The normal defect of some classes of matrices (Q1947115)

An \(n \times n\) matrix \(A\) has a normal defect of \(k\), \(\operatorname{nd}(A)=k\), if there exists an \((n+k) \times (n+k)\) normal matrix \(A_{ext}\) with \(A\) as a leading principal submatrix and \(k\) minimal. \(A_{ext}\) is called a minimal normal completion of \(A\). In a previous paper, it was shown that the normal defect of a matrix \(A\) is bounded by \[ \max{\{i_+[A,A^*], i_-[A,A^*]\}} \leq \operatorname{nd}(A) \leq \operatorname{rank}(\|A\|^2I_n-A^*A), \] where \([A,A^*]=AA^*-A^*A\) and \(i_{\pm}(M)\) refers to the number of positive/negative eigenvalues of \(M=M^*\). In this paper, the authors compute the normal defect of a special class of \(4 \times 4\) matrices, namely matrices whose only nonzero entries lie on the superdiagonal, and they provide details for constructing minimal normal completion matrices \(A_{ext}\). Specifically, they prove that for this type of matrices \(\operatorname{nd}(A)=\max{\{i_+[A,A^*], i_-[A,A^*]\}}\). In addition, the authors also present results on the normal defect of block diagonal matrices. Given a block diagonal matrix \(A=\operatorname{diag}(A_i)_{i=1}^m\), where \(A_i \in \mathbb{C}^{n_i \times n_i}\), they obtain necessary and sufficient conditions that guarantee \(\operatorname{nd}(A)=\sum_{i=1}^m \operatorname{nd}(A_i)= \max{\{i_+[A,A^*], i_-[A,A^*]\}}\). Finally, they show an example of a \(6 \times 6\) block diagonal matrix having the property that its normal defect is strictly less than the sum of the normal defects of each of its blocks, and they provide sufficient conditions for when the normal defect of a block diagonal matrix is equal to the sum of the normal defects of each of its blocks.