High order parameter uniform numerical method for singular perturbation problems (Q2371483): Difference between revisions

A numerical method is proposed for the following singularly perturbed linear two-point boundary-value problem \[ -\varepsilon (a(x)y')' + b(x)y = f(x), \quad (0,1), \quad y(0) = \eta_0,\quad y(1) = \eta_1. \] This equation is reduced to standard form by simple substitution used by the author himself and other researchers in the past. Then the proposed scheme is applied on a layer resolving piecewise-uniform Shishkin mesh. The author does not mention any advantage of their proposed scheme over the existing scheme of \textit{G. Sun} and \textit{M. Stynes} [Numer. Math. 70, No. 4, 487--500 (1995; Zbl 0824.65078)] which is also almost fourth order convergent and much simpler than the present scheme, and applicable for semi-linear problems. In addition, it seems that the results given in Lemma 5.4 are valid only for fixed mesh, and there is no explanation about its validity in the case of variable mesh. Further, in the rates of convergence given in Table 6, the uniform rate of convergence corresponding to \(r_5\) is \(0.24E+01\), whereas the error estimates of Theorem 5.5 is of order \(O(N^{-4})\).