A fourth-order spline method for singular two-point boundary-value problems (Q1097009): Difference between revisions

Two methods are described for the solution of the (weakly) singular two- point boundary-value problem \(x^{\alpha}(x^{-\alpha}y')'=f(x,y),\) \(0<x<1\), \(0<\alpha <1\), \(y(0)=A\), \(y(1)=B\). Consider the uniform mesh \(x_ i=ih\), \(h=1/N\), \(i=0(1)N\). Define the linear functionals \(L_ i(y)=y(x_ i)\) and \(M_ i(y)=(x^{- \alpha}(x^{\alpha}y')')|_{x=x_ i}.\) In both these methods a piecewise spline solution is obtained in the form \(s(x)=s_ i(x)\), \(x\in [x_{i-1},x_ i]\), \(i=1(1)N\), where in each subinterval \(s_ i(x)\) is in the linear span of a certain set of (non-polynomial) basis functions in the representation of the solution y(x) of the two-point boundary value problem and satisfies the interpolation conditions: \(L_{i- 1}(s)=L_{i-1}(y),\) \(L_ i(s)=L_ i(y),\) \(M_{i-1}(s)=M_{i-1}(y),\) \(M_ i(s)=M_ i(y).\) The unknown parameters \(y_ i\) and \(M_ i(y)\), \(i=1(1)N-1\) satisfy conditions of the form \[ (*)\quad -\frac{1}{J_ i}y_{i-1}+(\frac{1}{J_ i}+\frac{1}{J_{i+1}})y_ i- \frac{1}{J_{i+1}}y_{i+1}+k_{i,i-1}M_{i-1}(y)+k_{i,i}\quad M_ i(y)+k_{i,i+1}\quad M_{i+1}(y)=0,\quad i=1(1)N-1. \] The first method consists in replacing \(M_ i(y)\) by \(f(x_ i,y_ i)\) and solving (*) to obtain the values \(y_ i\). In the second method, the solution \(\bar y{}_ i\) is generated at the nodal points by adapting a fourth-order method due to the first author and using the continuity conditions (*) to obtain the corresponding smoothed approximations for \(M_ i(y)\).