CLASS-PT: nonlinear perturbation theory extension of the Boltzmann code CLASS

This is a modification of the CLASS code that computes the non-linear power spectra of dark matter and biased tracers in one-loop cosmological perturbation theory, for both Gaussian and non-Gaussian initial conditions.

This version is based on CLASS v3 and includes all upstream improvements (updated BBN, Halofit/HMcode, spectral distortions, etc.). A GitHub Actions workflow automatically checks for new upstream CLASS releases and opens a pull request when updates are available.

CLASS-PT can be interfaced with the MCMC sampler MontePython using the custom-built likelihoods found here.

Getting started

Read these instructions for the installation details. See also this troubleshooting guide.

The installation instructions for CLASS can be found on the official code webpage.

Dependencies

• C/C++ compiler (gcc/g++ recommended, with -fopenmp support)
• OpenBLAS (path configured in Makefile via OPENBLAS)
• Python 3 with NumPy and Cython (for the Python wrapper)

Compilation

First, edit the Makefile to set the OPENBLAS variable to the path of libopenblas.a on your system.

Then ensure the OpenBLAS and GCC shared libraries are on your library path:

export LD_LIBRARY_PATH=/path/to/openblas/lib:/path/to/gcc/lib64:$LD_LIBRARY_PATH

Build and install:

# Build the CLASS executable and static library
make class

# Build and install the Python wrapper
make classy

# Or build the wrapper in-place (for development)
cd python && python setup.py build_ext --inplace && cd ..

# Clean build artifacts
make clean

After installation, verify:

from classy import Class
c = Class()
c.set({'output': 'mPk'})
c.compute()
print(c.pk(0.1, 0))  # should print a positive number

Quick example

Once you are all set, check out this jupyter notebook for examples of working sessions. Here's a simple example of computing the galaxy power spectrum multipoles with CLASS-PT:

# Import modules
from classy import Class
import numpy as np

# Set usual CLASS parameters
z_pk = 0.5
cosmo = Class()
cosmo.set({'A_s':2.089e-9,
           'n_s':0.9649,
           'tau_reio':0.052,
           'omega_b':0.02237,
           'omega_cdm':0.12,
           'h':0.6736,
           'YHe':0.2425,
           'N_ur':2.0328,
           'N_ncdm':1,
           'm_ncdm':0.06,
           'z_pk':z_pk
          })
# Set additional CLASS-PT settings
cosmo.set({'output':'mPk',
           'non linear':'PT',
           'IR resummation':'Yes',
           'Bias tracers':'Yes',
           'cb':'Yes', # use CDM+baryon spectra
           'RSD':'Yes',
           'AP':'Yes', # Alcock-Paczynski effect
           'Omfid':'0.31', # fiducial Omega_m
           'PNG':'No' # single-field inflation PNG
         })
cosmo.compute()

# Define some wavenumbers and compute spectra
khvec = np.logspace(-3,np.log10(1),1000) # array of k in 1/Mpc
cosmo.initialize_output(khvec, z_pk, len(khvec))

# Define nuisance parameters and extract outputs
b1, b2, bG2, bGamma3, cs0, cs2, cs4, Pshot_nbar, a0_nbar, a2_nbar, b4 = 2., -1., 0.1, -0.1, 0., 30., 0., 3000., 0., 0., 10.
pk_g0 = cosmo.pk_gg_l0(b1, b2, bG2, bGamma3, cs0, Pshot_nbar, a0_nbar, a2_nbar, b4)
pk_g2 = cosmo.pk_gg_l2(b1, b2, bG2, bGamma3, cs2, a2_nbar, b4)
pk_g4 = cosmo.pk_gg_l4(b1, b2, bG2, bGamma3, cs4, b4)

You can also use the Mathematica notebook 'read_tables.nb' to read the code output. We also provide a technical summary of the fNL implementations here.

Using the code

You can use CLASS-PT freely, provided that in your publications you cite at least the code paper arXiv:2004.10607. Feel free to cite the other companion papers devoted to new large-scale structure analysis methodologies!

Authors

• Mikhail (Misha) Ivanov
• Anton Chudaykin
• Marko Simonovic
• Oliver Philcox
• Giovanni Cabass