Symbolic asymptotics

Publication date 14 December 2004

Published in Algorithms and Computation in Mathematics (Search for Journal in Brave)

exp-log algorithm exp-log functions function towers Hardy fields Liouvillian functions multiseries nested expansions nested forms Pfaffian functions star products zero-equivalence

Mathematics Subject Classification ID

Research exposition (monographs, survey articles) pertaining to field theory (12-02) Asymptotic approximations, asymptotic expansions (steepest descent, etc.) (41A60) Research exposition (monographs, survey articles) pertaining to computer science (68-02) Symbolic computation and algebraic computation (68W30)

Review text

The book is about the automatic computation of the asymptotic behavior of functions of a real variable. This means that limits of functions are computed, and also some measure is given of how rapidly a given function approaches its limit. The methods used in the book are at least potentially implementable. However the book is really about algorithms and the mathematical principles that underline them, rather than of implementations. A function is considered as an object which is built from constants and the identity function by means of certain operations. Thus a function may be represented by a tree whose leaves are constants and nodes are operations. Another useful concept is a tower of sets. A tower of sets of functions is a finite chain \({\mathcal F}_0\subset {\mathcal F}_1\subset\ldots\subset {\mathcal F}_n\) where \({\mathcal F}_0\) usually is a field or a ring. Each \({\mathcal F}_{i+1}\) is generated from \({\mathcal F}_i\) by addition of a single element \(f_{i+1},\) satisfying a differential equation over \({\mathcal F}_i\) of the first order. If each \(f_{i+1}\) is a logarithm or an exponential then the set of functions defined by means of tower of sets is the field of exp-log functions. If \(f_{i+1}\) is also allowed to be a real root of an algebraic equation over \(F_i,\) then we obtain Hardy's class of L-functions. If we further allow \(f_{i+1}\) to be any integral of an element in \({\mathcal F}_i,\) we obtain the class of Liouvillian functions. Finally if \(f_{i+1}\) is allowed to be defined by any equations of first order and degree the class of Pfaffian functions is obtained. Now let us consider how the asymptotic behavior will be expressed. The traditional approach here is to use asymptotic power series. In the book three more types of expansion are presented. There are multiseries, nested expansions and star products. By using them it is possible to express the asymptotic behavior of exp-log functions, Liouvillian functions, inverse functions and implicit functions. The traditional stance in asymptotics is to regard asymptotic series as the prime object of study. By contrast, the author regards functions as the prime object. A lot of algorithmic problems are dealt with in the book. We have to work with the expressions, rather than directly with the functions. So it is necessary to determine whether two different expressions represent the same function. This is the problem of zero equivalence. Some conjectures such as Schanuel conjecture, uniformity conjecture are used in the book. The mathematical technics employed in the book include the theory of Hardy fields, tools from differential algebra, some ideas from asymptotic series, many of the standard technics of computer algebra and some basic ideas from measure theory.

Recommendations

Cited in

(12)

This page was built for publication: Symbolic asymptotics

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1889976)