Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 2.19 (Peano) — existence of a solution for continuous f on [t₀, t₀ + T₀]

Proved
TeschlODE.IVP.peano_existence

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

existenceordinary-differential-equationsp2o-batch-books5p2o-gran-per-chapterp2o-plan-bookp2o-v1peano

Let f:R×Rn→Rnf : \mathbb{R} \times \mathbb{R}^n \to \mathbb{R}^nf:R×Rn→Rn, t0∈Rt_0 \in \mathbb{R}t0​∈R, x0∈Rnx_0 \in \mathbb{R}^nx0​∈Rn and T,δ>0T, \delta > 0T,δ>0. Suppose fff is continuous on

V=[t0,t0+T]×Bˉδ(x0),Bˉδ(x0)={x:∣x−x0∣≤δ},V = [t_0, t_0 + T] \times \bar B_\delta(x_0), \qquad \bar B_\delta(x_0) = \{x : |x - x_0| \le \delta\},V=[t0​,t0​+T]×Bˉδ​(x0​),Bˉδ​(x0​)={x:∣x−x0​∣≤δ},

and let M=max⁡(t,x)∈V∣f(t,x)∣M = \max_{(t,x) \in V} |f(t,x)|M=max(t,x)∈V​∣f(t,x)∣. Then there exists at least one solution xxx of the initial value problem

x˙=f(t,x),x(t0)=x0(2.10)\dot x = f(t, x), \qquad x(t_0) = x_0 \qquad (2.10)x˙=f(t,x),x(t0​)=x0​(2.10)

on [t0,t0+T0][t_0, t_0 + T_0][t0​,t0​+T0​] which remains in Bˉδ(x0)\bar B_\delta(x_0)Bˉδ​(x0​), where

T0=min⁡{T,δM},δM=∞ if M=0.T_0 = \min\Bigl\{T, \frac{\delta}{M}\Bigr\}, \qquad \frac{\delta}{M} = \infty \text{ if } M = 0 .T0​=min{T,Mδ​},Mδ​=∞ if M=0.

The analogous result holds for the interval [t0−T0,t0][t_0 - T_0, t_0][t0​−T0​,t0​], with V=[t0−T,t0]×Bˉδ(x0)V = [t_0 - T, t_0] \times \bar B_\delta(x_0)V=[t0​−T,t0​]×Bˉδ​(x0​) and MMM the maximum of ∣f∣|f|∣f∣ there.

No Lipschitz condition is assumed, and uniqueness may fail (e.g. x˙=∣x∣\dot x = \sqrt{|x|}x˙=∣x∣​, x(0)=0x(0)=0x(0)=0). This is the book's existence theorem for merely continuous right-hand sides; it is proved by Euler polygons and the Arzelà–Ascoli theorem.

Formalization Note. The book writes the open ball Bδ(x0)B_\delta(x_0)Bδ​(x0​) and "the maximum of ∣f∣|f|∣f∣", and says the solution "remains in Bδ(x0)B_\delta(x_0)Bδ​(x0​)". A maximum over the open ball need not exist, and the solution can reach the sphere at t=t0+T0t = t_0 + T_0t=t0​+T0​ (e.g. n=1n = 1n=1, f≡1f \equiv 1f≡1, δ=1\delta = 1δ=1, T=2T = 2T=2 gives x(t)=x0+(t−t0)x(t) = x_0 + (t - t_0)x(t)=x0​+(t−t0​) with ∣x(t0+1)−x0∣=1|x(t_0 + 1) - x_0| = 1∣x(t0​+1)−x0​∣=1), so the statement uses the closed ball throughout, as the book's own proof does (∣xh(t)−x0∣≤M(t−t0)|x_h(t) - x_0| \le M(t - t_0)∣xh​(t)−x0​∣≤M(t−t0​)). MMM is given as the greatest element of {∣f(p)∣:p∈V}\{|f(p)| : p \in V\}{∣f(p)∣:p∈V}, which exists since VVV is compact and nonempty. T0T_0T0​ is if M = 0 then T else min T (δ / M), the book's convention δ/0=∞\delta/0 = \inftyδ/0=∞. "Solution on [t0,t0+T0][t_0, t_0 + T_0][t0​,t0​+T0​] remaining in Bˉδ(x0)\bar B_\delta(x_0)Bˉδ​(x0​)" means: x(t0)=x0x(t_0) = x_0x(t0​)=x0​ and, for every ttt in the interval, (t,x(t))∈V(t, x(t)) \in V(t,x(t))∈V and xxx has derivative f(t,x(t))f(t, x(t))f(t,x(t)) at ttt relative to the interval (one-sided at the endpoints). The book's V⊂UV \subset UV⊂U is not needed: fff is only required to be continuous on VVV. The forward and backward statements are the two conjuncts, each with its own VVV and MMM.

Preamble
import Mathlib
import Definitions.Def_TeschlODE_IVP_IsSolutionOn
Formal statement
namespace TeschlODE.IVP

theorem peano_existence {n : ℕ}
    (f : ℝ × EuclideanSpace ℝ (Fin n) → EuclideanSpace ℝ (Fin n))
    (t₀ T δ : ℝ) (x₀ : EuclideanSpace ℝ (Fin n)) (hT : 0 < T) (hδ : 0 < δ) :
    -- forward in time, on `V = [t₀, t₀ + T] × B̄_δ(x₀)`
    (ContinuousOn f (Set.Icc t₀ (t₀ + T) ×ˢ Metric.closedBall x₀ δ) →
      ∀ M : ℝ,
      IsGreatest ((fun p => ‖f p‖) '' (Set.Icc t₀ (t₀ + T) ×ˢ Metric.closedBall x₀ δ)) M →
      ∃ x : ℝ → EuclideanSpace ℝ (Fin n), x t₀ = x₀ ∧
        IsSolutionOn (Set.Icc t₀ (t₀ + T) ×ˢ Metric.closedBall x₀ δ) f
          (Set.Icc t₀ (t₀ + (if M = 0 then T else min T (δ / M)))) x) ∧
    -- backward in time, on `V = [t₀ - T, t₀] × B̄_δ(x₀)`
    (ContinuousOn f (Set.Icc (t₀ - T) t₀ ×ˢ Metric.closedBall x₀ δ) →
      ∀ M : ℝ,
      IsGreatest ((fun p => ‖f p‖) '' (Set.Icc (t₀ - T) t₀ ×ˢ Metric.closedBall x₀ δ)) M →
      ∃ x : ℝ → EuclideanSpace ℝ (Fin n), x t₀ = x₀ ∧
        IsSolutionOn (Set.Icc (t₀ - T) t₀ ×ˢ Metric.closedBall x₀ δ) f
          (Set.Icc (t₀ - (if M = 0 then T else min T (δ / M))) t₀) x) := by sorry

end TeschlODE.IVP
Source
Teschl, Ordinary Differential Equations and Dynamical Systems (author's preliminary version of AMS GSM 140, 2012), p. 56, Theorem 2.19
Read-back

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

The statement is a conjunction of two independent claims: a forward-in-time part and a backward-in-time part. Each part has its own continuity hypothesis and its own bound MMM.

Setting. The claims share these data:

  • n∈Nn \in \mathbb{N}n∈N is a dimension, and Rn\mathbb{R}^nRn is the Euclidean space with the usual Euclidean norm ∥⋅∥\|\cdot\|∥⋅∥.
  • f:R×Rn→Rnf : \mathbb{R} \times \mathbb{R}^n \to \mathbb{R}^nf:R×Rn→Rn is an arbitrary function defined everywhere. It takes a pair (t,y)(t, y)(t,y) with time ttt and state yyy. No regularity is assumed globally.
  • t0,T,δ∈Rt_0, T, \delta \in \mathbb{R}t0​,T,δ∈R are real numbers with T>0T > 0T>0 and δ>0\delta > 0δ>0.
  • x0∈Rnx_0 \in \mathbb{R}^nx0​∈Rn is a point.
  • Bˉδ(x0)={y∈Rn:∥y−x0∥≤δ}\bar B_\delta(x_0) = \{ y \in \mathbb{R}^n : \|y - x_0\| \le \delta \}Bˉδ​(x0​)={y∈Rn:∥y−x0​∥≤δ} is the closed Euclidean ball.

The two claims use these rectangles:

V+=[t0, t0+T]×Bˉδ(x0),V−=[t0−T, t0]×Bˉδ(x0).V^{+} = [t_0,\, t_0 + T] \times \bar B_\delta(x_0), \qquad V^{-} = [t_0 - T,\, t_0] \times \bar B_\delta(x_0).V+=[t0​,t0​+T]×Bˉδ​(x0​),V−=[t0​−T,t0​]×Bˉδ​(x0​).

The predicate IsSolutionOn\mathrm{IsSolutionOn}IsSolutionOn. The conclusions use IsSolutionOn(V,f,I,x)\mathrm{IsSolutionOn}(V, f, I, x)IsSolutionOn(V,f,I,x), taken from the imported module Definitions.Def_TeschlODE_IVP_IsSolutionOn. Its definition is not included in the code given here, so its meaning cannot be stated from this artifact. Its arguments are, in order:

  • a set V⊆R×RnV \subseteq \mathbb{R} \times \mathbb{R}^nV⊆R×Rn;
  • the function fff;
  • a set of times I⊆RI \subseteq \mathbb{R}I⊆R;
  • a curve x:R→Rnx : \mathbb{R} \to \mathbb{R}^nx:R→Rn, defined on all of R\mathbb{R}R.

Whatever the statement asserts about xxx solving the equation x′=f(t,x)x' = f(t, x)x′=f(t,x), it asserts only through this predicate.

Forward claim. Suppose fff is continuous on V+V^{+}V+, meaning continuity of the restriction of fff to V+V^{+}V+; nothing is assumed about fff outside V+V^{+}V+. Then for every real MMM that is the greatest element of

{ ∥f(t,y)∥:(t,y)∈V+ },\{\, \|f(t,y)\| : (t,y) \in V^{+} \,\},{∥f(t,y)∥:(t,y)∈V+},

there is a function x:R→Rnx : \mathbb{R} \to \mathbb{R}^nx:R→Rn with:

  • x(t0)=x0x(t_0) = x_0x(t0​)=x0​, and
  • IsSolutionOn(V+,f,[t0, t0+τM], x)\mathrm{IsSolutionOn}\big(V^{+}, f, [t_0,\, t_0 + \tau_M],\, x\big)IsSolutionOn(V+,f,[t0​,t0​+τM​],x),

where

τM={Tif M=0,min⁡ ⁣(T, δ/M)if M≠0.\tau_M = \begin{cases} T & \text{if } M = 0, \\ \min\!\big(T,\ \delta / M\big) & \text{if } M \neq 0. \end{cases}τM​={Tmin(T, δ/M)​if M=0,if M=0.​

"Greatest element" means two things: MMM is attained as ∥f(t,y)∥\|f(t,y)\|∥f(t,y)∥ at some point of V+V^{+}V+, and M≥∥f(t,y)∥M \ge \|f(t,y)\|M≥∥f(t,y)∥ at every point of V+V^{+}V+. So MMM is the maximum of ∥f∥\|f\|∥f∥ on V+V^{+}V+. It is unique, and it is at least 000.

Backward claim. This is the same statement with V−V^{-}V− in place of V+V^{+}V+:

  • continuity of fff on V−V^{-}V− is assumed;
  • MMM is the maximum of ∥f∥\|f\|∥f∥ over V−V^{-}V−, which may differ from the forward MMM;
  • the curve satisfies x(t0)=x0x(t_0) = x_0x(t0​)=x0​ and IsSolutionOn(V−,f,[t0−τM, t0], x)\mathrm{IsSolutionOn}\big(V^{-}, f, [t_0 - \tau_M,\, t_0],\, x\big)IsSolutionOn(V−,f,[t0​−τM​,t0​],x), with τM\tau_MτM​ defined as above.

The curves in the two claims are separate existential witnesses. Nothing says one curve works in both directions.

Degenerate cases.

  • The maximum. Since T>0T > 0T>0 and δ>0\delta > 0δ>0, both rectangles are nonempty. If a maximum MMM exists, then M≥0M \ge 0M≥0.
  • M=0M = 0M=0. This happens exactly when fff vanishes on the whole rectangle. The time interval is then the full [t0,t0+T][t_0, t_0 + T][t0​,t0​+T], or [t0−T,t0][t_0 - T, t_0][t0​−T,t0​] backward. No division by zero occurs, because the case M=0M = 0M=0 is split off before δ/M\delta / Mδ/M is formed.
  • M>0M > 0M>0. Then τM=min⁡(T,δ/M)\tau_M = \min(T, \delta/M)τM​=min(T,δ/M), which lies in (0,T](0, T](0,T], so the interval always has positive length.
  • Maximum not attained. If the set of values ∥f∥\|f\|∥f∥ on the rectangle has no greatest element, the claim for that direction holds vacuously. This happens, for example, when fff is not continuous there. Under the continuity hypothesis the rectangle is compact, so a maximum always exists.
  • n=0n = 0n=0. R0\mathbb{R}^0R0 has a single point 000. Then x0=0x_0 = 0x0​=0, the ball is {0}\{0\}{0}, ∥f∥≡0\|f\| \equiv 0∥f∥≡0, and so M=0M = 0M=0 and τM=T\tau_M = TτM​=T. The curve xxx is the constant 000, and the content reduces to whatever IsSolutionOn\mathrm{IsSolutionOn}IsSolutionOn says in that case.
  • Outside the interval. Nothing in the code constrains xxx at times outside [t0,t0+τM][t_0, t_0 + \tau_M][t0​,t0​+τM​] or [t0−τM,t0][t_0 - \tau_M, t_0][t0​−τM​,t0​], except through the unseen IsSolutionOn\mathrm{IsSolutionOn}IsSolutionOn.
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