Baker–Davenport reduction
TheoremNearest-integer estimate in the Baker–Davenport reduction#✓fully proved
For a real number \(x\) write \(\lVert x \rVert = |x - \operatorname{round}(x)|\) for the distance from \(x\) to a nearest integer. Let \(q\), \(M\), \(u\) be natural numbers with \(u \le M\), let \(v\) be an integer, and let \(\kappa\) and \(\mu\) be real numbers. If \(\varepsilon = \lVert \mu q \rVert - M \lVert \kappa q \rVert\), then \(\varepsilon \le q \, |u\kappa - v + \mu|\).
theorem nearestInteger_reduction_core (q M u : ℕ) (kappa mu : ℝ) (v : ℤ)
(hu : u ≤ M)
(epsilon : ℝ)
(heq : epsilon = |mu * (q : ℝ) - (round (mu * (q : ℝ)) : ℝ)| -
(M : ℝ) * |kappa * (q : ℝ) - (round (kappa * (q : ℝ)) : ℝ)|) :
epsilon ≤ (q : ℝ) * |(u : ℝ) * kappa - (v : ℝ) + mu|
Sources
- Nearest-integer Baker–Davenport reduction core: a named reusable arithmetic prerequisite stage extracted from Lemma (lines 285--292) of Eric F. Bravo, Carlos A. Gómez, and Florian Luca, "Product of Consecutive Tribonacci Numbers With Only One Distinct Digit", Journal of Integer Sequences 22 (2019), Article 19.6.3. https://cs.uwaterloo.ca/journals/JIS/VOL22/Gomez/gomez3.tex. Scope: the printed lemma defines nearest-integer distance by
‖x‖ = |x - round x|andε := ‖μq‖ - M‖κq‖; this stage records the reverse-triangle inequality used by the reduction argument. It does not claim to formalize the full lemma. The source's irrationality, continued-fraction,q > 6M, and logarithmic hypotheses belong to the later wrapper and are deliberately omitted from this arithmetic stage, as are the source's0 < qand0 < ε, which the inequality does not need. The explicitheqequality hypothesis is necessary: a prior default-valued epsilon binder did not assert the definition and made the statement false.
TheoremBaker–Davenport reduction for an arbitrary denominator, absolute-value form#✓fully proved
For a real number \(x\) write \(\lVert x \rVert = |x - \operatorname{round}(x)|\) for the distance from \(x\) to a nearest integer. Let \(M\) and \(q\) be natural numbers with \(q > 0\), and let \(\kappa\), \(\mu\), \(A\), \(B\) be real numbers with \(A > 0\) and \(B > 1\). Put \(\varepsilon = \lVert \mu q \rVert - M \lVert \kappa q \rVert\) and suppose that \(\varepsilon > 0\). Then there are no natural numbers \(r\), \(s\), \(w\) with \(r \le M\), \(w \ge \log(Aq/\varepsilon) / \log B\) and \(|r\kappa - s + \mu| < A B^{-w}\).
theorem bakerDavenport_reduction_absolute_general
(M q : ℕ) (hq : 0 < q)
(kappa A mu B epsilon : ℝ)
(hA : 0 < A) (hB : 1 < B)
(hepsilon : epsilon = |mu * (q : ℝ) - (round (mu * (q : ℝ)) : ℝ)| -
(M : ℝ) * |kappa * (q : ℝ) - (round (kappa * (q : ℝ)) : ℝ)|)
(hepsilonpos : 0 < epsilon) :
¬∃ (r s w : ℕ), r ≤ M ∧
(w : ℝ) ≥ Real.log (A * (q : ℝ) / epsilon) / Real.log B ∧
|(r : ℝ) * kappa - (s : ℝ) + mu| < A * (B⁻¹) ^ w
Uses 1 · Used by 1
TheoremBaker–Davenport reduction lemma, absolute-value form#✓fully proved
For a real number \(x\) write \(\lVert x \rVert = |x - \operatorname{round}(x)|\) for the distance from \(x\) to a nearest integer. Let \(M\) be a positive integer and \(\kappa\) an irrational real number, and let \(q\) be the denominator of the \(j\)-th convergent of the continued fraction of \(\kappa\) for some natural number \(j\), with \(q > 6M\). Let \(A\), \(B\), \(\mu\) be real numbers with \(A > 0\) and \(B > 1\). Put \(\varepsilon = \lVert \mu q \rVert - M \lVert \kappa q \rVert\) and suppose that \(\varepsilon > 0\). Then there are no positive integers \(r\), \(s\), \(w\) with \(r \le M\), \(w \ge \log(Aq/\varepsilon) / \log B\) and \(0 < |r\kappa - s + \mu| < A B^{-w}\).
Of the hypothesis \(q > 6M\) only \(q > 0\) is used. The hypotheses that \(M > 0\), that \(\kappa\) is irrational and that \(q\) is a convergent denominator, and the positivity of \(r\), \(s\), \(w\) and of \(|r\kappa - s + \mu|\), are not needed: the statement follows from the version for an arbitrary positive denominator.
MetaMathlibExt.bakerDavenport_reduction_absolute · 15 lines
theorem bakerDavenport_reduction_absolute
(M : ℕ) (hM : 0 < M)
(kappa : ℝ) (hkappa : Irrational kappa)
(j : ℕ) (q : ℕ)
(hq : q = (kappa.convergent j).den)
(hqM : 6 * M < q)
(A mu B epsilon : ℝ)
(hA : 0 < A) (hB : 1 < B)
(hepsilon : epsilon = |mu * (q : ℝ) - (round (mu * (q : ℝ)) : ℝ)| -
(M : ℝ) * |kappa * (q : ℝ) - (round (kappa * (q : ℝ)) : ℝ)|)
(hepsilonpos : 0 < epsilon) :
¬∃ (r s w : ℕ), 0 < r ∧ 0 < s ∧ 0 < w ∧ r ≤ M ∧
(w : ℝ) ≥ Real.log (A * (q : ℝ) / epsilon) / Real.log B ∧
0 < |(r : ℝ) * kappa - (s : ℝ) + mu| ∧
|(r : ℝ) * kappa - (s : ℝ) + mu| < A * (B⁻¹) ^ w
Sources
- Eric F. Bravo, Carlos A. Gomez, and Florian Luca, "Product of Consecutive Tribonacci Numbers With Only One Distinct Digit," Journal of Integer Sequences 22 (2019), Article 19.6.3, Lemma
bd, lines 285–292. Source URL https://cs.uwaterloo.ca/journals/JIS/VOL22/Gomez/gomez3.tex.
TheoremBaker–Davenport reduction for an arbitrary denominator, one-sided form#✓fully proved
For a real number \(x\) write \(\lVert x \rVert = |x - \operatorname{round}(x)|\) for the distance from \(x\) to a nearest integer. Let \(N\) and \(q\) be natural numbers with \(q > 0\), and let \(\gamma\), \(\mu\), \(A\), \(B\) be real numbers with \(A > 0\) and \(B > 1\). Put \(\xi = \lVert \mu q \rVert - N \lVert \gamma q \rVert\) and suppose that \(\xi > 0\). Then there are no natural numbers \(u\), \(v\), \(w\) with \(u \le N\), \(w \ge \log(Aq/\xi) / \log B\) and \(0 < u\gamma - v + \mu < A B^{-w}\).
theorem bakerDavenport_reduction_oneSided_general
(N q : ℕ) (hq : 0 < q)
(γ A μ B ξ : ℝ)
(hA : 0 < A) (hB : 1 < B)
(hξ : ξ = |μ * (q : ℝ) - (round (μ * (q : ℝ)) : ℝ)| -
(N : ℝ) * |γ * (q : ℝ) - (round (γ * (q : ℝ)) : ℝ)|)
(hξpos : 0 < ξ) :
¬∃ (u v w : ℕ), u ≤ N ∧
(w : ℝ) ≥ Real.log (A * (q : ℝ) / ξ) / Real.log B ∧
0 < (u : ℝ) * γ - (v : ℝ) + μ ∧
(u : ℝ) * γ - (v : ℝ) + μ < A * (B⁻¹) ^ w
Used by 1
TheoremBaker–Davenport reduction lemma, one-sided form#✓fully proved
For a real number \(x\) write \(\lVert x \rVert = |x - \operatorname{round}(x)|\) for the distance from \(x\) to a nearest integer. Let \(N\) be a positive integer and \(\gamma\) an irrational real number, and let \(q\) be the denominator of the \(j\)-th convergent of the continued fraction of \(\gamma\) for some natural number \(j\), with \(q > 6N\). Let \(A\), \(B\), \(\mu\) be real numbers with \(A > 0\) and \(B > 1\). Put \(\xi = \lVert \mu q \rVert - N \lVert \gamma q \rVert\) and suppose that \(\xi > 0\). Then there are no positive integers \(u\), \(v\), \(w\) with \(u \le N\), \(w \ge \log(Aq/\xi) / \log B\) and \(0 < u\gamma - v + \mu < A B^{-w}\).
Of the hypothesis \(q > 6N\) only \(q > 0\) is used. The hypotheses that \(N > 0\), that \(\gamma\) is irrational and that \(q\) is a convergent denominator, and the positivity of \(u\), \(v\) and \(w\), are not needed: the statement follows from the version for an arbitrary positive denominator.
MetaMathlibExt.bakerDavenport_reduction_oneSided · 15 lines
theorem bakerDavenport_reduction_oneSided
(N : ℕ) (hN : 0 < N)
(γ : ℝ) (hγ : Irrational γ)
(j : ℕ) (q : ℕ)
(hq : q = (γ.convergent j).den)
(hqN : 6 * N < q)
(A μ B ξ : ℝ)
(hA : 0 < A) (hB : 1 < B)
(hξ : ξ = |μ * (q : ℝ) - (round (μ * (q : ℝ)) : ℝ)| -
(N : ℝ) * |γ * (q : ℝ) - (round (γ * (q : ℝ)) : ℝ)|)
(hξpos : 0 < ξ) :
¬∃ (u v w : ℕ), 0 < u ∧ 0 < v ∧ 0 < w ∧ u ≤ N ∧
(w : ℝ) ≥ Real.log (A * (q : ℝ) / ξ) / Real.log B ∧
0 < (u : ℝ) * γ - (v : ℝ) + μ ∧
(u : ℝ) * γ - (v : ℝ) + μ < A * (B⁻¹) ^ w
Sources
- Bilizimbéyé Edjeou, Amadou Tall, and Mohamed Ben Fraj Ben Maaouia, "Powers of Two as Sums of Three Lucas Numbers," Journal of Integer Sequences 23 (2020), Article 20.8.8, https://cs.uwaterloo.ca/journals/JIS/VOL23/Edjeou/edj4.tex, Lemma, lines 200–211.
Loading dependency explorer…
The dependency explorer needs JavaScript and its graph data.