How small can nonzero commutators be?

From MaRDI portal





The problem of the value of a commutator \(AB-BA\) for \(A,B \in {\mathcal G}\), where \({\mathcal G}\) denotes a group of complex matrices, is investigated. This problem leads to the question if \(\rho (AB-BA)\), where \(\rho\) means the spectral radius, can be arbitrarily small compared to \(\rho (A) \rho (B)\) for all pairs \(A, B\) without being identically zero. It is shown that if \(\rho (AB - BA) \leq \epsilon = \sqrt{3}\) on a compact or connected group \({\mathcal G}\) then \({\mathcal G}\) is triangularizable and for the compact case \({\mathcal G}\) is abelian. In this case the bound \(\epsilon\) on the spectral radius of commutators is zero if it is less then \(\sqrt{3}\). In case of more general groups or semigroups the modified sufficient condition of the form \(\rho(AB-BA) \leq \epsilon \rho(A) \rho(B)\) should be satisfied for positive \(\epsilon\) and for all \(A\) and \(B\) which would imply triangularizability or at least reducibility. It is shown that for a group \({\mathcal G}\) any \(\epsilon < \sqrt{3}\) implies that the commutator subgroup \({\mathcal G'}\) is triangularizable and thus \({\mathcal G}\) is solvable. Certain infinite-dimensional analogues of the given cases are also briefly discussed.











This page was built for publication: How small can nonzero commutators be?

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4681346)