A new algorithm for computing symbolic limits using hierarchical series

Abstract

We describe an algorithm for computing symbolic limits, i.e. limits of expressions in symbolic form, using hierarchical series. A hierarchical series consists of two levels: an inner level which uses a simple generalization of Laurent series with finite principal part and which captures the behaviour of subexpressions without essential singularities, and an outer level which captures the essential singularities. Once such a series has been computed for an expression at a given point, the limit of the expression at the point is determined by looking at the most significant term of the series. By this method one can compute, for example, the limit of:

$$\mathop {\lim }\limits_{n \to \infty } \frac{{e^{e^{\sqrt {\ln \ln n} } } }}{{e^{\sqrt {\ln n} } }} - n^{\sqrt {\ln n} } .$$

This algorithm solves the limit problem for a large class of expressions.

Extended Abstract