Paper 4, Section I, B

Numerical Analysis | Part IB, 2011

Consider the multistep method for numerical solution of the differential equation y=f(t,y)\mathbf{y}^{\prime}=\mathbf{f}(t, \mathbf{y}) :

l=0sρlyn+l=hl=0sσlf(tn+l,yn+l),n=0,1,\sum_{l=0}^{s} \rho_{l} \mathbf{y}_{n+l}=h \sum_{l=0}^{s} \sigma_{l} \mathbf{f}\left(t_{n+l}, \mathbf{y}_{n+l}\right), \quad n=0,1, \ldots

What does it mean to say that the method is of order pp, and that the method is convergent?

Show that the method is of order pp if

l=0sρl=0,l=0slkρl=kl=0slk1σl,k=1,2,,p\sum_{l=0}^{s} \rho_{l}=0, \quad \sum_{l=0}^{s} l^{k} \rho_{l}=k \sum_{l=0}^{s} l^{k-1} \sigma_{l}, \quad k=1,2, \ldots, p

and give the conditions on ρ(w)=l=0sρlwl\rho(w)=\sum_{l=0}^{s} \rho_{l} w^{l} that ensure convergence.

Hence determine for what values of θ\theta and the σi\sigma_{i} the two-step method

yn+2(1θ)yn+1θyn=h[σ0f(tn,yn)+σ1f(tn+1,yn+1)+σ2f(tn+2,yn+2)]\mathbf{y}_{n+2}-(1-\theta) \mathbf{y}_{n+1}-\theta \mathbf{y}_{n}=h\left[\sigma_{0} \mathbf{f}\left(t_{n}, \mathbf{y}_{n}\right)+\sigma_{1} \mathbf{f}\left(t_{n+1}, \mathbf{y}_{n+1}\right)+\sigma_{2} \mathbf{f}\left(t_{n+2}, \mathbf{y}_{n+2}\right)\right]

is (a) convergent, and (b) of order 3 .

Typos? Please submit corrections to this page on GitHub.