Documentation

Oseledets.Krieger.KeaneSerafin

The Keane–Serafin inductive assembly (Krieger sub-problem A, dynamical core) #

This file develops the inductive assembly of the Keane–Serafin / Downarowicz construction (Keane–Serafin, On the countable generator theorem, Fund. Math. 157 (1998), 255–259; Downarowicz, Entropy in Dynamical Systems, Thm 4.2.3, first half): the construction that turns the per-step Keane–Serafin refinement lemma into the KeaneSerafinData of Oseledets.Krieger.Generator, whence the finite-entropy countable two-sided generator (exists_countable_twoSided_generator) becomes unconditional.

The construction (Keane–Serafin 1998) #

Keane–Serafin produce, inductively, an increasing sequence Q₀ ⪯ Q₁ ⪯ ⋯ of finite partitions with Q₀ = {X}, where at step k one applies the Keane–Serafin Lemma with ε = 2^{-(k+1)}, P = Qₖ, A = Bₖ := natGeneratingSequence α k to get Q_{k+1}. The Lemma guarantees:

  1. Qₖ ⪯ Q_{k+1} (refinement);
  2. Bₖ ∈ ⋁ₙ eⁿ σ(Q_{k+1}) mod 0 (the two-sided itinerary of Q_{k+1} recovers Bₖ); and
  3. H(Q_{k+1}) ≤ H(Qₖ) + gₖ + 2^{-(k+1)} where gₖ = h(e, Q̄ₖ) − h(e, Qₖ) is the dynamical entropy gap (Q̄ₖ := Qₖ ∨ {Bₖ, Bₖᶜ}).

Telescoping (3) with the Pinsker bound h(e, Q̄ₖ) − h(e, Qₖ) ≤ H(Q_{k+1}) − H(Qₖ)(*) and the finite KS entropy h := h(e) < ∞ (the load-bearing finiteness shortcut) gives the uniform static bound sup_k H(Qₖ) ≤ h + 1 < ∞. The limit countable partition Q := ⋁ₖ Qₖ then has finite static entropy, and its two-sided itinerary recovers every Bₖ, hence two-sidedly generates mod 0.

(*) In the elementary Keane–Serafin write-up the entropy increment of step k is controlled directly by the count bound Jₖ ≤ e^{(gₖ+2δ)m} on the number of surviving sub-atoms per row, which comes from the Shannon–McMillan–Breiman convergence in probability ("most atoms of the m-block join have measure ≈ e^{−hm}"). This is the one genuinely analytic ingredient.

The single dynamical residual: the per-step Keane–Serafin Lemma #

The per-step Lemma (3) is not re-derived here: its proof needs the in-probability Shannon–McMillan–Breiman theorem (the equipartition μ(atom) ∈ [e^{−(h+δ)m}, e^{−(h−δ)m}] for most m-block atoms), which the repository's SMB development (Oseledets.Krieger.SMBSharp) currently isolates as the open R5 (Chung -domination) residual — only the integral-level rate identity ksEntropyPartition_eq_condEntropy_iSup and the crude name-count upper bound are proved. The in-probability equipartition is genuinely load-bearing for the count bound Jₖ ≤ e^{(gₖ+2δ)m}, hence for the entropy economy of the step, hence for summable_index_mass.

We therefore expose the per-step output as the honest hypothesis bundle KeaneSerafinStep, capturing exactly what the SMB-driven Lemma produces for the whole enumerated limit family:

The inductive assembly keaneSerafinData_of_step then produces KeaneSerafinData from a KeaneSerafinStep unconditionally and sorry-free. The hard analytic content (SMB + Rokhlin producing the geometric mass envelope) is the remaining dynamical residual, isolated in KeaneSerafinStep, exactly mirroring how Oseledets.Krieger.Krieger isolates the M1+M2 combinatorics in KriegerCodingData.

Why a hypothesis bundle and not an outright construction #

The outright construction of KeaneSerafinData (no hypotheses beyond ergodic/aperiodic measure-preserving e of finite entropy) needs the in-probability SMB, which is BLOCKED upstream. Bundling the SMB output is the faithful reduction: it keeps the structural content — turning two-sided recovery into mod-0 generation, and the geometric mass envelope into finite static Shannon entropy (Downarowicz Fact 1.1.4, cHμ_summable_of_summable_index_mul) — proved here, while the open analytic residual is named, not faked. See the module note at the end for the precise residual.

References #

The per-step Keane–Serafin output bundle #

KeaneSerafinStep captures, in faithful form, the output of the (SMB-driven) Keane–Serafin inductive construction for the whole enumerated limit family Q := ⋁ₖ Qₖ: a countable family of measurable cells whose two-sided itinerary recovers every standard-Borel generating set mod 0, and whose cell measures obey an enumerated geometric mass envelope μ(Qᵢ) ≤ Cᵢ with ∑ i·Cᵢ < ∞.

structure Oseledets.Krieger.KeaneSerafinStep {α : Type u_1} [ : MeasurableSpace α] [StandardBorelSpace α] (μ : MeasureTheory.Measure α) (e : α ≃ᵐ α) :
Type u_1

The Keane–Serafin step output (the faithful SMB-residual hypothesis bundle for sub-problem A). It bundles the enumerated limit partition Q : ℕ → Set α of the Keane–Serafin construction together with the two facts the SMB-driven per-step Lemma + Rokhlin tower guarantee:

  • cells_measurable: each cell is measurable;
  • recovers: the two-sided Q-itinerary recovers each standard-Borel generating set natGeneratingSequence α m mod 0 — the generation witness;
  • a summable index-mass envelope mass_envelope/mass_le: there is C : ℕ → ℝ with μ(Qᵢ).toReal ≤ Cᵢ and ∑ i, i·Cᵢ < ∞ — the finite static entropy witness in the directly summable_index_mass-feeding form (Downarowicz Fact 1.1.4 input). The geometric envelope is what the row-by-row organization of the Keane–Serafin construction (per-step measure ≤ 2^{−k}, geometric decay) produces; it is strictly the SMB content, isolated here.
  • Q : Set α

    The enumerated limit partition Q = ⋁ₖ Qₖ.

  • cells_measurable (i : ) : MeasurableSet (self.Q i)

    Each cell of the limit partition is measurable.

  • recovers (m : ) : ∃ (t : Set α), MeasurableSet t MeasurableSpace.natGeneratingSequence α m =ᵐ[μ] t

    The two-sided Q-itinerary recovers each standard-Borel generating set mod 0.

  • C :

    The geometric mass envelope Cᵢ ≥ μ(Qᵢ) (the SMB-produced per-row geometric decay).

  • mass_le (i : ) : (μ (self.Q i)).toReal self.C i

    Each cell measure is bounded by its envelope value.

  • envelope_summable : Summable fun (i : ) => i * self.C i

    The envelope is index-summable: ∑ i, i·Cᵢ < ∞.

Instances For

    The inductive assembly: KeaneSerafinStep ⟹ KeaneSerafinData #

    The assembly is the unconditional structural step: the enumerated geometric mass envelope μ(Qᵢ) ≤ Cᵢ with ∑ i·Cᵢ < ∞ yields the summable_index_mass field Summable (fun i => i·μ(Qᵢ).toReal) by termwise comparison (each i·μ(Qᵢ) ≤ i·Cᵢ for i ≥ 0), and recovers/cells_measurable carry over verbatim.

    theorem Oseledets.Krieger.summable_index_mass_of_envelope {α : Type u_1} [ : MeasurableSpace α] {μ : MeasureTheory.Measure α} [StandardBorelSpace α] {e : α ≃ᵐ α} (S : KeaneSerafinStep μ e) :
    Summable fun (i : ) => i * (μ (S.Q i)).toReal

    The Keane–Serafin index-mass comparison. A geometric mass envelope μ(Qᵢ).toReal ≤ Cᵢ with ∑ i·Cᵢ < ∞ certifies the summable_index_mass field directly: the index-weighted measure sequence i ↦ i·μ(Qᵢ).toReal is dominated termwise by the summable i ↦ i·Cᵢ (and is nonnegative), so it is summable. This is the directly-KeaneSerafinData-feeding form of the entropy economy.

    The Keane–Serafin data from a step. The enumerated limit partition of a KeaneSerafinStep, together with its recovery witness and the index-mass comparison summable_index_mass_of_envelope, assembles into the KeaneSerafinData of Oseledets.Krieger.Generator — the dynamical input to the finite-entropy countable two-sided generator. Unconditional and sorry-free.

    Equations
    • S.toData = { Q := S.Q, cells_measurable := , recovers := , summable_index_mass := }
    Instances For

      The finite-entropy countable two-sided generator from a Keane–Serafin step. A KeaneSerafinStep produces a countable (-indexed) partition Q that two-sidedly generates (α, e, μ) mod 0 and has finite static Shannon entropy. This is exists_countable_twoSided_generator_of_keaneSerafinData fed the assembled data — the headline of sub-problem A, conditional only on the (faithful, SMB-residual) KeaneSerafinStep.

      The multi-level structural assembly #

      The Keane–Serafin construction is inductive: it produces an increasing sequence of finite partitions Qₖ, with the level-k two-sided itinerary recovering the k-th generating set Bₖ mod 0, and the level cells placed (after the row-organization) with geometrically decaying measure. This subsection assembles such per-level data into a single KeaneSerafinStep (hence a KeaneSerafinData), unconditionally:

      The level cells Q k i are the enumerated cells of the Keane–Serafin construction (e.g. the genuine join atoms placed in rows); the only SMB-dependent input is the geometric envelope mass_le/envelope_summable, which the row-organization supplies.

      theorem Oseledets.Krieger.ctwoSidedSat_mono_of_range_subset {α : Type u_1} [ : MeasurableSpace α] {κ : Type u_2} {κ' : Type u_3} (e : α ≃ᵐ α) {Q : κSet α} {Q' : κ'Set α} (hsub : Set.range Q Set.range Q') :

      Range-monotonicity of the countable two-sided saturation. If range Q ⊆ range Q' then the two-sided itinerary σ-algebra grows: ctwoSidedSat e Q ≤ ctwoSidedSat e Q'. Indeed generateFrom (range Q) ≤ generateFrom (range Q') by generateFrom_mono, and comap and the -supremum are monotone. This is the engine of recovery-lifting: a recovery witness measurable in a sub-family's saturation is a fortiori measurable in the larger family's saturation.

      theorem Oseledets.Krieger.recovers_union_of_levelRecovers {α : Type u_1} [ : MeasurableSpace α] {μ : MeasureTheory.Measure α} [StandardBorelSpace α] {e : α ≃ᵐ α} (Q : Set α) (hrec : ∀ (m : ), ∃ (t : Set α), MeasurableSet t MeasurableSpace.natGeneratingSequence α m =ᵐ[μ] t) (m : ) :

      Recovery of the union family from per-level recovery. If, for each m, the level-m two-sided saturation recovers natGeneratingSequence α m mod 0, then the union family Q⁺ p := Q p.1 p.2 (over ℕ × ℕ) also recovers each generating set mod 0: each level-m saturation embeds into the union family's saturation (the union family's range contains the level-m range), so a level-m witness lifts. This is the recovers-field form of the reduction isGeneratingTwoSidedMod0c_sigmaUnion_of_levelRecovers (which states the generation conclusion; here we keep the per-m witnesses, as KeaneSerafinStep.recovers requires).

      structure Oseledets.Krieger.KeaneSerafinLevels {α : Type u_1} [ : MeasurableSpace α] [StandardBorelSpace α] (μ : MeasureTheory.Measure α) (e : α ≃ᵐ α) :
      Type u_1

      The per-level Keane–Serafin data (the inductive form of KeaneSerafinStep). It bundles an increasing sequence of finite partitions, presented as a ℕ × ℕ-indexed family Q : ℕ → ℕ → Set α (level k, cell i), with:

      • cells_measurable: each cell Q k i is measurable;
      • levelRecovers: the level-k two-sided itinerary recovers natGeneratingSequence α k mod 0; and
      • a per-cell geometric envelope C : ℕ → ℕ → ℝ, μ(Q k i).toReal ≤ C k i, whose ℕ × ℕ-sum ∑_{k,i} (enumeration index)·C k i is summable.

      The assembly KeaneSerafinLevels.toStep re-indexes ℕ × ℕ → ℕ to produce a KeaneSerafinStep.

      Instances For

        The Keane–Serafin step from per-level data. Re-indexing the union family Q⁺ p := Q p.1 p.2 through the pairing equivalence ℕ ≃ ℕ × ℕ (Nat.unpair), the per-level recovery lifts to the enumerated family (recovers_union_of_levelRecovers), and measurability and the geometric envelope carry over verbatim, producing a KeaneSerafinStep. Unconditional and sorry-free.

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

          The finite-entropy countable two-sided generator from per-level Keane–Serafin data. The inductive KeaneSerafinLevels (an increasing sequence of finite partitions with per-level recovery and a per-cell geometric envelope) produces a countable partition Q two-sidedly generating (α, e, μ) mod 0 with finite static Shannon entropy — sub-problem A, conditional only on the SMB-residual per-level data.

          The remaining dynamical residual (precise) #

          Everything above is unconditional and sorry-free. The single open residual for sub-problem A is the construction of a KeaneSerafinStep (equivalently KeaneSerafinLevels) from the raw dynamical hypotheses — an ergodic, aperiodic, measure-preserving e of a standard-Borel probability space with finite Kolmogorov–Sinai entropy h := (ksEntropy he).toReal < ∞. Concretely, the residual is the per-step Keane–Serafin Lemma:

          Given a finite partition P, a measurable set A, and c > 0, set P̄ := P ∨ {A, Aᶜ} and g := h(e, P̄) − h(e, P). Then there is a finite partition Q with P ⪯ Q, A ∈ ⋁ₙ eⁿ σ(Q) mod 0, and H(Q) ≤ H(P) + g + c.

          Its proof is the row-organized Rokhlin-tower marker-painting of Keane–Serafin §2. The rokhlin_tower lemma (Oseledets.Krieger.RokhlinTower) supplies the tower; the genuine missing ingredient is the in-probability Shannon–McMillan–Breiman theorem: for δ > 0 and m large, most atoms of the m-block join ⋁₀^{m−1} eⁿ P̄ have measure in [e^{−(h̄+δ)m}, e^{−(h̄−δ)m}] (and likewise for P), which yields the per-row count bound Jᵢ ≤ e^{(g+2δ)m} driving the entropy increment H(Q') ≲ g. The repository's SMB development (Oseledets.Krieger.SMBSharp) currently proves only the integral-level rate identity ksEntropyPartition_eq_condEntropy_iSup and the crude name-count upper bound (Oseledets.Krieger.SMB.ae_limsup_div_infoFun_le_log_card); the pointwise / in-probability equipartition is its open R5 (Chung -domination) leaf. The in-probability statement is genuinely load-bearing here (it gives the lower deviation bound keeping the surviving sub-atoms from being too small, hence caps the per-row count), so it is the sharp, non-fakeable residual. Until R5 lands, KeaneSerafinStep / KeaneSerafinLevels is the faithful hypothesis boundary, and exists_countable_twoSided_generator_of_levels is the unconditional reduction of sub-problem A to it.

          The summable_index_mass strengthening (vs. the weaker H(Q) < ∞) is not an extra obstacle: the row-organized construction places the Jᵢ ≈ e^{gm} cells of row i (measures ≈ e^{−h̄m}) at consecutive enumeration indices, so the per-index envelope Cᵢ decays geometrically and ∑ i·Cᵢ < ∞ is the natural output, fed directly into KeaneSerafinStep.envelope_summable.