On a corrected Fejér quadrature formula of the second kind (Q281228)

The integral \(I(f) = \int_{-1}^1 f(t)\,dt\) is computed by an interpolatory quadrature formula, where the function \(f:[-1,1]\to {\mathbb R}\) is given. Motivated by \textit{T. Hasegawa} and \textit{H. Sugiura} [ibid. 130, No. 1, 135--149 (2015; Zbl 1316.65033)], the author uses a corrected Fejér quadrature formula \(Q_n(f)\) having as nodes \(\cos \frac{j \pi}{n+1}\) \((j=1,\dots,n)\) and the additional points \(\pm \cos \frac{\pi}{2(n+1)}\). Then \(Q_n(f)\) has positive weights given explicitly. Further, the author determines the precise degree of exactness and optimal error bounds for \(Q_n(f)\). The convergence of \(Q_n(f)\) to \(I(f)\) for \(n\to \infty\) is shown for Riemann integrable functions \(f:[-1,1]\to {\mathbb R}\) and for functions \(f\in C([-1,1))\) with monotonic singularity at \(1\), i.e., \(f\) is monotonic in some neighborhood of \(1\) such that the improper integral \(I(f)\) exists. Thus \(Q_n(f)\) has essentially the same rate of convergence of the known Clenshaw-Curtis method (cf. [\textit{C. W. Clenshaw} and \textit{A. R. Curtis}, ibid. 2, 197--205 (1960; Zbl 0093.14006)]). Numerical examples are given, too.