Suborthogonality and orthocentricity of matrices (Q958023)

The paper is closely related to the author's former work. In the first part of the paper, some results on submatrices of orthogonal and unitary matrices are presented. For instance, the following result is very interesting. Theorem 2.2. Every \(p \times q\) submatrix of a unitary (or, orthogonal) \(n \times n\) matrix has at least \(p + q - n\) singular values equal to one and the remaining singular values less than one. Conversely, if \(A\) is a \(p \times q\) matrix that has \(k\) singular values equal to one and the remaining \(\min(p, q) - k\) singular values less than one, then for every \(m \geq p + q - k\) there exists a unitary (or, orthogonal) \(m \times m\) matrix containing \(A\) as a submatrix, and for no \(m\) smaller than \(p + q - k\) does such matrix exist. The second part of the paper is devoted to orthocentric matrices. These matrices are introduced in this paper. Theorem 2.6 gives four equivalent statements for a real \(n \times (n+1)\), \(n\geq 2\), matrix and every matrix satisfying these conditions is said to be orthocentric. We repeat here a part of Theorem 2.6 as a definition. A real \(n \times (n+1)\), \(n\geq 2\), matrix \(A\) is orthocentric if there exists a nonsingular diagonal matrix \(D\) such that \((AD)(AD)^T=I_n\) and the nonzero vector \(y\) for which \(Ay=0\) has all coordinates different from zero. A simplex in the Euclidean space is said to be orthocentric if altitudes meet in one point, the orthocenter. If the orthocenter is an interior point of the simplex, then the simplex is called a positive orthocentric simplex. The following result (Theorem 3.5) explains the relation between orthocentric matrices and simplex geometry in Euclidean space. A real \(n \times (n + 1)\) matrix \(A\) is orthocentric if and only if it is a matrix of coordinates of \(n + 1\) directions, listed as columns, of the \(n + 1\) normals to \((n - 1)\)-dimensional faces of some positive orthocentric \(n\)-simplex.