Theorem 7.16 — generalized Poincaré–Bendixson theorem
ProvedTeschlODE.Planar.generalized_poincare_bendixsonLet be an open subset of and . Fix and , and suppose is compact, connected, and contains only finitely many fixed points. Then one of the following holds:
This classifies all compact connected limit sets of planar flows with finitely many equilibria; it is the capstone of Chapter 7.
Formalization Note. is a Bool (true = ). (i) is with . (ii) is with periodic and . (iii) is: every with lies on a non-closed orbit ( not periodic) with , and , for fixed points (the book's "", a one-point set identified with its point). "Only finitely many fixed points" is Set.Finite of ; "connected" is IsPreconnected together with the separate nonemptiness hypothesis.
import Mathlib import Definitions.Def_TeschlODE_Planar_orbit import Definitions.Def_TeschlODE_Planar_omegaLimitSet import Definitions.Def_TeschlODE_Planar_IsPeriodicPoint
namespace TeschlODE.Planar
/-- Teschl, Theorem 7.16 (generalized Poincaré–Bendixson, pp. 223–224): let `M ⊆ ℝ²` be open,
`f ∈ C¹(M, ℝ²)`, `x ∈ M`, `σ ∈ {±}`, and suppose `ω_σ(x)` is nonempty, compact, connected and
contains only finitely many fixed points. Then
(i) `ω_σ(x)` is a fixed orbit `{x₀}` with `f x₀ = 0`; or
(ii) `ω_σ(x) = γ(y)` is a regular periodic orbit (`y` periodic, `f y ≠ 0`); or
(iii) every point `y ∈ ω_σ(x)` is a fixed point or lies on a non-closed orbit `γ(y) ⊆ ω_σ(x)`
whose limit sets `ω₊(y)` and `ω₋(y)` are single fixed points of `ω_σ(x)`. -/
theorem generalized_poincare_bendixson {M : Set (Fin 2 → ℝ)} (hM : IsOpen M)
{f : (Fin 2 → ℝ) → Fin 2 → ℝ} (hf : ContDiffOn ℝ 1 f M) (σ : Bool)
{x : Fin 2 → ℝ} (hx : x ∈ M)
(hne : (omegaLimitSet f M σ x).Nonempty) (hcpt : IsCompact (omegaLimitSet f M σ x))
(hconn : IsPreconnected (omegaLimitSet f M σ x))
(hfin : {y | y ∈ omegaLimitSet f M σ x ∧ f y = 0}.Finite) :
(∃ x₀, f x₀ = 0 ∧ omegaLimitSet f M σ x = {x₀}) ∨
(∃ y, IsPeriodicPoint f M y ∧ f y ≠ 0 ∧ omegaLimitSet f M σ x = orbit f M y) ∨
(∀ y ∈ omegaLimitSet f M σ x, f y = 0 ∨
(orbit f M y ⊆ omegaLimitSet f M σ x ∧ ¬ IsPeriodicPoint f M y ∧
∃ a ∈ omegaLimitSet f M σ x, ∃ b ∈ omegaLimitSet f M σ x, f a = 0 ∧ f b = 0 ∧
omegaLimitSet f M true y = {a} ∧ omegaLimitSet f M false y = {b})) := by sorry
end TeschlODE.Planar
Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Throughout, points of the plane are pairs . The statement uses three notions defined elsewhere and imported: omegaLimitSet, orbit and IsPeriodicPoint. Their bodies were not given with this declaration, so this read-back treats them as opaque. It writes:
- for
omegaLimitSet f M σ x, a subset of that depends on , , a Boolean flag and a point ; - for
orbit f M y, a subset of that depends on , and ; - " is periodic" for
IsPeriodicPoint f M y, a property of that depends on and .
The code alone does not show which Boolean value ( or ) corresponds to forward time and which to backward time, if either does. It also does not show how these notions behave at points outside .
Data and hypotheses. Let:
- be an open set;
- be a function defined on all of that is continuously differentiable (class ) on . Nothing is assumed about outside .
- ;
- .
Assume all four of the following:
- is nonempty.
- is compact.
- is preconnected: it cannot be split into two disjoint relatively open nonempty pieces. Together with nonemptiness, this is the same as connected.
- The set of zeros of inside is finite:
Conclusion. At least one of the following three alternatives holds (inclusive "or").
- (i) There is a point with and
The code does not separately require .
-
(ii) There is a point such that:
- is periodic (in the sense of
IsPeriodicPoint f M); - ;
- .
- is periodic (in the sense of
-
(iii) For every , either , or all of the following hold:
- ;
- is not periodic;
- there exist points with , and
Here and are not required to be distinct. Both limit sets of are taken with the same and .
Degenerate cases. The dimension is fixed at 2, so there is no zero-dimensional or one-element-type case. cannot be empty, because .
- If is empty for some (whether that can happen depends on the unseen definition), the hypotheses fail and the theorem says nothing about that .
- Alternative (iii) is automatically true at every that is a zero of . So if every point of is a zero of , (iii) holds with no further content. By finiteness and connectedness, is then a single zero, which also satisfies (i).
- The alternatives are not stated to be mutually exclusive.
- Whether a zero of counts as "periodic" depends on the unseen definition. (ii) excludes such points explicitly by requiring .
- The hypotheses do not require . If the definitions allow points of outside , the conditions or at those points involve values of that are not constrained by the hypothesis. Likewise, , "periodic" and at such points take whatever values the definitions assign outside .
- Whether any of the hypotheses can actually be satisfied, or is vacuous for some inputs, cannot be determined without the bodies of
omegaLimitSet,orbitandIsPeriodicPoint.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.