Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-generatedPipeline-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.

PCHIP slopes are second-order accurate on uniform nonstationary grids

Statement

Let a<b, let f be C3 on an open interval containing [a,b], and suppose ∣f′(x)∣≥c>0 for every x∈[a,b]. For an integer N≥2, put h=(b−a)/N>0 and xj=a+jh for j=0,…,N. Let mj be the slopes selected by PCHIP slope selection from the values f(xj). There are constants C,h0>0, depending on f, c, and [a,b], such that for every such grid with h<h0, max⁡0≤j≤N∣mj−f′(xj)∣≤Ch2.

Facts & Assumptions

Given: Reals a<b, a C3 real function on an open neighborhood of [a,b], a positive lower bound for ∣f′∣ on [a,b], an integer N≥2, and the uniform grid with h=(b−a)/N>0.

[F1]

PCHIP slope selection: interior slopes use the weighted harmonic formula when neighboring secants have the same nonzero sign; the one-sided endpoint estimate is ((2h0+h1)d0−h0d1)/(h0+h1), followed by a sign clip and an opposite-sign cap.

[F2]

PCHIP slope selection has no zero denominator: for admissible inputs every selected slope is a finite real number and every formula denominator is nonzero.

[F3]

Continuity and derivative on a real interval: C3 on an open interval means derivatives through order three exist and are continuous there.

[F4]
[F6]

Taylor's theorem with an integral remainder: if derivatives through order r+1 are continuous, the degree-r Taylor polynomial at c has remainder bounded by M∣x−c∣r+1/(r+1)! when ∣f(r+1)∣≤M on the segment.

[F7]

A continuous real function on a compact subset of R is bounded: a continuous real function on a compact subset of R is bounded there.

[F8]

Heine-Borel by bisection: every closed bounded interval [a,b] is compact: every closed interval [u,v] with u≤v is compact.

[F9]

Every nonempty finite set of reals has a maximum and a minimum: every finite list of real numbers has a maximum.

Proof

1.1F1F3F4F5F7F8given

By [F3], f′ is continuous. It cannot take both a positive and a negative value on [a,b], since [F5] would then give a zero of f′, contradicting ∣f′∣≥c. Thus f′ has one sign throughout. On each grid cell [F4] gives dj=f′(ξj) for some interior ξj, so all secants have that sign and ∣dj∣≥c. Since a<b and h=(b−a)/N>0, the grid is admissible for [F1], and [F8] makes [a,b] compact. Since f′, f′′, and f′′′ are continuous by [F3], [F7] supplies M1,M2,M3≥0 with ∣f′∣≤M1, ∣f′′∣≤M2, and ∣f′′′∣≤M3 on [a,b].

1.2F3F6givenalgebra

Fix an interior node xk and set s=f′(xk) and q=f′′(xk). Applying [F6] with degree two at xk to xk−h and xk+h and dividing the remainders by h gives dk−1=s−hq/2+e− and dk=s+hq/2+e+, where ∣e−∣,∣e+∣≤M3h2/6. Since ∣s∣≥c, for sufficiently small h both secants stay bounded away from zero. The identity 1/(s+u)=1/s−u/s2+u2/(s2(s+u)) then gives, with a uniform constant, 1/dk−1=1/s+hq/(2s2)+O(h2) and 1/dk=1/s−hq/(2s2)+O(h2).

2.1F1F2step 1.2givenalgebra

On a uniform grid the two weights in [F1] are both 3h, so the interior slope equals 2/(1/dk−1+1/dk). The first-order terms in step 1.2 cancel, giving 1/dk−1+1/dk=2/s+O(h2). Since ∣s∣≤M1, the leading term has magnitude at least 2/M1; thus the denominator stays bounded away from zero for sufficiently small h, and its reciprocal yields mk=s+O(h2) with a constant uniform in k.

2.2F1F2F6step 1.1givenalgebra

At the left endpoint, [F6] at a gives d0=f′(a)+hf′′(a)/2+O(M3h2) and d1=f′(a)+3hf′′(a)/2+O(M3h2), by expanding f(a+h) and f(a+2h). Therefore the raw endpoint estimate in [F1] is m^0=(3d0−d1)/2=f′(a)+O(M3h2). For small h it has the sign of f′(a); step 1.1 gives that same sign for d0,d1, so neither the sign clip nor the opposite-sign cap changes it. Reflecting the expansions at b gives the identical O(h2) error for mN.

3.1step 2.1step 2.2F9given∎

The interior and endpoint estimates above use the bounds M1,M2,M3 and c−1 on the fixed interval; their constants and smallness thresholds are therefore uniform in the node index. Taking a common maximum for the error constants and a common positive minimum for the thresholds gives C,h0 valid at every interior node and both endpoints. The finite list ∣mj−f′(xj)∣, 0≤j≤N, has a maximum by [F9], and the uniform pointwise estimate bounds it by Ch2.

Depends on

Used by

Dependency tree · two levels

50 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