convergence
 

I am solving transient problems with semi-implicit scheme.But my codes is always divergent at one instance.It is possible that at that time, the physical quantitity changes dramatically.To solve this problem , I tried to decrease the time step,but after my time step decreased from 1e-11 to 1e-12, the problem still exists. Since my codes need to run unitill t=1e-5 second, I cannot decrease the time step any more. I also tried to use SUR, but I think the effect is same as decreasing the time step. Is it right? Can anyone give me some advice to solve this problem.

thanks !