Adding ETHOS dark matter power spectrum to the code, plus smooth-k window function for making HMF #252

charlottenosam · 2021-07-26T18:40:33Z

This PR adds the ETHOS DM power spectrum as a function of new AstroParams h_PEAK and log10_k_PEAK (see Bohr+2020, Mason+in prep).

This parameterized power spectrum spans WDM (h_PEAK=0) to CDM (log10_k_PEAK --> inf), and everything in between has dark acoustic oscillations. log10_k_PEAK [1-10] is the scale at which the matter power spectrum has its first oscillation (or cut-off for WDM) and h_PEAK [0-1] is the relative height of that oscillation. This should enable some easy explorations of a range of DM models.

Major code changes:

Add new AstroParams h_PEAK and log10_k_PEAK. These are really cosmology parameters but they won't effect ICs on large scales -- they are really only changing the HMF, so it's fine to treat them as AstroParams
Add USE_ETHOS flag to FlagOptions
Move HMF window function FILTER from GlobalParams to FlagOptions so that it can be easily changed after ICs are calculated
Add smooth-k window function for HMFs (see Leo+2018), which works better than tophat + sharp-k for truncated matter power spectra. The default parameters were calibrated to fit CDM, WDM and ETHOS DM models well at z=5-12.
Some broadcasting of these new variables to functions that need them when calculating HMFs etc.

I guess we don't include any example notebooks though - would you guys like a notebook before we do the PR?

Including ETHOS DM models in 21cmFAST

With astro parameters and flag options for SFRD integrals

To avoid filter in ICs. Sigma8 takes a tophat always

into ETHOS_DM

add ETHOS transfer function to ETHOS branch

codecov · 2021-07-26T19:19:07Z

Codecov Report

Merging #252 (885b728) into master (7c81ba3) will decrease coverage by 0.14%.
The diff coverage is 71.42%.

@@            Coverage Diff             @@
##           master     #252      +/-   ##
==========================================
- Coverage   86.24%   86.10%   -0.15%     
==========================================
  Files          12       12              
  Lines        2778     2800      +22     
==========================================
+ Hits         2396     2411      +15     
- Misses        382      389       +7

Impacted Files	Coverage Δ
src/py21cmfast/inputs.py	`87.61% <69.23%> (-2.28%)`	⬇️
src/py21cmfast/outputs.py	`92.13% <100.00%> (ø)`

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update bb049dc...885b728. Read the comment docs.

steven-murray

Thanks so much @charlottenosam, this is really impressive.

I've made a couple of small comments on the code -- nothing major. The only major things to do, I think, are:

Update the changelog
Add a section to a tutorial (or a whole tutorial if you wish)
Add tests (perhaps just an extra test in the integration test suite)

src/py21cmfast/inputs.py

steven-murray · 2021-07-29T14:12:02Z

src/py21cmfast/inputs.py

+ h_PEAK, log10_k_PEAK: double, optional
+ ETHOS parameters for dark acoustic oscillations (DAOs) in the matter power spectrum.
+ Location (10^log10_k_PEAK h/Mpc-1) and height (h_PEAK from 0 to 1) of the first DAO peak.
+ Warm dark matter (WDM) corresponds to h_PEAK=0, and log10_k_PEAK changes its mass. See Ref.~YY. (TODO!)


Just noting that there's a TODO here. Is there a plan to publish first then merge?

steven-murray · 2021-07-29T14:16:37Z

src/py21cmfast/src/UsefulFunctions.c

@@ -151,27 +151,30 @@ double HI_ion_crosssec(double nu);


 /* R in Mpc, M in Msun */
-double MtoR(double M){
+double MtoR(double M, struct FlagOptions *flag_options){


I think it would be cleaner to have MtoR depend on the int variable filter rather than all of flag_options.

steven-murray · 2021-07-29T14:18:56Z

src/py21cmfast/src/ps.c

+ //add the suppression and DAOs due to ETHOS DM-DR interactions
+ // TODO: add debugging
+ // LOG_DEBUG("T=%e",ETHOS_DAOs(k, astro_params_sfrd));


Noting another TODO here. remove if unnecessary.

steven-murray · 2021-07-29T14:20:49Z

src/py21cmfast/src/ps.c

@@ -4292,3 +4370,51 @@ void FreeTsInterpolationTables(struct FlagOptions *flag_options) {
 LOG_DEBUG("Done Freeing interpolation table memory.");
 interpolation_tables_allocated = false;
 }
+
+double ETHOS_DAOs(double k, struct AstroParams *astro_params_sfrd){


again, better to have it depend just on h_PEAK and log10_k_PEAK, then pass those in directly when calling (since it's just two variables) in my opinion.

BradGreig

Hey @charlottenosam, this all looks pretty good to me.

My main comment basically extends the suggestion by @steven-murray. In passing through the couple extra parameters as function arguments, then you shouldn't need the Broadcast_struct_global_SFRD function.

src/py21cmfast/inputs.py

BradGreig · 2021-07-30T06:26:27Z

src/py21cmfast/src/21cmFAST.h

@@ -190,6 +195,7 @@ int CreateFFTWWisdoms(struct UserParams *user_params, struct CosmoParams *cosmo_
 void Broadcast_struct_global_PS(struct UserParams *user_params, struct CosmoParams *cosmo_params);
 void Broadcast_struct_global_UF(struct UserParams *user_params, struct CosmoParams *cosmo_params);
 void Broadcast_struct_global_HF(struct UserParams *user_params, struct CosmoParams *cosmo_params, struct AstroParams *astro_params, struct FlagOptions *flag_options);
+void Broadcast_struct_global_SFRD(struct UserParams *user_params, struct CosmoParams *cosmo_params, struct AstroParams *astro_params, struct FlagOptions *flag_options);


I think you should remove this function, and follow @steven-murray's suggestion of just passing the FILTER, h_peak and log10_k_PEAK parameters explicitly to the functions expecting them.

The idea of these functions are to broadcast everything to all functions in the respective PS (ps.c), UF (UsefulFunctions.c) and HF (heating_helper_progs.c) files. Thus this is essentially a duplication of Broadcast_struct_global_PS. As you only need to pass a couple of specific arguments once or twice, explicitly adding those as function arguments would be preferable.

charlottenosam · 2022-04-19T11:22:54Z

I am finally working on this again (slowly..!), so have converted to draft and will make the changes you have suggested :)

steven-murray · 2022-09-20T21:35:05Z

@charlottenosam bumping this :-)

JulianBMunoz and others added 24 commits March 9, 2021 10:56

First commit

64ab9b2

Including ETHOS DM models in 21cmFAST

Broadcasted SFRD struct

fc6e267

With astro parameters and flag options for SFRD integrals

Add sigma8 flag to dsigma_dk

cec70ca

To avoid filter in ICs. Sigma8 takes a tophat always

add ETHOS transfer function

2637b93

fixed hanging bracket

e9dfd7b

renamed variables so that code compiles

083ea8e

Merge branch 'ETHOS_DM' of https://github.com/charlottenosam/21cmFAST-v3

e2b010a

into ETHOS_DM

linting and check ETHOS_DAO has k variable always

d1362ce

made flake8 happy in inputs.py docstrings

0c86cf0

Merge branch 'master' into ETHOS_DM

b529d64

Merge pull request #216 from charlottenosam/ETHOS_DM

f003021

add ETHOS transfer function to ETHOS branch

docstring update to test push

b167b04

make sure USE_ETHOS is a flag

a1304e9

k_peak --> log10_k_peak and informative errors for its range

eaeb594

Merge branch 'master' into ETHOS_DM

ed14a0c

restore param checks in inputs.py that I accidentally overwrote...

ee9e55d

fix merge from master to ps.c and inputs.py with better errors

0372548

fix ps.c so install works

9f8f86c

fixed T_ETHOS to take absolute value of the whole thing

189a4b3

add clobber to output.save

3334c13

Merge branch 'master' into ETHOS_DM

79da4e0

fix smooth-k ST parameters (p was 1/2 what it should be)

ef32bb8

merge with master

6a6a0e9

reference placeholders

9398fa3

charlottenosam added context: python wrapper Predominantly affecting the Python wrapper context: C backend Changes occur predominantly in the C code type: feature: physical New physical feature to be included type: feature: code New ability in the code (user-side) labels Jul 26, 2021

steven-murray reviewed Jul 29, 2021

View reviewed changes

BradGreig requested changes Jul 30, 2021

View reviewed changes

charlottenosam added 3 commits August 11, 2021 14:49

workaround for SECOND_ORDER_LPT_CORRECTIONS

0580d28

Merge branch 'master' into ETHOS_DM

3150d27

add ValueError for FAST_FCOLL_TABLES with USE_ETHOS

885b728

charlottenosam marked this pull request as draft April 19, 2022 11:22

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Adding ETHOS dark matter power spectrum to the code, plus smooth-k window function for making HMF #252

Adding ETHOS dark matter power spectrum to the code, plus smooth-k window function for making HMF #252

charlottenosam commented Jul 26, 2021

codecov bot commented Jul 26, 2021 •

edited

Loading

steven-murray left a comment

steven-murray Jul 29, 2021

steven-murray Jul 29, 2021

steven-murray Jul 29, 2021

steven-murray Jul 29, 2021

BradGreig left a comment

BradGreig Jul 30, 2021

charlottenosam commented Apr 19, 2022

steven-murray commented Sep 20, 2022

Adding ETHOS dark matter power spectrum to the code, plus smooth-k window function for making HMF #252

Are you sure you want to change the base?

Adding ETHOS dark matter power spectrum to the code, plus smooth-k window function for making HMF #252

Conversation

charlottenosam commented Jul 26, 2021

codecov bot commented Jul 26, 2021 • edited Loading

Codecov Report

steven-murray left a comment

Choose a reason for hiding this comment

steven-murray Jul 29, 2021

Choose a reason for hiding this comment

steven-murray Jul 29, 2021

Choose a reason for hiding this comment

steven-murray Jul 29, 2021

Choose a reason for hiding this comment

steven-murray Jul 29, 2021

Choose a reason for hiding this comment

BradGreig left a comment

Choose a reason for hiding this comment

BradGreig Jul 30, 2021

Choose a reason for hiding this comment

charlottenosam commented Apr 19, 2022

steven-murray commented Sep 20, 2022

codecov bot commented Jul 26, 2021 •

edited

Loading