Large time behaviour of linear functional differential equations (Q1417438): Difference between revisions

The authors consider linear autonomous functional-differential equations of the type \[ {d\over dt} Dx_t= Lx_t,\quad t\geq 0,\quad x_0= \phi, \] with \(x_t(\theta)= x(t+ \theta)\), \(-r\leq\theta\leq 0\), \[ L\phi= \int^0_{-r} d\eta(\theta\phi(\theta),\quad D\phi= \phi(0)- \int^0_{-r} d\mu(\theta)\phi(\theta). \] Here, \(\theta\), \(\mu\) are \(n\times n\)-matrices. Let the \(C_0\) semigroup \(T(t)\) be defined by \(T(t)\phi= x_t(\cdot;\phi)\) where \(x(\cdot;\phi)\) is the solution of the FDE and let \(A\) be the infinitesimal generator of \(T(t)\). Let \(\lambda\) be an eigenvalue of \(A\) and let \(P_\lambda\) denote the spectral projection onto the corresponding eigenspace \(M_\lambda\). The authors use Dunford calculus including a formula for the resolvent of \(A\) to represent the spectral projection. In the case where \(A\) has a dominant simple eigenvalue this is used to estimate the large time behavior of the solution. A similar approach using Dunford calculus is applied to the linear scalar periodic equation \[ {dx(t)\over dt}= a(t)x(t)+ \sum^m_{j=1} b_j(t) x(t- r_j), \] for \(t\geq s\) and \(x_s=\phi\), where \(a\), \(b_j\) are periodic with period \(\omega\) and \(r_j= j\omega\).