Higher-rank numerical ranges and compression problems (Q855575)

Higher-versions of the standard numerical ranges for matrices are considered. Namely, for an \(N\times N\) complex matrix \(T\) and \(k\geqslant 1\), the rank-\(k\) numerical range of~\(T\) is defined as \(\Lambda_k(T)=\{\lambda\in {\mathbb C}\mid PTP = \lambda P\) for some rank-\(k\) projection \(P\}\). An explicit characterization of \(\Lambda_k(\cdot)\) is obtained for Hermitian matrices; the similar problem remains open for normal matrices. A central motivation for this investigation comes from quantum error correction.