Differential analysis of matrix convex functions (Q865452)

The authors analyze by differential methods matrix convex functions of a fixed order defined on a real interval \(I.\) In the first section they prove necessary global, and sufficient local conditions for \(n\)-convexity in terms of the (semi-)positive definiteness of the matrix \[ K_n(f;t)=\left( f^{i+j}(t)\over (i+j)!\right)^n_{i,j=1} \] at each point \(t\in I,\) or at a point \(t_0\in I.\) The sufficient conditions prove to be global if \(n=2.\) Moreover, they investigate the gaps existing between classes of matrix convex functions of successive orders, and they provide explicit examples of the type of functions contained in these gaps. Indeed, they prove that while there exist \(n\)-concave (or \(n\)-convex) and \(n\)-monotone polynomials of degree \(m\) if \(m\geq 2n,\) there are no \(n\)-convex polynomials of degree \(m\) in \(I,\) for \(m=3,\dots,2n-1.\) In Section 2 they conjecture that \(n\)-convexity is a local property; they prove, under suitable assumptions, a characterization of \(2\)-convex functions, and they obtain as a corollary the proof of their conjecture in the case \(n=2.\) In the last section they introduce a fractional transformation that relates the set of matrix convex functions of each order \(n\) with the set of matrix convex functions of the following order \(n+1.\)