Alphabeta Math
TheoremStatement: AI-adaptedProof: AI-adaptedPipeline-generatedprecheck passjudge pass (gpt-6-luna)
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.

A validated rational residual-tube certificate encloses an exact ODE solution

Statement

If Finite checker for rational residual-tube certificates returns certified on a finite rational residual-tube certificate for the rational ODE field f and initial value y0, then the IVP has a unique exact solution y on the entire interval [a,T]. On every panel, ∥y(t)−Uj(t)∥2<Rj(t),y(t)∈Xj, where Uj, Rj, and Xj are the certified reconstruction, positive radius, and state box. The residual is evaluated from the stored polynomial Uj itself, so any interpolation, stage, or storage error already present in those coefficients is included.

Facts & Assumptions

Given: A finite rational certificate and a certified return from the checker in Finite checker for rational residual-tube certificates.

[F1]

Finite rational residual-tube certificate for an ODE trajectory: a certificate stores exact rational panels, polynomial reconstructions, boxes, radius functions, residual envelopes, and strict supersolution bounds.

[F2]

Outward-rounded operations are inclusion-isotone and enclose DAG evaluations: every finite interval DAG enclosure contains each exact value assigned to its nodes; nonfinite statuses assert no finite enclosure.

[F3]

Symbolic derivative recursion for a rational expression DAG is correct: the rational field DAG is C1 where its divisions are nonzero, its formal derivatives are correct, and it is locally Lipschitz there.

[F4]

Choice-free residual error bound for piecewise-C1 ODE reconstructions: for a piecewise-C1 reconstruction, piecewise-continuous one-sided Lipschitz coefficient ℓ, and residual bound ρ, the exact solution error is at most its integrating-factor residual bound.

[F5]

A maximal ODE solution cannot remain in a compact tube to a finite endpoint: a maximal solution cannot end at a finite right endpoint while its graph remains in a compact subset of the open ODE domain.

[F8]

Every Picard–Lindelöf initial value problem has one maximal solution on an open interval: a locally Lipschitz IVP has a unique maximal solution through its initial data.

[F10]

The second fundamental theorem: if G is differentiable on [a,b] with G′=f and f is integrable, then ∫abf=G(b)−G(a): for a C1 function on a closed interval, with one-sided endpoint derivatives, integrating its continuous derivative recovers the endpoint increment.

[F12]

The Euclidean inner product ⟨x,y⟩=∑k<nxkyk on Rn: the Euclidean inner product is bilinear, and ∥w∥22=⟨w,w⟩.

[F13]

Squaring is monotone on the nonnegatives: for nonnegative r,s, r2≤s2 implies r≤s.

Proof

1.1F1F2F12F13given

A certified return means each interval operation was finite and every listed inclusion and strict inequality passed. By [F2], all exact rational-DAG values and their derivative-DAG values lie in the checked intervals. Thus on each panel f is defined on [tj,tj+1]×Xj. The residual interval tests bound each component's absolute value by its checked supremum, whose squared sum is at most ρj2; [F12] and [F13] give ∥Uj′−f(t,Uj)∥2≤ρj. Define ∥A∥F:=(∑i,kAik2)1/2; the Jacobian interval tests bound this sum by Lj2, so [F13] gives ∥Dyf∥F≤Lj. The checker also proves Uj(t)+[−Rj(t),Rj(t)]d⊆Xj. Its exact initial squared comparison and [F12]–[F13] give ∥U0(a)−y0∥2≤d0<R0(a). For each coordinate, ∣U0(a)i−(y0)i∣2≤∑k∣U0(a)k−(y0)k∣2=∥U0(a)−y0∥22, so [F13] places y0 in U0(a)+[−R0(a),R0(a)]d⊆X0. Hence (a,y0) lies in the checked open ODE domain.

2.1F3F4F6F7F12step 1.1algebra

Fix w=u−v and define ϕ(s)=⟨w,f(t,v+sw)⟩ for 0≤s≤1. The segment stays in Xj by convexity. By [F3] and bilinearity in [F12], ϕ′(s)=⟨w,Dyf(t,v+sw)w⟩. The bound in step 1.1 and rowwise Cauchy–Schwarz give ∥Dyf w∥2≤∥Dyf∥F∥w∥2≤Lj∥w∥2, and [F7] then gives ∣ϕ′(s)∣≤Lj∥w∥22. [F6] on [0,1] yields ϕ(1)−ϕ(0)=⟨u−v,f(t,u)−f(t,v)⟩≤Lj∥u−v∥22. Thus the one-sided Lipschitz hypothesis in [F4] holds with ℓ=Lj whenever the exact and reconstructed states lie in the certified box.

3.1F1F3F4F8F10F11F13step 1.1step 2.1algebra

Picard–Lindelöf applies locally by [F3], and [F8] gives a unique maximal solution through (a,y0). On any interval where it exists and lies in the checked panel boxes, apply [F4] panel by panel. Write δj(t):=Rj′(t)−LjRj(t)−ρj, which is strictly positive by the checker. Integrating the scalar identity Rj′=LjRj+ρj+δj with the same integrating factor shows that the radius dominates the error bound strictly: the initial gap is positive because d0<R0(a), and each panel adds the positive integral of e∫stLjδj(s), using positivity of the exponential [F11] and the fundamental theorem [F10]. For e=y−Uj, each coordinate satisfies ∣ei∣2≤∑k∣ek∣2=∥e∥22, so [F13] gives ∣ei∣≤∥e∥2; hence ∥e∥2<Rj(t) implies y(t)∈Uj(t)+[−Rj(t),Rj(t)]d⊆Xj. The comparison therefore applies up to any first exit and gives a strict radius bound there, ruling out that exit. At each panel endpoint the stored reconstructions and radii agree, so the strict inequality continues across the finite partition.

4.1F3F5F9step 3.1step 1.1∎

If the maximal solution ended before or at T, the no-first-exit conclusion in step 3.1 would keep its graph inside the finite union of the closed bounded boxes [tj,tj+1]×Xj, which is compact by [F9] and lies in the rational field's open domain. All divisions stay away from zero there by step 1.1. This contradicts [F5], so the maximal solution reaches T. Applying the panelwise error estimate on the whole interval gives ∥y(t)−Uj(t)∥2<Rj(t) and hence y(t)∈Xj for every panel time. Uniqueness follows from local Lipschitz continuity in [F3] and the local uniqueness clause of Picard–Lindelöf.

Depends on

Used by

Dependency tree · two levels

94 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources