Eigenvectors and eigenvalues: a new formula? (Q2193419)

Let \(A\) be a Hermitian \(n\times n\) matrix and \(V=[v_{ij}]\) be a unitary matrix such that \(V^{\ast}AV=\operatorname{diag}(\lambda_{1},\dots,\lambda_{n})\). Suppose that \(M_{i}\) is the principal minor of \(A\) obtained by deleting the \(i\)-th row and column and let \(\Sigma_{i}\) denote the list of \(n-1\) eigenvalues of \(M_{i}\). Then using the relationship between a matrix and its adjugate, the authors give a proof of the eigenvector-eigenvalue identity: for all \(i,j\) we have \[ \left\vert v_{ij}\right\vert ^{2}\prod_{k\neq i}(\lambda_{i}-\lambda_{k})=\prod_{\mu\in\Sigma_{j}}(\lambda_{i}-\mu). \] Equivalently, if \(f(\lambda)\) and \(f_{j}(\lambda)\) are the characteristic polynomials for \(A\) and \(M_{j}\), respectively, then \(\left\vert v_{ij}\right\vert ^{2}f^{\prime}(\lambda_{i})=f_{j}(\lambda_{i})\). The authors refer to the recent preprint [\textit{P. B. Denton} et al., ``Eigenvectors from eigenvalues: a survey of a basic identity in linear algebra'', Preprint, \url{arXiv:1908.03795}] which examines the history of this identity. Using cooperative crowdsourcing the authors of the preprint located at least 39 references to the identity in papers dating back to 1934. The preprint considers the many different proofs and tries to explain why such an elementary result which has applications in physics, statistics, numerical analysis and graph theory is considered just as folklore in some mathematical communities. Indeed, it rarely appeared in textbooks and is apparently almost invisible to usual literature searches.