fastGRAHSP 1.0.2

Fast neural network simulator for fluxes of galaxies and quasars.

fastGRAHSP

Fast neural network simulator for fluxes of galaxies and quasars.

Generates observed fluxes for common photometric filters based on intrinsic galaxy properties.

fastGRAHSP can be used for simulating surveys extremely rapidly.

Usage

You need to prepare a model parameter array (see example below):

• log10 tau: time-scale of starformation in Myr

• log10 age: age of oldest stars in Myr

• log10 AFeII: amplitude of lines

• log10 Alines: amplitude of lines

• linewidth: width of broad emission lines

• Si: strength of 12µm silicate feature (positive: emission, negative: absorption)

• fcov: covering factor, based on the ratio of 12µm to 5100A luminosity ratio

• COOLlam: central wavelength in µm of cold torus component

• COOLwidth: width (in dex) of cold torus component

• log10 HOTfcov: peak luminosity ratio of hot to cold torus component

• HOTlam: central wavelength in µm of hot torus component

• HOTwidth: width (in dex) of hot torus component

• plslope: AGN power law slope

• plbendloc: wavelength of UV bend.

• log10 plbendwidth: width of UV bend (in dex).

• log10 EBV: E(B-V) attenuation coefficient of entire system.

• log10 EBV_AGN: E(B-V) attenuation coefficient of nucleus.

• alpha: dust slope of Dale+ model

• z: redshift

• M: log10 of galaxy stellar mass in solar masses

• L5100A: log10 of AGN luminosity at 5100A in erg/s.

For example, you can simulate the fluxes of 10000 galaxies like so:

N = 10000
df = dict(
    tau=np.random.uniform(1, 10000, size=N),
    age=np.random.uniform(1, 10000, size=N),
    AFeII=10**np.random.uniform(-1, 1, size=N),
    Alines=10**np.random.uniform(-1, 1, size=N),
    linewidth=np.random.uniform(1000, 30000, size=N),
    Si=np.random.uniform(-5, 5, size=N),
    fcov=np.random.uniform(0, 1, size=N),
    COOLlam=np.random.uniform(12, 28, size=N),
    COOLwidth=np.random.uniform(0.3, 0.9, size=N),
    HOTfcov=np.random.uniform(0, 10, size=N),
    HOTlam=np.random.uniform(1.1, 4.3, size=N),
    HOTwidth=np.random.uniform(0.3, 0.9, size=N),
    plslope=np.random.uniform(-2.6, -1.3, size=N),
    plbendloc=10**np.random.uniform(1.7, 2.3, size=N),
    plbendwidth=10**np.random.uniform(-2, 0, size=N),
    uvslope = np.zeros(N),
    EBV=10**np.random.uniform(-2, -1, size=N),
    EBV_AGN=10**np.random.uniform(-2, 1, size=N),
    alpha=np.random.uniform(1.2, 2.7, size=N),
    z=np.clip(np.random.normal(size=N)**2, 0, 6),
    M=10**np.random.uniform(7, 12, size=N),
    L5100A=10**np.random.uniform(38, 46, size=N),
)

emulator_args = np.transpose([
    np.log10(df['tau']),
    np.log10(df['age']),
    np.log10(df['AFeII']),
    np.log10(df['Alines']),
    df['linewidth'],
    df['Si'],
    df['fcov'],
    df['COOLlam'],
    df['COOLwidth'],
    df['HOTlam'],
    df['HOTwidth'],
    np.log10(df['HOTfcov']),
    df['plslope'],
    df['plbendloc'],
    np.log10(df['plbendwidth']),
    df['uvslope'],
    np.log10(df['EBV']),
    np.log10(df['EBV_AGN']),
    df['alpha'],
    df['M'],
    df['L5100A'],
    df['z'],
    df['EBV'] + df['EBV_AGN'],
])
results = predict_fluxes(emulator_args)
total_fluxes, total_columns, GAL_fluxes, GAL_columns, AGN_fluxes, AGN_columns = results
i = total_columns.index('WISE1')
print(total_fluxes[:, i])  # WISE1 flux in mJy

X-ray flux

There is also a module to convert given:

• Intrinsic AGN luminosity in the 2-10keV rest-frame energy range

• Column density N_H in cm^2

• Photon Index

• Redshift

Here is an example:

from fastGRAHSP.xray import XFluxModel
model = XFluxModel('hard', lumdistfunc)
nH = np.linspace(20.01, 25.9, 100)
Gamma = 2 + np.zeros(100)
z = 2.0 + np.zeros(100)
L = 45 + np.zeros(100)
logflux = model.photon_log10flux_for(L, Gamma, nH, z)

The observed X-ray bands available are:

• ‘soft’: Chandra 0.5-2keV

• ‘full’: Chandra 0.5-7keV – actually

• ‘hard’: Chandra 2-7keV – can be compared to results from a ECF of 2.625E-11 corresponding to a PhoIndex=1.4

• ‘ehard’: eROSITA 2.3-5keV – can be compared to results from a ECF of (1 / 1.102e+11 * 1.56 * 1.8) corresponding to a PhoIndex=2

• ‘emain’: eROSITA 0.6-2.3keV – can be compared to results from a ECF of (1 / 1.054e+12 * 0.29 * 1.8) corresponding to a PhoIndex=2

• ‘uhrd’: NuSTAR 3-24keV

• ‘bat’: BAT 14-195keV

Where indicated above, the fluxes are actually converted to net source counts, and then converted back to fluxes with a ECF. This gives a more realistic comparison to real data. Responses used are from COSMOS (NuSTAR), CDF-S 4Ms (Chandra, see BXA example source 179) and eFEDS (eROSITA).

In the example above, lumdistfunc is, if using astropy.cosmology:

from astropy.cosmology import FlatLambdaCDM
import astropy.units as u
cosmo2 = FlatLambdaCDM(H0=70, Om0=0.3)
def lumdistfunc2(z):
    return cosmo2.luminosity_distance(z).to(u.cm)

If using cosmolopy, it is:

import cosmolopy
cosmo = {'omega_M_0' : 0.3, 'omega_lambda_0' : 0.7, 'h' : 0.7}
cosmo = cosmolopy.distance.set_omega_k_0(cosmo)
dist = cosmolopy.distance.quick_distance_function(cosmolopy.distance.luminosity_distance, zmax=8, **cosmo)
def lumdistfunc1(z):
    return dist(z) * cosmolopy.constants.Mpc_cm

Licence

GPLv3 (see LICENCE file). If you require another license, please contact me.

Release Notes

1.0.0 (2025-06-11)

• Include X-ray flux computation

0.1.0 (2025-06-10)

• First version, multiwavelength fluxes