Behavior of best \(L_ p\) polynomial approximants on the unit interval and on the unit circle (Q804860): Difference between revisions

For a given \(f\in C([-1,1])\) we denote by \(p_ n\) its best uniform approximant out of \(P_ n\), the set of all algebraic polynomials of degree at most n. By a result of \textit{M. I. Kadec} [Uspekhi Mat. Nauk 15, No.1(91), 199-202 (1960; Zbl 0136.364)], if f is real-valued then there is an increasing subsequence \(\Lambda\) (f) of the positive integers such that for any subinterval \([a,b]\subset I:=[-1,1]\) (a\(\neq b)\) \[ \| f- p_ n\|_{[a,b]}/\| f-p_ n\|_ I=1,\quad n\in \Lambda (f), \] \(n\geq n([a,b])\). In other words \(\{p_ n\}\) does not approximate f better on any subinterval of I than it does on the whole interval \textit{A. Kroó} and \textit{E. B. Saff} [Proc. Am. Math. Soc. 103, No.1, 203-209 (1988; Zbl 0663.41027)] proved that such an effect also holds for complex-valued \(f\in C\subset I)\) and also for the analogous case of uniform approximation on the unit circle \(\partial \Delta\) of \(f\in C(\Delta)\cap H(int \Delta).\) In this paper the authors prove the analogues of the previous results for general \(L_ 2\) best approximation on I and \(\partial \Delta\). They also give an \(L_ p(p>0)\) version of the Jentzsch-Szegö type theorem on the distribution of zeros of best approximants proved earlier in the uniform approximation case by \textit{H. P. Blatt}, \textit{E. B. Saff} and \textit{M. Simkani} [J. Lond. Math. Soc., II. Ser 38, No.2, 307-316 (1988; Zbl 0684.41006)].