On complementary matrix algebras (Q909749): Difference between revisions

The following conjecture made in the book of \textit{I. Gohberg}, \textit{P. Lancaster} and the reviewer: Invariant subspaces of matrices with applications (1986; Zbl 0608.15004) is proved to be false in general: If \(A_ 1\) and \(A_ 2\) are algebras of linear operators on \({\mathbb{C}}^ n\) that are complementary (i.e., \(A_ 1\cap A_ 2=\{0\}\) and \(A_ 1+A_ 2\) is the algebra of all operators on \({\mathbb{C}}^ n)\) then there exist nontrivial complementary subspaces \(M_ 1\) and \(M_ 2\) of \({\mathbb{C}}^ n\) with \(M_ i\) invariant under \(A_ i\). Namely, for each integer \(n\geq 4\) (except possibly 7 and 11) there exist complementary algebras of operators on \({\mathbb{C}}^ n\) with no nontrivial complementary invariant subspaces. The conjecture is proved, under some additional hypotheses (for example, when \(A_ 2\) is the orthogonal complement to \(A_ 1)\).