Analysis and numerical solution of linear delay differential-algebraic equations (Q291911)

This clear and well-structured paper is devoted to the numerical analysis of delay differential algebraic equations, with one single delay parameter \(\tau\). The studied systems write \[ E(t)\dot{x}(t)=A(t)x(t)+B(t)x(t-\tau)+f(t) \] so are linear w.r.t. the unknown \(t \mapsto x(t)\). It is known, that a method of steps allows the numerical solution of causal DDEs. However it fails for noncausal ones, when the discretization takes into account future points to get the current points. In this paper, the theoretical part (on existence and uniqueness) relies on the analysis of the associated DAE (and its differential or strangeness index) \[ E(t)\dot{x}(t)=A(t)x(t)+T(t)\lambda(t)+f(t). \] The delayed term is considered as a parameter function \(\lambda(t)\), to study a pure DAE. A completion stage makes it strangeness-free : it needs formal differentiation of the initial DEs and introduces new constraints. However the remodeled DAE gets stable under differentiation. A second concept is a kind of stability after successive application of the forward shift operator: it corresponds to study the uniqueness of solutions of associated DAEs on shifted intervals of width \(\tau\). Upon convenient hypotheses, a consistency theorem is given for retarded DDAEs. The main algorithmic part is the exposition of a a constructive method to solve general linear DDAEs numerically. It generalizes the method of steps for neutral or retarded DDAEs. The method is effective, based on linear algebra and it applies to causal linear DDAEs while a reformulation stage should be operated to handle noncausal DDAEs. As a conclusion this paper gives essential concepts, definitions and methods to understand the complexity of linear DDAEs and their numerical solutions.