Min matrices and mean matrices (Q642130)

One of the main topics of this beautiful paper is the so-called min matrix \(M = (m_{ij}) \in \mathbb{N}^{n \times n}, \; m_{ij} = \min (i,j)\), and some of its generalizations. The author starts with six proofs that the min matrix is positive semidefinite (psd). The tools used are from different parts of mathematics, e.g., linear algebra, analysis, operator theory, and differential equations. For every positive integer \(m\), the \(m\)th Hadamard power \(A^{\circ m} = (a_{ij}^m)\) of a psd matrix \(A\) is again psd. If the entries \(a_{ij}\) are nonnegative numbers, the \(m\)th fractional Hadamard power of \(A\) is given by \(A^{\circ 1/m} = (a_{ij}^{1/m})\). A nonnegative matrix \(A\) is said to be infinitely divisible if, for each \(m\), \(A^{\circ 1/m}\) is psd. It is shown that min matrices are infinitely divisible which is a stronger property than psd. Two of the above mentioned proofs motivate the explicit computation of the eigenvalues of the matrix \(M\). Attached are interesting historical background information as well as applications to vibrating strings and electric waves. In the third section, three families of matrices related to the (generalized) min matrix are investigated with respect to infinite divisibility. The final topic of the paper concerns so-called mean matrices, whose \((i,j)\) entry is a function of \(m(i,j)\), where \(m(.,.)\) denotes a mean function (e. g., arithmetic, geometric or harmonic means, to name just a few). If one takes the mean \(m(i,j) := \min (i,j),\) the min matrix \(M\) turns out to be a special mean matrix. Based on the previously derived properties, it is shown for a variety of means that the corresponding mean matrices are infinitely divisible.