Including Models for Population III Stars¶
One of the generic results of using More Realistic Galaxy Populations is that they tend to produce strong late-time absorption troughs in the global 21-cm signal (this was the point of Mirocha, Furlanetto, & Sun (2017). Our interpretation was that deviations from these predictions could indicate “new” source populations, like Population III stars and their remnants. Indeed, we found some evidence that such objects introduce asymmetries in the global 21-cm signal (see Mirocha et al. 2018 for details).
The PopIII stars in this paper are simple, and were designed to seamlessly integrate with ARES while capturing the general behavior of more detailed models (e.g., Mebane, Mirocha, & Furlanetto (2018). This section will describe how to make use of these models yourself.
The easiest way to tap into these models is via the ParameterBundle framework. To begin, the usual imports
[1]:
%pylab inline
import ares
import numpy as np
import matplotlib.pyplot as pl
Populating the interactive namespace from numpy and matplotlib
[2]:
pars = ares.util.ParameterBundle('mirocha2017:base') \
+ ares.util.ParameterBundle('mirocha2018:high')
############################################################################
## Parameter Bundle ##
############################################################################
## UPDATE: Setting pop_Tmin{0} -> None ##
## [ previously pop_Tmin{0} = 1.00e+04] ##
############################################################################
Parameter bundles are designed to be added together, so as to build-up more complex calculations piece by piece. The above snippet takes the default model from Mirocha, Furlanetto, & Sun (2017 and adds on the default PopIII model from Mirocha et al. 2018. The “high” suffix refers to the mass of the PopIII stars – in this case, high means \(\sim 100 \ M_{\odot}\). There are also bundles for “low” and “med” mass PopIII stars, which just changes the mass (and resultant spectra) according to the Schaerer (2002) models.
NOTE: The mirocha2017 models use some non-standard lookup tables by default. See Using Data From the Literature for more info.
You’ll notice that while the mirocha2017:base bundle contains parameters for two source populations – one that provides the UV emission and another that produces X-rays – the addition of the PopIII bundle adds two more populations, again one each for UV and X-ray. You can customize the properties of these sources further via the following parameters:
pop_sfr{2}The typical star formation rate (SFR) in PopIII halos (in \(M_{\odot} \ \mathrm{yr}^{-1}\)), in a mass-bin-averaged sense (i.e., we assume PopIII star formation occurs in discrete bursts so the SFR in any individual halo is ill-defined).pop_time_limit{2}The critical duration of the PopIII phase (in Myr).pop_bind_limit{2}The critical binding energy (in erg) at which point halos transition to PopII star formation.pop_temp_limit{2}The critical virial temperature (in K) at which point halos transition to PopII star formation.pop_rad_yield{3}The X-ray production efficiency in PopIII halos (in \(\mathrm{erg} \ \mathrm{s}^{-1} \ (M_{\odot} \ \mathrm{yr}^{-1})^{-1}\)).
NOTE: The duration of PopIII in individual halos will be determined by whichever of pop_time_limit, pop_bind_limit, and pop_temp_limit results in the smallest value (once converted to mass).
WARNING: When pop_time_limit is small (\(t \lesssim 20\) Myr) the difference between \(M_{\min}\) and \(M_{\max}\) approaches a single grid point in the halo mass function lookup table. As a result, the SFRD may suffer from numerical noise, which slows convergence.
It is possible to use a halo mass-dependent prescription for the PopIII SFR if you’d like. In that case, you’ll need to update pop_sfr_model{2} to be sfe-func. See More Realistic Galaxy Populations for a reminder on how to do that.
Note on Feedback¶
One of the defining features of PopIII sources is their susceptibility to the global Lyman-Werner (LW) background, which drives up the minimum halo mass for star formation and thus limits the PopIII star formation rate density (SFRD). Because the LW background depends on the SFRD, the introduction of PopIII sources means ARES calculations must be performed iteratively. As a result, you will notice that these models can be quite a bit slower than normal ares calculations (by a factor of a few up to an order of magnitude, typically.)
There is some control, here, however. If you’re not looking for much accuracy, you can change the default set of convergence criteria to accelerate things.
feedback_LWIf
False, LW feedback will be turned off.feedback_LW_mean_errIf
True, calculations will terminate as soon as the mean error in :math:M_{\min}or the PopIII SFRD satisfy the set tolerances (see next two parameters). By default, it isFalse.feedback_LW_sfrd_rtolThe relative tolerance needed in order for calculations to terminate. By default, this is
0.05.feedback_LW_sfrd_atolThe absolute tolerance needed in order for calculations to terminate. By default, this is
0.0(i.e., unused).feedback_LW_maxiterMaximum number of iterations allowed. By default, 50.
feedback_LW_mean_errIf
True, terminate calculation once mean error meets set tolerances. IfFalse(which is the default), require SFRD and/or \(M_{\min}\) to meet tolerance at all redshifts.feedback_LW_mixup_freqEvery
feedback_LW_mixup_freqiterations, use average of last two iterations rather than the prediction for the next step. This has been found to help speed-up convergence (see footnote #3 in the paper).
Performance Tricks¶
These models can be quite a bit more expensive than usual due to the iterative nature of the computation. To speed things up, you can run a grid of models and then use that grid to supply initial guesses to the solver for subsequent calculations, which generally reduces the number of iterations required dramatically, especially if the grid is finely sampled.
NOTE: For a more detailed intro to model grids, see Simple Parameter Study: 2-D Model Grid.
Let’s generate a very sparse grid just to get a feel for how this works.
[3]:
pars = ares.util.ParameterBundle('mirocha2017:base') \
+ ares.util.ParameterBundle('mirocha2018:high')
# Track the PopII and III SFRDs and Mmin values separately
blobs = ares.util.BlobBundle(**ares.util.ParameterBundle('mirocha2018:csfr_blobs'))
pars.update(blobs)
pars['feedback_LW'] = True
pars['feedback_LW_maxiter'] = 10
pars['pop_sfr{2}'] = 1e-5
pars['pop_time_limit{2}'] = 100.
############################################################################
## Parameter Bundle ##
############################################################################
## UPDATE: Setting pop_Tmin{0} -> None ##
## [ previously pop_Tmin{0} = 1.00e+04] ##
############################################################################
[4]:
# Initialize a ModelGrid object.
pars['verbose'] = False
mg = ares.inference.ModelGrid(**pars)
# Just vary one parameter for now.
mg.axes = {'pop_time_limit{2}': np.arange(100., 500., 100.)}
mg.run('popIII_grid', clobber=True, save_freq=1)
grid: N/A% | | ETA: --:--:--
Starting 4-element model grid.
Running 4-element model grid.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
grid: 25% |############ | ETA: 0:29:28
Checkpoint #1: Wed Feb 9 09:31:43 2022
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
grid: 50% |######################## | ETA: 0:18:55
Checkpoint #2: Wed Feb 9 09:40:49 2022
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
grid: 75% |#################################### | ETA: 0:09:18
Checkpoint #3: Wed Feb 9 09:49:50 2022
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
grid: 100% |################################################| ETA: 00:00:00
Checkpoint #4: Wed Feb 9 09:57:11 2022
grid: 100% |################################################| Time: 0:35:18
Processor 0: Wrote popIII_grid.*.pkl (Wed Feb 9 09:57:12 2022)
Calculation complete: Wed Feb 9 09:57:12 2022
Elapsed time (min) : 35.3
To speed this up, you could loosen the tolerance, e.g., feedback_LW_sfrd_rtol=0.2 instead of the default feedback_LW_sfrd_rtol=0.05, or further reduce feedback_LW_maxiter.
NOTE: You can enter the above in Python script and invoke with mpirun if you’ve got MPI and mpi4py installed.
To verify that this grid has run successfully, you could do something like
[5]:
anl = ares.analysis.ModelSet('popIII_grid')
############################################################################
## Analysis: Model Set ##
############################################################################
## ---------------------------------------------------------------------- ##
## Basic Information ##
## ---------------------------------------------------------------------- ##
## path : ./ ##
## prefix : popIII_grid ##
## N-d : 1 ##
## ---------------------------------------------------------------------- ##
## param #00: pop_time_limit{2} ##
############################################################################
[6]:
# Plot the time_limit vs. SFRD(z=105)
from ares.physics.Constants import rhodot_cgs
ax1 = anl.Scatter(['pop_time_limit{2}', 'popIII_sfrd_tot'],
ivar=[None, 10.], multiplier=[1, rhodot_cgs], fig=1)
ax1.set_yscale('log')
# Loaded popIII_grid.000.blob_1d.popIII_sfrd_tot.pkl
[7]:
# Just show all the PopIII SFRDs
ax2 = anl.ReconstructedFunction('popIII_sfrd_tot', samples='all', color='b', fig=2, multiplier=[rhodot_cgs])
ax2.set_yscale('log')
ax2.set_ylim(1e-8, 1e-2)
[7]:
(1e-08, 0.01)
Now, to use this model grid for further calculations, you can simply do:
[8]:
pars = ares.util.ParameterBundle('mirocha2017:base') \
+ ares.util.ParameterBundle('mirocha2018:high')
# Choose a value not represented in the grid.
pars['pop_time_limit{2}'] = 150.
# These are new!
pars['feedback_LW_guesses'] = 'popIII_grid'
pars['feedback_LW_guesses_from'] = ['pop_time_limit'] # no ID number needed
sim = ares.simulations.Global21cm(**pars)
sim.run()
############################################################################
## Parameter Bundle ##
############################################################################
## UPDATE: Setting pop_Tmin{0} -> None ##
## [ previously pop_Tmin{0} = 1.00e+04] ##
############################################################################
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
############################################################################
## Analysis: Model Set ##
############################################################################
## ---------------------------------------------------------------------- ##
## Basic Information ##
## ---------------------------------------------------------------------- ##
## path : ./ ##
## prefix : popIII_grid ##
## N-d : 1 ##
## ---------------------------------------------------------------------- ##
## param #00: pop_time_limit{2} ##
############################################################################
# Loaded popIII_grid.000.blob_1d.popIII_Mmin.pkl
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
############################################################################
## ARES Simulation: Overview ##
############################################################################
## ---------------------------------------------------------------------- ##
## Source Populations ##
## ---------------------------------------------------------------------- ##
## sfrd sed radio O/IR Lya LW LyC Xray RTE ##
## pop #0 : sfe-func yes - - x x x - x ##
## pop #1 : sfrd->0 yes - - - - - x x ##
## pop #2 : sfr-func yes - - x x x - x ##
## pop #3 : sfrd->2 yes - - - - - x x ##
## ---------------------------------------------------------------------- ##
## Notes ##
## ---------------------------------------------------------------------- ##
## + pop_calib_lum != None, which means changes to pop_Z ##
## will *not* affect UVLF. Set pop_calib_lum=None to restore ##
## "normal" behavior (see S3.4 in Mirocha et al. 2017). ##
## ---------------------------------------------------------------------- ##
## Physics ##
## ---------------------------------------------------------------------- ##
## cgm_initial_temperature : 20000 ##
## clumping_factor : 3 ##
## secondary_ionization : 3 ##
## approx_Salpha : 3 ##
## include_He : True ##
## feedback_LW : True ##
## feedback_LW_Mmin : visbal2014 ##
## feedback_LW_fsh : None ##
############################################################################
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #1 complete.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #2 complete: mean_err=4.23e-02, max_err=1.06e+01, z(max_err)=6.8
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #3 complete: mean_err=8.18e-03, max_err=1.64e-01, z(max_err)=7.7
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #4 complete: mean_err=2.25e-03, max_err=4.03e-01, z(max_err)=6.8
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #5 complete: mean_err=1.92e-03, max_err=9.91e-01, z(max_err)=6.8
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #6 complete: mean_err=7.93e-04, max_err=3.29e-01, z(max_err)=6.8
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# LWB cycle #7 complete: mean_err=3.13e-04, max_err=1.09e-01, z(max_err)=6.8
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/inits/inits_planck_TTTEEE_lowl_lowE_best.txt.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
# Loaded $ARES/input/bpass_v1/SEDS/sed.bpass.constant.nocont.sin.z020
# WARNING: revisit scalability wrt fesc.
# WARNING: revisit scalability wrt fesc.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded /Users/jordanmirocha/Dropbox/work/mods/ares/input/optical_depth/optical_depth_planck_TTTEEE_lowl_lowE_best_He_1000x2158_z_5-60_logE_2.3-4.5.hdf5.
# Loaded $ARES/input/hmf/hmf_ST_planck_TTTEEE_lowl_lowE_best_logM_1400_4-18_z_1201_0-60.hdf5.
gs-21cm: 100% |#############################################| Time: 0:00:02
ARES will initialize a ModelGrid object using the popIII_grid* files in the current working directory. Importantly, we’ve told ARES to use the parameter pop_time_limit to set initial guesses, which seems obvious here but in general we could have more parameters in the grid and we could opt to only use a subset here.
If all goes according to plan, ARES will use the pop_time_limit{2}=100 grid point, pulling the result for the minimum mass evolution (which we saved as a blob) and using it as the initial guess. This should result in a SFRD very similar, but not identical to, the pop_time_limit{2}=100 model in our grid. And, most importantly, the pop_time_limit{2}=150 model will take far fewer iterations to complete than the original pop_time_limit{2}=100 run, at least if you set
feedback_LW_maxiter to a much larger number.
This kind of thing can be very useful if you want to vary parameters that do not affect the PopIII SFRD. In fact this is precisely what we did in Mirocha et al. (2018) <http://adsabs.harvard.edu/abs/2018MNRAS.478.5591M>_. In this case, we knew ahead of time that we would only survey the exact models represented in the grid, so we set:
[9]:
pars['feedback_LW_guesses_perfect'] = True
which told ARES not just to use results from the ModelGrid as first guesses, but to assume they are perfect, in which case no further iteration by the solver is required.
[ ]:
[ ]: