Extremal inverse eigenvalue problem for irreducible acyclic matrices

Abstract

In this paper, we study the inverse eigenvalue problem of constructing symmetric matrices whose graph is a tree, i.e. of constructing irreducible acyclic matrices from given eigendata consisting of the smallest and largest eigenvalues of their leading principal submatrices. We solve the problem completely for the case when the given eigenvalues are distinct and the subgraph of the tree induced by remains connected for . The result generalizes the results known for matrices described by some specific trees. The proofs of the main results are constructive and so provide an algorithm for computing the actual entries of the required matrix. Numerical solutions of the problem are then obtained with SCILAB by feeding the eigendata and plugging in the adjacency matrix as inputs. We also make an attempt to find the number of solutions for several special families of unlabelled trees.