32-digit values of the first 100 recurrence coefficients for a weight function with a logarithmic type singularity

Listed in Datasets

By Walter Gautschi

Purdue University

32-digit values of the first 100 recurrence coefficients for the weight function w(x)=-log(1-x^2) on [-1,1]

Version 1.0 - published on 22 Mar 2017 doi:10.4231/R7BR8Q6R - cite this Archived on 23 Apr 2017

Licensed under Attribution 3.0 Unported


32-digit values of the first 100 recurrence coefficients for orthogonal polynomials relative to the weight function w(x)=-log(1-x^2) on [-1,1] are computed by a modified-moment-based method using the routine sr_logtype(dig,32,100), where dig=36 has been determined by the routine dig_logtype(100,34,2,32), attesting to the high stability of the modified Chebyshev algorithm. For relevant computational details see Dirk P. Laurie and Laurette Rolfes, "Computation of Gaussian quadrature rules from modified moments --- Algorithm 015", J. Comput. Apppl. Math. 5 (1979), 235-243. (The 25-digit values of the Gauss nodes and weights in the table on p. 237 of the cited reference, however, seem to be accurate to only about 15-16 digits, judging from comparison with results obtained with our own recurrence coefficients and also from the fact that for the 5-point published Gauss formula, the sum of the weights reproduces the first moment only to about 16 digits, whereas with our data, for 5-, 10-, 20-, 30-point formulae we obtain agreement to all 32 digits.)

Cite this work

Researchers should cite this work as follows:



The dataset consists of one text file and five Matlab scripts.

The Purdue University Research Repository (PURR) is a university core research facility provided by the Purdue University Libraries, the Office of the Executive Vice President for Research and Partnerships, and Information Technology at Purdue (ITaP).