Average case weighted \(L_2\) Markov factors with doubling weights (Q2182914)

For a given normed linear space \(X\) with norm \(\|\cdot \|_X\), the worst case Markov factor \(M_n^{wor}(X)\) is defined by \[ M_n^{wor}(X):=\sup_{0\ne P\in \mathcal{P}_n}\frac{\|P' \|_X}{\|P \|_X}, \] where \(\mathcal{P}_n\) is the usual space of all algebraic polynomials of degree at most \(n\). A nonnegative integrable function \(\omega \) on \([-1,1]\) (a weight on \([-1,1]\)) is said to be a doubling weight if there exists a constant \(L>0\) depending only on \(\omega \) such that \[ \int_{2I\cap [-1,1]}\omega(x)dx\leq L\int _{I}\omega(x)dx \] holds for all subintervals \(I\subset [-1,1]\). The classical Markov inequality (for \(X=L_{\infty}\)) reads as \(M_n^{wor}(L_{\infty})=n^2\). For the usual space \(L_{p,\omega}\) with a doubling weight \(\omega\), \textit{G. Mastroianni} and \textit{V. Totik} [Constr. Approx. 16, No. 1, 37--71 (2000; Zbl 0956.42001)] proved \(M_n^{wor}(L_{p,\omega})\leq Cn^2\) where \(C=C(p,\omega)\). Recently, in [\textit{A. I. Aptekarev} et al., Proc. Am. Math. Soc. 143, No. 9, 3847--3862 (2015; Zbl 1332.41011)] and [\textit{V. Totik}, Proc. Am. Math. Soc. 147, No. 1, 153--166 (2019; Zbl 1404.26019)], asymptotically sharp constants for \(X=L_{2,\omega _{\alpha,\beta}}\) were obtained, where \(\omega _{\alpha,\beta}\) are the usual Jacobi weights. In this paper the authors consider the average case Markov factors, recently introduced from random polynomials in [\textit{L. Bos}, J. Approx. Theory 241, 1--10 (2019; Zbl 07043335)]. For \(a=(a_0,\ldots,a_n)^T\in \mathbb{R}^{n+1}\), a random polynomial is defined as \(P_a(x)=\sum_{i=0}^na_iP_i(x)\), where the coefficients \(a_i\) are independent \(N(0,\sigma ^2)\), and where \(\{P_j\}_{j=0}^n\) is an orthonormal basis of \(\mathcal{P}_n\) with respect to the inner product of \(L_{2,\omega }\) such that \(deg P_j = j\). In this respect, the average Markov factor \(M_n^{ave}(L_{2,\omega})\) is defined by \[ M_n^{ave}(L_{2,\omega}):=\mathbb{E}\left(\frac{\|P'_a\|_{2,\omega}}{\|P_a\|_{2,\omega}}\right), \] where \(\mathbb{E}\) denotes the mathematical expectation. The result of the paper extends one in the aforementioned Bos' paper, and reads as follows: {Theorem 1.1} Let \(\omega\) be a doubling weight on \([-1,1]\). Then we have \[ n^{-1/2}M_n^{wor}(L_{2,\omega})\ll M_n^{ave}(L_{2,\omega})\ll n^{3/2}. \] Here \(A_n\ll B_n\) means that there exists a constant \(C>0\) independent of \(n\) such that \(A_n\leq CB_n\).