Calculation of poles of meromorphic functions with q-d, r-s and \(\epsilon\)-algorithms. Acceleration of these processes (Q1094812)

Let f(z) be meromorphic for \(| z| <R\) and have poles \(\pi\) (r) (r\(\geq 1)\) in this disc and let \(f(z)=c_ 0+c_ 1z+c_ 2z^ 2+..\). for small z. Set \(\Pi (k)=\pi (1)...\pi (k)\) and \(\rho (n,k)=H_ k^{(n-1)}/H_ k^{(n)},\) where \(H_ k^{(n)}\) is the Hankel determinant formed from the \(c_ i\). With k fixed and subject to suitable conditions, (*) lim \(\rho\) (n,k)\(=\Pi (k)\) and \[ (**)\quad \lim \{\rho (n+1,k)-\Pi (k)\}/\{\rho (n,k)-\Pi (k)\}=\pi (k)/\pi (k+1) \] as \(n\to \infty\). The q-d, \(\epsilon\)- and other algorithms produce numbers expressible as determinantal quotients similar to \(\rho\) (n,k) for which relationships analogous to (*,**) hold. The sequence \(\{\rho\) (n,k)\(\}\) and others similar to it may be accelerated by use of Aitken's \(\delta^ 2\)-process to produce methods for estimating the poles \(\pi\) (r) (r\(\geq 1)\). The paper offers a commentary upon the above theory.