The Theory of Persistence
Théorème

N2 — Sieve self-consistency

The Eratosthenes sieve is the unique self-consistent multiplicative sieve on $\mathbb{N}_{\geq 2}$.

Statement

The set of primes P\mathbb{P} is the unique self-consistent multiplicative sieve on N2\mathbb{N}_{\geq 2}. Self-consistent means nS(P)n \in \mathcal{S}(\mathbb{P}) (i.e. nn survives) iff no prime p<np < n divides nn.

Théorème

Plain reading. The Eratosthenes sieve has a special property: what it keeps matches exactly what it was supposed to keep. No other multiplicative elimination algorithm has this consistency. The sieve defines itself.

Why it matters

N2 closes a subtle question: the internal consistency of the sieve. One might imagine another sieve (eliminating by other rules, keeping different residues) that would return its own result. N2 shows that the unique sieve whose output is consistent with its input rule is Eratosthenes’.

This self-consistency is what allows the cascade T0 → T6 to close without introducing an external parameter.

Proof — outline

  1. Existence: show P\mathbb{P} is self-consistent (a composite n=abn = ab with ana \leq \sqrt{n} has a prime factor pa<np \leq a < n, hence nn is eliminated).
  2. Uniqueness: assume a sieve C\mathcal{C} self-consistent and different from P\mathbb{P}. Derive a contradiction.

Detailed proof

Part 1 — Existence: P\mathbb{P} is self-consistent

Let nN2n \in \mathbb{N}_{\geq 2}. If nn is prime, by definition no prime p<np < n divides nn, and nn is kept by the sieve.

If nn is composite, write n=abn = ab with ana \leq \sqrt{n}. Then aa has a prime factor pa<np \leq a < n, hence pp divides nn. The Eratosthenes sieve eliminates nn as a multiple of pp. Self-consistency holds.

Part 2 — Uniqueness

Suppose another multiplicative sieve C\mathcal{C} is self-consistent. Let Cout\mathcal{C}_{\rm out} be the set of kept integers.

If CoutP\mathcal{C}_{\rm out} \neq \mathbb{P}, let nn be the smallest integer where the two differ: nCoutn \in \mathcal{C}_{\rm out} but nPn \notin \mathbb{P}, or vice versa.

Case 1: nCoutn \in \mathcal{C}_{\rm out}, nn composite. Then nn has a prime factor p<np < n. If pCoutp \in \mathcal{C}_{\rm out} (pp kept by C\mathcal{C}), then by self-consistency C\mathcal{C} must also keep nn (since C\mathcal{C} uses multiplicative rules). But then C\mathcal{C} keeps a composite — by construction of a strict multiplicative sieve, contradiction.

Case 2: nPn \in \mathbb{P} but nCoutn \notin \mathcal{C}_{\rm out}. Then a multiplicative divisor eliminated nn. But nn prime has only 1 and nn as divisors. None is strictly in {2,,n1}\{2, \ldots, n-1\} — contradiction.

In both cases, contradiction. So Cout=P\mathcal{C}_{\rm out} = \mathbb{P}.

Self-consistency is the property that lets the sieve dynamics close on itself. It gives the fixed point μ=15\mu^* = 15 its uniqueness (T5): it is the unique self-consistent solution of the cascade equation.

For the complete derivation, see chapter 2 of the monograph.

See also