Variable stepsize implementation of multistep methods for \(y\prime\prime =f(x,y,y^{\prime})\) (Q2491754)

The authors develop a procedure to obtain \(k\)-step Falkner methods (explicit and implicit) in their variable step-size versions, providing recurrence formulas to compute the coefficients efficiently, for the special differential equations \(y''=f(x,y)\) in connection with the Störmer-Conell methods. The first section is an introduction concerning the numerical solution of the second-order-initial value problem (IVP) \[ y''(x)=f(x,y(x)),\;y(x_0)=y_0,\;y'(x_0)=y'_0 \tag{1} \] and for the general second-order IVP with a dissipative term: \[ y''(x)=f(x,y(x),y'(x)),\;y(x_0)=y_0,\;y'(x_0)=y'_0 \tag{2} \] One presents a numerical method due to \textit{V. M. Falkner} [Phil. Mag. VII. Ser. 21, 624--640 (1936; Zbl 0013.27603)] for integrating the problem (2) directly; the algorithm is given by the following explicit formulas: \[ \begin{matrix} y_{n+1}=y_n+hy'_n+h^2\sum_{j=0}^{k-1}\beta_j\nabla^jf_n\\ y'_{n+1}=y'_n+h\sum_{j=0}^{k-1}\gamma_j\nabla^jf_n\\ \end{matrix}\tag{3} \] where \(h\) is the stepsize, \(y_n\) and \(y'_n\) are approximations of the solution and its derivative at \(x_n=x_0+nh,\) \(f_n=f(x_n,y_n,y'_n)\) and \(\nabla^jf_n\) is the standard notation for the backward differences, \(\beta_j,\,\gamma_j\) being real coefficients. There are also similar implicit formulas, both explicit and implicit having in common a fixed stepsize. The second section introduces some definitions about symmetric polynomials and Newton-divided differences, including generalized divided differences, that will be used later. In the third section one gives the main results for constructing the explicit variable \(k\)-step Falkner method of algebraic order \(k\). Section four develops a similar procedure to obtain an implicit variable-stepsize Falkner scheme. The fifth section describes reccurence formulas to obtain all the coefficients that appear in the explicit and implicit formulae given in the above sections. The analyses of stability are made in the sixth section. The procedure used to study stability for the explicit method is also presented, for the implicit one the proof being similar. The seventh section is devoted to illustrating the procedure in two particular cases: methods of algebraic orders 2 and 4, both explicit and implicit. The eighth section presents the results of some numerical tests in order to illustrate the performance of the method. The last section contains some conclusions like the fact that the formulation of the presented methods saves time and may, also, be extended in an easy way to differential equations of higher order, avoiding, thus, the computation of \(k\)-folds integrals.