Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Sharpness of the abc exponent (lim sup⁡L3≥1\limsup L_3 \ge 1limsupL3​≥1)

Open
NConjecture.abc_exponent_sharp

by LG · Nov 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

abc-conjecturenumber-theory

Browkin–Brzeziński (§1, p. 931) observe that the abc exponent 111 is sharp: "it is not difficult to show that there is a limit point of this set which is ≥1\ge 1≥1," i.e. lim sup⁡L(a,b,c)≥1\limsup L(a,b,c)\ge 1limsupL(a,b,c)≥1 over coprime triples. In ε\varepsilonε-form this says that for every ε>0\varepsilon>0ε>0 there are infinitely many coprime positive triples a+b=ca+b=ca+b=c whose quality exceeds 1−ε1-\varepsilon1−ε:

#{(a,b,c):a,b>0, gcd⁡(a,b)=1, a+b=c, c>rad⁡(abc) 1−ε}=∞.\#\{(a,b,c): a,b>0,\ \gcd(a,b)=1,\ a+b=c,\ c>\operatorname{rad}(abc)^{\,1-\varepsilon}\}=\infty.#{(a,b,c):a,b>0, gcd(a,b)=1, a+b=c, c>rad(abc)1−ε}=∞.

Here rad⁡(abc)\operatorname{rad}(abc)rad(abc) is the product of the distinct primes dividing abcabcabc. An explicit witness family is (1, 9k−1, 9k)(1,\,9^k-1,\,9^k)(1,9k−1,9k), whose quality is >1>1>1 for every kkk (Ramaekers, §1.1). This is the n=3n=3n=3 base case of the sharp lower bound lim sup⁡Ln≥2n−5\limsup L_n\ge 2n-5limsupLn​≥2n−5.

Formalization Note. A triple is stored as t=(a,b,c)∈N3t=(a,b,c)\in\mathbb N^3t=(a,b,c)∈N3 with c=t.2.2c=t.2.2c=t.2.2; the radical is Brocard.radical. The strict inequality c>rad⁡(abc)1−εc>\operatorname{rad}(abc)^{1-\varepsilon}c>rad(abc)1−ε is equivalent to L>1−εL>1-\varepsilonL>1−ε because rad⁡(abc)≥2\operatorname{rad}(abc)\ge 2rad(abc)≥2.

Preamble
import Mathlib
import Definitions.Def_Brocard_Defs
Formal statement
namespace NConjecture

theorem abc_exponent_sharp (ε : ℝ) (hε : 0 < ε) :
    Set.Infinite {t : ℕ × ℕ × ℕ | 0 < t.1 ∧ 0 < t.2.1 ∧ Nat.Coprime t.1 t.2.1 ∧
      t.1 + t.2.1 = t.2.2 ∧
      (Brocard.radical (t.1 * t.2.1 * t.2.2) : ℝ) ^ (1 - ε) < (t.2.2 : ℝ)} := by sorry

end NConjecture
Source
J. Browkin, J. Brzeziński, Some remarks on the abc-conjecture, Math. Comp. 62(206) (1994), 931–939. https://cr.yp.to/bib/1994/browkin.pdf §1 pp. 931–932 (limit point ≥ 1). Explicit family: C. Ramaekers, The abc-conjecture and the n-conjecture, TU/e (2009), §1.1. https://pure.tue.nl/ws/portalfiles/portal/67739846/657782-1.pdf
Read-back

What the Lean code literally says, in plain math · GLM (Z.ai), blind code-only pass — NOT an independent sub-agent

abc_exponent_sharp (ε : ℝ) (hε : 0 < ε). For every real ε\varepsilonε with 0<ε0<\varepsilon0<ε, the following set of triples of natural numbers is infinite:

{(a,b,c)∈N3: 0<a, 0<b, gcd⁡(a,b)=1, a+b=c, rad⁡(abc) 1−ε<c}.\{(a,b,c)\in\mathbb N^3:\ 0<a,\ 0<b,\ \gcd(a,b)=1,\ a+b=c,\ \operatorname{rad}(abc)^{\,1-\varepsilon}<c\}.{(a,b,c)∈N3: 0<a, 0<b, gcd(a,b)=1, a+b=c, rad(abc)1−ε<c}.

The triple is stored as ttt with a=t.1a=t.1a=t.1, b=t.2.1b=t.2.1b=t.2.1, c=t.2.2c=t.2.2c=t.2.2. The conditions are: aaa and bbb strictly positive; aaa and bbb coprime (Nat.Coprime); ccc equal to a+ba+ba+b (so c>0c>0c>0 follows, not separately assumed); and the real inequality rad⁡(a⋅b⋅c)1−ε<c\operatorname{rad}(a\cdot b\cdot c)^{1-\varepsilon}<crad(a⋅b⋅c)1−ε<c, where rad⁡(m)=∏p∈m.primeFactorsp\operatorname{rad}(m)=\prod_{p\in m.\mathrm{primeFactors}}prad(m)=∏p∈m.primeFactors​p is the product of distinct prime factors, cast to R\mathbb RR, raised to the real power 1−ε1-\varepsilon1−ε. When ε≥1\varepsilon\ge 1ε≥1 the exponent 1−ε≤01-\varepsilon\le 01−ε≤0, so rad⁡(abc)1−ε≤1<c\operatorname{rad}(abc)^{1-\varepsilon}\le 1<crad(abc)1−ε≤1<c holds automatically for every such triple.

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me