Richardson extrapolation. Practical aspects and applications (Q2401705)

This book is dedicated to Richardson extrapolation, which is one of the most well-known methods for the acceleration of a slowly converging sequences. It was first introduced by Lewis Fry Richardson (1881--1953) in 1910 and quickly became very popular among scientists and engineers, particularly in their treatment of advanced large-scale mathematical models. Indeed, Richardson extrapolation is a very reliable and robust technique, but here the authors point out that this is true \textit{only if and when} it is correctly applied. Therefore they concentrate their efforts on difficulties, which may arise when this mathematical tool is used. Actually, even though the implementation of this procedure seems to be very simple and straightforward, high efficiency and good results are often far from an easy task. The book consists of eight chapters. Chapter 1 introduces the definition of Richardson extrapolation for an arbitrary time-dependent numerical method for solving systems of ordinary differential equations (ODEs). The main conclusion is that the combination of any numerical method with Richardson extrapolation results in a new more accurate numerical method, \textit{only if} the new numerical method \textit{is stable}. Indeed the preservation of the stability of the computational process is a key issue of the authors. In Chapter 2 it is shown that \textit{for some classes} of explicit Runge-Kutta methods the application of Richardson extrapolation always results in new numerical methods with \textit{better} stability properties. In practice, these new numerical methods exhibit larger absolute stability regions than those of the underlying classical explicit Runge-Kutta methods, when the number of stages \(m\) is equal to the order of accuracy \(p\). Chapter 3 is devoted to linear multistep methods and predictor-corrector methods. The basic properties of these methods are presented and it is shown why the use of Richardson extrapolation in connection with linear multistep methods is the least difficult. However, it is shown also that the use of carefully chosen predictor-corrector schemes is similar to the result achieved by using Richardson extrapolation. Chapter 4 deals with the combination of some implicit numerical methods for solving stiff ODEs with Richardson extrapolation. Several more restrictive definitions of stability are introduced when they are related to the stability of stiff ODEs and some examples are presented to illustrate the usefulness of the obtained results when badly scaled, extremely ill-conditioned and very stiff atmospheric chemical schemes implemented in many well-known large-scale air pollution models are to be computed. Preserving the stability of the computational process is the central aim, as in the previous chapters. Chapter 5 treats the performance of Richardson extrapolation in connection with two well-known and commonly used splitting procedures. The authors are again mainly involved in the preservation of the stability properties. In particular, they suggest to carefully investigate the stability properties of the combined numerical methods since in this case the stability of the underlying methods cannot guarantee the stability of the combined methods when Richardson extrapolation is also used. Examples are reported to enlighten the stability issue. Chapter 6 describes the implementation of Richardson extrapolation in connection with some classes of systems of partial differential equations. A particular application of Richardson extrapolation is discussed within the context of some linear advection problems arising, for example, in air pollution modelling, but also in some other areas in fluid dynamics. Some appropriate numerical examples show the usefulness of the results obtained in this chapter for improving the accuracy of the calculated approximations. Whereas time-dependent problems are treated in the first six chapters, Chapter 7 reports some results on the application of Richardson extrapolation to the treatment of algebraic and trascendental equations or systems of such equations by iterative methods. This chapter also includes the application of Richardson extrapolation in numerical integration, where it reveals a powerful approach when the composite trapezoidal rule is the underlying method. Chapter 8 contains general conclusions and final remarks; in particular the authors emphasize that Richardson extrapolation may lead to unstable computations. The book is well-written, clear and rich in theoretical details, numerical tables and examples; even some research problems are listed at the end of each chapter to help researchers in their work. Indeed, this book may be of great help in improving the performance of Richardson extrapolation to solve any reader's particular task.