Positivity for explicit two-step methods in linear multistep and one-leg form (Q855277)

This paper is concerned with the preservation of the positivity of solutions of differential systems when they are integrated numerically by means of explicit linear two step methods. In the first part of the paper the authors consider the one parameter family of explicit linear two-step methods with second order accuracy given by \( w_{n+2}= \sum_{j=0}^1 [ - \alpha_j w_{n+j} + \beta_j \Delta t f_{n+j}]\) where \( f_{n+j}= f(w_{n+j})\), \( \alpha_0= 1 - \xi\), \( \alpha_1= \xi -2 \), \( \beta_j = (\xi/2)+ (-1)^{j+1}\) and \(\xi\) is a free parameter. Assuming that there exist some positive \( h^* >0\) such that \( v + \Delta t f(v) \geq 0\) for all \( v>0\) and \( 0 < h \leq h^*\) and that \( w_0>0\) and \(w_1\) is obtained by the forward Euler method, they prove that the above explicit two-step method preserves the positivity for positive step sizes such that \( \Delta t \leq \gamma (\xi ) h^*\) with some \(\gamma\)-function. A study of this function shows that the optimal method is obtained for \( \xi =2/3\) and then \( \gamma (2/3)= 1/2\). In the second part of the paper the same approach is applied to two-step one leg methods obtaining slightly larger bounds than in the case of linear multistep methods.