Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
Schinzel nd reducibility polynomials covering systems congruences
Schinzel, A., Reducibility of polynomials and covering systems of congruences. Acta Arith. 13 (1967), 91-101. The scan carries only the digitizer's icm mark with a copyright sign and prints no license line; the publisher's record labels the PDF download "Pobierz zgodnie z CC-BY", which the English site renders "Free download under CC-BY license", naming no version or license URL (https://www.impan.pl/get/doi/10.4064/aa-13-1-91-101, read 2026-10-02).
Motivated by a question of Turan on approximating an integer polynomial by an irreducible one, Schinzel studies when x^n + f(x) is irreducible over the rationals for infinitely many n, and links this to covering systems of congruences a_i mod m_i. Theorem 1 proves the equivalence of two propositions: (A) for every integer polynomial f with f(0) != 0, f(1) != -1 and f not identically 1 there is an arithmetic progression N with x^v + f(x) irreducible for every v in N; and (B) in every finite covering system with m_i > 1 at least one of the quotients m_j/m_i equals q^a for a prime q and an a >= 0, with a_j != a_i mod m_i and either q > 2 or m_i = 1 mod 2 or a_j != a_i mod (m_i/2). By Theorem 2, C implies B and B implies D, where (C) says every finite covering system with m_i > 1 has two equal moduli or an even modulus, and (D) says every such system has at least one modulus dividing another. The technique is polynomial-congruence machinery over Z[x]: Lemma 1 constructs f congruent to prescribed a_i(x) modulo pairwise coprime monic F_i(x) with degree below the product, and Lemma 2 shows for prime q and m <= n that the cyclotomic polynomials X_m and X_n are relatively prime mod q unless n/m = q^a (a >= 0), in which case X_n = X_m^{phi(n)/phi(m)} mod q. Here monic means leading coefficient +1 or -1. Schinzel singles out the consequence that Selfridge's conjecture, that no covering system has distinct odd moduli all greater than 1 (Problem 7), would make x^n + f(x) irreducible for infinitely many n for every integer f with f(0) != 0 and f(1) != -1.
Source: https://eudml.org/doc/204820.
Bears on. #7
Results to transcribe.
- Theorem 1: Proposition A (for every integer f with f(0) != 0, f(1) != -1, f not identically 1 there is an arithmetic progression N with x^v + f(x) irreducible for v in N) is equivalent to proposition B on finite covering systems, that some quotient m_j/m_i is a prime power q^a (a >= 0) with a_j != a_i mod m_i and either q > 2 or m_i odd or a_j != a_i mod (m_i/2).
- Theorem 2: C implies B and B implies D, where C says every finite covering system with all moduli > 1 has two equal moduli or an even modulus, and D says every such system has at least one modulus dividing another.
- Consequence of the Selfridge conjecture: If no finite covering system has distinct odd moduli greater than 1, then for every integer polynomial f with f(0) != 0 and f(1) != -1, x^n + f(x) is irreducible for infinitely many n.
- Lemma 1: If monic F_1,...,F_r in Z[x] are pairwise relatively prime modulo every prime, then there is f in Z[x] with f = a_i mod F_i for each i and deg f < deg(product of F_i).
- Lemma 2: For q prime and m <= n, the cyclotomic polynomials X_m and X_n are relatively prime mod q except when n/m = q^a (a >= 0), in which case X_n = X_m^{phi(n)/phi(m)} mod q.