The first kind Chebyshev--Lobatto quadrature rule and its numerical improvement (Q2491004)

In this paper the authors improve the \(n\)-nodes first kind Chebyshev-Lobatto quadrature rule by increasing its precision degree numerically from \(2n+1\) to \(2n+3\). In order to do this, they consider the integral bounds as two new unknown variables, so extending the basis space \(\{1,x,x^2,\ldots,x^{2n+1}\}\) to \(\{1,x,x^2,\ldots,x^{2n+3}\}\). The resulting nonlinear system has no analytical solution and it is solved numerically under some feasibility conditions. This means that the new rules are almost exact for the monomial basis \(f(x)=x^j,\;j=0,1,\ldots,2n+3\) and a study on the error value is presented. Many examples show the efficiency of such rules with respect to the first kind Chebyshev-Lobatto ones. In particular we point out that the improved 5-point formulas given in the paper can be compared in average with the 100002-point first kind Chebyshev-Lobatto formulas and the ten examples here considered show such an efficiency. In general, if \(f(x)\) has a rapidly convergent Taylor series, the new quadratures work better than the corresponding Gauss-Chebyshev rules.