Wiki
Wiki

Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.

Updated


Statement

Setting (Section 4, printed p. 302): α\alpha and β\beta are complex numbers such that (α+β)2(\alpha+\beta)^2 and αβ\alpha\beta are nonzero integers and α/β\alpha/\beta is not a root of unity, labelled so that ∣α∣≥∣β∣|\alpha|\ge|\beta|. For a positive integer mm, ω(m)\omega(m) is the number of distinct prime factors of mm.

Lemma 4.3 (printed p. 304). Let n>1n>1 be an integer, let pp be a prime not dividing αβ\alpha\beta, and let ℘\wp be a prime ideal of the ring of algebraic integers of Q(α/β)\mathbb Q(\alpha/\beta) that lies above pp and does not ramify. There is a positive number CC, effectively computable in terms of ω(αβ)\omega(\alpha\beta) and the discriminant of Q(α/β)\mathbb Q(\alpha/\beta), such that whenever p>Cp>C,

ord⁡℘ ⁣((αβ)n−1)<pexp⁡ ⁣(−log⁡p51.9log⁡log⁡p)log⁡∣α∣log⁡n.\operatorname{ord}_{\wp}\!\left(\left(\frac{\alpha}{\beta}\right)^{n}-1\right)< p\exp\!\left(-\frac{\log p}{51.9\log\log p}\right)\log|\alpha|\log n .

The paper says this lemma yields a crucial step in the proof of Theorem 1.1 (p. 295). In Section 5 it gives display (5.4): for n>c2n>c_2 and each prime pp dividing Φn(α,β)\Phi_n(\alpha,\beta) other than P(n/(3,n))P(n/(3,n)), it bounds ord⁡℘((α/β)n−1)\operatorname{ord}_{\wp}((\alpha/\beta)^n-1), which by (5.3) bounds ord⁡pΦn(α,β)\operatorname{ord}_p\Phi_n(\alpha,\beta) (p. 310). Theorem 1.2 follows from (6.5) on taking n=p−1n=p-1 in the lemma (p. 311).

Source and proof pointer

Cameron L. Stewart, On divisors of Lucas and Lehmer numbers, Acta Mathematica 211 (2013), 291--314, as identified on the source card. The lemma is stated on printed p. 304 (physical p. 14); its proof runs from p. 304 to p. 309 (physical pp. 14--19). In the arXiv:1008.1274v1 manuscript the same statement is Lemma 8, physical p. 10, with the proof on physical pp. 10--15.

In outline, the proof writes α/β\alpha/\beta as a root of unity times a 2v2^vth power θ2v\theta^{2^v} in Q(α/β)\mathbb Q(\alpha/\beta) with vv maximal, and then applies Yu's lower bound for linear forms in pp-adic logarithms (Section 3) to a linear form in which, as the paper itself remarks on p. 295, the number of terms is deliberately enlarged, to k=⌊log⁡p/(51.8log⁡log⁡p)⌋k=\lfloor\log p/(51.8\log\log p)\rfloor terms (display (4.9)) by bringing in small auxiliary primes. The proof is not transcribed here.

Read depth. Claims checked: the statement, its hypotheses, constants, label and page were read clause by clause on the printed page, together with the standing assumptions of Section 4. The proof was not checked line by line.

Bears on

  • Problem 977: the lemma supplies the bound (5.4) in the proof of Theorem 1.1, whose integer specialization (1.8) with a=2a=2, b=1b=1 gives P(2n−1)/n→∞P(2^n-1)/n\to\infty. The paper notes (p. 296) that Yamada's estimate (1.10) can take the lemma's place and gives the weaker bound (1.11), which also yields the limit.