The problem of SPECFEM3D stability

[1231231231232]

My research requires me to calculate for a long time. NSTEP is set to 800000, and the Maximum suggested time step = 2.04999997E-05, so DT is set to 0.00002. After the simulation calculation of 600000, the receiver value becomes larger. It's been getting bigger ever since. There's obviously instability, why is that, and how do you fix it? Ask for help.

[mnagaso]

Instability in Spectral Element Method (SEM) simulations for wave propagation after many timesteps is a common challenge. I would retry with smaller DT, a better mesh or an implicit time scheme.