Dirichlet orthogonal polynomials with Laguerre weight (Q2344304): Difference between revisions

Let \(\left\{\lambda_j\right\}_{j=1}^\infty\) be a sequence of distinct positive numbers. A \textit{Dirichlet polynomial} of degree at most \(m\) associated with this sequence of exponents has the following form \[ \sum_{n=1}^m a_n\lambda_n^{-it}=\sum_{n=1}^m a_n e^{-i\left(\log\lambda_n\right)t}, \] where \(\left\{a_n\right\}\subset{\mathbb C}\). The set of all such polynomials is denoted by \({\mathcal L}_m\). The article is devoted to study the Dirichlet orthogonal polynomials for the Laguerre weight, i.e., polynomials \(\psi_n\in{\mathcal L}_n\) having positive leading coefficient, and such that \[ \left(\psi_n,\psi_m\right)=\int_0^\infty\psi_n\left(t\right)\overline{\psi_m\left(t\right)}e^{-t}\,dt=\delta_{mn}, \] the Kronecker delta. The author obtains explicit representations for \(\psi_n\), as well as some related identities. In particular, it is shown that for \(\alpha\in\left(0,1\right)\) and \(t\in\left(0,\infty\right)\), we have \[ \psi_n\left(t\right)=-\frac{\Delta_n}{2\pi}e^{t\alpha} \int_{-\infty}^\infty e^{-its}R_n\left(-\alpha+is\right)\,ds, \] where \[ R_n\left(z\right)=\frac{1}{z-i\log\lambda_n}\prod_{j=1}^{n-1}\left(1+\frac{1}{z-i\log\lambda_j}\right), \] \[ D_n=\prod_{k=1}^{n-1}\left(1+\left[i\log\left(\lambda_k/\lambda_n\right)\right]^{-1}\right), \] and \[ \Delta_n=D_n/\left|D_n\right|. \] As a corollary of one of the explicit formulas (for the leading coefficient of \(\psi_n\left(t\right)\)), the distance from \(\lambda_n^{-it}\) to the \(\text{span}\left\{\lambda_j^{-it}\right\}_{j=1}^{n-1}\) is found: \[ \inf_{c_1,\dots,c_{n-1}}\left\|\lambda_n^{-it}-\sum_{j=1}^{n-1}c_j\lambda_j^{-it}\right\|_{L^2\left[0,\infty\right)}= \left(\prod_{k=1}^{n-1}\left(1+\left[\log\left(\lambda_n/\lambda_k\right)\right]^{-2}\right)\right)^{-1/2}. \] Some interesting representations are given in terms of the \(n\)-th reproducing kernel \[ K_n\left(u,v\right)=\sum_{j=1}^n\psi_j\left(u\right)\overline{\psi_j\left(v\right)}. \] Furthermore, the Christoffel functions come into play, and the representations mentioned above lead to interesting estimates. Let us cite two results. Theorem. {\parindent=6mm \begin{itemize}\item[(a)] Let \(n\geq 1\) and \(l\geq 0\). The function \[ e^{-x}\sum_{j=1}^n\left|\psi_j^{(l)}(x)\right|^2 \] is a decreasing function of \(x\in\left[0,\infty\right)\). \item[(b)] In particular, for \(x\in\left[0,\infty\right)\), \[ e^{-x}\sum_{j=1}^n\left|\psi_j(x)\right|^2\leq\sum_{j=1}^n\left|\psi_j(0)\right|^2=n; \] \[ e^{-x}\sum_{j=1}^n\left|\psi_j'(x)\right|^2\leq\sum_{j=1}^n\left|\psi_j'(0)\right|^2= \frac{n\left(n-1\right)\left(2n-1\right)}{6}+\sum_{j=1}^n\left(\log\lambda_j\right)^2. \] \end{itemize}} The following result is a Markov-Bernstein type inequality. Theorem. For \(n\geq 1\) and \(P\in{\mathcal L}_n\), we have \[ \left\|P'\right\|_{L^2\left[0,\infty\right)}\leq \left(\max_{1\leq j\leq n}\left|\log\lambda_j\right|+\sqrt{\frac{n\left(n-1\right)}{2}}\right) \left\|P\right\|_{L^2\left[0,\infty\right)}. \] The sharpness of these inequalities is also discussed. The results and methods developed in the article should be very interesting for specialists in approximation theory, harmonic analysis and several other areas of analysis and its applications.