The doubling map's binary join cells (dynamical crux) #
This module isolates the dynamical facts behind Rokhlin's entropy equality for the doubling map
T : y ↦ 2 • y on UnitAddCircle = AddCircle (1 : ℝ). Let α be the binary partition
{[0,1/2), [1/2,1)} (the mk-images of those half-open real intervals). Its n-fold join
⋁ₖ₌₀ⁿ⁻¹ T⁻ᵏ α is the partition of the circle into the 2ⁿ dyadic arcs of length 2⁻ⁿ.
The crux volume_binJoinCell states that every cell of that n-fold join has volume exactly
2⁻ⁿ. Feeding it into the abstract reduction ksEntropyPartition_of_uniform
(file AbstractEqui.lean) yields h(α, T) = log 2.
Structure of the argument #
ksJoinCells_succ_head_tail: the head/tail recursionCₙ₊₁,f = α_{f 0} ∩ T⁻¹(Cₙ, tail f).mem_binCell_iff: a circle point lies inα_iiff its[0,1)-representative lies inbinLift i = [i/2,(i+1)/2); this turns themk-image cell into a measurable preimage.volume_binCell_inter_preimage: the dynamical heart,volume (α_i ∩ T⁻¹ B) = volume B / 2. The doubling map restricted to the half-arcα_iis the affine2×-magnificationx ↦ 2x - ionto the whole circle, so it halves measures. The proof reduces — via the fundamental-domain measure formulavolume_eq_preimage_inter_IcoandReal.volume_preimage_mul_left(factor1/2) — to theℤ-translation invariance ofmk⁻¹' B(preimage_mk_add_int_eq).volume_binJoinCell: induction onnvia the head/tail recursion, halving the measure at each step fromvolume univ = 1.
For the integral side ∫ log|det D(2x)| dμ = ∫ log 2 = log 2 and the headline equality, see
Oseledets.Examples.Rokhlin.DoublingEquality.
The i-th cell of the binary partition of the unit circle: the mk-image of
[i/2, (i+1)/2).
Equations
Instances For
The canonical [0,1)-representative of a circle point: AddCircle.equivIco 1 0 followed by the
subtype coercion. It is the unique real in [0,1) mapping to the given point under mk.
Equations
- Oseledets.Examples.Rokhlin.rep y = ↑((AddCircle.equivIco 1 0) y)
Instances For
Characterization of a binary cell via the canonical representative. A circle point lies in
binCell i iff its [0,1)-representative lies in the lifted interval binLift i = [i/2,(i+1)/2).
This turns the mk-image cell into a preimage of a measurable real set under the measurable map
rep, and is the workhorse for measurability, disjointness, and cover.
Head/tail decomposition of an iterated-join cell. Splitting off the k = 0 factor,
⋂ₖ₌₀ⁿ T⁻ᵏ(α_{f k}) = α_{f 0} ∩ T⁻¹(⋂ₖ₌₀ⁿ⁻¹ T⁻ᵏ(α_{(tail f) k})). The shifted factors at
k = j+1 are T⁻¹ of the corresponding n-fold-join factor because T^[j+1] = T^[j] ∘ T.
Measurability of the binary cells #
The canonical representative map rep = Subtype.val ∘ equivIco 1 0 is measurable: it is the
forward map of the measurable equivalence AddCircle.measurableEquivIco 1 0 post-composed with the
measurable subtype coercion.
The dynamical half-scaling lemma #
The single genuinely-dynamical input: the two doubling-map branches each magnify by 2, so each
halves measures.
The doubling map in mk-coordinates: T (mk x) = mk (2 x).
mk⁻¹' B is invariant under translation by an integer (here i ∈ {0,1} ⊆ ℤ): mk (x + i) = mk x, so x + i ∈ mk⁻¹' B ↔ x ∈ mk⁻¹' B. We phrase it as a set translation identity.
Fundamental-domain measure formula with the half-open [0,1) domain (the Ico analogue of
AddCircle.add_projection_respects_measure): for measurable U, volume U = volume (mk⁻¹' U ∩ [0,1)). Obtained from the Ioc 0 1 version by measure_congr, since Ioc 0 1 =ᵐ Ico 0 1.
Half-scaling under the doubling map. For any measurable B ⊆ UnitAddCircle, the part of
the preimage T⁻¹ B lying in the binary half-arc α_i = binCell i has exactly half the measure of
B. Indeed T restricted to α_i = [i/2,(i+1)/2) is the affine map x ↦ 2x - i onto the whole
circle (a measure 2×-magnification), so it pulls B back to a set of half the measure inside
α_i.
The join cell measure #
Every binary n-fold-join cell of the doubling map has volume 2⁻ⁿ. Induction on n using
the head/tail recursion ksJoinCells_succ_head_tail and the half-scaling lemma
volume_binCell_inter_preimage: the base cell (n = 0) is the whole circle of measure 1, and
each step multiplies the measure by 1/2.