See below for a listing of the most important code and interface changes in Colossus, starting with version 1.1.0.
Fixes a bug in the Halo mass function module, where redshift was not correctly passed to the sigma function.
The changes in this version were largely inspired by a detailed comparison with the Core Cosmology Library (CCL) by the LSST-DESC.
Physical and astronomical constants were updated to IAU 2015 / PDG 2018 standard, including the definition of parsec/kpc/Mpc and the solar mass. Those changes translate into changes in the gravitational constant in astronomical units and the critical density of the universe, which in turn are used in numerous functions.
This change affects most outputs from Colossus, but only by factors up to 1E-4 or less. All stored pickles will automatically be recomputed following this change.
sugiyama95transfer function model.
When manually changing cosmology, all derived parameters are now automatically updated. Previously, changes to T_CMB0 and Neff did not have any effect. Thanks to Sebastian Bocquet for pointing out this issue!
The Halo mass function module now correctly passes additional arguments to the power spectrum, variance, and collapse overdensity functions. This only makes a difference to the results if the user passes additional parameters such as a tabulated power spectrum. Thanks to Wojciech Hellwing for finding this bug!
- Removed reference to packaging package by adding manual version comparison function.
- Added unit tests for versioning and storage.
- Added unit tests for derived constants.
- Added a new FAQ and troubleshooting page to the documentation.
Fixed small discrepancy in the unit system. The gravitational constant was adjusted by a factor of 4E-5, leading to the same discrepancy in the critical density of the universe. Thanks to Tom McClintock for pointing out this bug!
This change affects numerous outputs from Colossus, but only by factors of around 4E-5 (and much less in most cases).
Added a system to automatically delete outdated storage files. If files older than a certain version are found, a warning is displayed, the file is deleted, and the computations will be done from scratch.
Fixed bug in the Bocquet et al. 2016 mass function for the M200c and M500c mass definitions (thanks to Michelle Ntampaka for catching this!).
diemer18concentration model to
diemer19to match the publication date.
Changed the default concentration model from
Fixed bug in wCDM growth factor calculation.
Added the mass function model of Comparat et al 2017 to the Halo mass function module.
Added the bias models of Bhattacharya et al 2011 and Comparat et al 2017 to the Halo Bias module. Thanks to Johan Comparat for the suggestion!
This version corresponds to the published version of the code paper.
This change of the Gaussian filter represents a significant, not backward-compatible change. If you use the Gaussian filter in ANY of your calculations, please check your results – they will be affected. Before re-computing your results, please remove all temporary cosmology files in
~/.colossus/cache/cosmologyto make sure that the change has taken effect.
Due to the change in the Gaussian filter, the return of the
peakCurvature()function has changed. If you use this function, please check your results (and follow the procedure described in the note above).
Many small fixes to the documentation, thanks to Jerry Maggioncalda for his careful proofreading!
Activated continuous integration (i.e., automatically running the unit test suite after every commit). Thanks to Joseph Kuruvilla for setting that up!
- The Diemer & Joyce 2018 concentration model is presented in its published form. The routine was sped up through a pre-computed, stored interpolation table.
xDelta()function in the Navarro-Frenk-White profile module was restructured completely. It now uses an interpolation table instead of root finding which means that it now allows numpy arrays as input and makes it orders of magnitude faster (depending on the size of the input). The accuracy of the interpolation is better than 1E-7. The function interface has two fewer parameters.
- The cosmology of the Multidark-Planck simulations was added.
This version fixes several bugs and adds new features. Changes in the cosmology module include:
- Major bug fix: the growth factor was incorrect for \(w \neq -1\) cosmologies, an error that has been rectified in this release (thanks to Lehman Garrison for catching this bug).
- The redshift interpolation tables in the cosmology module are now spaced equally in \(\ln(1 + z)\) rather than \(z\). This change reduces the interpolation errors slightly and, more importantly, leads to less ringing in the first derivatives of some quantities, namely the linear growth factor. The new interpolation tables carry different names than the old ones, meaning that old cache files do not need to be deleted as the two tables can co-exist. Due to the changed tables (and the changes to the growth factor), some cosmology functions can exhibit differences of the order 0.1% compared to the previous version.
- The Planck 2018 cosmology was added (and can be used by setting
planck18-onlyfor the cosmology).
inverseoption was removed from the
angularDiameterDistance()function because the inverse is multi-valued and leads to an error.
Changes in the large-scale structure module:
- Three new bias models were added to the Halo Bias module, namely those of Jing 1998, Seljak & Warren 2004, and Pillepich et al. 2010.
- The function
powerSpectrumSlope()was added to the Peaks in Gaussian random fields module. This function evaluates the slope of the power spectrum or variance at a given peak height and is used in the bias and concentration modules.
- Bug fix: the
ps_argsparameter was not used in the
peakCurvature()functions (thanks to Michael Joyce for catching this bug).
Changes in the halo module:
- The halo concentration models of Ludlow et al. 2016, Child et al. 2018, and Diemer and Joyce 2018 were added.
- The Diemer and Kravtsov 2015 model was updated according to Diemer and Joyce 2018.
- The default concentation model remains the original Diemer & Kravtsov 2015 model, without the
improvements of Diemer and Joyce 2018. In a near-future release, the default concentration
model will switch to their new model which will influence a few functions such as
changeMassDefinitionCModel(). However, the numerical differences to the previous default model are small.
- The function
plotChainwas removed from the MCMC module to avoid including the
matplotliblibrary. The function is still available as part of the MCMC tutorial.
- Numerous small improvements were made in the documentation.
Version 1.2.1 is the version that coincided with the first publication of the code paper on arXiv.org. The following major changes were made:
- The documentation was reworked entirely.
- All functions and parameters that were deprecated in 1.1.0 have been removed from the code (rather than outputting warnings).
qyparameters in the
halo.splashbackmodule were renamed to
q_outto conform with the rest of the code. A number of other small inconsistencies in splashback radius interface were fixed.
Version 1.1.0 presents a major change to the Colossus interface, documentation, and tutorial system. The most important changes are that
- A new top-level module for large-scale structure, LSS, has been added, including functions previously housed in the cosmology module, the old halo bias module, and a new module for the halo mass function. The LSS module covers funtions that deal with peaks or halos as a statistical ensemble so that the cosmology module does no longer “know” anything about halos. Conversely, the halo module covers functions that apply to individual halos.
- The demo scripts have been converted to much more extensive Jupyter notebook Tutorials.
- A number of interfaces have been made more homogeneous.
- Wherever possible, deprecated function interfaces are still present for backward compatibility but issue a warning. These functions and parameters will be removed in the next version.
- This documentation has been reorganized and improved, and its location has shifted to https://bdiemer.bitbucket.io/colossus.
The following functions are now housed in the LSS module:
- Cosmology.lagrangianR() is now
- Cosmology.lagrangianM() is now
- Cosmology.collapseOverdensity() is now
- Cosmology.peakHeight() is now
- Cosmology.massFromPeakHeight() is now
- Cosmology.nonLinearMass() is now
- Cosmology.peakCurvature() is now
- The module halo.bias is now
- The LSS module contains a brand new module to compute the halo mass function,
The following changes apply to interfaces across modules:
- Any module that implements models (e.g., fitting functions for concentration), now features an
ordered dictionary called
modelsthat contains class objects with the properties of the respective models (which vary from module to module). This change affects the power spectrum, bias, halo mass function, concentration, and splashback modules. These new model dictionaries replace the previous
MODELSlists that were present in some of the modules.
- There is a new storage module as part of utilities. The storage parameter in the cosmology
module was renamed to persistence, as was the global setting
PERSISTENCE). The storage module can now be used by other modules or from outside of Colossus.
Changes in the cosmology module:
- Cosmology now allows for a non-constant dark energy equations of state. The implemented dark
energy models include a fixed or varying equation of state (see
Cosmologyclass for more information). As a result, the OL0, OL(), and rho_L() parameters and functions were renamed to
- The power spectrum models were extracted into a separate module,
cosmology.power_spectrum. The names of the available models were changed from
eisenstein98_zbto conform with other Colossus modules.
Pk_sourceparameter was renamed to
matterPowerSpectrum()function. In functions that call the power spectrum, the user can pass a
ps_argsdictionary containing kwargs that are passed to the power spectrum function.
matterPowerSpectrum()function now takes redshift as an optional parameter.
text_outputoption was removed from the cosmology object.
soundHorizon()function now returns the sound horizon in Mpc/h rather than Mpc in order to be consistent with the rest of the cosmology module.
Changes in the LSS module:
collapseOverdensity()function has been completely reworked. By default, it still returns the constant collapse overdensity threshold in an Einstein-de Sitter universe. If a redshift is passed, it applies small corrections based on the underlying cosmology. The previous parameters to this function will now cause an error. This change also affects all functions that rely on the collapse overdensity, such as
peakCurvature(). These functions now accept dictionaries of parameters that are passed to the collapse overdensity and
- The halo bias module was extended with two new models for halo bias.
- The input units to the
twoHaloTerm()function are now in comoving Mpc/h rather than physical kpc/h in order to conform to the unit system of the LSS module.
Changes in the halo module:
- The interface of the SO changing functions in
halo.mass_defshas changed. The function previously called pseudoEvolve is now called
evolveSO()to reflect its more general nature. The
pseudoEvolve()function is a wrapper for evolveSO, and has one fewer parameter than previously (no final mass definition).
DK14Profileconstructor does not take R200m as an input any more and instead computes it self-consistently regardless of what the other inputs are. In this new version, the redshift always needs to be passed to the constructor. These changes fix a bug with outer profiles that themselves rely on R200m as an input. Furthermore, the normalization of power-law outer profiles is no longer adjusted in order to maintain a constant amplitude of R200m changes. It is up to the user to ensure that the behavior of the outer profile makes sense physically.
klypin14_mconcentration models were renamed to
klypin16_mto maintain compatibility with the publication date of their paper.