Tabulated optically thin radiative cooling module with Townsend integration by dutta-alankar · Pull Request #349 · idefix-code/idefix

dutta-alankar · 2025-09-06T15:36:55Z

I have implemented optically thin radiative cooling module for Idefix. I have tested this for by allowing cooling of a collection of cells (initially at $2\times 10^6 \ \rm K$) in a uniform domain (and calculating temperature and $-\dot{e}/n_H^2$ from the code over time) which seems to exactly reproduce the cooling rate as supplied by the table read during runtime by the code (screenshot attached). The setup and analysis code is attached here for reference and further testing. I have also put in a sample Cloudy generated equilibrium cooling rate table for plasma at solar metallicity in presence of HM12 UV photo-ionizing background at redshift 0.

The integration method follows the Townsend algorithm (https://ui.adsabs.harvard.edu/abs/2009ApJS..181..391T/abstract), which is quite efficient in calculating the pressure drop due to radiative cooling. There is also a way to set a floor that switches off cooling below some temperature, as required by many popular setups. I have also included a small .rst documentation for this module.

Let me know if this is something worth integrating to the main repo and if something additional needs to be done.

Test setup:
cooling-test.tar.gz

Test result (purple: supplied table; green: code result):

dutta-alankar · 2025-10-20T14:53:19Z

@glesur I noticed that use of CGS quantities has now been incorporated into Idefix. If this radiative cooling module with Townsend integration is something you/developer community thinks worth being integrated into Idefix, then I would be happy to change the code to utilize the CGS units capability. Let me know.

Updated energy calculation in addSourceTerms.hpp to use energy per unit volume instead of specific energy.

dutta-alankar · 2026-06-19T09:31:12Z

I have now updated the Townsend radiative cooling module to utilize the code units to CGS units conversion as a runtime parameter from the ini file and the documentation has been updated accordingly along with the addition of a test problem of an advecting uniform box undergoing cooling in test/HD/CoolBox. @glesur @neutrinoceros Please let me know if you would like to have this feature in Idefix. I have been using this modified version of Idefix in simulating turbulent radiatively cooling mixing layers and also plan to use this in upcoming cloud-crushing simulations.

neutrinoceros · 2026-06-19T09:40:29Z

Hi @dutta-alankar
For the record, I have no say in what goes into idefix and it's not part of my job to work on it anymore.

dutta-alankar · 2026-06-19T09:43:59Z

Hi @dutta-alankar For the record, I have no say in what goes into idefix and it's not part of my job to work on it anymore.

@neutrinoceros I understand. Apologies for my lack of knowledge on the latest developments and dynamics in the Idefix development team.

neutrinoceros · 2026-06-19T09:46:29Z

No worries

glesur · 2026-06-22T07:08:50Z

@dutta-alankar thanks for the heads up and the additional test. I owe you an applogy for the delay in working on your PR...

I'll submit a review shortly, but in the meantime, I think you can probably start with fixing the pre-commit issues.

[FIX] `pre-commit` linting

Copilot

Pull request overview

Copilot reviewed 22 out of 24 changed files in this pull request and generated 6 comments.

dutta-alankar · 2026-06-29T10:09:29Z

+        // tabulated data does not enclose the temperature value
+        printf("RadCooling::TownsendIntegration Temperature out of range: T=%e, "
+               "valid range=[%e, %e]\n",
+               temperature, temperature_min_data, temperature_max_data);
+        Kokkos::abort("RadCooling::TownsendIntegration Temperature out of range");


It might be okay to let it pass.

Fix to not have radiative cooling in isothermal setup Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 22 out of 24 changed files in this pull request and generated 6 comments.

glesur · 2026-06-29T10:22:42Z

@dutta-alankar : good idea for the CIs. Drop me a message when you think this is ready so that I can trigger the workflow on our side.

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Copilot

Pull request overview

Copilot reviewed 26 out of 28 changed files in this pull request and generated 5 comments.

dutta-alankar · 2026-06-29T15:51:02Z

+  explicit Fluid_AddSourceTermsFunctor(Fluid<Phys> *hydro, real dt):
+    hydroin(hydro) {


Not related to my PR and I leave it to @glesur to consider this or not.

dutta-alankar · 2026-06-29T15:52:15Z

  IdefixArray3D<real> csIsoArr;
+  IdefixArray3D<real> delta_eng_cool;

+  Fluid<Phys> *hydroin;


Again not related to this PR and left to @glesur to decide

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

dutta-alankar · 2026-06-29T15:57:20Z

@dutta-alankar : good idea for the CIs. Drop me a message when you think this is ready so that I can trigger the workflow on our side.

@glesur I think now you can trigger the CI/CD workflow from your end.

dutta-alankar · 2026-07-05T09:44:41Z

@glesur In case you missed my previous comment. I think that this PR is now at a good state when you can trigger a CI/CD workflow.

The test runner introduced on develop (idfx_test_run.py) parses every testme.json during pytest collection and validates that the 'namings' list covers all parameters that vary across variants. The CoolBox testme.json predates the finalized schema: its variants vary 'mpi' while namings only listed 'ini', so collection raised Fail to generate tests from HD/CoolBox/testme.json : Naming parameter list not match the auto-detectection [...] Since one malformed testme.json aborts the whole collection, every CI test step of every job failed at its first test, regardless of -subdir. Two changes: * Rewrite the file following the canonical schema (cf. HD/FargoPlanet): shared settings move to a 'default' block (dec as integers), and the two runs are expressed as a single variant with mpi: [false, true] under namings "ini,mpi". Generates the same two configurations as intended. * Drop the check_file_produced entry: the runner evaluates it with a bare filename after restoring the repo-root cwd, so the check cannot see files written in the problem directory. The standard test (python/testidefix.py) already loads and validates analysis-time-temperature.dat, which covers the intent of the check.

dutta-alankar · 2026-07-06T14:02:35Z

@glesur Looks like the code can now clear the CI/CD tests.

dutta-alankar · 2026-07-06T17:45:36Z

@glesur I am now happy that all the tests passed. Although the test is classified in the Parabolic test group, I think this is a minor misclassification.

svalat

Hi @dutta-alankar,
we took some time to review your PR. Following some remarks and questions.

As said in the following comments, there are several goals here, if some can move in dedicated PRs it would be very nice to ease the tracking in the git history.

data file

The data file ./src/fuild/cooling/cooltable.dat is redundant with the one in the test case. As the one in the src is not used it would be nice the remove it.

svalat · 2026-07-07T07:30:25Z

+
+.. code-block:: ini
+
+   Cooling   Tabulated    cooltable.dat    Townsend    TcoolFloor    1.0e+04


As we might need to extend those arguments in the future and they have different meanings I would create a new section and set the parameters as distinct keys and enable when the section is present via input.CheckBlock().

An example (to adapt to what you think is better):

[Cooling] mode=tabulated table=cooltable.dat integration=Townsend # maybe 0 to disable as it is in K ? t_cool_floor=1.0e+04

svalat · 2026-07-07T07:44:04Z


    # transform all arguments from args into attributes of this instance
    self.__dict__.update(vars(args))
+    idefix_dir = os.path.realpath(os.path.abspath(self.idefixDir))


Is this patch required here ? I see no effect on the rest of the PR with those variables ?

If I well understand it is to call testme.py from anywhere by using abs paths ?

If it is that, can you make a dedicated PR for that as we need to look on side effects ?

svalat · 2026-07-07T07:47:40Z

+        // delta_eng(k,j,i) = eos.GetInternalEnergy(del_prs, Vc(RHO,k,j,i));
+      }
+      else {
+        while (T_indx_lo<=T_indx_hi) {


Why not using the interpolation of the LookupTable class you included ?

As it looks redundant with what does LookupTable.

Is it because you want to interpolate with log ? But in this case it can probably be done by putting the table in log and using the linear interpolation of LookupTable ? Or I miss-understood something ?

svalat · 2026-07-07T07:53:37Z

+  this->data = hydroin->data;
+  this->eos = *(hydroin->eos.get());
+
+  const int nCoolingParams = input.CheckEntry("Hydro","Cooling");


As said in the doc part, in term of API it would be better to put the options in a [Cooling] section and use keys for the arguments as they have different semantics and will probably evolve.

svalat · 2026-07-07T08:02:06Z

  sslogFileName << "idefix." << idfx::prank << ".log";

-  std::string logFileName(sslogFileName.str());
+  std::string logFileName;


The patch into this file as a different goal than this PR, can you make it in another PR or is it already what is in #270 ?

By the way, we should avoid rush on all MPI ranks to create the directory and make it only in master rank. This is done in your other PR if I well understood.

Same remark for the changes in the 3 next files which are related.

svalat · 2026-07-07T08:06:09Z

  set(Kokkos_ENABLE_CUDA_LAMBDA ON CACHE BOOL "Idefix requires lambdas on Cuda" FORCE)
  # CUDA_MALLOC_ASYNC disbaled by default in Kokkos 4.5, so not required here
  #set(Kokkos_ENABLE_IMPL_CUDA_MALLOC_ASYNC OFF CACHE BOOL "Disable Async malloc to avoid bugs on PSM2" FORCE)
+


This part of the PR target another goal, is it possible to make it in a dedicated PR ?

Also I would ask two questions here which might be good to give in the dedicated PR :

Is this a warning or a failure ?

What is the OS / lib conditions to reproduce ?

svalat · 2026-07-07T08:07:34Z

  )
-  find_package(HDF5 REQUIRED)
-  target_link_libraries(idefix "${HDF5_LIBRARIES}")
+  # Prefer imported targets (matches older working behavior), then fall back.


Same remark here, it targets another goal than the current PR, would it be possible to make it in a dedicated PR ?

Also if you remember the HDF5 version and OS version on which you had the issue it would be very nice.

dutta-alankar · 2026-07-07T11:26:35Z

@svalat I agree that multiple file changes happened over time (~1 yr) as I not only developed this module but also tried improving/fixing parts of the code on the way. This unfortunately resulted in code changes that do not exactly fit the main goal of this PR. I apologize for this and appreciate your effort in sorting through this. I will, over the next few days, issue separate PRs with separate branches for:

Cooling related implementation
- Separate Cooling block in ini
- Usage of built-in interpolation table (with log10 space interpolation)
- Perhaps extend the lookup table module to also parse column arranged data
log file directory creation
Addressing compile failure due to Cuda linking on CMake
Detection and linking of serial/parallel hdf5
Minor testme patch

glesur · 2026-07-07T12:01:12Z

note: for the log file, we already have #270, so no need to duplicate it.

dutta-alankar and others added 11 commits August 30, 2024 14:09

Tolerance updated in RegressionTest in test/HD/sod/testme.py

b36dafa

[FIX] Dump datatype agnostic file xdmf IO

58de06b

Merge branch 'idefix-code:master' into master

c7f83dd

[ADD] tabulated optically thin radiative cooling module

92992da

[FIX] additional tags for successful cmake compilation

1357703

[FIX] minor typo in documentation

b07f7ff

[FIX] decrement internal energy by radiative cooling

70d6c40

[FIX] cpplint edits

d4aa609

[DOC] radiative cooling module documentation

231951b

[DOC] unit conversion

16a3844

[FIX] reverse old change

6f7733f

dutta-alankar commented Jan 7, 2026

View reviewed changes

Comment thread src/fluid/addSourceTerms.hpp Outdated

dutta-alankar commented Jan 7, 2026

View reviewed changes

Comment thread src/fluid/fluid.hpp

dutta-alankar added 9 commits January 14, 2026 12:08

Merge branch 'idefix-code:master' into master

391b273

Merge branch 'master' into master

ff6dbfa

Change energy calculation to energy per unit volume

f625997

Updated energy calculation in addSourceTerms.hpp to use energy per unit volume instead of specific energy.

Usage of runtime code units in cooling

7bb5ef1

documentation update

7f8d399

Fix formatting of code block reference in radCooling.rst

c6040dc

Fix cooling to work on GPUs

781e56d

Fix non-existent library linking error (librt stub)

b45b6c6

Handle output log directory creation and log file dumping

02cd06d

dutta-alankar added 2 commits June 22, 2026 11:26

[FIX] pre-commit linting

6cb070a

Merge pull request #1 from dutta-alankar/pr-349

be83ab6

[FIX] `pre-commit` linting

dutta-alankar requested a review from Copilot June 29, 2026 09:57

Copilot started reviewing on behalf of dutta-alankar June 29, 2026 09:58 View session

Copilot AI reviewed Jun 29, 2026

View reviewed changes

Fix with HAVE_ENERGYguard on cooling

9350e90

Fix to not have radiative cooling in isothermal setup Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

dutta-alankar requested a review from Copilot June 29, 2026 10:11

Copilot started reviewing on behalf of dutta-alankar June 29, 2026 10:12 View session

Copilot AI reviewed Jun 29, 2026

View reviewed changes

dutta-alankar and others added 7 commits June 29, 2026 12:25

Apply suggestions from code review

7ea869d

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Added test for Cooling box

80140a9

Added test for Cooling box

ec0582c

pre-commit linting fix

ccff439

Cooling box test added to CI/CD workflow

c245f30

Fix to pass simulation run tests

d682ba9

Fix default executable directory

f27975d

dutta-alankar requested a review from Copilot June 29, 2026 14:47

Copilot started reviewing on behalf of dutta-alankar June 29, 2026 14:47 View session

Copilot AI reviewed Jun 29, 2026

View reviewed changes

Apply suggestions from code review

be68a22

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

dutta-alankar added 2 commits July 5, 2026 11:37

Minor rearrangement of cooling cases

445fe6c

Minor comment modification

0318e7b

dutta-alankar force-pushed the master branch from 14f1fa5 to ea0a21e Compare July 5, 2026 21:46

svalat requested changes Jul 7, 2026

View reviewed changes

		explicit Fluid_AddSourceTermsFunctor(Fluid<Phys> *hydro, real dt):
		hydroin(hydro) {


		.. code-block:: ini

		Cooling Tabulated cooltable.dat Townsend TcoolFloor 1.0e+04

Uh oh!

Conversation

dutta-alankar commented Sep 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dutta-alankar commented Oct 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dutta-alankar commented Jun 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

neutrinoceros commented Jun 19, 2026

Uh oh!

dutta-alankar commented Jun 19, 2026

Uh oh!

neutrinoceros commented Jun 19, 2026

Uh oh!

glesur commented Jun 22, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

dutta-alankar Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

glesur commented Jun 29, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Uh oh!

Uh oh!

Uh oh!

dutta-alankar Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

dutta-alankar commented Jun 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

dutta-alankar commented Jul 5, 2026

Uh oh!

dutta-alankar commented Jul 6, 2026

Uh oh!

dutta-alankar commented Jul 6, 2026

Uh oh!

svalat left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

dutta-alankar commented Sep 6, 2025 •

edited

Loading

dutta-alankar commented Oct 20, 2025 •

edited

Loading

dutta-alankar commented Jun 19, 2026 •

edited

Loading

dutta-alankar Jun 29, 2026 •

edited

Loading

dutta-alankar Jun 29, 2026 •

edited

Loading

dutta-alankar commented Jun 29, 2026 •

edited

Loading

dutta-alankar commented Jul 7, 2026 •

edited

Loading