Unitary dilation approach to contractive matrix completion. (Q1426312)

The authors discuss a contractive matrix completion that given a matrix \(M\) which has a block matrix representation \(M=(M_{ij})_{l\times l}\) characterizes all block lower triangular matrices of the form \[ T=\left [\begin{matrix} T_{11}&0&\cdots &0\\T_{21}&T_{22}&\cdots&0\\ \cdots &\cdots &\cdots &\cdots\\T_{l1}&T_{l2}&\cdots &T_{ll}\end{matrix}\right] \] satisfying \(\| M+T\| <1\). Many well-known statements which are equivalent to that there exists a matrix \(T\) such that \(\| M+T\| <1\) are given. For example, if all of the \(n\times n\) matrices are denoted by \({\mathcal{M}}(\mathbb{C}^n)\), there exists a matrix \(T\) such that \(\| M+T\| <1\) if and only if there exists a matrix \(P\in {\mathcal{M}}(\mathbb{C}^{2n})\) and \[ P=\left [\begin{matrix} P_{11}&P_{12}\\ P_{21}&P_{22}\end{matrix}\right], \] where \(P_{ij}\in {\mathcal{M}}(\mathbb{C}^n)\), \(1\leq i,j\leq 2\), with \(P_{12}\) and \(P_{21}\) are both invertible such that \[ \left[\begin{matrix} M+P_{11}&P_{12}\\ P_{21}&P_{22}\end{matrix}\right] \] is unitary. The main result in this paper is to give a new procedure to construct such a matrix \(P\). As pointed out by authors, the new procedure is numerically more efficient and reliable.