Alphabeta Math
ExampleConstruction: AI-generatedVerification: AI-generatedPipeline-generatedprecheck pass
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 two-node semilinear diffusion FAS cycle has an explicit local rate

Statement

On Ω=(0,3) with homogeneous endpoint values, the h=1 centered finite-difference discretization of the semilinear equation −u′′+u3=0 has fine residual F(u)=Au+(u13,u23),A=(2−1−12). Use P=(1,1)T, Iu=(u1+u2)/2, R=PT, the Galerkin coarse residual FH(v)=PTF(Pv), one pre- and one post-sweep S(u)=u−14F(u), and the exact selected coarse root. The unique fine root is u∗=0. The actual exact-branch FAS two-grid map is a contraction toward u∗ on B‾2(0,1/100) with error factor at most 27/400.

Facts & Assumptions

Given: The finite-difference model, transfer maps, Galerkin coarse operator, exact coarse branch, one pre/post update S, and exact real arithmetic specified in the Statement. Matrix and linear-map operator norms are induced by the Euclidean norm; for a bilinear map B, ∥B∥op:=sup⁡∥v∥2=∥w∥2=1∥B[v,w]∥2.

[F1]

Finite-budget nonlinear FAS two-grid cycle: the coarse target is FH(Iu)+R(−F(u)) and the corrected state is u+P(v−Iu).

[F2]

Derivative of the exact-branch FAS two-grid cycle: with invertible coarse Jacobian, the derivative of the actual two-grid map is the composition of the pre-smoother derivative, I−PDFH(Iu∗)−1RDF(u∗), and the post-smoother derivative.

[F3]

A smooth FAS cycle converges locally under its actual derivative bounds: if the actual cycle fixes the root, q=∥DE(u∗)∥<1, and DE is L-Lipschitz on a radius-r convex ball with q+Lr/2<1, then the cycle maps the ball into itself and contracts errors toward that root by q+Lr/2.

[F4]

On a convex open set, a uniform bound ∥Df(z)v∥2≤M∥v∥2 implies ∥f(y)−f(x)∥2≤M∥y−x∥2: on a convex open Euclidean set, a uniform bound on the total derivative of a map implies the corresponding Lipschitz bound between points of that set.

[F5]

Second derivative formula for a composed cycle map: if E=T∘C∘S, y=S(x), and z=C(y), then D2E(x)[h,k]=D2T(z)[DC(y)DS(x)h,DC(y)DS(x)k]+DT(z)(D2C(y)[DS(x)h,DS(x)k]+DC(y)D2S(x)[h,k]).

Example

Given: The mesh nodes 0,1,2,3, endpoint values u0=u3=0, the fine operator and transfers in the Statement, and exact coarse root selection.

Verification

1.1givenalgebra

The continuous model is −u′′+u3=0 on (0,3) with zero endpoint values, and this proof concerns the specified finite-difference residual. The vectors (1,1)T and (1,−1)T are eigenvectors of A with eigenvalues 1 and 3, so zTAz≥∥z∥22. For real s,t, (s−t)(s3−t3)=(s−t)2(s2+st+t2)≥0, since s2+st+t2=((2s+t)2+3t2)/4. Therefore, for a,b∈R2, ⟨a−b,F(a)−F(b)⟩=(a−b)TA(a−b)+∑i=12(ai−bi)(ai3−bi3)≥∥a−b∥22. Since F(0)=0, any other root a would give 0=⟨a,F(a)−F(0)⟩≥∥a∥22, so u∗=0 is the unique root. Direct differentiation gives DF(0)=A.

1.2givenF1constructalgebra

Direct multiplication gives AP=P, PTAP=2, and hence FH(v)=PTF(Pv)=2v+2v3. Writing m=Iu=(u1+u2)/2, [F1] gives the coarse target g(u)=FH(m)−PTF(u)=2m3−u13−u23. The scalar map ϕ(v)=2v+2v3 has ϕ′(v)=2+6v2≥2 and limits ±∞ at v→±∞, so for every g it has one exact root. An exact direct solver is v(g)=23sinh⁡ ⁣(13arsinh⁡ ⁣(334g)), because sinh⁡(3t)=3sinh⁡(t)+4sinh⁡3(t); thus ϕ(v(g))=g. The displayed formula is a composition of smooth real functions, so the selected branch, correction map C, and cycle map E are C2 on an open neighborhood of the ball. At g=0 the branch has v=0 and ϕ′(0)=2.

1.3F2algebra

The pre/post map has derivative at zero Bs=DS(0)=I2−14A=(1/21/41/41/2). The coarse-branch equation has right side g(u), a cubic with Dg(0)=0. Differentiating ϕ(v(u))=g(u) at zero gives 2Dv(0)=Dg(0)=0. The derivative of the coarse correction is Q=I2−PI=I2−12(1111), the orthogonal projection onto (1,−1). The symmetric vector (1,1) is an eigenvector of Bs with eigenvalue 3/4, and (1,−1) has eigenvalue 1/4. Thus for E=S∘C∘S, [F2] gives DE(0)=BsQBs=116Q,q=∥DE(0)∥2=116.

1.4givenalgebra

It remains to bound the derivative variation. Set r=1/100 and m(u)=(u1+u2)/2. On ∥u∥2≤r, ∣m∣≤r/2,∣g(u)∣<2r3,∥Dg(u)∥2<6r2,∥D2g(u)∥op<11r. The first inequality is Cauchy--Schwarz. The second follows from g=2m3−u13−u23 and ∣u1∣3+∣u2∣3≤∥u∥23. Differentiating gives Dg=6m2Dm−3(u12,u22) and D2g=12m(Dm⊗Dm)−diag⁡(6u1,6u2), which give the last two bounds. If ϕ(v(u))=g(u), then ∣v∣≤∣g∣/2<r3, ∥Dv∥≤∥Dg∥/2<3r2, and twice differentiating yields ∥D2v∥<12(11r+12r3(3r2)2)=5.5r+54r7. Writing C(u)=Qu+Pv(u), it follows that ∥DC∥<1.001 and ∥D2C∥<0.08 on this ball.

2.1F4F5step 1.4algebra

Since ∥Bs∥2=3/4, ∥S(u)∥2≤3r/4+r3/4<r for u in the radius-r ball. Orthogonality of Qu and Pv(u) gives ∥C(S(u))∥22≤r2+2r6<4r2. On the radius-2r ball, ∥DS∥2<0.751 and ∥D2S∥op≤0.03; on the radius-r ball, ∥D2S∥op≤0.015. The formula [F5] for E=S∘C∘S therefore gives ∥D2E(u)∥op≤0.03(1.001⋅0.751)2+0.751(0.08⋅0.7512+1.001⋅0.015)<0.07<1. For each unit w∈R2, apply [F4] on the open radius-r ball to the map u↦DE(u)w; the Hessian bound gives derivative norm less than 1. By continuity the resulting Lipschitz bound extends to the closed ball, so ∥DE(x)−DE(y)∥op≤∥x−y∥2 there.

3.1givenF1F3step 1.2step 1.3step 2.1algebra∎

The pre/post map S fixes zero because F(0)=0. At u=0, [F1] gives restricted value Iu=0 and coarse target FH(0)=0; by step 1.2, the exact selected coarse root is v=0, so the correction is zero and E(0)=0. By steps 1.3 and 2.1, q=1/16, L=1, and r=1/100, hence q+Lr/2=116+1200=27400<1. Applying [F3] proves invariance of the ball and the stated local error rate. This single fixed mesh example makes no mesh-independent claim.

Remarks

The diffusion matrix is the standard second-difference Dirichlet matrix for −u′′ on the four nodes, and the cubic is a monotone semilinear reaction. The monotonicity and all FAS convergence hypotheses used here are checked directly in finite dimensions; no parabolic compactness, finite-element projector stability, or continuum convergence theorem is imported.

Depends on

Used by

Nothing in the library uses this result yet.

Dependency tree · two levels

16 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