Will Ramanujan kill Baker-Gammel-Wills? (A selective survey of Padé approximation) (Q4699400)

The convergents \((\mu_n/\nu_n)\) of the continued fraction expansion \(c_0+{\mathbb K}_{k=1}^\infty{c_kz^{\ell_k} \over 1}\) (the \(c_k\) are complex and the \(\ell_k\) are positive integers) of the function \(f(z)\) are Padé approximants to that function. The Baker-Gammel-Wills (BGW) conjecture [\textit{G. A. Baker jun., J. L. Gammel} and \textit{J. G. Wills}, J. Math. Anal. Appl. 57, 323-339 (1961; Zbl 0102.05902)] says that if, except for poles, \(f(z)\) is analytic in \(|z|<1\), then there is an infinite subsequence of \((\mu_n/\nu_n)\) that converges to \(f\) uniformly in compact subsets of \(|z|<1\), excluding the poles of \(f\). This paper gives a survey of the convergence theory of these continued fractions and hence of Padé approximants including different types of convergence (in measure, in capacity) and the natural boundaries of convergence, all in functions of the solution of BGW. So far, the BWG has not been proved or disproved. However, the paper ends with a section on a continued fraction \(H_q(z)=1+{\mathbb K}_{k=1}^\infty {q^k z \over 1}\), given by Ramanujan. Strong arguments are given that suggest that for \(|q|=1\), but \(q\) not a root of unity, the latter Ramanujan continued fraction is a counter example for BGW. The only missing link in the proof to contradict the conjecture, is some asymmetry property for the zeros of a function associated with \(H_q(z)\). So far there is only computational evidence to confirm this. For the entire collection see [Zbl 0919.00054].