Keyboard shortcuts

Press ← or → to navigate between chapters

Press ? to show this help

Press Esc to hide this help

Metadata
slug: joshi-rust-2025-thue-morse-first-longest-start
bibkey: joshirust2025monochromatic
doi: 10.1016/j.tcs.2025.115391
url: https://arxiv.org/html/2501.05830v2
triage: theorem
motivation_gids:
  - D5/S1/Words/Complexity/ThueMorseReducedAbelianOdd

First longest Thue-Morse progressions at power-of-two differences

Problem

Joshi and Rust, Monochromatic arithmetic progressions in the Fibonacci, Thue-Morse, and Rudin-Shapiro words, Theoretical Computer Science 1050 (2025), 115391, arXiv:2501.05830v2, Section 3.2.2, Conjecture 3.8:

We have i(2^n+1)=3·2^(2n)-2^n-1, i(2^(2n)-1)=3·2^(4n)-2^(2n)+1, i(2^(2n+1)-1)=2^(2n+1)-1.

The display omits ranges. The contextual interpretation, from the preceding maximum-length formulas, is n>=2 in the first equation and n>=1 in the other two. Equivalently, for every natural e>=2, with q=2^e, i(q+1)=3q^2-q-1, i(q-1)=3q^2-q+1 for even e, and i(q-1)=q-1 for odd e. This range is an explicitly disclosed interpretation, not a printed quantifier. The previously known i(3)=45 excludes the plus-family extension to e=1; it is not a new refutation.

The word is the zero-indexed Thue-Morse word 0110100110010110..., with t(0)=0, t(2u)=t(u) and t(2u+1)=1-t(u). Definitions 2.2, 2.3 and 2.5 make A(d) the global maximum length over every nonnegative start and either letter, and i(d) the least start attaining it.

Motivation

This 2025 specialist conjecture concerns three unbounded families. The source explains that its Walnut method cannot handle the variable product needed when the maximum length varies with the difference. The existing ThueMorseReducedAbelianOdd module supplies the actual binary digit-parity word and its binary identities.

Gap

A progression at the proposed start alone does not establish maximality or firstness. The statement requires positive attainment, a bound on all lengths at all starts for either letter, and exclusion of every earlier start. The exceptional half-block possibilities at exponent two also require treatment.

Route

Write each start as aq+b, 0<=b<q. Binary block parity and complementary residues express progression letters through carries or borrows. Block recognition forces any length q+2 progression at difference q+1 to satisfy s+q+1=kq^2, k>=3, with an opposite next letter. For even e, a length q+4 progression at difference q-1 satisfies s+q=kq^2+1, k>=3, again with an opposite next letter. The explicit starts with k=3 attain those lengths.

For odd e, putting b=s mod q gives opposite letters at progression indices b and b+1. This bounds every length by q; the start q-1 attains q. For s<q-1, both opposite letters lie within that length, excluding an earlier maximum. Aedo, Grimm, Nagai and Staynova’s known recognition and maxima are prerequisites, not additional new results.

Falsifier

A natural e>=2 whose first attaining start differs from any applicable formula refutes the contextual conjunction. An earlier run of either letter attaining the stated maximum, a longer run at any start, or a mismatch with the actual zero-indexed word would invalidate the claimed source correspondence.

Evidence

  • Formal source: D5/S1/Words/ThueMorseMapFirstStart.lean.
  • Public declarations: MAP, FirstLongest, claim, and result : claim.
  • FirstLongest(d,s) is exists L>0, MAP(d,s,L) together with forall a N, MAP(d,a,N) -> N<=L and forall a<s, not MAP(d,a,L).
  • The source proves the full conjunction for every e>=2, using maximum lengths q+2, q+4, and q in the respective branches.
  • The result’s axiom closure is propext, Classical.choice, Quot.sound.
  • Exact source: https://arxiv.org/html/2501.05830v2#S3.Thmtheorem8.
  • Preregistered target and bounded literature audit: https://github.com/the-omega-institute/trureturing/issues/13307.

Triage

theorem. The source result establishes the contextual conjunction through attainment, global maximality and least-start exclusion. The binary carry/borrow and block-recognition mechanism supplies all-start exclusions; the known maximum formulas are recovered within that argument. The source does not assert a classification of all maximizing starts as a separate public theorem. Other Thue-Morse questions in Joshi and Rust, including Question 3.7, remain unresolved by this result.

ASSUMED-UNVERIFIED

The preregistration’s bounded audit on 5 October 2026 found no identified exact resolution in the checked adjacent literature, OEIS A342827, arXiv and OpenAlex searches, formal-conjectures, and repository ownership checks. The unavailable thesis and unreadable search/index results supplied no negative evidence. Coverage and freshness are bounded; worldwide priority is unverified. The parameter range is the contextual interpretation explained above.