Reduced stability of parameter-dependent matrices (Q1375099)

This paper examines stability properties of parameter-dependent linear systems \(\dot z=A (\varepsilon)z\) where \(A(0)\) is a block diagonal \((n+m) \times (n+m)\) matrix consisting of an \(n+n\) zero matrix and an \(m+m\) stable matrix \(F\). The authors partition \(A(\varepsilon)\), all of whose entries are smooth functions of \(\varepsilon\), into blocks: diagonal blocks \(C(\varepsilon)\) and \(F(\varepsilon)\), and off-diagonal blocks \(E(\varepsilon)\) and \(D(\varepsilon)\). Their aim is then to obtain stability conditions on \(A(\varepsilon)\) near \( \varepsilon=0\) in terms of the \(n\times n\) reduced matrix \(G(\varepsilon) =C- DF^{-1}E\), which is assumed to be nonsingular for \(0\leq \varepsilon< \varepsilon_0\) for some \(\varepsilon_0\). The matrix \(G(\varepsilon)\) is the Schur complement of \(F\) in \(A\). The authors provide sufficient conditions for \(A(\varepsilon)\) and \(G(\varepsilon)\) to have the same stability properties; these depend on an asymptotic expansion of \(G(\varepsilon)\). Explicit necessary and sufficient conditions are provided when \(F=-I\) and \(n=2\) in terms of the location of the spectrum of \(G(\varepsilon)\).