The generalized Bernstein problem on weighted lacunary polynomial approximation (Q2569475)

The paper deals with the generalized Bernstein problem on weighted lacunary polynomial approximation. Let \(M(\Lambda)\) denote the space of lacunary polynomials which are finite linear combinations of the system \(\{t^\lambda: \lambda\in\Lambda\}\), where \(\Lambda: \{\lambda_n: n= 1,2,\dots\}\) is a sequence of increasing nonnegative integers. Let a weight \(\alpha\) be an even nonnegative function continuous on \(R\) such that \(\lim_{|t|\to\infty}{\alpha(t)\over\log|t|}= \infty\) and \[ L^p_\alpha= \Biggl\{f: \| f\|_{p,\alpha}= \Biggl(\int^\infty_{-\infty}|f(t)^{-\alpha(t)}|^p\,dt\Biggr)^{1/p}<\infty\Biggr\},\quad 1\leq p<\infty. \] The main result of the authors is the following Theorem: Suppose \(\alpha(e^t)\) is a convex function on \(R\). Then \(M(\Lambda)\) is dense in \(L^p_\alpha\) \((1\leq p<\infty)\) if and only if \[ \int^\infty_1\alpha(\exp\{\lambda_1(t)- a\})/t^2\,dt= \int\text{ and }\int^\infty_1\alpha\Biggl(\exp\Biggl\{{\alpha_2(t)- a\over t^2}\Biggr\}\Biggr)\,dt= \infty \] for each \(a\in R\), where \[ \begin{aligned} \lambda_1(r) &= 2\sum_{\lambda_n\leq r,\lambda_n\text{odd}}{1\over\lambda_n},\quad r\geq \lambda_1;\quad \lambda_1(r)= 0,\quad r<\lambda_1,\\ \lambda_2(r) &= 2\sum_{\lambda_n\leq r,\lambda_n\text{even}}{1\over\lambda_n},\quad r\geq \lambda_1;\quad \lambda_2(r)= 0,\quad r< \lambda_1.\end{aligned} \]