Multiplicative number theory · Unrefereed preprint

The odd-part ratio under slow variationA note on the printed Hilberdink problem

Alec Kriebel · ORCID · Version 1.0.1 · 23 September 2026

The limit exists under the hypothesis printed in the original report. Its value follows from a short monotonicity argument.

Read the 3-page paperSource & verificationZenodo upload kit

The result

Let f be a nonzero nonnegative multiplicative function. Let F(x) sum its values over all positive integers up to x, and let F2(x) sum only over odd integers.

If F(2x)/F(x) → 1, then

limx→∞ F2(x)/F(x) = 1 / ∑k≥0 f(2k).

The reciprocal is zero when the series diverges. Bounded summatory functions are included.

Writing every integer uniquely as 2k times an odd integer gives F(x) = ∑k≥0 f(2k) F2(x/2k). If AK is the first K + 1 terms of the local sum, monotonicity gives AKF2(x) ≤ F(2Kx). For fixed K, division by F(x) and slow variation give a limiting upper bound of 1/AK. If the full sum A is finite, F(x) ≤ AF2(x) supplies the matching lower bound.

Why the wording matters

OWR-15956-012 and Hilberdink’s original Problem 3 ask for existence under slow variation: F(λx)/F(x) → 1. The original report then predicts a different, weighted constant. A simple example exposes the conflict:

For f(n) = 1/nValue
Actual odd-part limit1/2
Report’s weighted prediction3/4

Exact scope. This proves the printed limit-existence assertion and corrects its constant. The surrounding formulas suggest a possible intended hypothesis F(λx)/F(x) → λ. This paper does not resolve that different formulation or establish the proposer’s intent.

Easy to check

The analytic proof is the mathematical certificate. The accompanying script checks the finite identities and bounds with exact rational arithmetic, including examples that are multiplicative but not completely multiplicative.

python3 verification/verify.py

Run from the extracted source package with Python 3.9 or later. No extra packages or network access. Normal and optimized runs passed. These finite checks do not constitute a formal proof of the limiting theorem.

Priority and attribution

A bounded literature audit found no exact earlier resolution. Closely related results by Yeats (2002, 2003) have different hypotheses. Some related full texts were inaccessible, and earlier published or folklore observations cannot be excluded. This is a qualified priority finding, not a certification of originality.

The paper was prepared from an AI-generated candidate supplied by the author. OpenAI Codex assisted with drafting, exact checks, and separate AI audits. No external human referee was consulted. The Zenodo kit is ready for manual deposit; no DOI has been assigned by this project.