Tridiagonal C 1-collocation (Q1106631)

Let \(\Omega\) be a finite or infinite interval in the real line and L a differential operator of order not greater than 2; let \(\{x_ k\}\) be the finite or infinite set of nodes, in a uniform partition of \(\Omega\) into subintervals and let the differential equation \(Lu=f\) be approximated at the interior nodes \(x=x_ k\) of \(\Omega\) by the tridiagonal family of schemes \((a\cdot v)_ k=\alpha_{-1}v_{k-1}- \alpha_ 0v_ k+\alpha_ 1v_{k+1}=b(k),\) where \(v_ k\) is expected to be an approximation of \(u_ k\), the data b(k) being a linear combination of values of f at a selected set of points in the two subintervals adjacent to the interior nod \(x_ k\). The algebraic equation obtained from the discretization is \(Au=b.\) It is shown that several known schemes are members of the new method and the properties of the matrix A are listed. A new formulation of the collocation method is proposed. Some hints are given for the cases of nonconstant coefficients and nonlinear problems.