Superconvergence of piecewise polynomial collocations for nonlinear weakly singular integral equations (Q1127699): Difference between revisions

Using regularity results established by the authors in [Z. Anal. Anwend. 13, No. 3, 463-476 (1994; Zbl 0803.45007)], they present a complete and detailed analysis of the (global and local) convergence rates of piecewise polynomial collocation solutions (on suitably graded meshes) for nonlinear integral equations of Urysohn type \[ u(x)= \int^b_0 K(x,y,u(y))dy+ f(x),\quad x\in I:=[0,b]. \] Here, the kernel \(K(x,y,u)\) either contains a weak singularity (of algebraic or logarithmic type) along the diagonal \(x= y\) of the domain \(I\times I\) or is bounded but has a weakly singular derivative at these points. In addition to optimal \(L^\infty\) error estimates on \(I\), analogous estimates are derived for the collocation error at the collocation points, depending on the type of singular behavior of \(K\) and on the choice of the scaling parameter \(r\geq 1\) which determines the mesh \[ x^{(N)}_j= (b/2)(j/N)^r,\quad j= 0,1,\dots,N;\quad x^{(N)}_{N+ j}:= b- x^{(N)}_{N- j},\quad j=1,\dots, N. \] The theory is illustrated by two numerical examples using piecewise linear collocation at the Gauss points.