Tao Lemma 4.3 (4.4): S_{eta,1}(x,0) = x*int(eta) + O*(||eta'||_1 x / (40 log cx))
ProvedTaoFivePrimes.smoothedExpSum_zero_asymptoticFor a cutoff , a modulus , a scale and a frequency , write
being the von Mangoldt function. Let be smooth and supported in for some , and let satisfy . Then
where denotes a quantity of absolute value at most . For the non-negative cutoffs used in the paper the main term is the of the source.
The estimate is what converts the prime-counting content of into the analytic quantity ; together with the trivial bound (4.3) of the same lemma it underlies the mass computations of Sections 4 and 8.
Quoted input The proof in the source invokes the explicit Chebyshev estimate of Rosser and Schoenfeld, for , where . That estimate is not available in the ambient library, so it appears as an explicit hypothesis, stated exactly in the range and the form in which it is used.
Formalization Note The sum is complex-valued; at it is a real number embedded in , and the error is measured by the complex absolute value.
import Mathlib import Definitions.Def_TaoFivePrimes_SmoothedExpSum open MeasureTheory open scoped ArithmeticFunction.vonMangoldt
theorem TaoFivePrimes.smoothedExpSum_zero_asymptotic
(eta : ℝ → ℝ) (hsm : ContDiff ℝ (⊤ : ℕ∞) eta) (c x : ℝ)
(hc0 : 0 < c) (hc1 : c ≤ 1) (hx : 1 ≤ x)
(hsupp : ∀ t : ℝ, t < c ∨ 1 < t → eta t = 0)
(hcx : (10 : ℝ) ^ 8 ≤ c * x)
(hpsi : ∀ y : ℝ, c * x ≤ y → y ≤ x →
|Chebyshev.psi y - y| ≤ y / (40 * Real.log (c * x))) :
‖TaoFivePrimes.smoothedExpSum eta 1 x 0 - (((∫ t : ℝ, eta t) * x : ℝ) : ℂ)‖
≤ 1 / (40 * Real.log (c * x)) * (∫ t : ℝ, |deriv eta t|) * x := by sorry