The paper studies a triangular grid of side length n, which contains n-squared small triangular cells. The central question is how many ways you can choose 3 cells from this grid such that no two of the chosen cells are "neighbors," meaning they neither share an edge nor share a corner. Because the triangular grid has natural symmetries (you can rotate it by 120 or 240 degrees, or reflect it), many such selections are essentially the same arrangement seen from a different angle. The paper counts these selections up to those symmetries, treating configurations that can be rotated or reflected into each other as identical.
To do this counting, the authors use a classical tool called Burnside's lemma, which turns the problem of counting distinct arrangements under symmetry into a problem of counting arrangements that stay unchanged when a particular symmetry is applied. This requires carefully working out how many 3-cell non-adjacent selections look the same before and after each rotation or reflection. A key structural insight simplifies the work considerably: when adjacency means sharing an edge, the resulting neighbor graph turns out to be bipartite, which means no two neighbors of a cell are neighbors of each other, eliminating one of the harder cases. For corner-sharing adjacency, the sets of mutually adjacent triples have a clean geometric description that also makes counting tractable.
The main payoff is rigorous proofs of three generating functions that had been conjectured in July 2026 by Sean Irvine and posted to the Online Encyclopedia of Integer Sequences without proof. The paper not only confirms those conjectures but also derives explicit closed-form formulas for all four related counting sequences. Each formula is a quasi-polynomial of degree 6, meaning it behaves like a degree-6 polynomial but with coefficients that cycle with a small period (either 3 or 6) depending on n. As a bonus, the same techniques settle a fourth related sequence that had lacked a proof since 2017.