Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedprecheck passjudge pass (gpt-5.6-terra)audited 2026-09-04
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.

Bounded variation gives one-sided Dirichlet integrability

Statement

Assume the Axiom of Countable Choice.

Let 0<δ<1/2, and let u:[0,δ]→R have bounded variation. Assume that u(0)=0 and lim⁡t↓0u(t)=0. Then

∫0δu(t)sin⁡((2N+1)πt)sin⁡(πt) dt→0as N→∞.

Facts & Assumptions

Given: The Axiom of Countable Choice, a real δ with 0<δ<1/2 and a bounded-variation function u:[0,δ]→R such that u(0)=0 and lim⁡t↓0u(t)=0.

[L1]

Jordan decomposition writes u=u(0)+Pu−Nu with Pu,Nu nondecreasing, normalized by Pu(0)=Nu(0)=0, and minimal among such decompositions (Jordan decomposition for functions of bounded variation).

[L2]

The variation identity is Var⁡[0,η](u)=Pu(η)+Nu(η) for every η∈[0,δ] (The positive and negative variations are nondecreasing and give the Jordan identities).

[L3]

The Dirichlet kernel satisfies DN(t)=sin⁡((2N+1)πt)/sin⁡(πt) for t∉Z (Closed form and size bounds for the Dirichlet kernel).

[L6]

Assuming the Axiom of Countable Choice, Fourier coefficients of an L1(T) function tend to 0 at infinity (Riemann-Lebesgue lemma for Fourier coefficients).

Proof

technique · direct
1.1L1L2givencontradiction

By [L1], write u=Pu−Nu with Pu(0)=Nu(0)=0 and Pu,Nu nondecreasing. Let p:=inf⁡0<t≤δPu(t)=lim⁡t↓0Pu(t),q:=inf⁡0<t≤δNu(t)=lim⁡t↓0Nu(t). Since u(t)→0, one has p−q=0. If p=q>0, define P~(0)=N~(0)=0 and P~(t)=Pu(t)−p, N~(t)=Nu(t)−p for t>0. Then P~,N~ are again nondecreasing, nonnegative, normalized, and satisfy u=P~−N~, but P~(t)<Pu(t) for every t>0, contradicting the minimality in [L1]. Hence p=q=0, so Pu(t)→0,Nu(t)→0,Var⁡[0,t](u)=Pu(t)+Nu(t)→0 as t↓0 by [L2].

1.2L3L4L5algebra

For 0≤t≤δ, define HN(t):=∫0tDN(s) ds. Using [L3] and the change of variables y=(2N+1)πs, HN(t)=∫0(2N+1)πtsin⁡y⋅gN(y) dy, where gN(y):=1(2N+1)πsin⁡(y/(2N+1)). If (2N+1)πt≤1, then 0<y/(2N+1)≤1<2, so [L5] and ∣sin⁡y∣≤y give ∣sin⁡y gN(y)∣≤3/π on [0,(2N+1)πt], hence ∣HN(t)∣≤3/π. If (2N+1)πt>1, the same estimate controls the part on [0,1], while on [1,(2N+1)πt] the function gN is positive and decreasing because 0<y/(2N+1)≤π/2. Bonnet's theorem [L4] therefore gives a point ξ with ∫1(2N+1)πtsin⁡y gN(y) dy=gN(1)∫1ξsin⁡y dy+gN((2N+1)πt)∫ξ(2N+1)πtsin⁡y dy. The absolute value is at most 2gN(1)+2gN((2N+1)πt)≤4gN(1)≤12π by monotonicity of gN and [L5]. Thus ∣HN(t)∣≤15π(0≤t≤δ, N≥0).

2.1L2L4step 1.1step 1.2givenchoosealgebra

Fix ε>0. By step 1.1, choose η∈(0,δ) such that Var⁡[0,η](u)<πε60. Applying [L4] to the monotone functions Pu and Nu on [0,η], and using Pu(0)=Nu(0)=0, step 1.2 yields ∣∫0ηPu(t)DN(t) dt∣≤30πPu(η),∣∫0ηNu(t)DN(t) dt∣≤30πNu(η). Therefore ∣∫0ηu(t)DN(t) dt∣≤30π(Pu(η)+Nu(η))=30πVar⁡[0,η](u)<ε2.

3.1L3step 2.1algebra

On [η,δ], define ψ(t):=1[η,δ](t) u(t)eiπtsin⁡(πt). Because sin⁡(πt) is bounded away from 0 on [η,δ] and u is bounded on the compact interval [η,δ], one has ψ∈L1(T). By [L3], ∫ηδu(t)DN(t) dt=Im⁡∫01ψ(t)e2πiNt dt=Im⁡ ψ^(−N).

4.1L3L6step 2.1step 3.1choosealgebra∎

By [L6], ψ^(−N)→0. So step 3.1 gives ∫ηδu(t)DN(t) dt→0. Choose N0 such that the absolute value of this integral is below ε/2 for every N≥N0. Combining with step 2.1 shows that, for N≥N0, ∣∫0δu(t)DN(t) dt∣<ε. Since DN(t)=sin⁡((2N+1)πt)/sin⁡(πt) on (0,δ] by [L3], the stated limit follows.

Depends on

Used by

Dependency tree · two levels

49 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