Documentation

MathlibNt.SieveTheory.LinearSieve.Suzuki.SuzukiEquation1053MinusExclusion

The analytic core preceding Suzuki (10.53) #

This file internalizes the exact calculus and pairing identities used before the asymptotic estimates in (10.47)--(10.53). No derivative sign, adjacent-value estimate, or Equation1053MinusExclusion is assumed.

noncomputable def Section10Equation1053.phiMinus (ξ : ) (c s : ) :

The phase φ₋(s)=∫₁ˢ ξ(t)dt-cs.

Equations
Instances For
    noncomputable def Section10Equation1053.psiMinus (r ξ : ) (c s : ) :

    The exponent ψ₋(s)=φ₋(s)-log r(s+1) in (10.46)--(10.53).

    Equations
    Instances For
      noncomputable def Section10Equation1053.envelopeMinus (R ξ : ) (c s : ) :

      Weighted minus envelope, in a form convenient for the pairing identity.

      Equations
      Instances For
        @[simp]
        theorem Section10Equation1053.exp_neg_psiMinus {r ξ : } {c t : } (hr : 0 < r (t + 1)) :
        Real.exp (-psiMinus r ξ c t) = r (t + 1) * Real.exp (-phiMinus ξ c t)

        The adjoint equation alone gives the exact first derivative of r. For a=2,b=1 this is the non-asymptotic precursor of source (10.41).

        theorem Section10Equation1053.log_adjoint_shift_hasDerivAt {R : } {β : } (h : Section10Lemma1028FirstCrossing.FirstCrossingDDEApparatus R β) {s : } (hs : 0 < s + 1) (hr : 0 < h.adjoint (s + 1)) :
        HasDerivAt (fun (t : ) => Real.log (h.adjoint (t + 1))) ((h.adjoint (s + 1) + h.adjoint (s + 2)) / ((s + 1) * h.adjoint (s + 1))) s

        Exact derivative of log r(s+1), before the asymptotic replacement (10.41).

        theorem Section10Equation1053.psiMinus_hasDerivAt {R ξ : } {β c s : } (h : Section10Lemma1028FirstCrossing.FirstCrossingDDEApparatus R β) ( : Section10Lemma1028.Proposition1020Xi ξ) (hs : 0 < s + 1) (hr : 0 < h.adjoint (s + 1)) :
        HasDerivAt (psiMinus h.adjoint ξ c) (ξ s - c - (h.adjoint (s + 1) + h.adjoint (s + 2)) / ((s + 1) * h.adjoint (s + 1))) s

        Exact version of the first line used in (10.47). The paper next replaces its adjoint quotient by (λ-1)/s + O(1/s²).

        theorem Section10Equation1053.canonicalXi_unit_secant {ξ : } ( : Section10Lemma1028.Proposition1020Xi ξ) {s t : } (hs : 3 s) (ht : t Set.Icc (s - 1) s) :
        0 ξ s - ξ t ξ s - ξ t 2 * (s - t)

        A pointwise Taylor/secant estimate for the canonical phase on a unit interval. This is the part of (10.47) supplied solely by Proposition 10.20.

        theorem Section10Equation1053.exp_neg_integral_by_parts {ψ p q : } {a b : } ( : tSet.uIcc a b, HasDerivAt ψ (p t) t) (hp : tSet.uIcc a b, HasDerivAt p (q t) t) (hp0 : tSet.uIcc a b, p t 0) (hmain : IntervalIntegrable (fun (t : ) => Real.exp (-ψ t)) MeasureTheory.volume a b) (hrem : IntervalIntegrable (fun (t : ) => Real.exp (-ψ t) * q t / p t ^ 2) MeasureTheory.volume a b) :
        (t : ) in a..b, Real.exp (-ψ t) = Real.exp (-ψ a) / p a - Real.exp (-ψ b) / p b - (t : ) in a..b, Real.exp (-ψ t) * q t / p t ^ 2

        Exact integration-by-parts identity underlying (10.53). It deliberately keeps the remainder integral explicit; bounding it is precisely where source (10.41)--(10.42), (10.47)--(10.52), and the asymptotics of Proposition 10.20 enter.

        theorem Section10Equation1053.pairing_zero_weighted_identity {R ξ : } {β c s : } (h : Section10Lemma1028FirstCrossing.FirstCrossingDDEApparatus R β) (hs : β < s) (hr : tSet.Icc (s - 1) s, 0 < h.adjoint (t + 1)) :
        s * h.adjoint s * R s = (t : ) in s - 1..s, envelopeMinus R ξ c t * Real.exp (-psiMinus h.adjoint ξ c t)

        Pairing-zero rewritten in the weighted variables used in (10.54), with no monotonicity hypothesis.