Zeros of Padé approximants for entire functions with smooth Maclaurin coefficients (Q1340501): Difference between revisions

Let \(f(z)= \sum_{j=0}^ \infty a_ j z^ j\) be an entire function whose coefficients satisfy Lubinsky's smoothness condition, i.e., \(\eta_ j= (a_{j+1} a_{j-1}) /a_ j^ 2 \to\eta\) as \(j\to\infty\). Let \(s_ n(z)\) be the \(n\)-th partial sum of the expansion for \(f\) and \(\pi_{n,m}(z)= P_{n,m} (z)/ Q_{n,m} (z)\) its \((n,m)\) Padé approximant. This paper studies the convergence of the sums \(s_ n\) and the numerators \(P_{n,m}\) as \(n\to\infty\). More precisely, these quantities are normalized by using \(z= ua_ n/ a_{n+1}\) as their argument, so that the unit circle becomes the natural boundary for convergence. The convergence proved is in Chebyshev norm and holds locally uniformly outside or inside the unit disk excluding the origin, depending on \(|\eta |\) being 1 or less than 1. The convergence for the numerators requires that \(\eta_ j\) converges to \(\eta\) smoothly enough. When \(\eta\) is not a root of unity, similar convergence results for the difference \(s_ n- \pi_{n,m}\) hold everywhere in \(\mathbb{C}\) except for a finite number of points (depending on \(m\)). Since these convergence properties contain explicit expressions describing the asymptotic behavior, one can extract from the numerator convergence, some conclusions about the asymptotic behavior of the zeros of \(\pi_{n,m}\).