Small polynomials with integer coefficients

Abstract: We study the problem of minimizing the supremum norm, on a segment of the real line or on a compact set in the plane, by polynomials with integer coefficients. The extremal polynomials are naturally called integer Chebyshev polynomials. Their factors, zero distribution and asymptotics are the main subjects of this paper. In particular, we show that the integer Chebyshev polynomials for any infinite subset of the real line must have infinitely many distinct factors, which answers a question of Borwein and Erd'{e}lyi. Furthermore, it is proved that the accumulation set for their zeros must be of positive capacity in this case. We also find the first nontrivial examples of explicit integer Chebyshev constants for certain classes of lemniscates. Since it is rarely possible to obtain an exact value of integer Chebyshev constant, good estimates are of special importance. Introducing the methods of weighted potential theory, we generalize and improve the Hilbert-Fekete upper bound for integer Chebyshev constant. These methods also give bounds for the multiplicities of factors of integer Chebyshev polynomials, and lower bounds for integer Chebyshev constant. Moreover, all the mentioned bounds can be found numerically, by using various extremal point techniques, such as weighted Leja points algorithm. Applying our results in the classical case of the segment

[0, 1]

, we improve the known bounds for the integer Chebyshev constant and the multiplicities of factors of the integer Chebyshev polynomials.

Review text

Let \(E\) be a compact subset of the plane and let \(\| \cdot\| _E\) denote the supremum norm on \(E\). The integer Chebyshev constant \(t_{\mathbb Z}(E)\) is the limit as \(n \to \infty\) of the \(n\)th root of the infimum of \(\| P\| _E\) over all non-zero polynomials of degree at most \(n\) with integer coefficients. The extremal polynomials \(Q_n\) for this problem are called integer Chebyshev polynomials. The interesting case is that of \(\text{ cap}(E) < 1\) since otherwise \(t_{\mathbb Z}(E) = 1\). The author computes the exact value of \(t_{\mathbb Z}(E)\) for a certain class of lemniscates. One of the author's main results is that the integer Chebyshev polynomials for any infinite subset of the real line must have infinitely many distinct factors (over the integers). The case \(E = [0,1]\) has been much studied. The author improves many of the known results in this case, for example showing that \[ Q_n(x) = (x(1-x))^{[\alpha_1 n]} (2x-1)^{[\alpha_2 n]} (5x^2 - 5x + 1)^{[\alpha_3 n]} R_n(x), \] where \(0.31 \leq \alpha_1 \leq 0.34, 0.11 \leq \alpha_2 \leq 0.14\) and \(0.035 \leq \alpha_3 \leq 0.057\). He obtains the following improvement on previous estimates: \(0.4213 < t_{\mathbb Z}([0,1]) < 0.4232\). The techniques used are from weighted potential theory [\textit{E. B. Saff} and \textit{V. Totik}, Logarithmic Potentials with External Fields, Grundlehren der Mathematischen Wissenschaften. 316, Berlin: Springer (1997; Zbl 0881.31001)].

Recommendations

Cites work

Cited in

(25)

This page was built for publication: Small polynomials with integer coefficients

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q816475)