OAI/NumberTheory/JointDickman/Arithmetic/RoughMomentBounds.lean

GitHub source file

7 declarations, 0 contain sorry, 0 declared as axioms

Uniform bounds for the moving roughness correction

The exponent shift 1 / log P costs only a fixed power of log P. This supplies the moment in Rankin's argument from Mertens' estimate, rather than assuming a cutoff-dependent bound.

  1. theorem OAI.JointDickman.shifted_reciprocal_le line 20
  2. theorem OAI.JointDickman.shifted_prime_sum_le line 31
  3. theorem OAI.JointDickman.smoothCorrection_shifted_bound line 43

    Source documentation

    Mertens controls the shifted absolute correction moment by a fixed power of the logarithm, uniformly over all finite truncations.

    theorem smoothCorrection_shifted_bound
        (hM : PublishedInputs.PrimeReciprocalMertensInput) :
        ∃ A : ℝ, 0 < A ∧ ∀ (P : ℕ) (S : Finset ℕ) (z : ℝ),
          2 ≤ P → 1 ≤ Real.log P → 0 ≤ z → z ≤ 1 / 2 →
          (∑ n ∈ S, |smoothCorrection (Nat.primesLE P) z n| /
            (n : ℝ) ^ (1 - 1 / Real.log P)) ≤
            A * (Real.log P) ^ (Real.exp 1) := by
    …

    Mentions: OAI.JointDickman.PublishedInputs.PrimeReciprocalMertensInput, OAI.JointDickman.smoothCorrection

  4. theorem OAI.JointDickman.roughConvolutionTail_shifted_bound line 86

    Source documentation

    The same constant controls the actual large-divisor convolution error.

    theorem roughConvolutionTail_shifted_bound
        (hM : PublishedInputs.PrimeReciprocalMertensInput) :
        ∃ A : ℝ, 0 < A ∧ ∀ (P N V : ℕ) (z : ℝ),
          2 ≤ P → 1 ≤ Real.log P → 0 < V → 0 ≤ z → z ≤ 1 / 2 →
          |roughConvolutionTail (Nat.primesLE P) z N V| ≤
            (N : ℝ) / (V : ℝ) ^ (1 / Real.log P) *
              (A * (Real.log P) ^ (Real.exp 1)) := by
    …

    Mentions: OAI.JointDickman.PublishedInputs.PrimeReciprocalMertensInput, OAI.JointDickman.roughConvolutionTail

  5. theorem OAI.JointDickman.log_pow_le_rpow line 101
  6. theorem OAI.JointDickman.smoothCorrection_log_moment_le line 116

    Source documentation

    Exponential tilting controls every logarithmic correction moment.

    theorem smoothCorrection_log_moment_le (E S : Finset ℕ) (k : ℕ)
        (z : ℝ) {ε : ℝ} (hε : 0 < ε) :
        (∑ n ∈ S, |smoothCorrection E z n| / (n : ℝ) * (Real.log n) ^ k) ≤
          ((k.factorial : ℝ) / ε ^ k) *
            ∑ n ∈ S, |smoothCorrection E z n| / (n : ℝ) ^ (1 - ε) := by
    …

    Mentions: OAI.JointDickman.smoothCorrection

  7. theorem OAI.JointDickman.smoothCorrection_log_moments_bound line 138

    Source documentation

    The absolute constant in the roughness bound works for every fixed moment order; only the factorial and extra logarithmic power change.

    theorem smoothCorrection_log_moments_bound
        (hM : PublishedInputs.PrimeReciprocalMertensInput) :
        ∃ A : ℝ, 0 < A ∧ ∀ (P : ℕ) (S : Finset ℕ) (z : ℝ) (k : ℕ),
          2 ≤ P → 1 ≤ Real.log P → 0 ≤ z → z ≤ 1 / 2 →
          (∑ n ∈ S, |smoothCorrection (Nat.primesLE P) z n| / (n : ℝ) *
            (Real.log n) ^ k) ≤
            A * (k.factorial : ℝ) * (Real.log P) ^ (Real.exp 1 + k) := by
    …

    Mentions: OAI.JointDickman.PublishedInputs.PrimeReciprocalMertensInput, OAI.JointDickman.smoothCorrection