Title:Potts partition function and isomorphisms of weighted trees

Abstract:We consider a question pertaining to the topic of graph reconstruction, which entertains links with the $W$-polynomial and (theoretical) statistical physics. Motivated by several open questions, we slightly deviate from the usual approaches to study, in the context of weighted trees, whether a given data (which can be obtained from the $W$-polynomial) distinguishes non-isomorphic weighted trees. We prove that this is the case if one restricts to any \emph{good class} of vertex-weighted trees. Good classes are rich: letting $\mathcal{C}$ be the class of all vertex-weighted trees, one can obtain for each weighted tree $(T,w)$ a weighted tree $(T',w')$ in polynomial time, so that $\mathcal{C}':=\left\{(T',w')\,:\,(T,w)\in \mathcal{C}\right\}$ is good and two elements $(A,b)$ and $(X,y)$ of $\mathcal{C}$ are isomorphic if and only if $(A',b')$ and $(X',y')$ are.