We discuss two notions of hyperbolicity for finite-time linear differential equations. The first notion (Dhyperbolicity) is based on the dynamic (or EPH) partition, the second (M-hyperbolicity) is motivated by exponential dichotomies. We study conditions under which D-hyperbolicity implies M-hyperbolicity.
D-hyperbolicity and M-hyperbolicity on finite-time intervalsWith t − , t + ∈ R satisfying t − < t + , let I = [t − , t + ] and consider the finite-time linear differential equatioṅLet Φ :Define strain tensor and strain acceleration tensor S, M : I → R d×d of (1) by
(t) + A(t) , M(t) :=Ṡ(t) + S(t)A(t) + A(t) S(t) for all t ∈ I.The zero-strain tensor of (1) is defined asWith these quantities, recall two notions of hyperbolicity (see [1][2][3] for details, as well as [5,6] for more general background information on finite-time dynamics and its applications). Definition 1.1 (D-Hyperbolicity) System (1) is called D-hyperbolic on I if, for all t ∈ I, S(t) is indefinite and non-degenerate, and ξ, M (t)ξ > 0 for all ξ ∈ Z(t). Definition 1.2 (M-Hyperbolicity) System (1) is called M-hyperbolic on I if there exist constants α < 0 < β and an invariant projector P , i.e., the mapping P : I → R d×d is projection-valued with P (t)Φ(t, s) = Φ(t, s)P (s) for all t, s ∈ I, such that
Main ResultWe show that M-hyperbolicity follows from D-hyperbolicity if all eigenvalues of the symmetric part of (1) stay strictly separated and only one single eigenvalue has positive (or negative) sign for all t ∈ I.
Theorem 2.1 Suppose that there exists a differentiable orthogonal transformation
. , λ d (t) and S(t) has, for each t ∈ I, one positive eigenvalue and d − 1 negative eigenvalues, or vice versa. Then (1) is M-hyberbolic whenever it is D-hyperbolic.
the eigenvalues of the symmetric matrix S(t).Since (1) is D-hyperbolic, S(t) is not degenerate for all t ∈ I, and it follows that λ i (t) > 0 or λ i (t) < 0 for all 1 ≤ i ≤ d and t ∈ I. We assume that S(t) has 1 negative eigenvalue and d − 1 positive eigenvalues, i.e. λ d (t) < 0 < λ d−1 (t) ≤ . . . ≤ λ 1 (t) for all t ∈ I. For convenience, we divide the proof into four steps.Step 1: With the evolution operator Φ of (1) we define X(t) := diag λ 1 (t), . . . , λ d−1 (t), −λ