Some remarks on the monotonicity of primary matrix functions on the set of symmetric matrices (Q2630978): Difference between revisions

Let \(\mathrm{Sym}(n)\) denote the set of real symmetric matrices of order \(n\). Let \(I\) be an open interval of \(\mathbb{R}\) and let \(S_I\) denote the set of all symmetric matrices with no eigenvalues outside \(I\). For \(A \in \mathrm{Sym}(n)\), let \(A=Q^TDQ\) be an orthogonal diagonalization, where \(D\) is a diagonal matrix with diagonal entries \(\lambda_1, \lambda_2\dots,\lambda_n\) arranged in decreasing order. A continuously differentiable function \(f:S_I \rightarrow \mathrm{Sym}(n)\) is said to be a \textit{primary matrix function} if \(f(A):=Q^TEQ\), where \(E\) is the diagonal matrix whose diagonal entries are \(f(\lambda_1), f(\lambda_2), \dots, f(\lambda_n)\) arranged in that order. The main result of the authors establishes a close relationship between different concepts of monotonicity for primary matrix functions. It is the equivalence of the statements: {\parindent=0.6cm \begin{itemize}\item[(a)] \(f\) is a strictly increasing function on \(I\); \item[(b)] \(f\) is \textit{Hilbert space monotone} (meaning that \(\langle f(B)-f(A), B-A \rangle > 0\) whenever \(A,B \in \mathrm{Sym}(n)\) with \(A \neq B\)); \item[(c)] \(\langle f(A+H)-f(A), H \rangle > 0\) whenever \(A \in \mathrm{Sym}(n)\) and \(H \in \mathrm{Sym}(n)\) is also positive definite. \item[(d)] It is shown that each of these conditions is implied by the statement: \(f(B)-f(A)\) is positive definite whenever \(B-A\) is positive definite. \end{itemize}}