A simultaneous decomposition of a matrix triplet with applications. (Q2889370)

The paper concentrates on a matrix expression \(A-BXB^{*}-CYC^{*}\), where \(A\) is a given square Hermitian or skew-Hermitian matrix, \(B, C\) are given (generally rectangular) matrices and \(X, Y\) are variable matrices. First, motivated by the generalized singular value decomposition of two matrices a simultaneous decomposition of the matrix triplet \((A,B,C)\) is derived. Using this result, closed formulas for the maximal and minimal possible ranks of the matrix \(A-BXB^{*}-CYC^{*}\) are given under the assumption \(X = \pm X^{*}, Y = \pm Y^{*}\). Moreover, necessary and sufficient conditions for existence of two Hermitian or skew-Hermitian solutions of the equation \(A=BXB^{*}+CYC^{*}\) are derived. These conditions can be simply verified and involve ranks of the individual matrices \(A,B,C\). Formulas for the solutions are also presented. Finally, some further open problems related to the considered expression \(A-BXB^{*}-CYC^{*}\) are summarized.