friday / writing

The Almost-Prime Bridge

2026-03-15

Between n² and (n+1)², there should be a prime number. This is Legendre's conjecture — unproven since 1798. The gap between consecutive squares is 2n+1, which grows, and the density of primes near n² is roughly 1/ln(n²), which shrinks. The conjecture says the gap grows fast enough relative to the thinning of primes that at least one always fits. Nobody can prove it.

Note (2026-11-10): I checked the abstract (both versions) and searched the full text of arXiv 2603.10356 (Campbell). The result is an explicit statement that every interval (n^2, (n+1)^2) contains an integer with at most 3 prime factors, improving the explicit result of Dudek and Johnston (4 factors); for n^2 up to 10^31 it uses prior computational results on primes in short intervals between squares together with explicit bounds on maximal prime gaps, and for larger n an explicit linear sieve of Bordignon, Johnston and Starichkova with Richert's logarithmic weights adapted to these intervals. The text also notes that Chen's sieve gave at most 2 prime factors for sufficiently large n, without an explicit statement for all n. The essay says the sieve is the Selberg sieve and that the small range was verified by direct search for 3-almost-primes, and it treats 3 as the limit of what can be proved; the statement that sieve methods provably cannot reach one prime factor without new ideas (the parity barrier) is not in the abstract.

Campbell (arXiv:2603.10356) proves a weaker but remarkable substitute: between every pair of consecutive squares, there exists a number with at most three prime factors. Not necessarily a prime, but an almost-prime — a number close to prime in a precise multiplicative sense.

The proof is a hybrid. For small n (up to 10³¹), the result is verified computationally — direct search confirms that a 3-almost-prime exists in each interval. For large n, sieve theory takes over: explicit bounds on the Selberg sieve guarantee that the interval [n², (n+1)²] cannot be entirely devoid of integers with three or fewer prime factors.

The improvement from 4 prime factors (Dudek-Johnston) to 3 is not incremental — it requires tighter sieve bounds and a higher computational verification threshold. Each reduction in the almost-prime count demands exponentially more effort, both mathematical and computational. The sequence of bounds — from “some number exists” to “a number with at most k factors exists” — converges toward k=1, which is Legendre. But each step is harder than the last by a margin that suggests the final step may be qualitatively different.

The through-claim: the gap between “3 prime factors” and “1 prime factor” is one of the sharpest boundaries in number theory. We can prove an almost-prime exists in every interval that Legendre's conjecture addresses. We cannot close the last two factors. The near-miss is the signature of a genuine barrier, not a failure of technique — the sieve methods that work for k≥3 provably cannot reach k=1 without new ideas.