[FLASH-USERS] [EXT] Re: E-MHD Aniso conductivity BUG

[Hansen, Eddie]

Thank you for pointing that out Mathieu. There may be a units issue for transport coefficients when using non-default UnitSystem (something other than ‘none’). The transport routines assume that the B-field has 1/sqrt(4*pi) built-in already. We will look into this more carefully for all transport processes and include any fixes in a patch and/or in a subsequent minor release version.

Jiri,
I also noticed a mistake in the inequality in your email. The paper you referenced indeed shows that kappa_perp / gamma_perp > 3. So the general relation should be:
kappa_perp > gamma_perp.

--
Eddie Hansen
Research Scientist
Flash Center for Computational Science


From: Mathieu Bailly-Grandvaux <mbaillygrandvaux at eng.ucsd.edu>
Date: Monday, May 15, 2023 at 3:51 PM
To: Hansen, Eddie <ehansen at pas.rochester.edu>
Cc: Loffelmann, Jiri <loffejir at fjfi.cvut.cz>, flash-users at flash.rochester.edu <flash-users at flash.rochester.edu>
Subject: [EXT] Re: [FLASH-USERS] E-MHD Aniso conductivity BUG
Hi Eddie,

I chime in here about the B-field units, in case it is a relevant point of concern with the new routines for heat transport in the last released version of FLASH. What I used to do is to declare the flag "UnitSystem = 'cgs'" to avoid the 4pi normalization factor such as B can be expressed directly in Gauss (compared to the default "UnitSystem='none'", for which the 4pi factor is needed). Is it still correct and compatible with the new routines?

Best,
Mathieu

On Mon, May 15, 2023, 12:41 PM Hansen, Eddie <ehansen at pas.rochester.edu<mailto:ehansen at pas.rochester.edu>> wrote:
Hello,

The ‘chi’ that you are referring to is different between those two routines on purpose, but it’s not because of different B-field units…

The fits for the kappa’s from Ji & Held, Phys. Plasmas 20, 042144, 2013 use a collision time tau that is multiplied by an extra factor of Z. So with chi = omega*tau, the chi‘s in JiHeld/Conductivity_anisoFullState vs DaviesWen/Thermoelectric_fullState are different by a factor of Z.

The flux limiters for the Thermoelectric effects (e.g., Nernst) are not well-tested yet. They were implemented because we knew they might be needed in the future. This topic is addressed in Davies, et al., Phys. Plasmas 22, 112703 (2015).

--
Eddie Hansen
Research Scientist
Flash Center for Computational Science


From: flash-users <flash-users-bounces at flash.rochester.edu<mailto:flash-users-bounces at flash.rochester.edu>> on behalf of Loffelmann, Jiri <loffejir at fjfi.cvut.cz<mailto:loffejir at fjfi.cvut.cz>>
Date: Monday, May 15, 2023 at 1:43 PM
To: flash-users at flash.rochester.edu<mailto:flash-users at flash.rochester.edu> <flash-users at flash.rochester.edu<mailto:flash-users at flash.rochester.edu>>
Subject: [FLASH-USERS] E-MHD Aniso conductivity BUG
Hello FLASH users,

I have two questions for you.

#1
I noticed an error in the computation of anisotropic thermal conduction coefficients inside the new FLASH version 4.7 and I would like to check that my belief is correct.

In my E-MHD simulations, using all the newly added MHD terms (Seebeck, Nernst, CrossNernst, aniso. thermal cond.), the magnetic advection velocity corresponding to the Nernst term is in some situations similar or even greater than the advection of temperature due to grad(T) in the direction perpendicular to the magnetic field.
Using the formalism from this paper https://iopscience.iop.org/article/10.1088/1741-4326/ac25c1<https://urldefense.com/v3/__https:/iopscience.iop.org/article/10.1088/1741-4326/ac25c1__;!!Mih3wA!DuALZDn2Aiv1otRU3Bm64LhM3F-Oxfd7veokgy1ikyGQzwuH1akCr1srtyZRzD8LEimISKKkDrSKzSaG-q-LbgQm$>:
      gamma_perp >= kappa_perp
In addition, in the paper in Figure 2, there is clearly visible, that the ratio of these transport coefficients, kappa_perp / gamma_perp, should be always greater than 3.
Therefore, I suspect an implementation error in the computation of the kappa_pepr coefficient.

When reading through the two files that compute kappa_perp and gamma_perp coefficients:
/FLASH4.7/source/physics/materialProperties/Conductivity/ConductivityMain/JiHeld/Conductivity_anisoFullState.F90
/FLASH4.7/source/physics/materialProperties/Thermoelectric/ThermoelectricMain/DaviesWen/Thermoelectric_fullState.F90
I noticed that the computation is very similar and truly, for the same values of magnetization ('chi'), it gives the coefficients with the correct ratio.

Finally, I discovered that both codes compute different values of magnetization, with a difference of sqrt(4.0*PI).
I believe that in Conductivity_anisoFullState, the magnetic field is in Gaussian units and in Thermoelectric_fullState it is in different physical units. And since the computation of the magnetization is precisely the same, it differs in the end. I am not really sure why the units are different, as I am not as familiar with all of the code, but either way, I believe the magnetization should be the same...

I propose a quick fix:
add line 'xmag = xmag / sqrt(4.0*PI)' after line number 133, in file Conductivity_anisoFullState.F90.

I am happy to provide further details and the setup of the simulation where the incorrect fraction appears.


2#
Secondly, I would like to ask about the usage of flux limiters for the Nernst and CrossNernst terms in the E-MHD model inside FLASH.
I would like to use E-MHD for the laser-target problems and I am wondering what values to use for 'nernstflcoef' and 'crossfieldflcoef'?
Do you have any references or experience with this?
Should I use the same value as for thermal diffusion (diff_eleFlCoef)? When for laser-target problems, it is usually reasonable to set diff_eleFlCoef = 0.05-0.1...

Regards,
Jiří Löffelmann

_______________________________________________
flash-users mailing list
flash-users at flash.rochester.edu<mailto:flash-users at flash.rochester.edu>

For list info, including unsubscribe:
https://urldefense.com/v3/__https://flash.rochester.edu/mailman/listinfo/flash-users__;!!Mih3wA!DuALZDn2Aiv1otRU3Bm64LhM3F-Oxfd7veokgy1ikyGQzwuH1akCr1srtyZRzD8LEimISKKkDrSKzSaG-o0CDY7M$<https://urldefense.com/v3/__https:/flash.rochester.edu/mailman/listinfo/flash-users__;!!Mih3wA!DuALZDn2Aiv1otRU3Bm64LhM3F-Oxfd7veokgy1ikyGQzwuH1akCr1srtyZRzD8LEimISKKkDrSKzSaG-o0CDY7M$>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://flash.rochester.edu/pipermail/flash-users/attachments/20230515/47e55443/attachment-0001.htm>