On collision of multiple eigenvalues for matrix-valued Gaussian processes

From MaRDI portal





The authors consider Gaussian processes \((X(t))_{t\in \mathbb R^N}\) defined on \(\mathbb R^N\) and taking values in the space of \(d\times d\) symmetric or Hermitean matrices. These processes are dynamical versions of the GOE and GUE ensembles in the sense that, for each fixed time \(t\in \mathbb R^N\), \(X(t)\) is a GOE/GUE random matrix. Apart from the restriction imposed by symmetry, the matrix entries are i.i.d. copies of some Gaussian process. The authors study the random set of times \(t\) for which the matrix \(X(t)\) has \(k\) colliding eigenvalues. They provide conditions under which the probability that this set is empty is \(0\) or \(1\). If the set is non-empty, they compute its Hausdorff dimension.











This page was built for publication: On collision of multiple eigenvalues for matrix-valued Gaussian processes

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2033127)