Multiplicative maps preserving the higher rank numerical ranges and radii (Q967489)

Given a positive integer \(k\), the \(k\)-numerical range, \(\Lambda_k(A)\), of a complex \(n\)-by-\(n\) matrix \(A\) is the set of all scalars \(\alpha\) such that \(PAP=\alpha P\) for some projection \(P\) with rank \(k\). This extends the classical numerical range and has broad applications in quantum information theory. The maximal modulus of scalars from \(\Lambda_k(A)\) is called \(k\)-numerical radius, \(r_k(A)\), with the convention that \(r_k(A)=-\infty\) if \(\Lambda_k(A)=\emptyset\). Motivated with applications in quantum error correcting codes, the authors classify multiplicative maps \(\Phi\), from various classical subgroups of \(M_n(\mathbb C)\) into \(M_n(\mathbb C)\), which preserve the \(k\)-numerical radius in a sense that \(r_k(A)=r_k(\Phi(A))\). The subgroups studied in the article include GL\(_n\), SL\(_n\), unitary and special unitary groups, and also a semigroup of matrices with a bounded rank. They then proceed to show that multiplicative maps, defined on the sub(semi)groups mentioned above, which preserve \(k\) numerical range, \(k\in\{1,\dots,n-1\}\), are unitary similarities. The main idea of the proofs is to utilize the known forms of multiplicative maps on most of the above mentioned sub(semi)groups. The authors add to the list a classification of nonconstant multiplicative maps from a special unitary group into \(M_n(\mathbb C)\). They show that every such multiplicative map extends to an additive isomorphism on \(M_n(\mathbb C)\).