Some numerical results on best uniform rational approximation of \(x^ \alpha\) on [0,1] (Q1192647): Difference between revisions

Recently \textit{H. Stahl} [Best uniform approximation of \(| x|\) on \([-1,1]\), Mat. Sb. to appear)]\ proved a conjecture due to \textit{R. S. Varga}, \textit{A. Ruttan} and \textit{A. J. Carpenter} [Mat. Sb. 182, 1523- 1541 (1991; Zbl 0739.65010)]: \[ \lim_{n\rightarrow\infty} e^{\pi\sqrt{2n}} E_{n,n}(\sqrt{x};[0,1])=8. \] Using 3300 cpu hours on two mainframes at the Argonne National Laboratory, the authors calculated to six decimals the limits \[ \lambda(\alpha)=\lim_{n\rightarrow\infty} e^{\pi\sqrt{4\alpha n}} E_{n,n} (x^{\alpha};[0,1])=8 \] \(\alpha=j/8,\;1\leq j\leq 8\) (intermediate results had to be calculated to 200 decimal places; the main tools were the Remez algorithm and Richardson's extrapolation). The numerical evidence strongly supports the conjecture \[ \lambda(\alpha)=2^{2(\alpha+1)} |\sin(\alpha\pi)|, \] \noindent and it is the authors firm belief that this can be proved using the methods employed in the paper by H. Stahl quoted above.