Rational approximation. II (Q1249667): Difference between revisions

Let \(\pi_m\) denote the class of all real polynomials of degree at most \(m\) and \(\pi_{m,n}\) denote the collection of all rational functions \(r_{m,n}(x)=\frac{p(x)}{q(x)}, p\in\pi_m, q\in\pi_n\). Let \(\lambda_{m,n}\equiv\lambda_{m,n}(f^{-1})=\inf_{r_{m,n}\in\pi_{m,n}}\left\|\frac1{f(x)}-\gamma_{m,n}(x)\right\|_{l_{\infty}[0,\infty]}\) where \(f\), given by \(f(z)=\sum_{k=0}^{\infty}a_kz^k\), is an entire function with all non-negative coefficients. In Part I [\textit{P.Erdős} and \textit{A.R.Reddy}, Adv. Math. 21, 78-109 (1976; Zbl 0334.00019)] , the authors mainly reviewed and proved certain results concerning \(\lambda_{0,n}\). In the present paper, which contains 22 theorems, the authors devote themselves to show that for certain classes of entire functions the error obtained by rational functions of degree \(n\) in approximating on \([0,\infty)\) under the uniform norm is much smaller than the error obtained by recipocals of polynomials of degree \(n\). For example, they show in Theorem 10 that if \(f\) is an entire function of order \(\rho(1\leq\rho< \infty)\), type \(\tau\) and lower type \(\omega(0< \omega\leq\tau< \infty)\), then for every polynomial \(P_n(x)\) of degree \(n\) and all large \(n\), there exist positive constants \(a\) and \(b\) for which \(\left\|\frac{x+1}{f(x)}-\frac 1{P_n(x)}\right\|_{L_{\infty}[0,\infty)}\geq a\exp((-bn^{1-1/3\rho})\) whereas in Theorem 17 they establish that for such functions there is some \(\beta (0< \beta< 1)\) such that \(\lambda_{1,n}\left(\frac{1+x}{f(x)}\right)\leq\beta^n\). It has also been shown that for certain entire functions, for example \(f(z) = e^{e^{z}}\), there is little difference between the errors abtained by rational functions and the errors obtained by recipocals of polynomials. Incidentally, the following interesting results has also been obtained: \[ \lim_{n\to\infty}\left[\lambda_{0,n}(xe^{-x})\right]^{1/2\log n}=e^{-1}=\lim_{n\to\infty}\left[\lambda_{0,n}((1+x)e^{-x})\right]^{{1/(2n)}^{2/3}}. \]