RUNGE–KUTTA METHODS FOR RETARDED FUNCTIONAL DIFFERENTIAL EQUATIONS (Q5700340)

This paper is concerned with the numerical solution of initial value problems for retarded functional differential equations \( x'(t)= f(t, x_t),\) where the \(d\)-dimensional vector valued function \( x_t\) is defined by \(x_t(\theta)= x(t+\theta),\) \(\theta \in [-r,0]\) with some constant \( r \geq 0\), and \(f\) is assumed to be sufficiently smooth so that for all initial conditions \((\sigma, \varphi),\) \( \sigma \in {\mathbb R},\) \(\varphi \in C([-r,0], {\mathbb R}^d)\) there exist a unique (maximal) solution \( x =x(\sigma, \varphi): [\sigma - r, t) \to {\mathbb R}^d\). The methods under consideration are \(s\)-stage Runge-Kutta methods defined by the Butcher coefficients \( A(\cdot)\in {\mathbb R}^{s \times s},\) \( b(\cdot) \in {\mathbb R}^s,\) \( c \in {\mathbb R}^s, (c_i \geq 0)\), where \( A, b \) are polynomial matrix and vector functions, respectively, and \(c\) a constant vector, so that starting from \( t_0=0\) the method \( ( A,b,c )\) defines a continuous approximation \(\eta ( \alpha h), \alpha \in [0,1]\) to the exact solution for \( t \in [0,h]\) given by \( \eta (\alpha h) = \varphi (0) + h \sum_{i=1}^s b_i( \alpha) K_i \) with \( K_i= f( \sigma +c_i h, Y_{h c_i}^i)\) and \( Y^i : [-r,c_i h] \to {\mathbb R}^d\) defined by \( Y^i( \theta )= \varphi (\theta ),\) \( \theta \in [-r,0],\) with \( Y^i ( \alpha h) = \varphi (0) + h \sum_{j=1}^s a_{ij} (\alpha) K_j,\) \( \alpha \in [0,c_i]\). Note that in contrast with earlier approaches to retarded functional equations with continuous Runge-Kutta methods in which only \( b( \cdot)\) is a polynomial function, here \(A\) is allowed to have the same type of dependence. In this context the authors establish a theory of (discrete and uniform) order for the proposed methods and derive some particular methods of low order.