Dear Christophe, Can we use some explicit method at each time step instead of solving a big linear system while solving wave equation in GetDP? This will save a lot of computation time especially when I have thousand of time steps. If it can, do we have much trade off in accuracy? By the way, thank you for your comments on the Perfectly macthing absorbing layer. Lin Ji -- Dept. of Math., RPI 518-276,6904 (work)