Criteria for exponential dichotomy for triangular systems (Q2347153)

It is well-known that linear nonautonomous ordinary differential equations \(\dot x=A(t)x\) can be brought into (w.l.o.g.) upper-triangular form \[ (\ast)\qquad\dot y=R(t)y \] by means of a Lyapunov transformation. This kinematic similarity leaves the property of an exponential dichotomy (ED for short), and consequently also the Sacker-Sell spectrum \(\Sigma\) invariant. While it is not difficult to determine dichotomy properties (and, in turn, \(\Sigma\)) for diagonal systems, an interesting question is to relate corresponding properties of \((\ast)\) and its diagonal part. This is particulary relevant for numerical schemes to determine \(\Sigma\) (see, for instance, \textit{L. Dieci} and \textit{E. S. Van Vleck} [J. Dyn. Differ. Equations 19, No. 2, 265--293 (2007; Zbl 1130.34027)]). The authors of this interesting paper study upper-block-triangular systems of linear ODEs. On the half line it is shown that an ED of the entire system extends to the diagonal systems (Theorem 1) and conversely (Theorem 2, provided the super-diagonal block is bounded). Also the spectrum of a block-triangular system coincides with its diagonal part. The situation on the whole real line is more subtle: An ED of the entire system is equivalent to (i) dichotomies of the diagonal blocks, (ii)~the indices of the diagonal bocks sum to \(0\) and (iii) the nonexistence of nontrivial bounded entire solutions. The whole line spectrum lies between the union of the two half line spectra and the whole line spectrum of its diagonal. Finally, a procedure to determine the whole line spectrum is given.