[FLASH-USERS] Jet ignition problems: split/unsplit Hydro (+ mass_scalar)

[dongwook at flash.uchicago.edu]

Dear Massimo,

I am the author of the unsplit hydro solver in FLASH3, and thanks for
using and testing it. I haven't had a chance to write it up yet,
but a full 3D paper will arrive in a near future.

I would like to answer your questions regarding to the unsplit solver.
My answers are inlined.

> 3) To try another way I've begun to use unsplit hydro solvers. This way
> the jet can be regulated
> every timestep. What kind of solver do you suggest for Jet ignition (I
> have not much knowledge of them)?
> I'm using HLLC and with 0.3 CFL the sims never crashes, also with high vel
> and resolution.
> How can it be possible such a difference between split and unsplit method?

I've seen couple of different code behaviors between the split and unsplit
solvers. Those differences absolutely depend on various numerical
approaches in each solver.

I can easily name key differences between them
  (listed as "ones in split" vs. "ones in unsplit"):
  (1) split vs. unsplit time update
  (2) orders of spatial accuracy
       (e.g., 3rd order PPM vs. 2nd order MUSCL-Hancock)
  (3) Riemann solvers
      (e.g., two-shock Riemann solver vs. HLL* family, Roe, LLF, etc)

There are absolutely more than these, but as you can see that it all can
be combined and generate different results in general.
This is just an example I'd want to give you some feelings that differences
in each numerical components can readily make big differences in
code stability, accuracy, and efficiency.

So, you sometimes need to do some research little bit in order to find a
right solver for your simulations. And I am glad to hear that the unsplit
solver can do more for you. By the way, your choice of HLLC solver
should work very well in most cases.

> 4) How much are the unsplit solvers in FLASH "reliable"?
> Have you done some tests or are they still in a "beta" phase?
> Again, are they adequate to study a jet ignition?

Of course, the version you have should be reliable. The code has been
tested, verified, benchmarked, and compared with a series of test
problems at the Flash Center before release. And clearly, there have
been many updates to the code since the release.
If you are more interested in getting more updated unsplit code
(before any official release), please let me know.

> 5) Last point. I need also the evolution of a MASS_SCALAR and it seems
> that it is not
> evolving with the flow with the unsplit solvers (HLLC). How can I use
> them? (In Flash.h it is present)
> Is it possible? If not, is it easy to implement (or someone else has done
> this)?

As Alexander already mentioned, this can be done without too much difficulty.
And I believe this already has been implemented since the last release.
I will double check this in the unsplit code and can get back to you later.

Best,
Dongwook


> Thank you very much in advance for any help.
> In literature I've seen many flash-users implementing jets; hope that
> someone has interesting way to attach
> the problems I mentioned (maybe yet solved), or also completely different
> approaches for jets.
>
>
>
> Max
>
>
> (Massimo Gaspari, University of Bologna, ITA)
>
> _________________________________________________________________
> Naviga al sicuro. Scarica gratis Internet Explorer 8 per MSN
> http://www.microsoft.com/italy/windows/internet-explorer/msn.aspx