slug: oeis-a092028-least-self-divisor-power-minus-one bibkey: firoozbakht2004a092028 doi: null url: https://oeis.org/A092028 triage: theorem motivation_gids:

• D5/S3/Arith/Congruence/FiroozbakhtLeastSelfDivisorPowerMinusOne

The least self-divisor exponent in OEIS A092028

Problem

OEIS A092028, NAME (%N, verbatim):

a(n) is the smallest m > 1 such that m divides n^m-1

COMMENTS (%C, verbatim; Farideh Firoozbakht, Mar 26 2004):

Conjecture 1: All terms of this sequence are primes.

Conjecture 2: a(n) is the smallest prime factor of n-1 or for n>2, A092028(n) = A020639(n-1).

The literal formal claim is ∀ n > 2, a(n) = minFac(n − 1), where a(n) = sInf {m ∈ ℕ | 1 < m ∧ m ∣ n^m − 1}. In the natural numbers, sInf ∅ = 0; for every n > 2, the defining set is nonempty. Conjecture 1 is a corollary because minFac(n − 1) is prime, and is not formalized as a separate public theorem.

Motivation

The equality identifies the least solution of a self-referential divisibility condition for every natural input above two. It proves the unbounded OEIS conjecture rather than extending the entry’s finite data.

Gap

The searches recorded in preregistration issue #7526 and its probe on 2026-09-13 found that OEIS history entries #1 through #16, including revision #16, still carry both conjectures. The sibling plus-sign entry A092067 was proved by Neder in 2019; it supplied a search pattern only and is not imported. Exact searches returned MathOverflow 0, Crossref 0, and GitHub issues 0. arXiv, Semantic Scholar, and OpenAlex were rate-limited, while GitHub code search was unauthenticated. Those unavailable surfaces are ASSUMED-UNVERIFIED. The completed searches are bounded and establish no priority claim.

Route

For the upper bound, let p = minFac(n − 1). The number p is prime and p ∣ n − 1, hence n^p ≡ 1 (mod p) and p belongs to the defining set. For the lower bound, take any m > 1 with m ∣ n^m − 1 and let q = minFac(m). The multiplicative order of (n : ZMod q) divides both m and q − 1. Minimality of q gives gcd(m, q − 1) = 1, so the order is one. Thus q ∣ n − 1, and minFac(n − 1) ≤ q ≤ m. Together the two bounds give the claimed equality.

Falsifier

One natural number n > 2 for which the least m > 1 dividing n^m − 1 differs from minFac(n − 1) would contradict the theorem.

Evidence

• Lean module: D5/S3/Arith/Congruence/FiroozbakhtLeastSelfDivisorPowerMinusOne.lean.
• Main theorem: firoozbakht_a092028.
• Axioms: std3 (propext, Classical.choice, Quot.sound).
• The orchestrator checked every n with 3 ≤ n ≤ 3000 and found zero mismatches.
• The probe checked all 9998 cases with 3 ≤ n ≤ 10000 and found zero mismatches. For n = 3..12, the values are 2, 3, 2, 5, 2, 7, 2, 3, 2, 11.

Triage

theorem. The formal result proves the equality for every natural n > 2.

ASSUMED-UNVERIFIED

arXiv, Semantic Scholar, and OpenAlex were rate-limited, and GitHub code search was unauthenticated. The successful literature searches were bounded. No claim of exhaustive literature coverage or first-publication priority is made.