scientific article; zbMATH DE number 1232440

From MaRDI portal
Publication:4223085





The book is the first part of a large monograph devoted to numerical methods for solving singularly perturbed equations, i.e. differential equations with a small parameter at the highest derivatives. The presence of a small parameter implies that, in a solution to the problem, along with a smooth component, a boundary layer component also occurs varying fast in a small neighborhood about the boundary; the latter is called a boundary layer. The authors describe grid methods for solving linear singularly perturbed first-order ordinary differential equations and boundary value problems for second-order ordinary differential equations with selfadjoint operator. In Chapter 1, the authors deal with the Cauchy problem \[ \varepsilon u'+a(x) u=f(x), \quad x \in (0,1), \qquad u(0)=u_0, \] where \(\varepsilon\) is a small parameter. In Chapter 2, finite-difference methods for the following boundary value problem are studied: \[ -\varepsilon^2 u+a(x) u=f(x), \quad x \in (0,1),\qquad u(0)=u_0, \quad u(1)=u_1. \] In Chapter 3, a finite-element approach is presented for the latter problem. Of course, from the viewpoint of modern mathematical physics, these problems are primitive, but, nevertheless, they bear on themselves some features of the equations of parabolic and elliptic types and have some characteristic features of a singular degeneration. The simplicity of the equations under consideration gives the possibility of realizing a detailed theoretical analysis for the numerical methods under study. Standard finite-difference methods or finite-element methods on uniform grids give an unsatisfactory accuracy in view of huge error of approximation inside a boundary layer; therefore, some special methods are required. Assuming that the coefficients of the differential equations are smooth, the authors construct and analyze grid methods that converge both with high order and uniformly with respect to the singular perturbation parameter. In each chapter of the book, before considering these methods, the authors study the differential properties of a solution and of the functions of a boundary layer. Then the derivation is subsequently presented of an acceleration of convergence of finite-difference and finite-element schemata. The consideration of these methods is organized in decreasing order of a priori information about a solution. The exhaustion method for a boundary layer is used under the greatest information about a solution, when the necessary terms of the boundary layer component are known, and the problem is reduced to finding the smooth component by means of standard methods. In the situation when the form of the boundary layer component is known, but some of its parameters are unknown, the exponentially fitted method is used for finding both components of the solution. Finally, under the least information about a solution, when only the degree of growth of the derivatives is known, various variations of the method of adaptive grids are used. Similar methods were studied earlier, for example, in the book by \textit{H.-G. Roos}, \textit{M. Stynes} and \textit{L. Tobiska} [Numerical methods for singularly perturbed differential equations. Convection-diffusion and flow problems. Springer Series in Computational Mathematics. 24. Berlin: Springer (1996; Zbl 0844.65075)]. The book serves both as an introduction to the topic and as an overview of available results. The book is well written and will be valued both by graduate students and experts in this field as a handy monograph.











This page was built for publication:

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4223085)