PGSCM: A family of \(P\)-stable boundary value methods for second-order initial value problems (Q433923)

This paper is concerned with the numerical solution of initial value problems for special second order equations: \[ y''(t) = f(t, y(t)), y(t_0)= y_0, \quad y'(t_0)= y'_0, \quad t \in [t_0, t_0+T] \] having a periodic or oscillatory solution. The discretization methods are linear multistep methods with \( \nu \geq 2\) steps that are assumed to be be \(P\)-stable due to the special properties of the solution. More specifically for a uniform grid \( t_j = t_0 + j h,\) \( j=0, \dots ,N\), \( h = T/N\) with step size \(h\) in the integration interval \([t_0, t_0+T]\), the approximations to the solution \( y_j \simeq y(t_j)\) are defined by the discrete equations \[ y_{n+1} - 2 y_n + y_{n-1} = h^2 \sum_{j= - \nu}^{\nu} \beta_{j + \nu}^{( 2 \nu)} f(t_j, y_{n+j}) \] with \( n= \nu, \nu +1, \dots , N-\nu\), with suitable real coefficients \(\beta_{j + \nu}^{( 2 \nu)}\), together with \( \nu-1\) linear initial conditions at the left part of the grid and \( \nu +1\) conditions at the end part of the grid so that the resulting \(( 2 \nu)\)-step method has order \( 2 \nu\) and is \(P\)-stable. These type of methods are usually referred to as boundary value methods and are extensively studied by \textit{L. Brugnano} and \textit{D. Trigiante} [Solving ODEs by linear multistep initial and boundary value methods. Amsterdam: Gordon and Breach (1998; Zbl 0934.65074)]. The results of some numerical experiments for the 2-D test problem known as Kramarz's system are presented to show the error and stability behavior for several values of \( \nu\) and different frequencies.