Two-sided solution of ODE's via a posteriori error estimates
Let y(x) be the solution of the initial value problem \(y'=f(x,y)\), \(y(0)=y_ 0\), \(x\in [0,1]\) where \(y_ 0\in [y_ 0,\bar y_ 0]\), f(x,y)\(\in [\underline f(x,y),\bar f(x,y)]\) and assume that it is solved by a discretization method computing approximations to the solution at the nodes of a grid in the integration interval. Let \(S_ h(x)\) be a cubic spline that interpolates the numerical solution at the grid points. With these assumptions, the aim of this paper is to prove that there exists a two-sided solution y I given by \(y\quad I=S_ h+[-1,1]S\quad v_ h+[-\alpha,\alpha]S\quad u_ h,\) where \(S\) \(u_ h\) and \(S\) \(v_ h\) are also cubic splines and \(\alpha\) is a constant which can be computed solving numerically some variational-type equations. Thus, a two-sided solution is given as a linear combination of cubic splines. Finally, the author presents an example in which this technique is applied to obtain a two-sided solution of a third order linear stiff problem.
This page was built for publication: Two-sided solution of ODE's via a posteriori error estimates
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1105995)
