Alphabeta Math
TheoremStatement: AI-adaptedProof: AI-adaptedPipeline-generated
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

Compact continuation for locally Lipschitz ODEs: a solution whose graph stays in a compact subset of the domain near a finite endpoint extends past it

Statement

Let n≥1, let D⊆R×Rn be open, and let F:D→Rn be continuous and locally Lipschitz in the state variable, locally uniformly in time, in the sense of Local Lipschitz continuity in the state variable, locally uniform in time and parameters. Let t0<b be reals, let x0∈Rn with (t0,x0)∈D, and let x:[t0,b)→Rn be a solution of the initial value problem

x′(t)=F(t,x(t)),x(t0)=x0

on [t0,b) (First-order systems, initial value problems, and solutions on intervals). Suppose there is a compact set K⊆D and a real δ>0 such that

(t,x(t))∈Kfor every t∈(b−δ,b)∩[t0,b).

Then:

  1. there is ε>0 and a solution x~:[t0,b+ε)→Rn of the same initial value problem with x~(t)=x(t) for all t∈[t0,b); in particular x~(b)=lim⁡t↑bx(t) exists and lies in the state projection of K;
  2. if moreover D=R×U for an open U⊆Rn and F is autonomous, F(t,z)=G(z) with G:U→Rn continuous and locally Lipschitz in the state variable (Local Lipschitz continuity in the state variable, locally uniform in time and parameters), and if there is a compact K0⊆U with x(t)∈K0 for every t∈(b−δ,b)∩[t0,b), then the extension conclusion of clause 1 holds as well.

Facts & Assumptions

Given: The data of the statement: open D⊆R×Rn, a continuous F that is locally Lipschitz in the state variable in the sense of [L2], a solution x of the initial value problem on [t0,b), and a compact K⊆D with (t,x(t))∈K for all t∈(b−δ,b)∩[t0,b).

[L1]

At every point (t∗,y∗)∈D, the initial value problem has a unique local solution on some interval around t∗; any two solutions through the same initial data agree on every common subinterval containing the initial time (Picard-Lindelöf local existence and uniqueness for first-order systems, First-order systems, initial value problems, and solutions on intervals).

[L2]

In the cylinder terminology of Local Lipschitz continuity in the state variable, locally uniform in time and parameters, this item uses C=[a,c]×B‾(z,r)⊆D, where a,c∈R, a≤c, z∈Rn, r>0, and B‾(z,r)={w∈Rn:∥w−z∥2≤r} (Open ball, closed ball and sphere in a metric space). F is locally Lipschitz in the state variable when every such cylinder carries a finite L with ∥F(t,u)−F(t,w)∥2≤L∥u−w∥2 for (t,u),(t,w)∈C (Local Lipschitz continuity in the state variable, locally uniform in time and parameters).

[L4]

For s<t and an integrable curve g with values in Rn one has ∥∫stg∥2≤∫st∥g∥2, and for a real integrable u≤M on [s,t] one has ∫stu≤M(t−s) (For a≤b and f:[a,b]→Rm integrable when a<b, ∥∫abf∥2≤∫ab∥f∥2; for a<b, ∥f∥2 is integrable, If f≤g on [a,b] and both are integrable then ∫abf≤∫abg; and m(b−a)≤∫abf≤M(b−a)).

[L5]

The Euclidean norm is a norm on Rn, so it satisfies the triangle inequality ∥u+w∥2≤∥u∥2+∥w∥2 (Cauchy-Schwarz ∣⟨x,y⟩∣≤∥x∥2∥y∥2 with its equality case, the triangle inequality for ∥⋅∥2, the parallelogram law and polarisation) and the reverse triangle inequality ∣∥u∥2−∥w∥2∣≤∥u−w∥2 (The reverse triangle inequality in a normed space); moreover (Rn,d2) is complete: every Cauchy sequence in Rn converges (R and Rn for n≥1 with the Euclidean metric are complete, componentwise from the Cauchy criterion in R, Complete metric space: every Cauchy sequence converges in the space, Vector-valued functions f:A→Rm, their limits and continuity, with the dictionary to the metric notions).

[L8]

A continuous real-valued function on a closed interval [s,t], s<t, is integrable; hence a continuous curve into Rn on [s,t] is integrable componentwise (A continuous function on [a,b] is Riemann integrable, by Heine-Cantor and Riemann's criterion, The derivative and the Riemann integral of a vector-valued function: an intrinsic derivative and a componentwise integral).

Proof

technique · direct
1.1givenL5L6L7

If δ>b−t0 replace δ by b−t0; then the interval (b−δ,b)∩[t0,b) is nonempty. Choose any t in it; the hypothesis gives (t,x(t))∈K, so K is nonempty. The norm function z↦∥F(z)∥2 is continuous on the compact metric subspace K: given z0∈K and ε>0, continuity of F at z0 [L7] gives δ0>0 with ∥F(z)−F(z0)∥2<ε whenever d2(z,z0)<δ0, and the reverse triangle inequality [L5] then gives ∣∥F(z)∥2−∥F(z0)∥2∣≤∥F(z)−F(z0)∥2<ε. Hence [L6] gives M:=max⁡{∥F(z)∥2:z∈K}<+∞.

2.1step 1.1givenL3L4L7L8

For s,t with b−δ<s<t<b: since x solves the initial value problem, [L3] gives x(t)=x0+∫t0tF(u,x(u)) du and x(s)=x0+∫t0sF(u,x(u)) du, hence x(t)−x(s)=∫stF(u,x(u)) du; the integrand is continuous in u on [s,t] because u↦F(u,x(u)) is a composition of continuous maps (the components of x are continuous by differentiability, [L7]) and (u,x(u))∈K on this interval, so its components are integrable on [s,t] by [L8]. By [L4], ∥x(t)−x(s)∥2≤∫st∥F(u,x(u))∥2 du≤M(t−s).

3.1step 2.1L5L7algebra

The limit xb:=lim⁡t↑bx(t) exists. Indeed, choose k0≥1 with 1/k0<min⁡{δ,b−t0} [L7] and put yk:=x(b−1/k) for k≥k0; by step 2.1, ∥yk−yl∥2≤M ∣1/k−1/l∣ for all k,l≥k0, so (yk) is a Cauchy sequence in Rn and converges, by [L5], to some xb∈Rn. Given η>0, choose k≥k0 with 1/k<η/(4(M+1)) and ∥yk−xb∥2<η/2; then for every t∈(max⁡{b−δ,b−η/(4(M+1))},b), step 2.1 gives ∥x(t)−xb∥2≤∥x(t)−yk∥2+∥yk−xb∥2≤M(∣t−b∣+1/k)+η/2<M⋅η/(2(M+1))+η/2<η. Hence lim⁡t↑bx(t)=xb.

4.1step 3.1L6

The point (b,xb) lies in K, hence in D: for t∈(b−δ,b)∩[t0,b) the points (t,x(t)) lie in K and converge to (b,xb) as t↑b (in the product metric, the first coordinate trivially and the second by step 3.1); since K is closed [L6], its complement is open, so a point outside K would have a ball around it disjoint from K, which is impossible for (b,xb) because every such ball contains some (t,x(t)).

5.1step 2.1step 3.1step 4.1givenL3L4L6L8algebra

Extend x to the continuous curve xˉ:[t0,b]→Rn by xˉ(t):=x(t) for t<b and xˉ(b):=xb; it is continuous by step 3.1. Let 0<h0<δ. For s∈(b−h0,b) and t∈(s,b), step 2.1 and additivity of the integral [L3] give (x(t)−x(s))−∫sbF(u,xˉ(u)) du=−∫tbF(u,xˉ(u)) du; since (u,xˉ(u))∈K for every u∈[s,b] (for u<b by hypothesis, for u=b by step 4.1), one has ∥F(u,xˉ(u))∥2≤M there, and the norm inequality [L4] gives ∥(x(t)−x(s))−∫sbF(u,xˉ(u)) du∥2≤∫tb∥F(u,xˉ(u))∥2 du≤M(b−t)→0 as t↑b. With x(t)→xb (step 3.1) this yields xb−x(s)=∫sbF(u,xˉ(u)) du. Hence the left difference quotient satisfies ∥xˉ(b)−xˉ(s)b−s−F(b,xb)∥2≤sup⁡s≤u≤b∥F(u,xˉ(u))−F(b,xb)∥2⟶0(s↑b) because for every η>0 continuity of xˉ at b and of F at (b,xb) makes the displayed integrand smaller than η for all u∈[s,b] once s is sufficiently close to b; [L6] ensures each supremum is finite. So xˉ is differentiable at every point of [b−h0,b] as a function on that interval, with xˉ′(t)=F(t,xˉ(t)) for t<b and left derivative F(b,xb) at t=b; it is therefore, for every such h0, a solution of the initial value problem z′=F(t,z), z(b)=xb on [b−h0,b] in the sense of [L1] and [L3].

5.2step 4.1L1L2

Apply the local existence conclusion of Picard-Lindelöf [L1] at (b,xb)∈D, whose hypotheses on F are the assumed continuity and local Lipschitz continuity in the state variable [L2]. It gives a solution through (b,xb) on some interval around b; choose h>0 so [b−h,b+h] is contained in that interval and denote the restricted solution by y. Then y is a solution on [b−h,b+h] through (b,xb).

6.1step 5.2step 5.1L1algebra

Let h′:=min⁡{h,δ/2}>0, where h is the Picard-Lindelöf half-length of step 5.2; by step 5.1 the curve xˉ is a solution on [b−h′,b] through (b,xb), and the restriction of y is a solution on [b−h,b]⊇[b−h′,b] through the same point; by the uniqueness clause of [L1] they agree on [b−h′,b]. Define x~:[t0,b+h)→Rn by x~(t):=xˉ(t) for t≤b and x~(t):=y(t) for b<t<b+h. Then x~ is continuous, and it is differentiable with x~′(t)=F(t,x~(t)) at every t∈[t0,b+h): for t<b and for t>b this is the respective solution property, while at t=b the left difference quotient tends to F(b,xb) by step 5.1 and the right difference quotient is the two-sided derivative y′(b)=F(b,xb) of the solution y; the two agree, so the derivative of x~ at b exists and equals F(b,xb)=F(b,x~(b)). With ε:=h this is an extension as asserted in clause 1.

7.1step 1.1step 2.1step 3.1step 4.1step 5.2step 5.1step 6.1givenL6algebra∎

Under the hypotheses of clause 2, the argument of steps 1.1-6.1 applies with K replaced by the given compact K0⊆U: the bound of step 1.1 is M:=max⁡{∥G(z)∥2:z∈K0}<+∞ by [L6], the Lipschitz estimate of step 2.1 becomes ∥x(t)−x(s)∥2≤M(t−s) because F(u,x(u))=G(x(u)) and x(u)∈K0 for u∈(b−δ,b)∩[t0,b), the limit point of step 4.1 lies in K0 since K0 is compact hence closed [L6], and (b,xb)∈R×U=D because xb∈K0⊆U; the local solution of step 5.2 and the gluing of step 6.1 are then carried out for the same field F(t,z)=G(z) at (b,xb). This proves clause 2 with the same ε=h.

Depends on

Dependency tree · 2 levels

Sources