Braiding of the attractor and the failure of iterative algorithms (Q1108637)

A purely iterative algorithm assigns to any complex monic polynomial of degree d a complex rational function of degree k such that the coefficients of an image rational function depend rationally on the polynomial. A fundamental result of the author's states that for \(d\geq 4\), there is no purely iterative algorithm T such that (T(p)) n(z) converges to the set of roots of the polynomial p for ``almost'' all (p,z). [\textit{C. McMullen}: Ann. Math., II. Ser. 125, 467-493 (1987; Zbl 0634.30028)]. Theorem 1.2 gives a localized version of this result: For any open connected subset V of all degree d polynomials fulfilling a certain tangledness condition, any purely iterative algorithm fails to converge to the set of roots for polynomials in V. This applies e.g. for a neighborhood of X d (d\(\geq 4)\) and still holds if the algorithm is only a continuous mapping into the set of expanding rational functions. The method of proof consists in an analysis of the braid constituted by the motion of the roots over a loop in an attractive family of rational functions.