The Kung–Traub Barrier: Sharp Polynomial Information Bounds for Complex Root-Finding
Year: 2026
Type: preprint
Abstract: We study the polynomial order of the information available for approximating a simple root of a complex-valued function using n evaluations of the function or its derivatives. Nodes and derivative orders may be chosen from previous responses, without requiring continuity or computability of the rules. We work with a uniform class of entire functions sufficiently close to the identity and consider their unique distinguished roots in the unit disk. Polynomial order is defined using families of functions that yield the same information and differ from a base function by polynomials of degree at most n. We prove that, among all causal information procedures using exactly n ≥ 1 evaluations and locally valid on this class, the maximal polynomial information order is 2^(n−1). For the upper bound, we construct a fixed entire function and polynomial perturbations of degree at most n, whose coefficients tend to zero, that preserve all query responses exactly while separating the roots sufficiently to exclude every larger exponent. The construction uses two systems of polynomial kernels, concentration of Wronskian zeros, and diagonalization. For the reverse inequality, we construct an explicit information procedure based on inverse Hermite interpolation and establish a uniform bound on the separation of the roots of any two functions in the class yielding identical information, without requiring their difference to be polynomial. Consequently, the Kung–Traub barrier is attained for the polynomial information order and the mathematical model considered here.
Author: Castro Marín, Rodrigo Alberto
Institution: University of Valparaíso
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7236479633
