Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 8.8 — Liouville's formula for volumes, V˙(t)=∫U(t)div⁡f dx\dot V(t) = \int_{U(t)} \operatorname{div} f\,dxV˙(t)=∫U(t)​divfdx (8.28)

Proved
TeschlODE.HigherDim.volume_hasDerivAt

by mikedeng1 · 1 vote · Sep 28, 2026 · Mathlib 0df444a (Lean v4.33.1)

divergencedynamical-systemsliouville-formulaordinary-differential-equationsp2o-batch-books5p2o-gran-per-chapterp2o-plan-bookp2o-v1

Let f:Rn→Rnf : \mathbb{R}^n \to \mathbb{R}^nf:Rn→Rn be C1C^1C1 and let Φ\PhiΦ be the (local) flow of the dynamical system x˙=f(x)\dot x = f(x)x˙=f(x) on Rn\mathbb{R}^nRn. Let UUU be a bounded open subset of Rn\mathbb{R}^nRn, write U(t)=Φ(t,U)U(t) = \Phi(t, U)U(t)=Φ(t,U) and V(t)=∣U(t)∣=∫U(t)dxV(t) = |U(t)| = \int_{U(t)} dxV(t)=∣U(t)∣=∫U(t)​dx (Lebesgue measure). Let t0t_0t0​ be a time at which every point of the closure U‾\overline UU is still alive, t0∈Ixt_0 \in I_xt0​∈Ix​ for all x∈U‾x \in \overline Ux∈U. Then V(t0)<∞V(t_0) < \inftyV(t0​)<∞, div⁡f\operatorname{div} fdivf is integrable on U(t0)U(t_0)U(t0​), and VVV is differentiable at t0t_0t0​ with

V˙(t0)=∫U(t0)div⁡(f(x)) dx.(8.28)\dot V(t_0) = \int_{U(t_0)} \operatorname{div}(f(x))\,dx. \qquad (8.28)V˙(t0​)=∫U(t0​)​div(f(x))dx.(8.28)

This is the book's generalization of Liouville's formula (3.91) from linear systems to volumes transported by a nonlinear flow. It gives V(t)=Ve−(1+σ+b)tV(t) = V e^{-(1+\sigma+b)t}V(t)=Ve−(1+σ+b)t for the Lorenz equation and, since a Hamiltonian vector field has zero divergence, Liouville's theorem 8.10.

Formalization Note. The book states the lemma for "a dynamical system on Rn\mathbb{R}^nRn" under the chapter's standing assumption that the flow is complete, so that U(t)U(t)U(t) is defined for all ttt. The statement keeps the local flow and requires instead that U‾\overline UU be alive at t0t_0t0​; this is implied by completeness, and it is needed: if only the points of UUU are alive, U(t0)U(t_0)U(t0​) can have infinite volume (for x˙=x2\dot x = x^2x˙=x2, U=(0,1)U = (0, 1)U=(0,1), t0=1t_0 = 1t0​=1). f∈C1f \in C^1f∈C1 is the book's standing regularity (Chapter 6) and what the first variational equation in the proof needs. The two finiteness/integrability conjuncts rule out the junk values of ENNReal.toReal and of the Bochner integral. VVV is read as the real number (volume(Φ(t,U)))(\text{volume}(\Phi(t, U)))(volume(Φ(t,U))).toReal, which is finite for ttt near t0t_0t0​.

Preamble
import Mathlib
import Definitions.Def_TeschlODE_HigherDim_IsIntegralCurve
import Definitions.Def_TeschlODE_HigherDim_IsMaximalFlow
import Definitions.Def_TeschlODE_HigherDim_divergence

open MeasureTheory
Formal statement
namespace TeschlODE.HigherDim

/-- Teschl, Lemma 8.8, p. 236, (8.28): for a `C¹` vector field on `ℝⁿ` with flow `Φ`, a bounded
open `U` and a time `t₀` at which every point of the closure `Ū` is still alive, the volume
`V(t) = |Φ(t, U)|` is finite at `t₀`, `div f` is integrable on `U(t₀) = Φ(t₀, U)`, and
`V̇(t₀) = ∫_{U(t₀)} div f(x) dx`. -/
theorem volume_hasDerivAt {n : ℕ}
    (f : EuclideanSpace ℝ (Fin n) → EuclideanSpace ℝ (Fin n)) (hf : ContDiff ℝ 1 f)
    (I : EuclideanSpace ℝ (Fin n) → Set ℝ)
    (Φ : ℝ → EuclideanSpace ℝ (Fin n) → EuclideanSpace ℝ (Fin n))
    (hΦ : IsMaximalFlow f Set.univ I Φ)
    (U : Set (EuclideanSpace ℝ (Fin n))) (hU : IsOpen U) (hUb : Bornology.IsBounded U)
    (t₀ : ℝ) (ht₀ : ∀ x ∈ closure U, t₀ ∈ I x) :
    volume (Φ t₀ '' U) < ⊤ ∧ IntegrableOn (divergence f) (Φ t₀ '' U) ∧
      HasDerivAt (fun t : ℝ => (volume (Φ t '' U)).toReal)
        (∫ x in Φ t₀ '' U, divergence f x) t₀ := by sorry

end TeschlODE.HigherDim
Source
Teschl, Ordinary Differential Equations and Dynamical Systems (author's preliminary version of AMS GSM 140, 2012), p. 236, Lemma 8.8, Eq. (8.28)
Read-back

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

Inputs. Fix the following:

  • a natural number nnn;
  • a map f:Rn→Rnf : \mathbb{R}^n \to \mathbb{R}^nf:Rn→Rn (Euclidean Rn\mathbb{R}^nRn) that is continuously differentiable (C1C^1C1) on all of Rn\mathbb{R}^nRn;
  • time sets I(x)⊆RI(x) \subseteq \mathbb{R}I(x)⊆R;
  • a map Φ:R×Rn→Rn\Phi : \mathbb{R} \times \mathbb{R}^n \to \mathbb{R}^nΦ:R×Rn→Rn.

Hypothesis on (I,Φ)(I, \Phi)(I,Φ). (I,Φ)(I,\Phi)(I,Φ) is a maximal flow of fff on M=RnM = \mathbb{R}^nM=Rn. This means that for every x∈Rnx \in \mathbb{R}^nx∈Rn all of the following hold:

  • I(x)I(x)I(x) is open and order-connected and contains 000;
  • Φ(0,x)=x\Phi(0,x) = xΦ(0,x)=x;
  • t↦Φ(t,x)t \mapsto \Phi(t,x)t↦Φ(t,x) is differentiable at every t∈I(x)t \in I(x)t∈I(x) with derivative f(Φ(t,x))f(\Phi(t,x))f(Φ(t,x));
  • whenever ψ:R→Rn\psi : \mathbb{R} \to \mathbb{R}^nψ:R→Rn is differentiable with ψ′(t)=f(ψ(t))\psi'(t) = f(\psi(t))ψ′(t)=f(ψ(t)) at every ttt of an open order-connected set J∋0J \ni 0J∋0, and ψ(0)=x\psi(0) = xψ(0)=x, then J⊆I(x)J \subseteq I(x)J⊆I(x) and ψ=Φ(⋅,x)\psi = \Phi(\cdot,x)ψ=Φ(⋅,x) on JJJ.

Further hypotheses.

  • U⊆RnU \subseteq \mathbb{R}^nU⊆Rn is open and bounded.
  • t0t_0t0​ is a real number such that t0∈I(x)t_0 \in I(x)t0​∈I(x) for every xxx in the closure U‾\overline{U}U.

Notation. Write:

  • Φt(U)={Φ(t,x):x∈U}\Phi_t(U) = \{\Phi(t,x) : x \in U\}Φt​(U)={Φ(t,x):x∈U} for the image of UUU at time ttt;
  • λ\lambdaλ for Lebesgue measure on Rn\mathbb{R}^nRn, with values in [0,∞][0,\infty][0,∞];
  • div⁡f(x)=∑i=1n(Df(x)ei)i\operatorname{div} f(x) = \sum_{i=1}^n (Df(x)e_i)_idivf(x)=∑i=1n​(Df(x)ei​)i​ for the trace of the Fréchet derivative of fff at xxx.

Conclusion. All three of the following hold:

  1. λ(Φt0(U))<∞\lambda(\Phi_{t_0}(U)) < \inftyλ(Φt0​​(U))<∞.
  2. div⁡f\operatorname{div} fdivf is Lebesgue-integrable on Φt0(U)\Phi_{t_0}(U)Φt0​​(U).
  3. The real-valued function
V(t)=λ(Φt(U)) converted to a real number,V(t) = \lambda\big(\Phi_t(U)\big) \text{ converted to a real number},V(t)=λ(Φt​(U)) converted to a real number,

where the value ∞\infty∞ is converted to 000, is differentiable at t0t_0t0​ (as a two-sided derivative) with

V′(t0)=∫Φt0(U)div⁡f(x) dx.V'(t_0) = \int_{\Phi_{t_0}(U)} \operatorname{div} f(x)\, dx .V′(t0​)=∫Φt0​​(U)​divf(x)dx.

For ttt near t0t_0t0​, V(t)V(t)V(t) is computed from the values Φ(t,x)\Phi(t,x)Φ(t,x), x∈Ux \in Ux∈U, whatever they are. This includes values at times t∉I(x)t \notin I(x)t∈/I(x), which the flow hypothesis does not constrain. The statement does not assert that the sets Φt(U)\Phi_t(U)Φt​(U) are measurable.

Degenerate cases.

  • If U=∅U = \emptysetU=∅, the hypothesis on t0t_0t0​ is vacuous. The conclusion then reads λ(∅)=0<∞\lambda(\emptyset) = 0 < \inftyλ(∅)=0<∞, integrability over ∅\emptyset∅, and the constant function 000 having derivative 0=∫∅0 = \int_\emptyset0=∫∅​.
  • If n=0n = 0n=0, R0\mathbb{R}^0R0 is a single point of measure 111, div⁡f≡0\operatorname{div} f \equiv 0divf≡0, and UUU is either empty or that point.
Human review
  • Endorsed by Shuze Chen · Oct 2, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 2, 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