arXiv subjects
Publications and source records attributed to Wolfram Koepf.
In this article we will describe the \Maple\ implementation of an algorithm presented in~\cite{Koe92}--\cite{Koeortho} which computes an {\em exact\/} formal power series (FPS) of a given function. This procedure will enable the user to reproduce most of the results of the extensive bibliography on series~\cite{Han}. We will give an overview of the algorithm and then present some parts of it in more detail.
Formal Laurent-Puiseux series are important in many branches of mathematics. This paper presents a {\it Mathematica} implementation of algorithms developed by the author for converting between certain classes of functions and their equivalent representing series. The package {\tt PowerSeries} handles functions of rational, exponential, and hypergeometric type, and enables the user to reproduce most of the results of Hansen's extensive table of series. Subalgorithms of independent significance generate differential equations satisfied by a given function and recurrence equations satisfied by a given sequence.
In this article we present a method to implement orthogonal polynomials and many other special functions in Computer Algebra systems enabling the user to work with those functions appropriately, and in particular to verify different types of identities for those functions. Some of these identities like differential equations, power series representations, and hypergeometric representations can even dealt with algorithmically, i.\ e.\ they can be computed by the Computer Algebra system, rather than only verified. The types of functions that can be treated by the given technique cover the generalized hypergeometric functions, and therefore most of the special functions that can be found in mathematical dictionaries. The types of identities for which we present verification algorithms cover differential equations, power series representations, identities of the Rodrigues type, hypergeometric representations, and algorithms containing symbolic sums. The current implementations of special functions in existing Computer Algebra systems do not meet these high standards as we shall show in examples. They should be modified, and we show results of our implementations.
We consider the family B-tilde of bounded nonvanishing analytic functions f(z) = a_0 + a_1 z + a_2 z^2 + ... in the unit disk. The coefficient problem had been extensively investigated, and it is known that |a_n| <= 2/e for n=1,2,3, and 4. That this inequality may hold for n in N, is know as the Kryż conjecture. It turns out that for f in B-tilde with a_0 = e^-t, f(z) << e^{-t (1+z)/(1-z)} so that the superordinate functions e^{-t (1+z)/(1-z)} = sum F_k(t) z^k are of special interest. The corresponding coefficient function F_k(t) had been independently considered by Bateman [3] who had introduced them with the aid of the integral representation F_k(t) = (-1)^k 2/pi int_0^pi/2 cos(t tan theta - 2 k theta) d theta . We study the Bateman function and formulate properties that give insight in the coefficient problem in B-tilde.
In \cite{Koe92}--\cite{Koe93c} the first author published an algorithm for the conversion of analytic functions for which derivative rules are given into their representing power series $\sum\limits_{k=0}^{\infty}a_{k}z^{k}$ at the origin and vice versa, implementations of which exist in {\sc Mathematica} \cite{Wol}, (s.\ \cite{Koe93c}), {\sc Maple} \cite{Map} (s.\ \cite{GK}) and {\sc Reduce} \cite{Red} (s.\ \cite{Neun}). One main part of this procedure is an algorithm to derive a homogeneous linear differential equation with polynomial coefficients for the given function. We call this type of ordinary differential equations {\sl simple}.