[CIG-SEISMO] Asking SPECFEM2D code blowing up

[Max Rietmann]

Dear Jungrak,

You are running into the well-known Courant--Friedrichs--Lewy (CFL) 
condition. The size of the time step DT, is proportional to the size of 
the smallest element (dx_min).  When you make the mesh very fine, you 
must make the time step very small as well.

Usually in the parameter file, if you decrease the size of the 
time-step, you must also increase the number of steps you take, in order 
for the simulated time to be the same.

For example:

* make smaller mesh: dx_min from 4m -> 2m
* now smaller time-step: DT from 0.04 -> 0.02
* now more steps: NSTEP from 1000 -> 2000

hopefully this helps,

Max

On 07/24/2013 02:29 AM, Jungrak SON wrote:
> Dear CIG,
>
>   This is Jungrak Son from South Korea, who attended CIG workshop in 
> alaska. Thanks gain to you all who have prepared great workshop, 
> posters and presentations. As a master student, a beginner of SPECFEM, 
> it was a great great time and impressive very much.
>
> I have tried to understand SPECFEM2D code, but I found an error which 
> I could not overcome. When I try to run the code with very fine mesh , 
> dx = 2m, even homogeneous model becomes unstable and stop running with 
> below message.
>
> " Max norm of vector field in solid (elastic) =9.76E+144
> code become unstable and blew up in solid (elastic) "
>
> When I use very small delta_t, it becomes ok, but very short running 
> time was useless. Let me ask this to great CIG researchers. What 
> should I do to overcome this problem? What kind of parameter may I 
> change? or How can I understand relevant stability equation of SPECFEM2D?
>
> Thanks very much for your consideration and have a nice summer.
>
> - Jungrak Son ( J.Son )
>
>
> _______________________________________________
> CIG-SEISMO mailing list
> CIG-SEISMO at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo


-- 
---------------------------
PhD Student
Institute of Computational Science
USI Lugano, Switzerland
+41 (0)44 633 3566

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://geodynamics.org/pipermail/cig-seismo/attachments/20130724/938db3c7/attachment.html>