Hyperbolicity of the maximal invariant set of a transverse plug gluing, given one metric, uniqueness, C¹ time-t maps and the flow properties
OpenAnosovPlugs.plugGluing_hyperbolic_of_pieces_of_flowPropertyLet and be plugs: compact Hausdorff smooth 3-manifolds with boundary carrying nonsingular C¹ vector fields transverse to the boundary. Let and be unions of connected components of the exit and entrance boundaries, and let be a C¹ diffeomorphism (the mission's IsBoundaryDiffeo). Let be a plug gluing (the mission's IsPlugGluing): is a compact Hausdorff smooth 3-manifold with boundary, and are C¹ embeddings with injective derivatives whose images cover and meet exactly along the seam , , and is the vector field on with and . The field is not assumed to be C¹. Assume in addition that and are hyperbolic plugs (their maximal invariant sets , are hyperbolic sets), and that the gluing is transverse: at every point the pushed exit leaf and the entrance leaf of through are transverse curves in (the mission's CurvesTransverseAt), where and are the exit and entrance laminations. A set of a 3-manifold is a hyperbolic set of a vector field on (the mission's IsHyperbolicSet F Λ) if there are a continuous Riemannian metric , line fields , on and constants , such that at every : , the two line fields are invariant under the derivatives of the time- maps of the flow of for all , and for , , and for , . The time- map is the mission's flowMap F t, which sends to for a chosen integral curve of on with when one exists, and to otherwise. We write for the time- map of , the mission's flowMap Z t: it sends to for a chosen integral curve of on with when one exists (the mission's FlowDefined Z y t), and to otherwise. Assume the maximal invariant set of is described (the hypothesis hΛ, proved in the mission as plugGluing_maxInvSet_subset and plugGluing_maxInvSet_superset):
Assume further the following. Each is the conclusion of another mission theorem, applied to the data above:
- (one metric for both pieces) a continuous Riemannian metric on together with line fields and constants, separately for each piece, that make and hyperbolic sets of with respect to this same (
isHyperbolicSet_of_metricapplied toplugGluing_pieces_hyperbolic); - (uniqueness) integral curves of on are determined by their starting point, for every (
plugGluing_integralCurveOn_unique); - (C¹ time- maps) for every point of the maximal invariant set (the points that lie on an integral curve of defined for all times), is an interior point of , and for every there is an open neighbourhood of such that the orbit of every is defined on and is of class C¹ on (
plugGluing_flowMap_contMDiffOn); - (flow properties) for every : ; ; ; ; (
flowMap_flowProperty_of_regularity); - (invariant pieces) and for all (
plugGluing_pieces_flowMap_invariant).
Then the whole maximal invariant set is hyperbolic:
In words: this is the fifth sentence of the paper's proof of Proposition 1.1, which says that, by a classical consequence of the hyperbolic theory, the orbit of inherits a hyperbolic structure, with the paper's description of the bundles at a connecting point : the stable bundle is and the unstable bundle is , pushed into by .
Formalization Note This theorem is plugGluing_hyperbolic_of_pieces_of_regularity with two added hypotheses, each the verbatim conclusion of a mission theorem (hypotheses 4 and 5). The flow properties of flowMap Z and the invariance of the pieces are now hypotheses. The transport of the derivative of flowMap Z through the seam, by the derivatives of , and , is still inside this theorem. Two classical theorems remain to be proved inside this theorem, and Mathlib (at the pinned version) has neither. The first is the stable manifold theorem for the hyperbolic sets and : the points whose forward orbit stays in carry a contracted line field, and a C¹ curve inside a leaf of is tangent to the sum of this line and the direction of the field; without it, the transversality hypothesis on the leaves gives no information on the bundles. The second is the extension of hyperbolicity over transverse heteroclinic orbits (cone fields near the two pieces, their transport along the connecting orbits, uniform constants). The paper's description of the bundles at connecting points is not asserted, because IsHyperbolicSet is existential in the bundles. The field is not assumed C¹; the time- map flowMap Z t is junk-valued where no integral curve exists.
import Mathlib import Definitions.Def_AnosovPlugs_Gluing open scoped Manifold ContDiff Topology open Set
namespace AnosovPlugs
theorem plugGluing_hyperbolic_of_pieces_of_flowProperty
{U : Type} [TopologicalSpace U] [ChartedSpace (EuclideanHalfSpace 3) U]
[IsManifold I3 ∞ U] [T2Space U] [CompactSpace U]
{V : Type} [TopologicalSpace V] [ChartedSpace (EuclideanHalfSpace 3) V]
[IsManifold I3 ∞ V] [T2Space V] [CompactSpace V]
{W : Type} [TopologicalSpace W] [ChartedSpace (EuclideanHalfSpace 3) W]
[IsManifold I3 ∞ W] [T2Space W] [CompactSpace W]
(X : (x : U) → TangentSpace I3 x) (Y : (y : V) → TangentSpace I3 y)
(hX : IsHyperbolicPlug X) (hY : IsHyperbolicPlug Y)
(Tout : Set U) (Tin : Set V) (hTout : IsUnionOfComponents Tout (outBoundary X))
(hTin : IsUnionOfComponents Tin (inBoundary Y)) (φ : U → V)
(hφ : IsBoundaryDiffeo φ Tout Tin)
(htransv : ∀ p ∈ φ '' (exitLamination X ∩ Tout) ∩ entranceLamination Y,
CurvesTransverseAt (pushLeaf φ Tout (exitLeaf X) p) (entranceLeaf Y p) p)
(Z : (w : W) → TangentSpace I3 w) (iU : U → W) (iV : V → W)
(hglue : IsPlugGluing X Y Tout φ Z iU iV)
(hΛ : maxInvSet Z = iU '' maxInvSet X ∪ iV '' maxInvSet Y ∪
{w | ∃ x ∈ exitLamination X ∩ Tout, φ x ∈ entranceLamination Y ∧
∃ γ : ℝ → W, γ 0 = iU x ∧ IsMIntegralCurve γ Z ∧ w ∈ range γ})
(g : RiemannianMetric3 W)
(hU : ∃ (Es Eu : (x : W) → Submodule ℝ (TangentSpace I3 x)) (C lam : ℝ), 0 < C ∧ 0 < lam ∧
∀ x ∈ iU '' maxInvSet X,
Module.finrank ℝ (Es x) = 1 ∧ Module.finrank ℝ (Eu x) = 1 ∧
Es x ⊔ Submodule.span ℝ {Z x} ⊔ Eu x = ⊤ ∧
(∀ t : ℝ, (Es x).map (mfderiv I3 I3 (flowMap Z t) x).toLinearMap = Es (flowMap Z t x)) ∧
(∀ t : ℝ, (Eu x).map (mfderiv I3 I3 (flowMap Z t) x).toLinearMap = Eu (flowMap Z t x)) ∧
(∀ t : ℝ, 0 ≤ t → ∀ v ∈ Es x,
g.norm (flowMap Z t x) (mfderiv I3 I3 (flowMap Z t) x v)
≤ C * Real.exp (-lam * t) * g.norm x v) ∧
(∀ t : ℝ, 0 ≤ t → ∀ v ∈ Eu x,
g.norm (flowMap Z (-t) x) (mfderiv I3 I3 (flowMap Z (-t)) x v)
≤ C * Real.exp (-lam * t) * g.norm x v))
(hV : ∃ (Es Eu : (x : W) → Submodule ℝ (TangentSpace I3 x)) (C lam : ℝ), 0 < C ∧ 0 < lam ∧
∀ x ∈ iV '' maxInvSet Y,
Module.finrank ℝ (Es x) = 1 ∧ Module.finrank ℝ (Eu x) = 1 ∧
Es x ⊔ Submodule.span ℝ {Z x} ⊔ Eu x = ⊤ ∧
(∀ t : ℝ, (Es x).map (mfderiv I3 I3 (flowMap Z t) x).toLinearMap = Es (flowMap Z t x)) ∧
(∀ t : ℝ, (Eu x).map (mfderiv I3 I3 (flowMap Z t) x).toLinearMap = Eu (flowMap Z t x)) ∧
(∀ t : ℝ, 0 ≤ t → ∀ v ∈ Es x,
g.norm (flowMap Z t x) (mfderiv I3 I3 (flowMap Z t) x v)
≤ C * Real.exp (-lam * t) * g.norm x v) ∧
(∀ t : ℝ, 0 ≤ t → ∀ v ∈ Eu x,
g.norm (flowMap Z (-t) x) (mfderiv I3 I3 (flowMap Z (-t)) x v)
≤ C * Real.exp (-lam * t) * g.norm x v))
(huniq : ∀ (γ γ' : ℝ → W) (t : ℝ), IsMIntegralCurveOn γ Z (uIcc 0 t) → IsMIntegralCurveOn γ' Z (uIcc 0 t) →
γ' 0 = γ 0 → ∀ s ∈ uIcc 0 t, γ' s = γ s)
(hreg : ∀ w ∈ maxInvSet Z, I3.IsInteriorPoint w ∧ ∀ t : ℝ, ∃ O : Set W, IsOpen O ∧ w ∈ O ∧
(∀ y ∈ O, FlowDefined Z y t) ∧ ContMDiffOn I3 I3 1 (flowMap Z t) O)
(hflow : ∀ w ∈ maxInvSet Z,
(∀ t : ℝ, flowMap Z t w ∈ maxInvSet Z) ∧
(∀ s t : ℝ, flowMap Z t (flowMap Z s w) = flowMap Z (s + t) w) ∧
(∀ v : TangentSpace I3 w, mfderiv I3 I3 (flowMap Z 0) w v = v) ∧
(∀ (s t : ℝ) (v : TangentSpace I3 w), mfderiv I3 I3 (flowMap Z (s + t)) w v =
mfderiv I3 I3 (flowMap Z t) (flowMap Z s w) (mfderiv I3 I3 (flowMap Z s) w v)) ∧
(∀ t : ℝ, mfderiv I3 I3 (flowMap Z t) w (Z w) = Z (flowMap Z t w)))
(hinv : (∀ w ∈ iU '' maxInvSet X, ∀ t : ℝ, flowMap Z t w ∈ iU '' maxInvSet X) ∧
(∀ w ∈ iV '' maxInvSet Y, ∀ t : ℝ, flowMap Z t w ∈ iV '' maxInvSet Y)) :
IsHyperbolicSet Z (maxInvSet Z) := by sorry
end AnosovPlugs