A new inversion-free iterative method for solving a class of nonlinear matrix equations (Q2081680)

Assume that \(p\ge 1\) is a positive integer, \(A\) and \(B\) are Hermitian positive semidefinite matrices and \(M\in\mathbb{C}^{n\times n}\) is an arbitrary \(n\times n\) complex matrix. Consider the following matrix equations: \[ X^p-MXM^*=Q, \] \[ X^p=MXM^*\text{ and }X^p=A+MX^{-1}M^*, \] where \(Q\) is a Hermitian positive definite matrix. In this paper the authors generalize the above equations as follows: \[ X^p=A+M(B+X^{-1})^{-1}M^*, \] and provide a new algorithm that works for all choices of the positive semidefinite matrix \(B\) even if it is singular. Using a matrix inequality obtained in the paper, the convergence of the proposed algorithm is established.