On the dynamics of Halley’s method

Gang Liu, Soumen Pal, Saminathan Ponnusamy · 2025

Read On the dynamics of Halley’s method by Gang Liu, Soumen Pal, Saminathan Ponnusamy (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-00631.txt.

Introduction

Gang Liu, Soumen Pal, and Saminathan Ponnusamy study the long-run behavior of Halley's method. Their paper, from twenty twenty-five, looks at a fast root-finding rule used on complex polynomials. The rule's paths split the plane into basins that flow to each root. The edges of those basins form a fractal set with rich structure.

The team proves the method settles for broad classes of polynomials. For many of them the fractal edge is one connected piece. The basins that touch each root stretch out without bound. The symmetry of the method matches that of the polynomial. One further result shows a basin can also be bounded, a twist worth note.

Catalog record

CREATORGang Liu, Soumen Pal, Saminathan Ponnusamy
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE59 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-00631

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

On the dynamics of Halley’s method Gang Liu, Soumen Pal, Saminathan Ponnusamy Abstract In this article, we study the global dynamics of Halley’s method applied to complex polynomials. Specifically, we analyze the structure and connectivity of the Julia set of this method. The convergence behavior, symmetry properties, and topological features of the corresponding Fatou and Julia sets are studied for various classes of polynomials, including unicritical, cubic, and quartic polynomials with non-trivial symmetry groups. In particular, we prove that Halley’s method H p H_{p} is convergent, its Julia set is connected, the immediate basins are unbounded, and the symmetry group of it coincides with that of the polynomial whenever p p belongs to one of the above classes. We further extend our results to a broader class of polynomials. It is shown that the immediate basin of Halley’s method H p H_{p} corresponding to a root of p p can be bounded. We also make some remarks on the dynamics of Halley’s method applied to a cubic polynomial in general. Keyword: Fatou and Julia sets; Halley’s method; Convergent classes; Symmetry in Julia set. AMS Subject Classification: 37F10, 65H05 1 Introduction Complex dynamics is the study of iteration of functions in the extended complex plane \widehat{\mathbb{C}}=\mathbb{C}\cup\{\infty\} , mainly dealing with rational or entire functions. For a given rational function R:\widehat{\mathbb{C}}\to\widehat{\mathbb{C}} , the dynamical behavior of the sequence \{R^{n}(z)\} as n\to\infty can exhibit intricate and rich structures, where R^{n} denotes the n -fold composition of R with itself. The Fatou set \mathcal{F}(R) consists of points in \widehat{\mathbb{C}} where the family \{R^{n}\} is normal in some neighborhood. Its complement, the Julia set \mathcal{J}(R)=\widehat{\mathbb{C}}\setminus\mathcal{F}(R) , represents the chaotic and highly sensitive dynamics of R . These sets form the core objects of study in complex dynamics. By definition, the Fatou set is open, and a component of it is called a Fatou component. The Julia set is therefore closed. However, the Julia set is not always connected, and a connected subset of it known as a Julia component. A fundamental aspect in characterizing these two sets is the classification of the fixed points of a rational function R . For a fixed point z_{0}\in\widehat{\mathbb{C}} (i.e., R(z_{0})=z_{0} ), its multiplier \lambda_{z_{0}} (or simply \lambda whenever the suffix is not required) is defined by the complex number R^{\prime}(z_{0}) whenever z_{0}\in\mathbb{C} , or, whenever R fixes \infty , its multiplier is G^{\prime}(0) , where G(z)=\frac{1}{R(1/z)} . The fixed point z_{0} is said to be attracting if |\lambda_{z_{0}}|<1 (in a special case when \lambda_{z_{0}}=0, z_{0} is said to be superattracting ), repelling if |\lambda_{z_{0}}|>1 , and indifferent if |\lambda_{z_{0}}|=1 . An indifferent fixed point is said to be parabolic or rationally indifferent if its multiplier is a root of unity, else, it is said to be irrationally indifferent . A point z^{*} is said to be a periodic point with period k if R^{k}(z^{*})=z^{*} . However, the equality does not hold for R^{i} , for any i=1,2,\dots,k-1 . A k -periodic point is classified as attracting, repelling, or indifferent by treating it as a fixed point of R^{k} . Note that, an attracting fixed point z_{0} of R is contained in \mathcal{F}(R) , and there is a set \mathcal{B}_{z_{0}} containing points whose successive iterations converge to z_{0} , i.e., \mathcal{B}_{z_{0}}=\{z\in\widehat{\mathbb{C}}:\lim\limits_{n\to\infty}R^{n}(z)=z_{0}\} . This set is called the basin of attraction corresponding to z_{0} . The component of the basin that contains z_{0} is known as the immediate basin , and we denote it as \mathcal{A}_{z_{0}} . For a rational function R , a Fatou component U can be of four types: • attracting domain if U contains an attracting fixed point. • parabolic domain if \partial U contains a parabolic fixed point \xi , and \lim\limits_{n\to\infty}R^{n}(z)=\xi for every z\in U . • Siegel disk if there is an analytic homeomorphism \phi:U\to\mathbb{D} such that \phi\circ R\circ\phi^{-1}(z)=e^{2\pi i\theta}z , for some irrational number \theta . Here \mathbb{D}=\{z:|z|<1\} . • Herman ring if there is an analytic homeomorphism \phi:U\to\mathbb{A}_{r}=\{z:1<|z|<r\} such that \phi\circ R\circ\phi^{-1}(z)=e^{2\pi i\theta}z , for some irrational number \theta . A Fatou component is said to be a rotation domain if it is either a Siegel disk or a Herman ring. Note that, for a complex function, there is another type of Fatou component, known as wandering domain. However, a rational function does not have a wandering domain. This is a remarkable result by Sullivan (see [ 2 , Theorem 8.1.2] ). For a detailed study on complex dynamics, we refer to [ 2 , 5 , 12 ] . A root-finding method is defined as a map F that assigns to each polynomial p a rational function F_{p} such that every root of p is an attracting fixed point of F_{p} . Iterative root-finding methods, such as Newton’s and Halley’s methods, not only provide practical schemes for locating roots of nonlinear equations but also give rise to interesting dynamical systems when studied in the complex plane. For a given point z_{0}\in\widehat{\mathbb{C}} (often called an initial guess ), the behavior of its orbits, i.e., the sequence \{F_{p}^{i}(z_{0})\}_{i\in\mathbb{Z}} , can be captured through the study of its Julia set \mathcal{J}(F_{p}) and Fatou set \mathcal{F}(F_{p}) . This viewpoint connects numerical analysis with the field of complex dynamics (see [ 1 , 3 , 11 ] and references therein). The order of convergence of F is n (we say F is n -th order convergent ) if the local degree of F_{p} is at least n at each simple root of p . Note that this notion describes the local rate of convergence near a root. For an example, Newton’s method is quadratic convergent. Given a polynomial p , F_{p} is said to be convergent if the Fatou set \mathcal{F}(F_{p}) is precisely the union of the basins of attraction of the roots of p , i.e., any point in the Fatou set will eventually converge to a root of p . This notion of convergence concerns the global behavior of the method in the complex plane. For a fixed natural number d\geq 2 , a root-finding method F is said to be generally convergent if F_{p} is convergent for all p belonging to a dense subset of the family of all polynomials of degree d . Newton’s method is generally convergent for d=2 . McMullen proved that there is no root-finding method which is generally convergent for d\geq 4 (see Theorem 1.1, [ 11 ] ). The dynamics of Newton’s method has been extensively studied and is now classical. Its rich literature includes [ 9 , 18 , 16 ] and references therein. However, comparatively less is known about the global dynamics induced by higher-order methods such as Halley’s method. This may be attributed to the increased degree of the resulting rational map when a higher-order convergent root-finding method is applied, which makes the associated dynamical study more challenging. Given a complex polynomial p , Halley’s method is defined by the rational map H_{p}(z)=z-\frac{2p(z)p^{\prime}(z)}{2(p^{\prime}(z))^{2}-p(z)p^{\prime\prime}(z)}. (1) This method is cubically convergent, i.e., the local degree of H_{p} at every simple root of p is at least three. Halley’s method is the second member of the well-known family of root-finding methods, namely, König’s methods, which is derived by the following formula K_{p,n}(z)=z+(n-1)\frac{\left(\frac{1}{p}\right)^{[n-2]}(z)}{\left(\frac{1}{p}\right)^{[n-1]}(z)}, where n\geq 2 is

Read the complete text