Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem (4.4) — F is flow-invariant for Lipschitz X iff X(x) ⊆ T_F(x) on F

Proved
ClarkeGradients.FlowInvariance.flowInvariant_iff_subset_tangentCone

by mikedeng1 · Sep 27, 2026 · Mathlib 0df444a (Lean v4.33.1)

differential-inclusionflow-invariancep2o-batch-p100bp2o-gran-per-chapterp2o-plan-paperp2o-v1tangent-coneviability

Let XXX be a multifunction from Rn\mathbb R^nRn to Rn\mathbb R^nRn whose values X(x)X(x)X(x) are nonempty and compact, and suppose XXX is Lipschitz in the sense of (4.2). Let FFF be a nonempty closed subset of Rn\mathbb R^nRn. The following are equivalent:

  1. FFF is flow-invariant for XXX (4.3): every trajectory (4.1) of x˙∈X(x)\dot x\in X(x)x˙∈X(x) starting in FFF stays in FFF on [0,1][0,1][0,1];
  2. for each x∈Fx\in Fx∈F, X(x)X(x)X(x) is tangent to FFF at xxx:
X(x)⊆TF(x)for all x∈F,X(x)\subseteq T_F(x)\qquad\text{for all } x\in F,X(x)⊆TF​(x)for all x∈F,

where TF(x)T_F(x)TF​(x) is the Clarke tangent cone (3.6).

The theorem shows that Clarke's tangent cone is exactly the right notion of tangency for invariance of closed sets under Lipschitz differential inclusions; it contains the classical invariance theorems of Bony and Brezis for Lipschitz vector fields.

Formalization Note "X(x)X(x)X(x) nonempty and compact for every xxx" is the standing assumption stated once at the top of §4 (p. 259) and is an explicit hypothesis. Trajectories are absolutely continuous on [0,1][0,1][0,1] with x˙(t)∈X(x(t))\dot x(t)\in X(x(t))x˙(t)∈X(x(t)) almost everywhere, encoded with HasDerivAt. TFT_FTF​ is the Clarke tangent cone (the polar of the normal cone (3.1)), not Mathlib's Bouligand tangentConeAt. The Lipschitz constant in (4.2) is global.

Preamble
import Mathlib
import Definitions.Def_ClarkeGradients_FlowInvariance_tangentCone
import Definitions.Def_ClarkeGradients_FlowInvariance_IsLipschitzMultifunction
import Definitions.Def_ClarkeGradients_FlowInvariance_FlowInvariant
Formal statement
namespace ClarkeGradients.FlowInvariance

/-- Clarke (1975), Theorem (4.4): let `X` be a Lipschitz multifunction (4.2) whose values `X(x)`
are nonempty and compact (standing assumption of §4), and let `F` be a nonempty closed subset of
`ℝⁿ`. Then `F` is flow-invariant for `X` (4.3) if and only if `X(x) ⊆ T_F(x)` for every `x ∈ F`,
where `T_F(x)` is the Clarke tangent cone (3.6). -/
theorem flowInvariant_iff_subset_tangentCone {n : ℕ}
    (X : EuclideanSpace ℝ (Fin n) → Set (EuclideanSpace ℝ (Fin n)))
    (hX : ∀ x, (X x).Nonempty ∧ IsCompact (X x)) (hLip : IsLipschitzMultifunction X)
    (F : Set (EuclideanSpace ℝ (Fin n))) (hF : F.Nonempty) (hFc : IsClosed F) :
    FlowInvariant X F ↔ ∀ x ∈ F, X x ⊆ tangentCone F x := by sorry

end ClarkeGradients.FlowInvariance
Source
Clarke, Generalized gradients and applications, Trans. Amer. Math. Soc. 205 (1975), p. 260, Theorem (4.4)
Read-back

What the Lean code literally says, in plain math · claude-opus-5-5

Let n∈Nn \in \mathbb{N}n∈N and write Rn\mathbb{R}^nRn for nnn-dimensional Euclidean space with the usual Euclidean norm. The statement quantifies over the following data:

  • a set-valued map X:Rn→2RnX : \mathbb{R}^n \to 2^{\mathbb{R}^n}X:Rn→2Rn, which assigns a subset X(x)⊆RnX(x) \subseteq \mathbb{R}^nX(x)⊆Rn to every point x∈Rnx \in \mathbb{R}^nx∈Rn;
  • a subset F⊆RnF \subseteq \mathbb{R}^nF⊆Rn.

It assumes:

  1. for every x∈Rnx \in \mathbb{R}^nx∈Rn (not only x∈Fx \in Fx∈F), the set X(x)X(x)X(x) is nonempty and compact;
  2. XXX satisfies the bundle predicate IsLipschitzMultifunction(X)\mathrm{IsLipschitzMultifunction}(X)IsLipschitzMultifunction(X);
  3. FFF is nonempty;
  4. FFF is closed in Rn\mathbb{R}^nRn.

Under these hypotheses it asserts the equivalence

FlowInvariant(X,F)  ⟺  for every x∈F:X(x)⊆TF(x),\mathrm{FlowInvariant}(X, F) \iff \text{for every } x \in F:\quad X(x) \subseteq T_F(x),FlowInvariant(X,F)⟺for every x∈F:X(x)⊆TF​(x),

where TF(x)T_F(x)TF​(x) is the bundle's tangentCone(F,x)\mathrm{tangentCone}(F, x)tangentCone(F,x). Three names in this statement are defined in the bundle, not in standard libraries: IsLipschitzMultifunction\mathrm{IsLipschitzMultifunction}IsLipschitzMultifunction, FlowInvariant\mathrm{FlowInvariant}FlowInvariant and tangentCone\mathrm{tangentCone}tangentCone. Their code was not provided with this declaration, so this read-back cannot say what they mean. In particular, it cannot confirm:

  • which Lipschitz-type condition on XXX is assumed (for example, which distance between sets it uses and whether the constant is global or local);
  • what notion of invariance "flow-invariant" means (for example, what kind of trajectories it quantifies over, on what time interval, and from which starting points);
  • which tangent cone TF(x)T_F(x)TF​(x) is used.

The code carries a comment describing these as Clarke's notions. That comment is documentation, not part of the checked statement, and the read-back does not rely on it. Both directions of the equivalence are asserted: the inclusion condition implies flow-invariance, and flow-invariance implies the inclusion condition. The inclusion X(x)⊆TF(x)X(x) \subseteq T_F(x)X(x)⊆TF​(x) is required for each point of FFF individually. Nothing is required of XXX outside FFF except hypotheses 1 and 2, which apply everywhere.

In the degenerate case n=0n = 0n=0, R0\mathbb{R}^0R0 is the one-point space {0}\{0\}{0}. Hypothesis 1 then forces X(0)={0}X(0) = \{0\}X(0)={0}, and hypotheses 3 and 4 force F={0}F = \{0\}F={0}. The statement reduces to FlowInvariant(X,{0})  ⟺  {0}⊆T{0}(0)\mathrm{FlowInvariant}(X, \{0\}) \iff \{0\} \subseteq T_{\{0\}}(0)FlowInvariant(X,{0})⟺{0}⊆T{0}​(0). Whether either side holds, and whether hypothesis 2 is satisfied here at all, depends on the unprovided definitions. The case F=RnF = \mathbb{R}^nF=Rn (the whole space) is allowed. There is no case F=∅F = \emptysetF=∅, because hypothesis 3 excludes it. The hypotheses X(x)≠∅X(x) \neq \emptysetX(x)=∅ and compactness are satisfiable, for example by a constant singleton map. Whether they can be satisfied together with IsLipschitzMultifunction(X)\mathrm{IsLipschitzMultifunction}(X)IsLipschitzMultifunction(X), so that the theorem is not vacuous, cannot be decided without that definition's code. The statement contains no division, natural-number subtraction, integral, supremum or extended-real arithmetic, so none of their default values arise in the displayed statement itself. Any such values inside the three bundle definitions cannot be assessed here.

Human review
  • Endorsed by Shuze Chen · Sep 28, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Sep 28, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me