Semidefiniteness without real symmetry (Q1971047)

Let \(A\) be an \(n{\times}n\) matrix with real entries. The authors show that a necessary and sufficient condition for \(A\) to have positive semidefinite or negative semidefinite symmetric part \(H(A) = \frac{1}{2}(A+A^T)\) is \[ \text{rank}[H(A)X] \leq \text{rank}[X^{T}AX]\quad \text{for all }X \in M_n({\mathbb R}). \] Further, if \(A\) has positive semidefinite or negative semidefinite symmetric part, and \(A^2\) has positive semidefinite symmetric part, then \[ \text{rank}[AX] = \text{rank}[X^{T}AX] \quad \text{for all }X \in M_n({\mathbb R}). \] This result implies the usual row and column inclusion property for positive semidefinite matrices. Finally, the authors show that if \(A, A^2, \ldots, A^k (k \geq 2)\) all have positive semidefinite symmetric part, then \[ \text{rank}[AX] = \text{rank}[X^{T}AX] = \cdots = \text{rank}[X^{T}A^{k-1}X]\quad \text{for all }X \in M_n({\mathbb R}). \]