The cosmology dependence of galaxy clustering and lensing from a hybrid N-body–perturbation theory model

Abstract

We implement a model for the two-point statistics of biased tracers that combines dark matter dynamics from N-body simulations with an analytic Lagrangian bias expansion. Using Aemulus, a suite of N-body simulations built for emulation of cosmological observables, we emulate the cosmology dependence of these nonlinear spectra from redshifts z = 0 to z = 2. We quantify the accuracy of our emulation procedure, which is sub-per cent at k = 1 hMpc−1 for the redshifts probed by upcoming surveys and improves at higher redshifts. We demonstrate its ability to describe the statistics of complex tracer samples, including those with assembly bias and baryonic effects, reliably fitting the clustering and lensing statistics of such samples at redshift z ≃ 0.4 to scales of kmax ≈ 0.6 hMpc−1. We show that the emulator can be used for unbiased cosmological parameter inference in simulated joint clustering and galaxy–galaxy lensing analyses with data drawn from an independent N-body simulation. These results indicate that our emulator is a promising tool that can be readily applied to the analysis of current and upcoming datasets from galaxy surveys.

Modelling nearest neighbour distributions of biased tracers using hybrid effective field theory

“…The value of 𝑏 𝑠 2 changes by ∼ 13% between the two. Kokron et al (2021) found that the tracer-tracer and tracermatter spectra are least sensitive to varying 𝑏 𝑠 2 . It is therefore the hardest to constrain with only these statistics, as evidenced by the simulated likelihood analyses carried out therein.…”

“…This spectrum is orders of magnitude smaller than other basis spectra at all scales (see fig. 1 of Kokron et al 2021).…”

“…We start by fitting the auto-correlation power spectrum 𝑃 ℎℎ , and the halo-matter cross-correlation spectrum 𝑃 ℎ𝑚 for the halo sampled defined above, using the procedure described in detail in Kokron et al (2021). We use 𝑘 max = 0.6ℎMpc −1 while performing the fit, to obtain the bias parameters 𝑏 𝑖 .…”

“…In Appendix A we report the effect of how 𝑏 𝑖 obtained from changing 𝑘 max affect the 𝑘NN-CDFs. We also note that, given the treatment of the shot noise term in Kokron et al (2021), the variance of 𝜖 (𝒒) is not determined separately from Poisson (or sampling) shot noise, and we set it to 0 in this subsection. This yields a set of bias values 𝑏 𝑖 , with which we construct the 𝑘NN-CDF predictions as outlined in Sec.…”

“…This was found to be especially true for moderately biased tracers at low redshifts, where the model worked well up to 𝑘 max ∼ 0.6 ℎMpc −1 , nearly doubling the range of scales from perturbation theory approaches. This approach, Hybrid Effective Field Theory (HEFT), has since been merged with an emulator framework (Kokron et al 2021;Zennaro et al 2021), and recently applied to data (Hadzhiyska et al 2021). While the existing studies all focused on 2-point clustering, the model applies the bias expansion at the field level, and then uses 𝑁-body evolution to compute the advected fields.…”