---
title: Traub's Root-Finding Algorithm
url: https://www.emergentmind.com/topics/traub-s-root-finding-algorithm
type: topic
---

# Traub's Root-Finding Algorithm

Traub's Root-Finding Algorithm is a multipoint iterative method for solving nonlinear equations, especially polynomials, that generalizes Newton's method by using additional function evaluations. Its most prominent form achieves cubic convergence for simple roots using only two function evaluations and one derivative evaluation per iteration, often with higher global reliability than standard Newton iterations. Traub's method is fundamental in the hierarchy of high-order root-finders and occupies a central place in the study of the dynamical properties of root-finding methods, especially in the context of holomorphic dynamics and the geometry of basins of attraction.

## 1. Definition and Formulation

Traub's root-finding method is commonly presented as an accelerated Newton-like iteration. For a polynomial $p(z)$ (degree $d$) and a damping parameter $\delta$, the family of damped Traub's methods is defined as

\[
T_{p, \delta}(z) = N_p(z) - \delta \frac{p(N_p(z))}{p'(z)},
\]
where
\[
N_p(z) = z - \frac{p(z)}{p'(z)}
\]
is the classical Newton’s step. For $\delta=0$, one recovers Newton’s method. For $\delta=1$, the standard Traub’s method is obtained:

\[
T_{p,1}(z) = N_p(z) - \frac{p(N_p(z))}{p'(z)}.
\]

This iteration is sometimes referred to as a "one and a half" Newton step: it computes the Newton iterate $N_p(z)$, then applies a correction based on the value and derivative at $z$.

## 2. Local Dynamical Properties and Order of Convergence

Traub's method is characterized by its high-order convergence at simple roots. Given a root $\alpha$ of $p$ with multiplicity $k$, write $p(z) = (z-\alpha)^k q(z)$ with $q(\alpha)\ne 0$. The derivative of $T_{p,\delta}$ at $\alpha$ is given by

\[
T_{p,\delta}'(\alpha) = \frac{k-1}{k} - \delta \frac{(k-1)^k}{k^{k+1}},
\]

so for simple roots ($k=1$):

\[
T_{p,\delta}'(\alpha) = 0,
\]

implying the fixed point is superattracting. Traub's method ($\delta=1$) is cubic for simple roots—order 3 convergence—whereas Newton's method is quadratic (order 2). For multiple roots, an attracting fixed point is guaranteed if

\[
\left| \frac{k-1}{k} - \delta \frac{(k-1)^k}{k^{k+1}} \right| < 1.
\]

This explicit formula for the local multiplier as a function of $\delta$ makes precise the improvement in convergence for the cubic case.

## 3. Global Dynamical Structure and Basins of Attraction

Since $T_{p,\delta}$ is a rational map of the complex plane, its global dynamics are governed by the interaction of the fixed points (roots of $p$), poles (zeros of $p'$), and free (non-fixed) critical points. The paper [2501.04450] emphasizes that, for generic degree $d$ polynomials, the point at infinity is also a fixed point of $T_{p,\delta}$ (for $\delta \neq d^d/(d-1)^{d-1}$), and its dynamical multiplier can be explicitly computed. The repelling or attracting character of infinity is determined by the size of $\delta$ relative to degree-dependent constants.

One of the most important global features is the geometry of the immediate basins of attraction, denoted $\mathcal{A}_{n,\delta}^*(\alpha)$ for root $\alpha$. For Newton's method, these basins are always simply connected and unbounded for simple roots. The authors of [2501.04450] conjecture—and prove in the quadratic and some special cases—that this also holds for Traub’s method and all interpolating $T_{p,\delta}$ with $\delta \in [0,1]$, so that every root's basin touches infinity and is topologically simple. This is crucial for establishing the existence of universal sets of initial conditions converging to every root, mirroring results for Newton's method.

## 4. Connection to Classical Methods and the Damped Traub Family

The damped Traub family $T_{p,\delta}$ interpolates smoothly between Newton’s method ($\delta=0$) and Traub’s method ($\delta=1$), providing a continuous tuning parameter for balancing the speed of local convergence and the global geometry of basins. The transition is analytically tractable: on polynomials with simple roots, the free critical points and their orbits under $T_{p,\delta}$ can be explicitly tracked (see Lemma 2.3 of [2501.04450]), leading to a detailed understanding of how global dynamical features evolve with $\delta$. For quadratic and certain monomial-based polynomials ($p(z)=z^n-\beta$), the structure of basins and their unboundedness are fully determined for all $\delta \in [0,1]$.

This family elucidates the connection between Newton and Traub's methods not only in convergence order but also in terms of the underlying holomorphic dynamics.

## 5. Visualization and Numerical Illustration

Numerical experiments based on explicit formulas (e.g., the Möbius-conjugated map $G_\delta$ for quadratics) showcase the differences and similarities in dynamical planes for Newton's and Traub's methods. For $\delta \in [0,1]$, in both the quadratic and cubic monic cases, all "free" critical points (those not artificially fixed by root or pole structure) eventually fall into the attracting basins of roots. For instance, in the quadratic case, $G_\delta(z) = z^2 \frac{z^2+2z+(1-\delta)}{(1-\delta)z^2+2z+1}$ is a Blaschke product preserving the unit circle, providing explicit analytic control over global basin topology.

Parameter planes and dynamical planes depict regions where all critical orbits converge to the roots, substantiating the claim of simple connectivity and unboundedness.

## 6. Implications for Practical Root Finding

The robust local convergence of Traub's method (cubic for simple roots) combined with its familiar global basin structure (when compared with Newton’s method) makes it a valuable scheme in practical root-finding for polynomials. The ability to retain simple connectivity and unboundedness of root basins supports the design of universal sets of starting values (as in the works of Hubbard, Schleicher, and Sutherland for Newton’s method), ensuring that all roots are accessible from an appropriate initialization. The tuning parameter $\delta$ can serve to address problematic convergence or sensitivity to initial conditions, offering both theoretical and algorithmic flexibility.

## 7. Summary Table: Key Differences Between Newton and Traub Methods

| Method          | Order at Simple Roots | Typical Number of Evaluations per Iteration | Basin Properties           |
|-----------------|----------------------|---------------------------------------------|---------------------------|
| Newton ($\delta=0$) | 2 (Quadratic)        | 1 value, 1 derivative                      | Simply connected, unbounded |
| Traub ($\delta=1$)  | 3 (Cubic)           | 2 values, 1 derivative                      | Simply connected, unbounded (for many cases) |

This table reflects the principal theoretical and dynamical distinctions drawn in [2501.04450], with the observed retention of favorable basin properties under Traub's method in all studied cases.

---

Traub’s root-finding algorithm, with its cubic convergence for simple roots, favorable global basin structure (in many studied cases), and explicit continuous family connection to Newton’s method, presents a powerful and theoretically grounded tool in both numerical analysis and the holomorphic dynamics of iterative schemes for polynomials [2501.04450].

Source: https://www.emergentmind.com/topics/traub-s-root-finding-algorithm