Stability and B-convergence of general linear methods (Q908660)

For an initial value problem in a Hilbert space X: \(x'(t)=f(t,x(t)),\quad x(0)=x_ 0\) a general linear scheme of the form \[ Y_ i^{(n)}=h\sum^{s}_{j=1}c^{11}_{ij} f(t_ n+\mu_ jh,\quad Y_ j^{(n)})+\sum^{r}_{j=1}c^{12}_{ij} y_ j^{(n-1)},\quad i=1,2,...,s \] \[ y_ i^{(n)}=\sum^{s}_{j=1}c^{21}_{ij} f(t_ n+\mu_ jh,\quad Y_ j^{(n)})+\sum^{r}_{j=1}c^{22}_{ij} y_ j^{(n-1)},\quad i=1,2,...,r \] \[ \xi_ n=\sum^{r}_{j=1}\beta_ jy_ j^{(n)},\quad n=1,2,...,N \] is considered. In this scheme \(Y_ i^{(n)}\), \(y_ i^{(n)}\), \(\xi_ n\) should approximate \(x(t_ n+\mu_ ih)\), \(H_ i(t_ n+\nu_ ih)\), \(x(t_ n+\eta h)\), respectively, where \(H_ i\) denote certain information about the solution x. Various notions of convergence (as for example optimal B-convergence, B- convergence), of consistency and order, and of stability are introduced. The author establishes several relations between these notions.