Power Additivity and Orthogonality (Q4243489)

Let \(A_{1},A_{2},\dots,A_{s}\) be a family of complex \(n\times n\) matrices with sum \(A\). The family is called power additive if \(A^{k}=A_{1} ^{k}+A_{2}^{k}+\dots+A_{s}^{k}\) for \(k=1,2\)\dots; it is rank additive if \(\text{rank}(A)= \text{rank}(A_{1})+ \text{rank}(A_{2})+\dots+ \text{rank}(A_{s})\); and is orthogonal if \(A_{i}A_{j}=0\) for all \(i\neq j.\) The authors explore the relationships between these three properties. Clearly orthogonality implies power additivity but not conversely, even in the \(2\times 2\) case. Neither rank additivity nor orthogonality implies the other in general, but a theorem of \textit{W. G. Cochran} [Proc. Cambridge Philos. Soc. 30, 178-191 (1934; Zbl 0009.12004); see \textit{P. Šemrl}, Linear Algebra Appl. 237/238, 477-487 (1996; Zbl 0847.62043)]shows that a rank additive family for which \(A\) is idempotent is an orthogonal family. The main results in the present paper are the simplified block matrix structures which can be obtained for the \(A_{i}\) under (simultaneous) similarity transformations assuming that one or more of the properties hold. The authors use these to obtain general relations between the properties. For example, if \(A\) has the property that every rank additive family with sum \(A\) is orthogonal, then \(A^{2}=\lambda A\) for some scalar \(\lambda\) (Theorem 2.1). It is also shown that orthogonality is equivalent to power additivity if each \(A_{i}\) has a group inverse (Theorem 3.1).