A proof of two conjectures of B. R. Smith on kneading cycles: kneading is conjugate to rotation
Description
In 2014 B. R. Smith (J. Integer Seq. 17, Article 14.11.8; arXiv:1408.4631) introduced an invertible operation on finite sequences of positive integers, kneading, which preserves the sum of the sequence and the parity of its length, and showed that through an explicit correspondence it is the Zagier reduction of indefinite binary quadratic forms of discriminant a^2 ± 4. He conjectured (Conjectures 6(i) and 7 of the published version; the conjectural comment in OEIS entry A000048) that every kneading cycle of sequences of even length and sum n has length l with (n−1)/l an odd integer, and that the number of such cycles of length l is (1/(2l)) Σ_{d|l, d odd} μ(d) 2^{l/d} = A000048(l), independently of n; he checked this for 2 ≤ n ≤ 31.
We prove both statements. The proof exhibits an explicit bijection Ψ from the compositions of n of even length onto the binary words of length n−1 with an odd number of 1's under which kneading becomes the cyclic shift: Ψ writes the parts of even index, then the parts of odd index, each part q as the block 1 0^(q−1), with the two end parts shortened by one letter, followed by a final letter, and with the first letter replaced by 0 when the first part is 1. The shift property reduces to 21 identities between strings, listed in the note and checked by machine; the bijectivity is a direct parsing argument; the counting follows by Möbius inversion over the odd divisors. The letter Ψ(c)_i records whether the first part of the i-th kneading of c is at least 2, which is how the bijection was found. As a remark, the total number of kneading cycles of the compositions of n of even length is the number of binary necklaces of length n−1 with an odd number of 1's (A000016). We also point out that the comment in A000048 must be read, as in Smith's paper, for discriminants of the form a^2+4: without that restriction it fails already for sum 7 and caliber 2. Smith's Conjecture 6(ii), on the imprimitivity of the forms in the cycles with (n−1)/l > 1, is not settled; we record the scaling law observed in the data (the continuant matrix of a cycle with (n−1)/l = 2r+1 is g·P_0 + h·J with g^2 − a g h − h^2 = 1, a unit of the order of discriminant a^2+4), which reduces it to an open problem.
The note was obtained with the author's methodology for open statements in the OEIS (subprocesses with time budgets, a log of obstacles with an algorithm added per obstacle, scripts that test the variants of a formula, literature search, compiled code for the blocking computations, independent verification and review by a second AI system), with the assistance of an AI system under the author's direction; the methodology section of the note describes the process. The verification programs are included: Python scripts (brute-force cycles, the closed form of Ψ for all compositions of n ≤ 22, the symbolic check of the 21 cases) and an independent Rust program that checks the conjugation, the bijectivity and the cycle counts for all 2^30 − 2 compositions of even length of 3 ≤ n ≤ 31, the range of Smith's own checks. Both PDFs (English and Spanish) are the same note.