Cluster algebras are a type of mathematical structure where you start with a set of variables and repeatedly apply a specific substitution rule called "mutation" to generate new variables. A remarkable property, called the Laurent phenomenon, says that every variable you can ever generate is actually a Laurent polynomial in your starting variables, meaning it can be written as a polynomial divided by a monomial (a product of powers of the starting variables). An even stronger observation, called positivity, says that the coefficients appearing in these Laurent polynomial expressions are always positive whole numbers, never negative. Positivity for cluster algebras was a major open problem for many years before being resolved around 2018.
Laurent phenomenon algebras, introduced by Lam and Pylyavskyy, are a broader generalization of cluster algebras that keep the Laurent phenomenon property but relax some of the strict rules governing how mutations work. One particularly interesting family of these algebras is built directly from directed graphs: you associate variables to vertices of a graph, and the graph's structure dictates the mutation rules. Lam and Pylyavskyy showed that the Laurent phenomenon holds for these graph-based algebras but conjectured that positivity should hold as well, meaning all the coefficients in the resulting Laurent polynomials should be nonnegative.
This paper proves that conjecture in full generality. For any directed graph without loops or repeated edges between the same pair of vertices, and starting from any valid initial set of variables, every generated variable expands as a Laurent polynomial with nonnegative integer coefficients. The proof works uniformly across all such graphs and all possible starting points, which is a strong and clean result. This extends positivity results from the cluster algebra world into this more general setting and confirms that the beautiful positivity phenomenon is not a quirk specific to cluster algebras but persists in this wider family of related algebraic structures.