BHMF inference code

This is the code to the paper

Multi-Generational Black Hole Population Analysis with an Astrophysically Informed Mass Function

Yannick Ulrich, Djuna Croon, Jeremy Sakstein, Samuel McDermott

arXiv:2406.06109

Usage

You can clone the repository by using

# read only
$ git clone --recursive https://github.com/samueldmcdermott/bhmf.git
# read & write
$ git clone --recursive git@github.com:samueldmcdermott/bhmf.git

Install requirements

You will need a reasonably modern fortran compiler (tested with gcc 10.2) and at least python 3.6. You may want to use a virtual environment for the required python packags

$ python -m venv bhmf
$ source bhmf/bin/activate
(bhmf) $ pip install -r requirements.txt

Next, you need to obtain the data from LVK. We use the GW Open Science API for this:

(bhmf) $ python get_data.py

Compile the code

Go to the directory multinest and run

$ cd multinest/
$ make

This should produce the multinest executable main.

Preparing the data

Before you can run our code, you need convert the data from the LVK format into our own. To do this, run

(bhmf) $ python prepare.py

This will load and merge all event files as well as the injection and create data.rec and inj.rec. These files are sufficient for inference and the full dataset from LVK is no longer required

Running the code

To set up a job, you need to prepare a run card describing the model, the priors, and the MCMC parameters. For example

m ppisn+planck+trivial
p  2.,  0.,  20.,  0.0, -4., -10., -7.0, -7.0,  20.,-4
P 10., 10., 120.,  0.5,  0.,   0., -0.3, -0.3, 150.,12.
n 100
t 0.5
e 0.3
o ppisn-h0/short
s

See below for a list of models and parameters. You can now run the MCMC

$ mpirun -np <number of jobs> ./main /path/to/run.card
$ python analyse.py /path/to/output

Models

• plp+flat+trivial+trivial: powerlaw+peak for the primary, flat for the secondary, no redshift or spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$

• plp+plp+trivial+trivial: powerlaw+peak for the primary and secondary, no redshift or spin. Primary and secondary are coupled through $q^\beta_q$. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $\beta_q$

• plp+pow+trivial+trivial: powerlaw+peak for the primary, power law for the secondary, no redshift or spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $k$

• plp+flat+planck+trivial: powerlaw+peak for the primary, flat for the secondary, fitting for $H_0$ but no spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $H_0$

• plp+pow+planck+trivial: powerlaw+peak for the primary, power law for the secondary, fitting for $H_0$ but no spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $k$, $H_0$

• plp+plp+plank+trivial: powerlaw+peak for the primary and secondary, fitting for $H_0$ but no spin. Primary and secondary are coupled through $q^\beta_q$. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $\beta_q$, $H_0$

• plp+flat+trivial+beta: powerlaw+peak for the primary, flat for the secondary, no redshift and $\beta$ distribution for spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $\alpha$, $\beta$

• plp+pow+trivial+beta: powerlaw+peak for the primary, power law for the secondary, no redshift and $\beta$ distribution for spin. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $k$, $\alpha$, $\beta$

• plp+plp+trivial+beta: powerlaw+peak for the primary and secondary, no redshift and $\beta$ distribution for spin. Primary and secondary are coupled through $q^\beta_q$. Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm max}$, $\mu$, $\sigma$, $\alpha$, $\lambda_p$, $\beta_q$, $\alpha$, $\beta$

• ppisn+flat+trivial+trivial: PPISN for the primary, flat for the secondary, no spin or redshift Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$

• ppisn+trivial+trivial: PPISN for the primary, physical for the secondary, no spin or redshift. Primary and secondary are coupled using $q^\beta_q$, Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$

• ppisn2P+trivial+trivial: PPISN for the primary, physical for the secondary, no spin or redshift. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$, Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q^{(0)}$, $\beta_q^{(1)}$

• ppisn+planck+trivial: PPISN for the primary, physical for the secondary, fitting for $H_0$ but not for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$, Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $H_0$

• ppisn+trivial+beta: PPISN for the primary, physical for the secondary, no redshift and $\beta$ distributions for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$, Parameters: $m_{\rm min}$, $\delta m$, $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $\alpha_1$, $\beta_1$, $\alpha_2$, $\beta_2$

• ppisn+trivial+beta-turnon: PPISN for the primary, physical for the secondary, no redshift and $\beta$ distributions for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$. The turnon parameters are fixed to their best fit of ppisn+trivial+trivial, i.e. $m_{\rm min}=4.09M_{\odot}$, $\delta m = 5.33M_{\odot}$ Parameters: $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $\alpha_1$, $\beta_1$, $\alpha_2$, $\beta_2$

• ppisn+trivial+gauss-turnon: PPISN for the primary, physical for the secondary, no redshift and normal distributions for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$. The turnon parameters are fixed to their best fit of ppisn+trivial+trivial, i.e. $m_{\rm min}=4.09M_{\odot}$, $\delta m = 5.33M_{\odot}$. We require $d < b$. Parameters: $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $\alpha_1$, $\beta_1$, $\alpha_2$, $\beta_2$

• ppisn+trivial+1beta-turnon: PPISN for the primary, physical for the secondary, no redshift and a single $\beta$ distributions for spin. Primary and secondary are coupled using $q^\beta_q$. The turnon parameters are fixed to their best fit of ppisn+trivial+trivial, i.e. $m_{\rm min}=4.09M_{\odot}$, $\delta m = 5.33M_{\odot}$ Parameters: $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $\alpha$, $\beta$,

• ppisn+trivial+1gauss-turnon: PPISN for the primary, physical for the secondary, no redshift and a signle normal distribution for spin. Primary and secondary are coupled using $q^\beta_q$. The turnon parameters are fixed to their best fit of ppisn+trivial+trivial, i.e. $m_{\rm min}=4.09M_{\odot}$, $\delta m = 5.33M_{\odot}$. We require $d < b$. Parameters: $m_{\rm BHMG}$, $a$, $b$, $d$, $\log_{10}\lambda_{21}$, $\log_{10}\lambda_{12}$, $\beta_q$, $\alpha$, $\beta$,

• ppisn+trivial+beta-mass: PPISN for the primary, physical for the secondary, no redshift and $\beta$ distributions for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$. The mass function parameters are fixed to their best fit of ppisn+trivial+trivial Parameters: $\alpha_1$, $\beta_1$, $\alpha_2$, $\beta_2$

• ppisn+trivial+gauss-turnon: PPISN for the primary, physical for the secondary, no redshift and normal distributions for spin. Primary and secondary are coupled using $q^\beta_q$ with two different $\beta$. The mass function parameters are fixed to their best fit of ppisn+trivial+trivial Parameters: $\alpha_1$, $\beta_1$, $\alpha_2$, $\beta_2$