Extreme points in spaces of polynomials (Q1429376): Difference between revisions

Let \({\mathcal P}_n\) denote the space of all polynomials with complex coefficients of degree not exceeding \(n\). Given a compact set \(E\subset\mathbb{C}\), one may treat \({\mathcal P}_n\) as a subspace of \(C(E)\), the space of continuous functions on \(E\), and equip it with the max norm \(\| P\| _\infty=\| P\| _{\infty,E}=\text{max}_{z\in E}| P(z)| \). The resulting space will be denoted by \({\mathcal P}_n(E)\). Working with the real polynomials on \([-1,1]\), \textit{A. G. Konheim} and \textit{T. J. Rivlin} [Am. Math. Mon. 73, 505--507 (1966; Zbl 0154.41402)] showed that an extreme point can be characterized by the number of zeroes of \(1-P^2\). In the first part of the article under review, the author characterizes the extreme points of the unit ball of \({\mathcal P}_n(\mathbb{T})\), where \(\mathbb{T}=\{z\in\mathbb{C}\), \(| z| =1\}\). For a polynomial \(P\), let \(\mu\) be the sum of the multiplicities of the zeroes of \(1-| P| ^2\) in \(\mathbb{T}\). It is shown that \(P\) is an extreme point if and only if either \(P\) is a monomial or a certain Wronski-type matrix associated with \(P\) has rank \(2(n-\mu+1)\). The second part of the article gives a similar type of result, characterizing the extreme points of the unit ball of \({\mathcal P}_n(K)\), with \(K\) a perfect compact subset of \(\mathbb{R}\). From these, it follows that an analogous result of Konheim and Rivlin [loc. cit.] does not hold. Indeed, the author finds examples of polynomials \(P_1\) and \(P_2\) for which \(1-| P_1| ^2\) and \(1-| P_2| ^2\) have the same zeroes, with the same multiplicities, but still one of them is extreme and the other is not. The article also presents a brief discussion of the extreme points of the spaces of (both complex-valued and real-valued) trigonometric polynomials of degree not exceeding \(n\).