Some implicit fourth and fifth order methods with optimum processes for numerical initial value problems (Q1081284)

For the scalar initial value problem \(y'=f(x,y)\), \(y(x_ 0)=y_ 0\) some implicit Runge-Kutta type methods of the form \[ y_{n+1}=y_ n+v(y_ n-y_{n+1})+h\phi (x_ n,x_{n+1},y_ n,y_{n+1};h), \] \[ \phi (x_ n,x_{n+1},y_ n,y_{n+1};h)=\sum^{r}_{i=1}w_{i-1}k_{i-1},\quad k_ 0=f(x_ n,y_ n),\;k_ 1=f(x_{n+1},y_{n+1}), \] \[ k_ i=f(x_{n+1}+a_ ih,y_{n+1}+b_ i(y_{n+1}-y_ n)+h\sum^{i- 1}_{j=0}b_{ij}k_ j),\quad (a_ i\neq 0), \] (i\(=2,3,...,r-1)\), which combine aspects of the Newton iteration scheme with Cash's methods are developed. The attainable order and the stability of the method as well as the choice of parameters are discussed, and the local truncation error is analysed. Some numerical tests are given. The algorithms seem to have certain advantages. For instance, refering to the local accuracy and the number of function evaluations per step, say stage number, one has \(p(3)=4\) and \(p(4)=5\) (where p(r) denotes the order of r-stage) contrary to the A-stable Cash-Moore method with \(p(3)=4\) only, and to the L-stable Cash method with \(p(4)=3\).