A fresh variational-analysis look at the positive semidefinite matrices world (Q438781)

This beautifully written paper reviews a very elementary notion, positive semidefiniteness of symmetric real matrices, using an original approach: variational analysis. Besides reviewing the most classical properties of positive semidefinite matrices, it also presents some less well known properties (for instance, the characterization of the positive semidefiniteness of a symmetric matrix by the nonnegativity of its principal invariants) and even an original result, namely, a necessary and sufficient condition on \(x^1,\dots,x^k,y\in \mathbb{R}^n\) for the existence of positive semidefinite symmetric \(n\times n\) matrices \(M_i\) \((i=1,\dots,k)\) such that \(M_{i}y=x^{i}\) for all \(i=1,\dots,k\) and \(\sum_{i=1}^{k}M_{i}=I_{n}\). A constructive proof of the characterization of the existence of positive definite matrices satisfying these conditions is also given. Another result for which a new proof is presented is the one due to \textit{S.-P. Han} and \textit{O. L. Mangasarian} [Linear Algebra Appl. 56, 89--103 (1984; Zbl 0529.15013)] stating that, given a closed convex cone \(K\) in \(\mathbb{R}^{n}\) with polar cone \(K^{o}\), a symmetric \(n\times n\) matrix \(A\) is positive definite if and only if \(\langle Ax,x\rangle >0\) for all \(x\in K\setminus \{0\}\) and \(\langle A^{-1}x,x\rangle >0\) for all \(x\in K^{o}\setminus \{0\}\). The new proof relies upon Moreau's decomposition theorem. An interesting section deals with the geometry of the set of positive semidefinite symmetric \(n\times n\) matrices, which is a closed convex cone in the space of symmetric \(n\times n\) matrices. The above is just a small sample of the many features of positive semidefiniteness discussed in the paper.