Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section5.Support.PercolationScale

PercolationScale #

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

The event Q, the light-path events, and the percolation scale #

Section 5.2 produces, for each maximal scale m, an ℕ-valued random variable Y_m below which every rescaled lattice path from □_{m-n} to the complement of □_{m-n+1} meets the good event Q at at least (3/4) 3^{m-n} of its sites. The reading is the strong one: for almost every sample a single Y_m precedes all scales and all paths.

Y_m is defined here as a total ℕ-valued function of the sample, finite in the sense of the source only almost surely: the defining property is available exactly on the set of samples for which the light-path events stop occurring, and the source asks only that this set carry full measure. The statements about Y_m are accordingly almost-everywhere statements, and the value off that set does not enter them; percolationScaleTotal_eq_zero_of_not records what that value is. Nothing is hidden behind a caller hypothesis.

Measurable Y is a strengthening of the source, which asks only for a random variable; it is stated explicitly and proved from the measurability of the events Q.

Main definitions #

References #

@[instance_reducible]

The classical decidability instance used by the Finset.filter below. The counted quantity does not depend on it: qSiteCount_eq_card_filter identifies the count with the filter formed by any other instance.

Equations
Instances For

    1. The event Q #

    The rescaled lattice point 3^{n-1} z.

    Equations
    Instances For

      The event Q(z+□_n, ε): the good cube event at z + □_n intersected with the large-scale event at threshold Cinj⁻¹ ε γ^{-1/2}.

      Equations
      • One or more equations did not get rendered due to their size.
      Instances For
        noncomputable def Algsuperdiff.Section5.Support.qSiteCount {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (n : ℤ) (ep : ℝ) (S : Finset (Fin d → ℤ)) (omega : Section3.Cutoff.CutoffSample d) :

        The number of sites of S at which Q occurs, at cube scale n and after the rescaling 3^{n-1} ℤ^d ↦ ℤ^d.

        Equations
        • One or more equations did not get rendered due to their size.
        Instances For
          theorem Algsuperdiff.Section5.Support.qSiteCount_eq_card_filter {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (n : ℤ) (ep : ℝ) (S : Finset (Fin d → ℤ)) (omega : Section3.Cutoff.CutoffSample d) [DecidablePred fun (z : Fin d → ℤ) => omega ∈ qEvent M Creg Cinj n (rescaledLatticePoint n z) ep] :
          qSiteCount M Creg Cinj n ep S omega = {z ∈ S | omega ∈ qEvent M Creg Cinj n (rescaledLatticePoint n z) ep}.card

          The count is insensitive to the decidability instance. Whatever instance a consumer's Finset.filter carries, the resulting cardinality is the one counted here.

          theorem Algsuperdiff.Section5.Support.qSiteCount_eq_sum_indicator {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (n : ℤ) (ep : ℝ) (S : Finset (Fin d → ℤ)) (omega : Section3.Cutoff.CutoffSample d) :
          ↑(qSiteCount M Creg Cinj n ep S omega) = ∑ z ∈ S, (qEvent M Creg Cinj n (rescaledLatticePoint n z) ep).indicator (fun (x : Section3.Cutoff.CutoffSample d) => 1) omega

          The site count as a sum of indicators. This is the form in which the count is compared with a sum over the sites of a path, both here and where the percolation estimate's count meets the count of the statement.

          2. The light-path events #

          The light-path event at scale k: some lattice path starting in the cube □_k and ending outside the cube □_{k+1} meets Q at fewer than (3/4) 3^k of its sites. The rescaling 3^{n-1} ℤ^d ↦ ℤ^d at the cube scale n = m - k turns the source's crossing of □_{m-1} to the complement of □_m into this crossing of the unit lattice.

          Equations
          • One or more equations did not get rendered due to their size.
          Instances For
            def Algsuperdiff.Section5.Support.pathExtend {d N : ℕ} (p : Fin (N + 1) → Fin d → ℤ) :
            ℕ → Fin d → ℤ

            The extension of a finite path to all of ℕ.

            Equations
            Instances For
              theorem Algsuperdiff.Section5.Support.pathExtend_apply {d N : ℕ} (p : Fin (N + 1) → Fin d → ℤ) {i : ℕ} (hi : i < N + 1) :
              pathExtend p i = p ⟨i, hi⟩
              theorem Algsuperdiff.Section5.Support.lightQPathEvent_eq_iUnion {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) (k : ℕ) :

              The path quantifier is countable. Only the restriction of a path to its first N + 1 sites occurs in the light-path event, so the existential over ℕ → Fin d → ℤ is a countable union over finite paths.

              3. The total percolation scale #

              The samples for which the light-path events eventually stop.

              Equations
              • One or more equations did not get rendered due to their size.
              Instances For
                noncomputable def Algsuperdiff.Section5.Support.percolationScaleTotal {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) (omega : Section3.Cutoff.CutoffSample d) :

                The percolation scale Y_m, as a total function of the sample: the least scale beyond which no light path occurs. On the exceptional set where light paths occur at arbitrarily large scales there is no such scale; the value there is 0 and no statement below uses it.

                Equations
                • One or more equations did not get rendered due to their size.
                Instances For
                  theorem Algsuperdiff.Section5.Support.percolationScaleTotal_spec {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) {omega : Section3.Cutoff.CutoffSample d} (h : ∃ (N : ℕ), ∀ (k : ℕ), N ≤ k → omega ∉ lightQPathEvent M Creg Cinj m ep k) (k : ℕ) :
                  percolationScaleTotal M Creg Cinj m ep omega ≤ k → omega ∉ lightQPathEvent M Creg Cinj m ep k

                  The defining property, on the set where it is available.

                  theorem Algsuperdiff.Section5.Support.percolationScaleTotal_eq_zero_of_not {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) {omega : Section3.Cutoff.CutoffSample d} (h : ¬∃ (N : ℕ), ∀ (k : ℕ), N ≤ k → omega ∉ lightQPathEvent M Creg Cinj m ep k) :
                  percolationScaleTotal M Creg Cinj m ep omega = 0

                  The value on the exceptional set. Nothing is hidden: where light paths occur at arbitrarily large scales the scale is 0. The almost-everywhere statements below do not use this value.

                  theorem Algsuperdiff.Section5.Support.le_percolationScaleTotal_iff {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) {omega : Section3.Cutoff.CutoffSample d} (h : ∃ (N : ℕ), ∀ (k : ℕ), N ≤ k → omega ∉ lightQPathEvent M Creg Cinj m ep k) (N : ℕ) :
                  N ≤ percolationScaleTotal M Creg Cinj m ep omega ↔ ∀ j < N, ¬∀ (k : ℕ), j ≤ k → omega ∉ lightQPathEvent M Creg Cinj m ep k

                  4. The percolation scale almost surely #

                  theorem Algsuperdiff.Section5.Support.percolationScaleTotal_ae_crossing {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) {P : MeasureTheory.Measure (Section3.Cutoff.CutoffSample d)} (hae : ∀ᵐ (omega : Section3.Cutoff.CutoffSample d) ∂P, omega ∈ eventuallyHeavyPaths M Creg Cinj m ep) :
                  ∀ᵐ (omega : Section3.Cutoff.CutoffSample d) ∂P, ∀ n ≤ m - ↑(percolationScaleTotal M Creg Cinj m ep omega), ∀ (N : ℕ) (x : ℕ → Fin d → ℤ), Section3.Provider.Percolation.IsLatticePath x N → x 0 ∈ Section3.Provider.Percolation.cubeAt (m - n).toNat 0 → x N ∉ Section3.Provider.Percolation.cubeAt ((m - n).toNat + 1) 0 → 3 / 4 * 3 ^ (m - n).toNat ≤ ↑(qSiteCount M Creg Cinj n ep (Percolation.pathSites N x) omega)

                  The crossing bound of the source, almost surely. For almost every sample and every scale n ≤ m - Y_m, every lattice path from □_{m-n} to the complement of □_{m-n+1} meets Q at at least (3/4) 3^{m-n} of its distinct sites.

                  5. Measurability #

                  theorem Algsuperdiff.Section5.Support.measurableSet_lightQPathEvent {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) (hq : ∀ (n : ℤ) (z : HCPolySupport.Vec d), MeasurableSet (qEvent M Creg Cinj n z ep)) (k : ℕ) :
                  MeasurableSet (lightQPathEvent M Creg Cinj m ep k)

                  The light-path events are measurable, given the measurability of the events Q. The countable-union decomposition reduces the path quantifier to a countable index, and on each finite path the count is a finite sum of indicators.

                  theorem Algsuperdiff.Section5.Support.measurableSet_eventuallyHeavyPaths {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) (hq : ∀ (n : ℤ) (z : HCPolySupport.Vec d), MeasurableSet (qEvent M Creg Cinj n z ep)) :
                  theorem Algsuperdiff.Section5.Support.measurable_percolationScaleTotal {d : ℕ} (M : Section3.ABKModel d) (Creg Cinj : ℝ) (m : ℤ) (ep : ℝ) (hq : ∀ (n : ℤ) (z : HCPolySupport.Vec d), MeasurableSet (qEvent M Creg Cinj n z ep)) :

                  The percolation scale is measurable. This is a disclosed strengthening of the source, which asks only that Y_m be a random variable.