a b s t r a c tThe maximum computing time of the continued fractions method for polynomial real root isolation is at least quintic in the degree of the input polynomial. This computing time is realized for an infinite sequence of polynomials of increasing degrees, each having the same coefficients. The recursion trees for those polynomials do not depend on the use of root bounds in the continued fractions method. The trees are completely described. The height of each tree is more than half the degree. When the degree exceeds one hundred, more than one third of the nodes along the longest path are associated with primitive polynomials whose low-order and high-order coefficients are large negative integers. The length of the forty-five percent highest order coefficients and of the ten percent lowest order coefficients is at least linear in the degree of the input polynomial multiplied by the level of the node. Hence the time required to compute one node from the previous node using classical methods is at least proportional to the cube of the degree of the input polynomial multiplied by the level of the node. The intervals that the continued fractions method returns are characterized using a matrix factorization algorithm. (W. Krandick). 1 Tel.: +1 919 602 4231.2 As the result of a tragic accident Werner Krandick passed away on March 26, 2012, after this article was accepted for publication. He was a major contributor to real root isolation of univariate polynomials with rational coefficients. He worked on both theoretical complexity and practically efficient algorithms and implementations. This article, reporting joint work with George Collins, provides a lower bound, conclusively resolving some long discusssions on comparisons between the continued fractions method and bisection-based methods. The article exemplifies Werner's work: precise, dedicated and cogent, giving generous credit to related results. He will be greatly missed by the JSC community.
0747-7171/$ -see front matterMignotte (1981, 1982), have a pair of close real roots, one on either side of 1/a. The distance between either root and 1/a is less than a −n/2−1 . Johnson concludes that the number of bisections required to separate the two roots dominates n, a fallacy in case 1/a is a bisection point. The polynomials S n (−x+1) have a pair of close real roots, one on either side of 1−1/a. This solves a perceived numbertheoretical problem: If the binary expansion of 1/a does not contain sufficiently many digits 1, then the expansion of 1 − 1/a will. Johnson asserts without proof that the bisection method will call itself for polynomials that are dense and have long coefficients. How dense? How many of the coefficients are long? Do their signs matter? There is no proof here, just speculation. Some statements do not make any sense, for instance the assertion that the length of an arbitrary level number in the recursion tree of the bisection method is codominant with the height of the tree. A few more errors can be discerned but much of the presentation is...