Two integer frieze patterns of width 12 whose rows 6 differ with exactly two sign changes #
Continuation of Formalizations.Combinatorics.Frieze.IntegerDominance (same conventions: entry a k i is
the continuant of a_i, …, a_{i+k-1}, indices modulo n = a.length). For width w = 12 (n = 15), row
6 = ⌊w/2⌋ lies inside the printed range 1 ≤ k ≤ [w/2] of Problem 1 of S. Tabachnikov
(A four-vertex theorem for frieze patterns?, Math. Intelligencer 40 (2018)). We record two integer
Conway–Coxeter friezes of width 12 whose row-6 differences have no zero and exactly two cyclic sign
changes, so the answer to the question is negative for integer friezes within the printed range too. An
exhaustive search found 480 such pairs at width 12 and none, in any row k ≤ ⌊w/2⌋, at smaller widths.
The verification is a finite computation checked by the kernel (decide +kernel).
The difference of row k of the friezes with quiddities b and a, at position i.
Equations
- Frieze.rowDiff a b k i = Frieze.entry b k i - Frieze.entry a k i
Instances For
The number of cyclic positions i < a.length where the row-k difference changes sign strictly.
Equations
- Frieze.signChanges a b k = (List.filter (fun (i : ℕ) => decide (Frieze.rowDiff a b k i * Frieze.rowDiff a b k ((i + 1) % a.length) < 0)) (List.range a.length)).length
Instances For
Control: the counter is not degenerate; row 5 of the same pair has five sign changes.
Two integer Conway–Coxeter friezes of width 12 whose rows 6 = ⌊12/2⌋ differ with no zero and
exactly two cyclic sign changes.