Ambiguities in Radiation Transport Documentation

[hwhitehead]

I have a series of questions concerning the non-relativistic radiation transfer module. There is some overlap with a previous thread here, but I feel that there is enough new content to warrant a second post.

Sigma versus Kappa

The code requires various "sigma" values be defined as opacities for each frequency bin. On the documentation page, this is stated to be $\sigma = \rho \kappa$, but in the pgen for the thermal_multigroup.cpp, the $\sigma$ value for electron scattering is set directly to kappa_es, is this code snippet in error?

athena/src/pgen/thermal_multigroup.cpp

Line 133 in ee2b2c7

pnrrad->sigma_s(k,j,i,ifr) = kappa_es;

Rosseland Mean Opacity

In the previous thread, I asked if there was a mistake in the documentation, which states that sigma_s + sigma_a is the Rosseland mean opacity. From the associated code paper it seems to me that sigma_s is the scattering opacity and sigma_a is the Rosseland opacity. This appears to be substantiated in radiation.hpp where the two arrays are defined

athena/src/nr_radiation/radiation.hpp

Lines 54 to 55 in ee2b2c7

	AthenaArray<Real> sigma_s, sigma_a; // opacity
	//scattering and fluxes weighted Rosseland mean

Should sigma_a be the Rosseland mean, or should sigma_s + sigma_a be the Rosseland mean?

Angular Grids

There is not a lot of information about the angular grid configuration available in the github documentation. If I wished to only track radiation along the cardinal directions of a Cartesian grid ($x$, $y$, $z$), how could this be configured using the nzeta and npsi input variables? I am assuming nzeta=2, npsi=3 to give n_ang = nzeta*npsi/2 = 3 in 3D, but I am not sure. Would n_oct also need to be set to 1? Including some diagrams in the documentation might make this a little clearer.

Outputting Data

There is currently no information in the output documentation concerning the radiation routine. Are users required to write their own output routines for radiation based quantities e.g. mean intensity in each cell, or is there support for naming variables in the <output> block of the input file?

The hdf5 output file appears to add variables to the dataset if radiation is enabled, but I am not so familiar with this area of the code and so it is not obvious to me what is being saved.

athena/src/outputs/athena_hdf5.cpp

Lines 143 to 144 in ee2b2c7

	if(NR_RADIATION_ENABLED || IM_RADIATION_ENABLED)
	num_variables[n_dataset] += 20 *pmb->pnrrad->nfreq;

Could additions be made to the documentation to make it clear what is being saved, or what needs to be handled by the user in uov?

Apologies for the number of questions, but I wanted to make sure I had a strong basis before I started running things.

[yanfeij]

On Tue, Nov 19, 2024 at 12:22 PM Henry Whitehead ***@***.***> wrote: I have a series of questions concerning the non-relativistic radiation transfer module. There is some overlap with a previous thread here <#605>, but I feel that there is enough new content to warrant a second post. Sigma versus Kappa The code requires various "sigma" values be defined as opacities for each frequency bin. On the documentation page <https://github.com/PrincetonUniversity/athena/wiki/Non-relativistic-Radiation-Transport>, this is stated to be $\sigma = \rho \kappa$, but in the pgen for the thermal_multigroup.cpp <https://github.com/PrincetonUniversity/athena/blob/master/src/pgen/thermal_multigroup.cpp>, the $\sigma$ value for electron scattering is set directly to kappa_es, is this code snippet in error?

density =1 for this test problem.

https://github.com/PrincetonUniversity/athena/blob/ee2b2c761df84aa863c38d2964854c4cf2c70145/src/pgen/thermal_multigroup.cpp#L133 Rosseland Mean Opacity In the previous thread <#605>, I asked if there was a mistake in the documentation <https://github.com/PrincetonUniversity/athena/wiki/Non-relativistic-Radiation-Transport>, which states that sigma_s + sigma_a is the Rosseland mean opacity. From the associated code paper <https://iopscience.iop.org/article/10.3847/1538-4365/ac9231/pdf> it seems to me that sigma_s is the scattering opacity and sigma_a is the Rosseland opacity. This appears to be substantiated in radiation.hpp where the two arrays are defined https://github.com/PrincetonUniversity/athena/blob/ee2b2c761df84aa863c38d2964854c4cf2c70145/src/nr_radiation/radiation.hpp#L54-L55 Should sigma_a be the Rosseland mean, or should sigma_s + sigma_a be the Rosseland mean?

Rosseland mean refers to the way to calculate the mean opacity. Both sigma_s and sigma_a are averaged opacities using rosseland mean. For a typical opacity table (such as OPAL), they will give you Rosseland mean opacity that refers to the sum of sigma_a and sigma_s. Hope this is clear.

Angular Grids There is not a lot of information about the angular grid configuration available in the github documentation. If I wished to only track radiation along the cardinal directions of a Cartesian grid ($x$, $y$, $z$), how could this be configured using the nzeta and npsi input variables? I am assuming nzeta=2, npsi=3 to give n_ang = nzeta*npsi/2 = 3 in 3D, but I am not sure. Would n_oct also need to be set to 1? Including some diagrams in the documentation might make this a little clearer.

The angle system typically does not set an angular grid that is exactly along the coordinate system. A typical angular grid will be saved to a file Rad_angles.txt if you run the code, where the file tells you the projections of all the angles to each coordinate (n_x, n_y, n_z). So in short, existing angular grid does not support angles just along x, y z coordinates.

Outputting Data There is currently no information in the output documentation <https://github.com/PrincetonUniversity/athena/wiki/Outputs> concerning the radiation routine. Are users required to write their own output routines for radiation based quantities e.g. mean intensity in each cell, or is there support for naming variables in the <output> block of the input file? The hdf5 output file <https://github.com/PrincetonUniversity/athena/blob/master/src/outputs/athena_hdf5.cpp> appears to add variables to the dataset if radiation is enabled, but I am not so familiar with this area of the code and so it is not obvious to me what is being saved. https://github.com/PrincetonUniversity/athena/blob/ee2b2c761df84aa863c38d2964854c4cf2c70145/src/outputs/athena_hdf5.cpp#L143-L144 Could additions be made to the documentation to make it clear what is being saved, or what needs to be handled by the user in uov?

If you specify prim or cons, the output will include moments of the specific intensities (radiation energy density E_r, radiation flux F_r, radiation pressure P_r tensor, opacities) all in code unit. The variables with _0 refer to co-moving frame quantities.

…

Apologies for the number of questions, but I wanted to make sure I had a strong basis before I started running things. — Reply to this email directly, view it on GitHub <#635>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACCWEJ6ZGBORNP7VTUKSRDD2BNXSPAVCNFSM6AAAAABSCUR3QKVHI2DSMVQWIX3LMV43ERDJONRXK43TNFXW4OZXGUYTINJYG4> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

-- Best, Yan-Fei Jiang Research Scientist, CCA, Flatiron Institute, 162 Fifth Avenue New York, NY 10010

[hwhitehead]

Thanks for the speedy reply, this clears a lot of things up. I think the Rosseland issue was a nomenclature confusion on my behalf, I've marked the previous thread as solved.

The angle system being detached from the coordinate system is interesting. Can you comment on the recommended number of angles? I assume this varies between simulation setups, but considering both the runtime and the memory requirements will be dependent on the angular resolution, I don't wish to over-do it. Is there a "minimum" number of angles/octants that you can recommend?