Accuracy of mathematical models. Dimension reduction, homogenization, and simplification (Q2193521)

The book under review addresses the question whether a given mathematical model (or a suitable derivation thereof) is ``accurate'': Given a mathematical formulation of a certain phenomenon then often this formulation leads to (partial) differential equations that are rather difficult to solve (numerically/efficiently). For example consider a three-dimensional model of a ``thin'' plate. Using engineering experiences, or concrete mathematical derivations, one develops a simplified model maybe a 2-dimensional one, which is -- as it is hoped -- characterised by a more efficient way to solve it. Another aspect of applications discussed in the book is the approximation of solutions of PDEs in low-dimensional function spaces. Then even though quite naturally rather seldom it is asked whether the solution of the simplification agrees well enough with the solution of the equation one started out with. This book focusses therefore precisely on this question. Moreover, the authors are particularly keen to derive estimates that provide quantitative bounds on the error relating the solutions of a certain original model to an approximation (in certain spaces) or to solutions of a simplified model. The book under review consists of 6 chapters and 1 Appendix. After having introduced some notation, the authors provide a comprehensive overview about inequalities related to differential operators such as Friedrichs, Poincaré or Korn type estimates. Given the authors' aim to derive estimates for potential errors as precise as possible, the overview over existing estimates is accompanied by references to works where the bounds in those estimates are as sharp as possible. In particular, constants involved in the above mentioned inequalities are also compared, where possible. The theoretical foundation of the book is laid in Chapter 2, where, based on error identities for saddle point problems, explicit bounds are derived. The diversity and applicability of the developed methods is demonstrated by many substantial applications; such as the various linear and nonlinear models for fluid dynamics. In Chapter 3, the authors apply their findings to compare models with different spatial dimensions. Most prominently, the Laplace equation and models from elasticity are analysed in the above mentioned situation of one dimension being `thin'. The authors thus provide computable quantified results that help to compare the 3D-solution with the 2D-solution. Chapter 4 focusses on the model simplification. In particular, the authors address continuous dependence of models with respect to the strong operator topology. Moreover, they treat geometric simplifications such as simplifying the boundary for Dirichlet and Neumann problems. As an example of highly oscillatory coefficients, the authors address elliptic homogenisation problems in Chapter 5. At the beginning of chapter 5, the authors briefly recall the homogenisation strategy via asymptotic expansion. Then they aim to derive estimates relating the fine-scale problem to the homogenised one. Quite naturally, these estimates need to be computable without computing the solutions to the problem with highly oscillatory coefficients. For this the authors also recall regularity results in elliptic theory. Penalisation and regularisation techniques for integral functionals are considered in Chapter 6 and various bounds relating the minimiser/solution of the penalised/regularised model to the other one are given. Appendix A contains a regularity result for elliptic solution theory. The book is a thorough treatise of error estimates in various situations and an excellent reference work. The authors made an effort to provide an accompanying, comprehensive guide to the literature.