Documentation

LeanPool.Superdiffusion.Algsuperdiff.Section5.Support.ConfinementScaleMinimality

ConfinementScaleMinimality #

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

Minimality at the confinement scale, and the level sets of the random scale #

The confinement scale m_t is the least scale n with L (1 + S_n) ≤ 3^n. Its defining inequality bounds the widened scale from above; what bounds the confinement scale itself from above is minimality, the failure of the same inequality one scale lower. This file turns that failure into a usable estimate and then into a bound on the probability of each level set of the random scale.

The first step is the recursion

S_{n-1} ≤ S_{n-1} + (1/9) S_n,

which holds because the widening weight 3^{-2(k-n)} loses exactly a factor of nine when the base scale drops by one. Combining it with the defining inequality at m_t itself, which bounds L S_{m_t} by 3^{m_t}, the failure one scale lower reads

3^{m_t - 1} < L (1 + S_{m_t-1} + (1/9) S_{m_t}) ≤ L (1 + S_{m_t-1}) + (1/9) 3^{m_t},

and hence 3^{m_t} ≤ (9/2) L (1 + S_{m_t-1}): the confinement scale is controlled by a single displacement scale, not by the widened supremum.

The scale index is random, so this is not yet a moment bound. What it gives is a bound on the probability of the level set {m_t = n}: on that event a single displacement scale is at least (2/27) 3^{n - m_0} - 1, where 3^{m_0} ≤ L is the deterministic lower bound for the scale, and Chebyshev's inequality at a uniform moment of the displacement scales turns this into a geometric decay in n - m_0. That decay is what makes the moments of 3^{m_t} and of any family indexed by the random scale summable.

References #

1. The recursion for the widened scale #

The widening recursion. Lowering the base scale by one adds the scale at the new index and divides the previous widened scale by nine.

2. Minimality #

theorem Algsuperdiff.Section5.Support.three_zpow_le_of_isConfinementScale {S : ℤ → ℝ} (hSnn : ∀ (k : ℤ), 0 ≤ S k) {L : ℝ} (hL : 0 < L) {m : ℤ} (hm : IsConfinementScale (fun (n : ℤ) => widenedScale S n) L m) :
3 ^ m ≤ 9 / 2 * L * (1 + S (m - 1))

Minimality at the confinement scale. Since the scale one below the confinement scale fails the defining inequality, and the widening recursion reduces the widened scale there to the single scale S_{m-1} plus a ninth of the widened scale at m, the confinement scale obeys 3^m ≤ (9/2) L (1 + S_{m-1}).

3. Measurability of the confinement scale #

theorem Algsuperdiff.Section5.Support.measurableSet_mem_confinementSet {Omega : Type u_1} [MeasurableSpace Omega] {S : ℤ → Omega → ℝ} (hmeas : ∀ (k : ℤ), Measurable (S k)) (L : ℝ) (n : ℤ) :
MeasurableSet {omega : Omega | n ∈ confinementSet (fun (j : ℤ) => widenedScale (fun (k : ℤ) => S k omega) j) L}

The defining set of the confinement scale is measurable at every fixed scale.

4. The level sets of the confinement scale #

theorem Algsuperdiff.Section5.Support.intLog_le_of_isConfinementScale {S : ℤ → ℝ} {L : ℝ} (hL : 0 < L) {m : ℤ} (hm : IsConfinementScale (fun (n : ℤ) => widenedScale S n) L m) :
Int.log 3 L ≤ m

The deterministic lower bound for the confinement scale.

theorem Algsuperdiff.Section5.Support.measure_isConfinementScale_eq_le {Omega : Type u_1} [MeasurableSpace Omega] (mu : MeasureTheory.Measure Omega) [MeasureTheory.IsProbabilityMeasure mu] {S : ℤ → Omega → ℝ} (hmeas : ∀ (k : ℤ), Measurable (S k)) (hSnn : ∀ (k : ℤ) (omega : Omega), 0 ≤ S k omega) {L : ℝ} (hL : 0 < L) {m : Omega → ℤ} (hm : ∀ᵐ (omega : Omega) ∂mu, IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L (m omega)) {C q : ℝ} (hC : 1 ≤ C) (hq : 0 < q) (hmom : ∀ (k : ℤ), ∫⁻ (omega : Omega), ENNReal.ofReal (S k omega) ^ q ∂mu ≤ ENNReal.ofReal C ^ q) (k : ℕ) :
mu {omega : Omega | m omega = Int.log 3 L + ↑k} ≤ ENNReal.ofReal (C * 3 ^ (3 - ↑k)) ^ q

The probability of a level set of the confinement scale. Writing the scale as m_0 + k with 3^{m_0} ≤ L < 3^{m_0+1}, on the event {m_t = m_0 + k} a single displacement scale is at least 3^{k-3}, so a uniform moment bound of order q for the displacement scales gives a geometric bound of ratio 3^{-q} for the probability of the event.

5. The measurable choice of the confinement scale #

noncomputable def Algsuperdiff.Section5.Support.confinementScale {Omega : Type u_1} (S : ℤ → Omega → ℝ) (L : ℝ) (omega : Omega) :

The confinement scale as a function of the sample. Where a confinement scale exists it is unique, and this is that scale; where none exists the value is 0.

The value off the existence event is immaterial. Every statement below about confinementScale is either conditioned on the existence of a confinement scale at the sample or holds almost surely, and the existence event has full measure whenever the displacement scales have one uniform moment bound (ae_exists_isConfinementScale); nothing in this development reads the value on its complement.

Equations
  • One or more equations did not get rendered due to their size.
Instances For
    theorem Algsuperdiff.Section5.Support.isConfinementScale_confinementScale {Omega : Type u_1} {S : ℤ → Omega → ℝ} {L : ℝ} {omega : Omega} (h : ∃ (m : ℤ), IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L m) :
    IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L (confinementScale S L omega)

    Where a confinement scale exists, the chosen value is one.

    theorem Algsuperdiff.Section5.Support.confinementScale_eq {Omega : Type u_1} {S : ℤ → Omega → ℝ} {L : ℝ} {omega : Omega} {m : ℤ} (hm : IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L m) :
    confinementScale S L omega = m

    The chosen value is the confinement scale, wherever there is one.

    theorem Algsuperdiff.Section5.Support.confinementScale_eq_zero {Omega : Type u_1} {S : ℤ → Omega → ℝ} {L : ℝ} {omega : Omega} (h : ¬∃ (m : ℤ), IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L m) :
    confinementScale S L omega = 0

    Where no confinement scale exists the chosen value is 0.

    theorem Algsuperdiff.Section5.Support.measurable_confinementScale {Omega : Type u_1} [MeasurableSpace Omega] {S : ℤ → Omega → ℝ} (hmeas : ∀ (k : ℤ), Measurable (S k)) (L : ℝ) :

    The chosen confinement scale is measurable. Each level set is the event that the scale is admissible intersected with countably many complements, and the level set at 0 carries in addition the event that no scale is admissible.

    theorem Algsuperdiff.Section5.Support.ae_isConfinementScale_confinementScale {Omega : Type u_1} [MeasurableSpace Omega] (mu : MeasureTheory.Measure Omega) [MeasureTheory.IsProbabilityMeasure mu] {S : ℤ → Omega → ℝ} (hmeas : ∀ (k : ℤ), Measurable (S k)) {C p L : ℝ} (hC : 0 ≤ C) (hp : 1 ≤ p) (hL : 0 < L) (hmom : ∀ (k : ℤ), ∫⁻ (omega : Omega), ENNReal.ofReal (S k omega) ^ p ∂mu ≤ ENNReal.ofReal C ^ p) :
    ∀ᵐ (omega : Omega) ∂mu, IsConfinementScale (fun (n : ℤ) => widenedScale (fun (k : ℤ) => S k omega) n) L (confinementScale S L omega)

    The chosen confinement scale is almost surely a confinement scale. One uniform moment bound for the displacement scales makes the defining set almost surely nonempty, and the infimum is then attained.