TY - JOUR

T1 - Global convergence and local superconvergence of first-kind Volterra integral equation approximations

AU - Brunner, Hermann

AU - Davies, Penny J.

AU - Duncan, Dugald Black

PY - 2011

Y1 - 2011

N2 - We present a comprehensive convergence analysis for discontinuous piecewise polynomial approximations of a first-kind Volterra integral equation with smooth convolution kernel, examining the attainable order of (super-) convergence in collocation, quadrature discontinuous Galerkin (QDG) and full discontinuous Galerkin (DG) methods. We introduce new polynomial basis functions with properties that greatly simplify the convergence analysis for collocation methods. This also enables us to determine explicit formulae for the location of superconvergence points (i.e., discrete points at which the convergence order is one higher than the global bound) for all convergent collocation schemes. We show that a QDG method, which is based on piecewise polynomials of degree m and uses exactly +1 quadrature points and nonzero quadrature weights, is equivalent to a collocation scheme, and so its convergence properties are fully determined by the previous collocation analysis and they depend only on the quadrature point location (in particular, they are completely independent of the accuracy of the quadrature rule). We also give a complete analysis for QDG with more than +1 quadrature points when the degree of precision (d.o.p.) is at least 2m+1. The behaviour (but not the approximation) is the same as that for a DG scheme when the d.o.p. is at least 2m+2. Numerical test results confirm that the theoretical convergence rates are optimal.

AB - We present a comprehensive convergence analysis for discontinuous piecewise polynomial approximations of a first-kind Volterra integral equation with smooth convolution kernel, examining the attainable order of (super-) convergence in collocation, quadrature discontinuous Galerkin (QDG) and full discontinuous Galerkin (DG) methods. We introduce new polynomial basis functions with properties that greatly simplify the convergence analysis for collocation methods. This also enables us to determine explicit formulae for the location of superconvergence points (i.e., discrete points at which the convergence order is one higher than the global bound) for all convergent collocation schemes. We show that a QDG method, which is based on piecewise polynomials of degree m and uses exactly +1 quadrature points and nonzero quadrature weights, is equivalent to a collocation scheme, and so its convergence properties are fully determined by the previous collocation analysis and they depend only on the quadrature point location (in particular, they are completely independent of the accuracy of the quadrature rule). We also give a complete analysis for QDG with more than +1 quadrature points when the degree of precision (d.o.p.) is at least 2m+1. The behaviour (but not the approximation) is the same as that for a DG scheme when the d.o.p. is at least 2m+2. Numerical test results confirm that the theoretical convergence rates are optimal.

U2 - 10.1093/imanum/drr029

DO - 10.1093/imanum/drr029

M3 - Article

SN - 1464-3642

VL - 32

SP - 1117

EP - 1146

JO - IMA Journal of Numerical Analysis

JF - IMA Journal of Numerical Analysis

IS - 3

ER -