Two theorems on Euclidean distance matrices and Gale transform (Q5956246): Difference between revisions

A Euclidean distance matrix (EDM, for short) \(D=( d_{ij})\) is an \(n \times n\) matrix whose entries are the squares of the distances of \(n\) points \(p^1, p^2,\dots ,p^n\) in some Euclidean space \(\mathbf R^r\); the smallest value \(r\) such that \(\mathbf R^r\) contains each point \(p^i\) is called the embedding dimension of \(D\). The Gale transform is a powerful technique used in the theory of polytopes [cf. \textit{D. Gale}, Ann. Math. Stud. 38, 255-263 (1956; Zbl 0072.37805) and \textit{B. Grünbaum}, Convex polytopes (1967; Zbl 0163.16603)] by means of which a set of \(n\) points of \(\mathbf R^{(n-1-r)}\) arises from a set of \(n\) points of \(\mathbf R^r\) and the new set reflects the affine dependencies of the first one. The authors use the Gale transform to characterize those EDM's which can be expressed as \(D=\lambda(E-C)\), where \(E\) is the matrix of all ones, \(C=( c_{ij})\) is a correlation matrix (i.e. a positive semidefinite symmetric matrix with every \(c_{ii} = 1\)) and \(\lambda\) is a nonnegative scalar. Firstly they prove that the entries of any column of \(DZ\) are constant, if \(D\) is an EDM and \(Z\) is a Gale matrix corresponding to \(D\); finally they characterize \( D=\lambda(E-C)\) as the EDM for which \(DZ=0\).