cubes of the partition #

              ABK26 e.bad.cubes.in.partition: the bad family ℐ = {□ ∈ 𝒲(□_m): 𝓑(□) occurs}.

              Equations
              • One or more equations did not get rendered due to their size.
              Instances For

                ABK26 ℐ_n = ℐ ∩ 𝒲(□_m, n): the bad cubes of layer n.

                Equations
                • One or more equations did not get rendered due to their size.
                Instances For
                  theorem Algsuperdiff.Section3.Provider.Whitney.badFamilyLayer_subset {d : ℕ} (M : ABKModel d) (m : ℤ) (hn : ℕ → ℕ) (n : ℕ) (omega : Cutoff.CutoffSample d) :
                  badFamilyLayer M m hn n omega ⊆ whitneyLayer m hn n

                  Adjacency, neighborhoods and components #

                  dist(□, □') = 0 in the manuscript's ℓ^∞ sense: the closed cubes intersect. On Vec d = Fin d → ℝ the ambient metric is the sup metric and Metric.closedBall (cubeCenter Q) (cubeRadius Q) is the closed realization of the triadic cube Q.

                  Equations
                  • One or more equations did not get rendered due to their size.
                  Instances For

                    ABK26 e.neighborhood.def: the neighborhood 𝒩(𝒥) = {□ ∈ 𝒲(□_m): dist(□, 𝒥) = 0} of a subfamily 𝒥 ⊆ 𝒲(□_m), read cubewise (see the module docstring).

                    Equations
                    • One or more equations did not get rendered due to their size.
                    Instances For