Advancing the matter bispectrum estimation of large-scale structure: fast prescriptions for galaxy mock catalogues

Abstract

We investigate various phenomenological schemes for the rapid generation of 3D mock galaxy catalogues with a given power spectrum and bispectrum. We apply the fast bispectrum estimator \MODALLSS{} to these mock galaxy catalogues and compare to $N$-body simulation data analysed with the halo-finder \texttt{ROCKSTAR} (our benchmark data). We propose an assembly bias model for populating parent halos with subhalos by using a joint lognormal-Gaussian probability distribution for the subhalo occupation number and the halo concentration. This prescription enabled us to recover the benchmark power spectrum from $N$-body simulations to within 1\% and the bispectrum to within 4\% across the entire range of scales of the simulation. A small further boost adding an extra galaxy to all parent halos above the mass threshold $M>2\times10^{14}\,h^{-1} M_\odot$ obtained a better than 1\% fit to both power spectrum and bispectrum in the range $K/3<1.1\,h\,\text{Mpc}^{-1}$, where $K=k_1+k_2+k_3$. This statistical model should be applicable to fast dark matter codes, allowing rapid generation of mock catalogues which simultaneously reproduce the halo power spectrum and bispectrum obtained from $N$-body simulations. We also investigate alternative schemes using the Halo Occupation Distribution (HOD) which depend only on halo mass, but these yield results deficient in both the power spectrum (2\%) and the bispectrum (>4\%) at $k,K/3 \approx 0.2\,h\,\text{Mpc}^{-1}$, with poor scaling for the latter. Efforts to match the power spectrum by modifying the standard four-parameter HOD model result in overboosting the bispectrum (with a 10\% excess). We also characterise the effect of changing the halo profile on the power spectrum and bispectrum.