Axler theta error, finite range: for
ProvedTaoFivePrimes.axler_theta_log_four_finite_certificateThe finite half of Axler's explicit fourth-power theta estimate.
Let
be the Chebyshev theta function, the sum running over the primes up to . This node asserts the pointwise estimate
It is the numerical half of Proposition 1 of Axler (2018), which is stated for all . Splitting that proposition at separates the two logically different ingredients: below the estimate is a finite verification (one has to control by explicit computation on the primes in the range), while above it is a consequence of analytic estimates for the zeros of . This node is the first ingredient.
Role. The constant and the threshold are exactly the ones in the target statement, and the range is precisely the range in which the target bound is not implied by the analytic estimate. The estimate is tight where it starts: at one has while , so the pointwise inequality holds with a margin of only about at the left endpoint, and no smaller constant would do at that point. This is why the numerical verification cannot be replaced by a weaker explicit bound: every published elementary estimate for (for instance , which is available on the platform for ) is off by a factor from the required accuracy, and is therefore useless here.
Formalization Note. The statement is unconditional, and it is stated for real ; the summand depends on only through , so the assertion is a finite statement about the primes up to . The companion node TaoFivePrimes.axler_theta_log_four_tail covers the complementary range .
import Mathlib open Finset
theorem TaoFivePrimes.axler_theta_log_four_finite_certificate (x : ℝ) (h1 : 70111 ≤ x) (h2 : x < 10 ^ 8) :
|(∑ p ∈ Nat.primesLE ⌊x⌋₊, Real.log (p : ℝ)) - x| <
100 * x / (Real.log x) ^ 4 := by sorry