← All problems

The conjecture that the Schur-degree threshold equals the Schur number

6 years old · traced to

For n≥2n\ge 2, let S(n)S(n) be the nnth Schur number. Let L(n)L(n) be the least length such that every sequence AA in N+\mathbb N_+ with ∣A∣=L(n)|A|=L(n) and average μ(A)≤n\mu(A)\le n satisfies sdeg⁡(A^)≥n\operatorname{sdeg}(\hat A)\ge n. The Schur-degree threshold conjecture.

L(n)=S(n−1)+1.L(n)=S(n-1)+1.

Equivalently, every sequence AA in N+\mathbb N_+ of length ∣A∣=S(n−1)+1|A|=S(n-1)+1 and average μ(A)≤n\mu(A)\le n satisfies sdeg⁡(A^)≥n\operatorname{sdeg}(\hat A)\ge n. The conjecture is known for n=2n=2 and 33, while the case n=4n=4 is explicitly left open in the source.

References
  1. Shalom Eliahou and Pastora Revuelta, “The Schur degree of additive sets”, arXiv:2006.01502 (2020).

Progress summary

Refreshed
Open · progress reported

The conjecture is proved in the two smallest cases, while an unverified October submission claims counterexamples beginning in the next two cases.

Eliahou and Revuelta (2020) proposed L(n)=S(n−1)+1L(n)=S(n-1)+1 for n≥2n\ge 2. They proved the cases n=2n=2 and n=3n=3, leaving the general conjecture open.

Known results

  • n=2n=2 and n=3n=3: proved by Eliahou and Revuelta (2020).
  • 14≤L(4)≤1614\le L(4)\le 16; the conjectured value is L(4)=14L(4)=14 (Eliahou and Revuelta, 2020).
  • The paper records S(n)≤nL(n)S(n)\le nL(n) and S(n)≤n(Rn−1(3)−1)S(n)\le n(R_{n-1}(3)-1).

Community submission (unverified)

Posted October 4, 2026: a submitted argument claims L(4)=16L(4)=16, using a length-1515 sequence, and claims 49≤L(5)≤6149\le L(5)\le 61 using a length-4848 construction. If correct, these would refute the conjecture for n=4n=4 and n=5n=5, but the submission has no independent corroboration.

Current status (as of November 2026): The cases n=2n=2 and n=3n=3 are settled; the claimed advances for n=4n=4 and n=5n=5 are unverified, and the general conjecture remains open.

Sources

Solutions 1

The conjecture predicts L(n) = S(n-1) + 1. Since S(3)=13 and S(4)=44 this predicts L(4)=14 and L(5)=45. We prove that L(4)=16 and L(5)>48 refuting the conjecture.See full solutionHide full solution

Theorem 1.2 (The degree-four threshold)

We have L(4)=16L(4)=16. More precisely, the sequence

A4=(3,1,3,4,3,1,3,4,3,1,3,4,3,1,3)A_4=(3,1,3,4,3,1,3,4,3,1,3,4,3,1,3)

has length 1515, total 4040, and every one of its prefixes has average at most 44 and blocksum Schur degree at most 33. This refutes the n=4n=4 case of the Conjecture and the conjectural value L(4)=14L(4)=14.

Theorem 1.3 (A degree-five interval)

We have

49≤L(5)≤61.49\leq L(5)\leq61.

Let

W=(3,3,3,3,3,3,1,3,3,3,3,3,3,4)W=(3,3,3,3,3,3,1,3,3,3,3,3,3,4)

and let A5A_5 be the concatenation of three copies of WW followed by six further entries equal to 33. Then A5A_5 has length 4848, total 141141, and every prefix has average at most 44 and blocksum Schur degree at most 44. Hence L(5)>48L(5)>48, refuting the n=5n=5 case of the Conjecture.

The colourings and details are written down in the attached paper.

The examples were found with the help of KI.

  • Schur_L4_L5_2026-09-20.pdf82,654 bytesOpen