[FLASH-USERS] Can high order time integration scheme work when WENO5 or WENO-Z is used in conjunction with with AMR grids provided by PARAMESH 4 in FLASH code in conjunction with

[Dongwook Lee]

Hi Weigang,

The answer is no.

The current releases of FLASH do not have any high-order RK method. Both
the split and the unspilt solvers implement a second-order characteristic
tracing-based method on both uniform and AMR grids.

This second-order time advancement has an advantage in terms of keeping the
entire code both computationally efficient and implementation-wise simple
in regards of exchanging guard cells, applying boundary conditions,
flux-conservations, especially on big parallel AMR runs. This is because
the higher the temporal accuracy, the more you have to do the above
mentioned tasks in general.

On the other sides, you don't get the expected high-order convergence fully
from the high-order spatial solvers (e.g., 5th order using the WENO
methods) over grid scales as the second-order temporal accuracy override
them. As a result, you are going to see second-order accuracy in smooth
flows at best.

Hope this helps.

Best,
Dongwook

On Tue, Jan 12, 2016 at 8:56 AM, 曾卫刚 <cengwg14 at mails.tsinghua.edu.cn> wrote:

> Hello everyone,
>
>
>       Can high order time integration scheme (such as 3rd or higher order
> Runge Kutta  method ) work when WENO5 or WENO-Z is used in conjunction with
> with AMR grids provided by a PARAMESH 4?
>
>   Thanks!
>  WeiGang
>
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20160112/b9194ea9/attachment-0001.htm>