If PML boundary conditions are used, the energy in the simulation region must die out, otherwise, the results may vary from the expected and give the wrong outcome. You should try increasing the number of PML layers and see if the energy in the simulation region is decreasing with time or not. Field time monitors can be used for that purpose.

Please check your system matrices for singularities. This may happen for the cases of singular matrices. Check, weather the inverses of your matrices exists? Also check weather your system is stable, controllable, reachable? If all these are okay, please check your theory and code dually. I hope this way you will solve your problem.

Good Luck.

Ali

Thank you very much, I'll try .