The dichotomy theorem for evolution bi-families (Q949635)

Let \(X\) be a Hilbert space and let \(G\) denote the closure of the differential operator \(-d/dt+D(t)\), acting on (a maximal subspace of) \(L^2(\mathbb R, H)\). The authors consider the order zero term in an additive perturbative form \(D(t) = A +B(t)\), where \(A\) is the generator of a bilateral semigroup and \(B(t)\) is a bounded piece-wise strongly continuous operator valued function. The natural assumption is imposed that, if a function \(f\) belongs to \(\ker G+\ker G^\ast\) and vanishes at a point \(t_0\), then \(f\) is identically zero. The main result of the article asserts that the operator \(G\) is Fredholm if and only if the not well-posed equation \(u'(t)=D(t)u(t)\), \(t\in\mathbb R\), has an exponential dichotomy of \([0,\infty)\) with projections \(P_+(t)\), \( \geq 0\), and an exponential dichotomy on \((-\infty,0]\) with projections \(P_-(t)\), \(t\leq 0\), and at \(t=0\) the pair of subspaces \((\text{Ran}\,P_+(0),\text{Ker}\,P_-(0))\) is Fredholm. Moreover, \[ \text{ind} G = \text{ind} (\text{Ran}\, P_+(0), \text{Ker}\, P_-(0)). \] A series of illuminating examples and particular cases are discussed in detail. The present article is a natural continuation of the authors' recent works, see [\textit{Y.\,Latushkin, A.\,Pogan} and \textit{R.\,Schnaubelt}, J.~Oper.\ Theory 58, No.\,~2, 387--414 (2007; Zbl 1134.47302)] and the references cited there, and contributions like [\textit{B.\,Sandstede} and \textit{A.\,Scheel}, Math.\ Nachr.\ 232, 39--93 (2001; Zbl 0994.35025)].