A sixth-order exponentially fitted method for the numerical solution of the radial Schrödinger equation (Q2638734)

The numerical solution of the one-dimensional Schrödinger equation \(y''(r)=[\ell (\ell +1)/r^ 2+V(r)-k^ 2]y(r)\) which can be written in the form \(y''(r)=f(r)\cdot y(r),\) \(r\in [0,\infty)\) is developed using a new sixth-order Runge-Kutta method. Standard library packages are not usable due to the fact that the problem is posed on an infinite interval. Many alternative methods have been used in an attempt to solve a wide class of problems of this type efficiently. The numerical method which is suggested here is a combination of the sixth-order Runge-Kutta type method and the exponential fitting. The numerical integration method suggested in this paper is of the form \[ y_{n+1}+\alpha y_ n+y_{n-1}=h^ 2[\beta_ 0(y''_{n+1}+y''_{n-1})+\beta_ 1(y''_{n+}+y''_{n-})+Vy''_ n] \] with \(y''_{n+1}\equiv f(r_{n+1})y_{n+1}\) and \(r_{n+1}=r_ n+h.\) The requirement of this method is to be exact for any linear combination of the functions: \(\{1,r,r^ 2,r^ 3,\exp (\pm \omega r)\), r exp(\(\pm \omega r)\}.\) Some numerical results are presented here testing the bound states problem and the resonance problem. Tables with comparative results obtained for seven numerical methods (absolute errors and real time of computation in CPUs) are given in this paper denoting that the errors related to the suggested method are the smallest ones, for a small increase in the CPU time. It is an interesting work presenting a new method more efficient and precise than previously developed finite difference methods.