The finite-step realizability of the joint spectral radius of a pair of \(d \times d\) matrices one of which being rank-one (Q445812): Difference between revisions

Let \(\mathbf{S}=\{S_1,\dots,S_K\}\subset\mathbb{R}^{d\times d}\), \(d,K\geq 2\), let \(\rho(A)\) denote the spectral radius of \(A\in\mathbb{R}^{d\times d}\), and denote \(\mathbb{K}=\{1,\dots,K\}\). The (generalized) spectral radius of~\(\mathbf{S}\) is defined by \[ \rho(\mathbf{S})=\limsup_{n\to\infty}\max_{(i_1,\dots,i_n)\in\mathbb{K}^n}\rho(S_{i_1}\cdots S_{i_n})^\frac{1}{n}. \] Motivated by \(\rho(A)=\rho(A^m)^\frac{1}{m}\) for all \(m\geq 1\), the spectral finiteness property (''sfp'' in the sequel) of~\(\mathbf{S}\) is defined as follows: There exists \((i_1^*,\dots,i_m^*)\in\mathbb{K}^m\) for some \(m\geq 1\) such that \[ \rho(\mathbf{S})=\rho(S_{i_1^*}\cdots S_{i_m^*})^\frac{1}{m}. \] Several authors disproved the conjecture that every~\(\mathbf{S}\) has the sfp. Finally \textit{K.~G.~Hare}, \textit{I.~D.~Morris}, \textit{N.~Sidorov} and \textit{J.~Theys} [Adv. Math.~226, 4667--4701 (2011; Zbl 1218.15005)] gave an explicit counterexample. \smallskip The present authors prove that if \(K=2\) and \(\mathrm{rank}\,S_1=1\) or \(\mathrm{rank}\,S_2=1\), then \(\mathbf{S}\) has the sfp. They also prove that under these assumptions \[ \rho(\mathbf{S})=\max_{l,m\in\mathbb{Z}_+}\rho(S_1^lS_2^m)^\frac{1}{l+m}, \] where \(\mathbb{Z}_+=\{0,1,2,\dots\}\). So the stability of the switched system induced by \(\{S_1,S_2\}\) is algorithmically decidable in this case.