a problem wit a trmporal variable in an astrophysical problem

[Miguel Caro]

Hello everybody,

i am trying to reproduce an experiment done in Ha's  article called "Numerical simulation of high mach number , Astrophysicial Jets with radioactive Cooling,J. Sci. Comput. 2005". Inthere it is reproduced the simulation of Mach 5 jet using CLAWPACK.

Well, i am trying to reproduce this example  of 2D euler equations with source term, but i have some errors  at the temporal variable. To illustrate thess errors, i put  the output from  terminal :

CLAW2... Step     1   Courant number = 0.046  dt =  0.5000D-03  t =  0.5000D-03

CLAW2... Step     2   Courant number = 0.475  dt =  0.5123D-02  t =         NaN

CLAW2... Step     3   Courant number = 0.475  dt =  0.5123D-02  t =         NaN

Any help i will appreciate

Regards.

Miguel.

--
"Calmaras la sed recogiendo el agua con tus propias manos".
Antiguo proverbio somalí.

Miguel A. Caro Candezano
Estudante de Doutorado
Universidade de São Paulo- ICMC

[Kyle Mandli]

What version of Clawpack are you using?  Have you tried turning of the source terms and seeing if the simulation still blows up?  From the paper it sounds like the Roe solver only works below Mach 6, are you trying at speeds below that?

Kyle

--
You received this message because you are subscribed to the Google Groups "claw-users" group.
To post to this group, send email to claw-...@googlegroups.com.
To unsubscribe from this group, send email to claw-users+...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/claw-users?hl=en.

[Miguel Caro]

Yes Kyle, the Roe Solver only works for Mach number below 6 in such problems.

I am working with  4.3 version.

2011/10/28 Kyle Mandli <kyle....@gmail.com>

[Kyle Mandli]

My guess is that either the initialization function is unphysical or the source term is not working (if it's enabled).  Have you tried running the shock bubble problem just to make sure something outside of your code is not working?

Kyle

[Miguel Caro]

Dear Kyle,

The source term is enabled.

I am going to write the initial conditions and the boundary conditions for this problem, maybe you can see something, sometimes when one see so much the same code lines one can not recognize the error, that i could not see.

The initial conditions :(rho,uv,p) = (0.5,0.0,0.4127)

boundary conditions at left (rho,uv,p) = (5,30,0,0.1427) if y \in [-0.05,0.05]

                                                (rho,uv,p) = (5,0,0,0.4127) otherwise.

I understand that the experiment in CLAWPACK only can reproduce with Mach number  below 5.

I activate the parameter in clawe2ez.data to work with source term:  method(5) =1

Regards 

Miguel.

2011/10/28 Kyle Mandli <kyle....@gmail.com>

[Miguel Caro]

Well Kyle,

i set CLAWPACK to solve the 2D  Euler equations  without source term.

I checked  the initial and boundary conditions and  I found  a "little" problem defining the domain. 

Resuming,  i solved the problem. But, it appears the NaN solution, probably for the existence of negative pressure and  negative density.

i was trying to replace  velocity u, at the left boundary, with small values to obtain a low Mach number.

I set u=5;  2 and  1.7;   and i obtained NaNs again.

Maybe the CLAWPACK version? I am actually using the V.4.3. Why? Because inthere it is the Euler 2D files.. ../applications/euler/2d/quadrants

In the last version, as 4.6, i did not see these directory, i mean the euler/2d/quadrants.

Miguel.

 

2011/10/29 Miguel Caro <mca...@gmail.com>

[Kyle Mandli]

I was wondering if negative pressure was the problem.  The HLLE solver should not do this if the initial and boundary conditions are correct so this might be an easy way to make sure that the Riemann solver is not producing negative pressures.  Another thing you might try is to check whether the calculated pressure becomes negative at the beginning of the Riemann solver and abort (or print the location in a warning) just to get an idea of where this is occurring.  If you can figure out which grid cell boundaries are producing the negative pressures you may get a better picture of what boundary condition is wrong.  

One last thing you could look at, run the 1D Euler problem with a Riemann problem defined by your boundary condition on the left and the initial condition on the right and see what happens.  If everything seems to be fine in that case, you could try resetting the boundary condition side back to what you are setting the values to at each time step (in b4step.f probably) and see what that produces.  I imagine you will find that for certain velocities that this will cause negative pressures to form.  The best way to do these sorts of boundary conditions is to figure out what flux you would like exactly at the boundary and calculate based on the current states what values of the ghost cells will give you the correct flux rather than just setting the states to what you want.

As for that version of Clawpack, I doubt that the older version of Clawpack is the problem but in any case we have not gotten around to updating all of the examples to the newest version of Clawpack unfortunately.  In reality though not much has changed inside the code as of 4.6 and that example could be modified to work with a little bit of effort.  Maybe someone on this list has already done the work (or has a similar example) that they could share?

Kyle