[FLASH-USERS] combining MHD and thermal conduction

[Slavin, Jonathan]

Hi Ernesto,

That worked! Just setting addThermalFlux to False prevents the crashes I
was getting and so far the results look good.  The run that I'm doing now
has the magnetic field set to zero, so I haven't tested what happens with a
non-zero field yet.  However, as I noted before, even a zero B field was
causing crashes previously.  So thank you very much for that tip!

Regards,
Jon

On Mon, Aug 14, 2017 at 5:50 PM, Slavin, Jonathan <jslavin at cfa.harvard.edu>
wrote:

> ​Hi Ernesto,
>
> I maybe should have mentioned that my runs have been in 2D (cylindrical).
> So have you done such runs?  Also for the conduction, I'm using unsplit
> diffusion with electron thermal conduction and power law conductivity.  You
> mention being careful with split vs. unsplit diffusion, but you don't say
> which one works - or what that means to be careful.​
>
> ​ Or is it just that parameter setting addThermalFlux? I'll try setting
> that to False to see if that helps.
>
> To answer your question, I'm not using super-time-stepping.  Didn't seem
> necessary since with Crank Nicholson diffusion you have unconditional
> stability.  So I haven't looked into how to use that.
>
> Thanks for the suggestions.
>
> Jon ​
>
>> ---------- Forwarded message ----------
>> From: ERNESTO ZURBRIGGEN <ezurbriggen at unc.edu.ar>
>> To: flash-users at flash.uchicago.edu
>> Cc:
>> Bcc:
>> Date: Mon, 14 Aug 2017 12:18:31 -0300
>> Subject: Re: [FLASH-USERS] combining MHD and thermal conduction
>> Hi Jon!
>>
>> >It made me wonder if there is some numerical way that the
>> implementations of thermal conduction and MHD interfere with each other.
>>
>> I have used MHD+(isotropic) Conduction without any problem of that sort.
>> One confusing thing is that you can select Diffusion/split and
>> Diffuse/unsplit. Be careful with that.
>>
>> Something curious is that if the simulation include unsplit/MHD +
>> (regardless of split or unsplit) Diffusion you have to also set the
>> runtime parameter addThermalFlux=.false. (.true. by default)
>> Otherwise, you are including the thermal conduction effect twice, i.e.,
>> heat diffusion via the HYPRE libraries decoupling the heat equation, and
>> adding the thermal flux to the total energy flux in the Hydro unit.
>> I think the most consistent configuration would be unsplit/MHD +
>> Diffusion/unsplit + addThermalFlux=.false.
>>
>>
>> >I tried reducing the conductivity further and found that only by
>> reducing it by many orders of magnitude (~8) was I able to get runs that
>> didn't crash
>>
>> Maybe you can try reducing the parameter dt_diff_factor.
>>
>> Are you using super-time-stepping?
>>
>> Best!
>> Ernesto.
>>
> --
________________________________________________________
Jonathan D. Slavin                 Harvard-Smithsonian CfA
jslavin at cfa.harvard.edu       60 Garden Street, MS 83
phone: (617) 496-7981       Cambridge, MA 02138-1516
cell: (781) 363-0035             USA
________________________________________________________
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20170815/85b55245/attachment.htm>