Harmonic number identities and Hermite-Padé approximations to the logarithm function (Q2575616): Difference between revisions

The author gives explicit formulae for the partial fraction decomposition of the functions \[ \left({n!\over (x)_{n+1}}\right)^{\lambda} \] with \(\lambda,\,n\) natural numbers and \((x)_n\) the standard Pochhammer symbol for the ascending factorial \[ (x)_0=1;\;(x)_n=x(x+1)\cdots (x+n-1),\;n\geq 1. \] An important role is played by the Faà di Bruno formula (lemma 1 in the paper; cf. [\textit{L. Comtet}, Advanced Combinatorics, (Dordrecht-Holland, The Netherlands), page 139, (1974; Zbl 0283.05001)], the partial sums of the general harmonic series \[ H_0^{(m)}=0,\;H_n^{(m)}=\sum_{j=1}^n\,{1\over j^m}\;(m,n=1,2,\ldots), \] and the Bell polynomials \[ \omega_{\ell}(\lambda,x)=\left[\left({d\over dx}\right)^{\ell} h(x)^{\lambda}\right]/h(x)^{\lambda}, \text{ where }h(x)={n!(x+k)\over (x)_{n+1}}. \] As corollaries several harmonic number identities are derived which, up to now, have either be conjectured or for which the proof has been given/announced using a Computer Algebra System. The final section of the paper uses the partial fraction decomposition to give the explicit form for the Latin polynomials (polynomials of type I) from the Hermite-Padé table for the system \(\{1,\ln{x},(\ln{x})^2,\ldots,(\ln{x})^m\}\) given as series around \(x=1\) [cf. \textit{H. Jager}, Nederl. Akad. Wet., Proc., Ser. A 67, 227- 249 (1964; Zbl 0151.04505) and \textit{K. Mahler}, Compos. Math. 19, 95--166 (1986; Zbl 0168.31303) (work done around 1938 in The Netherlands] for the theoretical framework of the generalized Hermite-Padé table and [\textit{K. Mahler}, Philos. Trans. Roy. Soc. London, Ser. A 245, 371--398 (1953; Zbl 0052.04404)]. Although the paper by Jager and the second paper by Mahler mentioned above are both in the list of references, the paper under review does not refer to these publications in the text. Moreover, because the system studied by the author uses \(\ln{x}\) around \(x=1\)---which is the `same' power series as \(\log{(1-x)}\) around \(x=0\)---the referee wonders whether the author has checked that his results are the same as those by Jager (the type I polynomials as explicitly given \({}_mF_{m-1}\) hypergeometric functions and the remainder as Meyer's \(G\)-function \(G_{m,m}^{m,0}\)) or not.