On growth of norms of Newton interpolating operators (Q624245): Difference between revisions

From the introduction: Let \(X=(x_k^{(n)})_{n=1}^{\infty}{}^n_{k=1}\) be an infinite triangular array of nodes in \([-1,1]\). Let \(\Lambda_N(X)\) denote the \(N\)-th Lebesgue constant, that is the uniform norm of the Lagrange interpolating operator defined by the points \(\left(x_k^{(N+1)}\right)_{k=1 }^{N+1}\). It is well-known that the sequence \((\Lambda_N(X))_{N=0}^{\infty}\) has at least logarithmic growth and that the Chebyshev array \(T\) is close to the optimal choice. Now suppose that the array \(X\) is nested, that is, any row of \(X\) consists of the previous row plus one more value. What is the growth of \((\Lambda_N(X))_{N=0}^{\infty}\) in this case? Starting from the classical papers by Runge, Bernstein and Faber, the problem of approximation properties of Lagrange interpolation has attracted attention of many mathematicians. A variety of results concerning asymptotic behavior of the Lebesgue constant were obtained from different arrays as well as for different metrics. How to choose a sequence \(X=(x_n)_{n=1}^{\infty}\subset [-1,1]\) with a moderate growth of \((\Lambda_N(X))_{N=0}^{\infty}\)? One can suppose that such a sequence must at least approximate the equilibrium arcsine distribution \(d\mu_e(x)={dx\over \pi \sqrt{1-x^2}}\). This means that the finite measures \({1\over n}\sum_{k=1}^{n}\delta _{x_k}\), where \(\delta _{x_k}\) stands for the unit mass located at \(x_k\), converge to \(d\mu _{e}\) in the weak* topology of measures. For example, the Leja sequence \(L\) has this approximating property, see e.g., [\textit{E. B. Saff} and \textit{V. Totik}, Logarithmic potentials with external fields. Grundlehren der Mathematischen Wissenschaften. 316. Berlin: Springer (1997; Zbl 0881.31001)]. \textit{R. Taylor} and \textit{V. Totik} [IMA J. Numer. Anal. 30, No.~2, 462--486 (2010; Zbl 1193.41003)] proved that the sequence \((\Lambda_N(L))_{N=0}^{\infty }\) has subexponential growth, that is, \({\log \Lambda_N(L)\over N}\rightarrow 0\) as \(N\rightarrow \infty \). The Leja sequence is a nested analog of the array \(F\) of Fekete points. Even for the set \([-1,1]\), the exact position of the Leja points is not known, whereas the corresponding \(n\)-th Fekete points are the zeros of the Jacobi polynomial \(P_{n-2}^{(1,1)}\) together with the points \(\pm 1\). It should be noted that asymptotically \(\Lambda_N(F)\geq N^{3/2}\). On the other hand, for the Chebyshev array \(T\), we have \(\Lambda_N(T)\leq (2/\pi )\log N+1\). Thus, a nested analog of the array \(T\) can be considered as a possible choice for the desired sequence \(X\). In the paper, the author arranges in a special manner the zeros of the Chebyshev polynomials \((T_{3^s})_{s=0}^{\infty}\) and show that \[ \Lambda_N(X)=O\left(\exp\left({\log^2N\over \log 3}+8\log N \right)\right) \] for such nodes. The constant involved in this estimate is not exact, but for a subsequence \((N_s)\) he obtains the lower bound \(\Lambda_{N_s}(X)\geq \exp \left({\log^2N_s\over \log 27}-5\log N_s \right)\).