Exponential fitting. With CD-ROM. (Q4679151)

Much of numerical analysis deals with approximations to functions for the purpose of interpolation, differentiation, quadrature and the solution of differential equations. The use of exponential and trigonometric functions as additions or alternatives to the classical use of polynomials in these approximations is now a well-established technique. Applications abound and range from the need to solve specific problems which fit naturally into the exponential fitting mode to modifications to existing approximations which naturally describe a possible exponential growth or periodicity. A notable early work in this area, and arguably a starting point for much of the later work, is the paper by \textit{W. Gautschi} [Numer. Math. 3, 381--397 (1961; Zbl 0163.39002)]. A central premise of this early and, of much that has followed, is that extensions of polynomial fitting to exponential and trigonometric basis functions is possible and, furthermore, it is often desirable. The authors of the book under review have each contributed significantly to the subject of exponential fitting. Furthermore they are deeply immersed in all aspects of the subject and the work is written with both authority and clarity. There is a nice balance between theory and applications and the results are supported both by algorithms in the form of Fortran codes and by the results of numerical experiments. As a starting point for each topic area, the classical theory is introduced and the exponential approach is presented within this context. The work is organized into six chapters starting with ``Introduction'', ``Mathematical properties'' and ``Construction of formulae for functions with oscillatory or hyperbolic variation''. Although some specific problems are introduced to motivate the third chapter, a road-map referred to as a six-step procedure is presented. The six steps referred to provide a general purpose approach to the construction of exponentially fitted approximations. Chapter 4, ``Numerical differentiation, quadrature and interpolation'', explains how to widen the usefulness of classical techniques for these problem areas. The remaining half of the book deals with initial value problems for ordinary differential equations with the chapters broken up along the traditional lines of ``Linear multistep solvers for ordinary differential equations'' and ``Runge-Kutta solvers for ordinary differential equations''. Many of the techniques discovered over the 40 or more years that exponential fitting has become popular have been discovered and rediscovered more than once but this book signals the maturity that this subject has now reached.