A Conditional Endpoint Exponent \(1-\varepsilon\) for Large Prime Factors of Shifted Primes
This mathematical note establishes a conditional consequence concerning the largest prime factor \(P^+(p-1)\) of shifted primes. Assuming the controlled-predecessor proposition stated in OpenAI's October 4, 2026 manuscript, "Primitive roots for every admissible integer base", we show that, for every fixed \(\varepsilon>0\), infinitely many primes \(p\) satisfy \(P^+(p-1)>p^{1-\varepsilon}\). More precisely, the hypothesis implies the quantitative lower bound \(\#\{p\in(x,2x):p\text{ prime},\ P^+(p-1)>p^{1-\varepsilon}\}\gg_{\varepsilon}\frac{x}{(\log x)^2}\). Consequently, \(\limsup_{\substack{p\to\infty\\p\text{ prime}}}\frac{\log P^+(p-1)}{\log p}=1\). The proof uses the factorization \(p-1=4rQ\), where \(Q>x^{0.9}\) is prime, together with restrictions on the prime factors of the complementary cofactor \(r\). A harmonic-tail estimate shows that large cofactors contribute negligibly, leaving a quantitatively large family of primes with \(r\le x^\eta\) for every fixed \(\eta>0\). This yields the endpoint exponent \(1-\varepsilon\), conditional on the stated hypothesis. The result concerns the classical problem of large prime factors of shifted primes, associated with earlier work of Fouvry, Baker--Harman, and subsequent improvements. Status and originality: This is a conditional mathematical exposition, not an independently established unconditional improvement. The endpoint deduction was previously presented in an October 23, 2026 mathematical exposition. The underlying OpenAI controlled-predecessor proposition has not been independently verified in this note. The document includes the conditional proof, mathematical attribution, references, and AI authorship disclosure.
Authors
- dr vengy
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-11-19
- DOI
- https://doi.org/10.5281/zenodo.24369176
- Primary Topic
- Analytic Number Theory Research
- Type
- article
- Field-Weighted Citation Impact
- 0.00