Determinant of the distance matrix of a tree with matrix weights (Q2497234)

If \(T\) is a tree with \(n\) vertices then, according to \textit{R. L. Graham} and \textit{H. O. Pollak} [Bell System Tech. J. 50, 2495--2519 (1971; Zbl 0228.94020)], the determinant of its distance matrix equals \((-1)^n(n-1)2^{n-2}\). Thus this determinant depends only on the number \(n\), but not on the tree itself. In the present paper this formula is generalised by assigning a weight to each edge of a given tree, the weights being \(s\times s\) matrices. The weighted distance of two vertices is then defined in a natural way (by summing up weights) as an \(s\times s\) matrix. The corresponding weighted distance matrix is given as an \(sn\times sn\) matrix which is composed of blocks of size \(s\times s\). Also, the following is shown: If all weight matrices are assumed to be positive definite then the weighted distance matrix has \(s\) positive and \(s(n-1)\) negative eigenvalues.