Numerical integration of functions with poles near the interval of integration (Q1379702): Difference between revisions

The author presents an automatic quadrature scheme for approximating integrals of the form \[ I(f; K)= \int^1_{-1} f(x)K(x)dx, \] where \(f(x)\) is a given smooth function on \([-1,1]\) and \(K(x)\) is a singular function having a double pole on the real axis or a complex conjugate pair of double poles near the interval of integration. The function \(f(x)\) is approximated by the sum \(p_N(x)= \sum^N_{k= 0} a^N_kT_k(x)\), of Chebyshev polynomials \(T_k(x)\), interpolating \(f(x)\) at the \(N+1\) abscissas \(\cos(\pi j/N)\), \(0\leq j\leq N\). The coefficients \(a^N_k\) are sufficiently computed by using the fast Fourier transform. An extrapolation method of Richardson type is used to evaluate the integral \(\int^1_{-1} T_N(x)K(x)dx\). Some numerical results show that the proposed quadrature method works.