A class of p-stable linear multistep numerical methods

From MaRDI portal





A new hybrid one-leg \(P\)-stable algorithm for approximate solution of the second-order ordinary differential equation \[ y= f(x,y)\tag{i} \] is derived. A two-step one-leg implicit algorithm for (i) has the form \[ \rho(E) y_n= h^2 f(\sigma(E)x_n, \sigma(E)y_n+ h^2\eta(E) f_n),\tag{ii} \] where \[ \rho(E)= \sum^2_{j= 1}\alpha_j E^j,\;\sigma(E)= \sum^2_{j=1} B_j E^j,\;\eta(E)= \sum^2_{j= 1}D_j E^j,\;Ey_n= y_{n+ 1}, \] \[ \alpha_2\neq 0,\quad \sum^2_{j= 1} B_j= 1,\quad f_k= f(x_k, y_k). \] The authors consider hybrid formulas which are a composition of two formulas like (ii) namely \[ y_{n+ r}+ \beta_1 y_{n+ 1}+ \beta_2 y_{n+ 2}= h^2 f(\sigma_1(E) x_n, \sigma_1(E) y_n+ h^2\eta_1(E) f_n),\tag{iii} \] \[ y_n- 2y_{n+ 1}+ y_{n+ 2}= h^2 f(\sigma_2(E) x_n, \sigma_2(E) y_n+ h^2\eta_2(E) f_n+ h^2 D^2_3 f_{n+ r}), \] \[ \sigma_1(E)= \sum^2_{j= 1} B^1_j E^j,\;\sigma_2(E)= \sum^2_{j= 1} B^2_j E^j,\;\eta_1(E)= \sum^2_{j= 1} D^1_j E^j,\;\eta_2(E)= \sum^2_{j= 1} D^2_j E^j, \] \[ \sum^2_{j= 1} B^1_j= \sum^2_{j= 1} B^2_j= 1,\quad 0\leq r\leq 2. \] Explicit values for the constants in (iii) are derived using the Padé approximation to the exponential function such that the formulas are \(P\)-stable. Results of a test computation with \(r= 0\) are presented.











This page was built for publication: A class of p-stable linear multistep numerical methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4702119)