A family of implicit Chebyshev methods for the numerical integration of second-order differential equations (Q1105997)

This paper is concerned with the numerical solution of initial value problems for special second-order differential equations \(y''=f(y,x)\). In order to derive the numerical methods, the authors start from an integral equation equivalent to the differential equation and then substitute the function f by their discrete Chebyshev expansion in the interval \([x- h,x+h]\). In this way, after some manipulations, they arrive to a formula \(y(x+h)-2y(x)+y(x-h)=...,\) whose right hand side is a linear combination of f(y(t),t) at the points \(x_{\pm j}=x\pm h(1+\cos \theta_ j)/2,\) \(j=0,1,...,n\), \(\theta_ j=(n-j)\pi /n.\) However, since \(y(x_ j)\) \(j=1,...,n-1\), are also unknown this formula must be supplemented with other implicit formulas of the same type to determine \(y(x_ j)\), \(j=1,...,n-1.\) Finally, some numerical experiments using as test problems the two- and nine-body problems with low eccentricity and Duffing's equations are presented. The above methods (for \(4\leq n\leq 20)\) are compared with Cowell, RK and RK-Nyström methods with fixed stepsize, and the authors conclude that although the computational cost per step of their methods is high, they allow large stepsizes and they are extremely accurate in large intervals of time.