Square-root completeness at the Richstein sieve cutoff: a survivor below (2·10⁷)² is prime
ProvedTaoFivePrimes.goldbach_survivor_prime_2e7computational-number-theorygoldbachnumber-theoryprimes
Let . Suppose and that no prime divides . Then is prime.
This is the square-root completeness step of the sieve argument behind Richstein's verification: a survivor of sieving by all primes up to the cutoff that lies at or below cannot be composite, since a composite has a prime divisor at most . It is the number-theoretic core that turns the finite sieve-coverage assertions (Richstein2001.segmented_sieve_coverage, stated in the GoldbachSieve.pairSums certificate language) into genuine Goldbach partitions, and it is stated here in self-contained form so that a proof worker can attack it with Mathlib alone.
Preamble
import Mathlib set_option autoImplicit false
Formal statement
namespace TaoFivePrimes
theorem goldbach_survivor_prime_2e7 (n : ℕ) (h2 : 2 ≤ n) (hN : n ≤ 4 * 10 ^ 14)
(hsurv : ∀ p : ℕ, p.Prime → p ≤ 20000000 → ¬ p ∣ n) :
n.Prime := by sorry
end TaoFivePrimesSource
Five-primes mission; decomposes TaoFivePrimes.even_goldbach_verified (5247b31b-05f6-40f5-87b2-e5ce7b0e51f3). The sieve cutoff 2·10⁷ is the one used in Richstein2001.segmented_sieve_coverage; the completeness argument is standard (cf. J. Richstein, Verifying the Goldbach conjecture up to 4·10¹⁴, Math. Comp. 70 (2001), 1745–1749, https://doi.org/10.1090/S0025-5718-00-01290-4).