Alphabeta Math
LemmaStatement: AI-generatedProof: AI-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.

A local minimum with a gradient inequality is stable

Statement

Let m≥1, let f∈C2(U) on open U⊆Rm, and let z∈U be a local minimum. Assume on a neighborhood of z that ∥∇f(y)∥≥c∣f(y)−f(z)∣ρ for c>0 and 0<ρ<1. Then z is a Lyapunov stable equilibrium for x′=−∇f(x). Here stability is local: for each sufficiently small ball about z, all solutions from a sufficiently smaller ball exist for all forward time and stay in the larger ball; this implies the usual condition for every tolerance.

Facts & Assumptions

[F4]

For a continuous field locally Lipschitz in the state variable, locally uniformly in time as defined in Local Lipschitz continuity in the state variable, locally uniform in time and parameters, Picard-Lindelöf local existence and uniqueness for first-order systems supplies local existence and uniqueness. Every Picard–Lindelöf initial value problem has one maximal solution on an open interval supplies the unique maximal solution and its open time interval.

[F5]

Conditional gradient inequality gives a length bound bounds the length of a downward gradient trajectory on a closed time interval inside a neighborhood where f≥f(z) and the assumed gradient inequality holds. Energy identity and displacement for Euclidean gradient trajectories bounds endpoint displacement by that length, with the closed-interval C1 convention stated there.

[F6]

Stability here uses the local autonomous ODE definition in Equilibria, Lyapunov stability and asymptotic stability of an autonomous first-order system: nearby maximal solutions must exist for every forward time and remain within each test ball compactly contained in the open state domain. The autonomous clause 2 of 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 extends a solution past a finite forward endpoint when its image stays in a compact subset of that domain. Both suppliers have domain mathematics. On a complete flow the resulting tolerance condition agrees with Point stability, attraction and stability of a set.

[F7]

Continuity and derivatives of positive-base real powers gives the derivative βsβ−1 for s>0; The exponential is positive and satisfies exp(-x)=1/exp(x) and the definition of positive-base powers make these powers positive. The exponent, product, quotient, and iterated-power laws for positive real bases and real exponents gives (au)v=auv for a>0. Only positive-base differentiation is used below; the zero-base case uses [F1].

Proof

Given: The data and hypotheses in the Statement.

Goal: Prove local forward existence and Lyapunov stability of z.

1.1F1F2given

Fix ε>0. Choose an open neighborhood W⊆U of z on which f≥f(z) and the gradient inequality both hold, and choose 0<r<ε with B‾(z,r)⊂W. The C2 hypothesis gives continuous partial derivatives, so f is totally differentiable; Fermat's theorem at its interior local minimum gives ∇f(z)=0. Hence z is an equilibrium.

1.2F1F2F3F4

The field −∇f is continuous and locally Lipschitz, as follows directly. On any closed ball K⊂U of positive radius, compactness and extreme values bound all ∣∂j∂if∣ by one M≥0. Each ∂if is totally differentiable by [F2]. For u,v∈K, the chain rule on u+t(v−u) bounds the derivative of ∂if(u+t(v−u)) by M∑j<m∣vj−uj∣≤Mm ∥v−u∥. The scalar mean value theorem on [0,1] gives ∣∂if(v)−∂if(u)∣≤Mm ∥v−u∥, hence ∥∇f(v)−∇f(u)∥≤mM∥v−u∥. These bounds are independent of time for the autonomous field on R×U, so the local and maximal ODE existence and uniqueness results apply.

2.1F1F2F7step 1.1

Put α=1−ρ>0 and η=(cαr/4)1/α>0. On positive bases, the derivative formula, positivity and the mean value theorem make s↦sα strictly increasing; the iterated-power law gives ηα=cαr/4. Continuity of f at z permits 0<δ<r/4 such that ∥y−z∥<δ implies 0≤f(y)−f(z)<η. Thus [f(y)−f(z)]α/(cα)<r/4, including f(y)=f(z) by the zero-base convention.

3.1F1F3F4F5step 1.1step 1.2step 2.1

Let ∥y−z∥<δ and let x be the maximal solution with x(0)=y, with forward interval [0,T), 0<T≤∞. If ∥x(t1)−z∥≥r for some t1<T, the set E={t∈[0,t1]:∥x(t)−z∥≥r} is nonempty and closed by continuity, hence compact. The minimum of the identity function on E gives τ>0. Continuity forces ∥x(τ)−z∥=r: a larger value would give an earlier time in E. On [0,τ] the trajectory stays in B‾(z,r)⊂W, so [F5] bounds its length and hence ∥x(τ)−y∥ by [f(y)−f(z)]α/(cα)<r/4. The triangle inequality gives ∥x(τ)−z∥<r/2, a contradiction. Consequently x(t)∈B(z,r) for 0≤t<T; applying the same length and displacement estimates to each [0,t], 0<t<T, gives ∥x(t)−z∥<r/2, also true at t=0. These restrictions are C1 in the closed-interval convention, since the ODE has continuous right-hand side.

4.1F3F4F6step 1.2step 3.1

If T<∞, the entire forward image lies in K0=B‾(z,r/2)⊂U, which is compact by Heine–Borel. The autonomous compact-continuation theorem applies to this continuous locally Lipschitz field and extends the solution past T, contradicting maximality. Therefore T=∞.

5.1F1F6step 1.1step 3.1step 4.1∎

For every ε>0 we have obtained δ>0 for which every maximal solution starting within δ of z exists for all forward time and satisfies ∥x(t)−z∥<r/2<ε. In particular this proves every test-ball condition of the local ODE stability definition in [F6], and hence Lyapunov stability. If a complete flow is available, it also proves its usual metric tolerance condition. No attraction to z is asserted.

Depends on

Used by · 1

Dependency tree · 2 levels

Each result is drawn at its shortest distance from this item. Arrows run from prerequisites to the items that use them; this item has a heavier outline. External source nodes link directly to the cited text. The view shows up to two dependency steps; open a prerequisite to continue.