The MICE Grand Challenge Lightcone Simulation I: Dark matter clustering

TL;DR

This work introduces the MICE Grand Challenge (MICE-GC), a 70-billion-particle N-body lightcone simulation in a (3 Gpc/h)^3 volume designed to capture structure growth from linear to highly non-linear scales. It validates 3D and 2D clustering statistics, quantifies mass-resolution effects by comparison with lower-resolution runs and analytic fits (RPT, Coyote Emulator, Halofit), and demonstrates percent-level BAO accuracy in the 3D power spectrum. The study also analyzes angular clustering in real and redshift space via all-sky lightcone maps, including Kaiser effects and non-linear RSD, and examines higher-order clustering through the 3-point function, highlighting resolution and realization impacts. Finally, the authors announce a public data release (MICECAT v1.0) and situate this work as Paper I in a series detailing halo catalogs and lensing maps (Papers II and III).

Abstract

We present a new N-body simulation from the MICE collaboration, the MICE Grand Challenge (MICE-GC), containing about 70 billion dark-matter particles in a (3 Gpc/h)^3 comoving volume. Given its large volume and fine spatial resolution, spanning over 5 orders of magnitude in dynamic range, it allows an accurate modeling of the growth of structure in the universe from the linear through the highly non-linear regime of gravitational clustering. We validate the dark-matter simulation outputs using 3D and 2D clustering statistics, and discuss mass-resolution effects in the non-linear regime by comparing to previous simulations and the latest numerical fits. We show that the MICE-GC run allows for a measurement of the BAO feature with percent level accuracy and compare it to state-of-the-art theoretical models. We also use sub-arcmin resolution pixelized 2D maps of the dark-matter counts in the lightcone to make tomographic analyses in real and redshift space. Our analysis shows the simulation reproduces the Kaiser effect on large scales, whereas we find a significant suppression of power on non-linear scales relative to the real space clustering. We complete our validation by presenting an analysis of the 3-point correlation function in this and previous MICE simulations, finding further evidence for mass-resolution effects. This is the first of a series of three papers in which we present the MICE-GC simulation, along with a wide and deep mock galaxy catalog built from it. This mock is made publicly available through a dedicated webportal, http://cosmohub.pic.es.

Figures (14)

• Figure 1: MICE-GC dark-matter lightcone simulation at $z=0.6$. The image shows the wide dynamic range, about 5 decades in scale, sampled by this Nbody simulation.
• Figure 2: State-of-the art in cosmological simulations: performance in terms of survey volume and particle mass, or equivalently, faintest $L_{min}$ galaxy luminosity (or absolute magnitude in r-band) reached according to an HOD galaxy assignment scheme to populate halos with $100$ or more dark-matter particles.
• Figure 3: Baryon Acoustic Oscillations (BAO) measured in MICE-GC (black symbols with error-bars) power spectrum compared to the theory prediction from Renormalized Perturbation Theory, RPT (blue line, see Crocce and Scoccimarro 2008) and the latest numerical fit from the Coyote Emulator (orange line, Heitmann et al. 2013) and the revised Halofit (green line, Takahashi et al. 2012). The RPT model at two loops reproduces very well the BAO in the simulation across redshifts (each panel is shown up to the maximum $k$ where RPT is valid). In turn at $z=0$ the Emulator yields a very good match with MICE-GC (except for the amplitude of the first peak with a difference $\lesssim 2\%$). At $z=0.5,1$ the broad-band power has the correct shape but is $2\%$ (systematically) above the N-body. The revised halofit also agrees with MICE-GC at the $2\%$ level at these redshifts but the amplitude of the oscillations are somewhat too large. Displayed error-bars assume Gaussian fluctuations, $\sigma_P = \sqrt{2/n_{\rm modes}} P_k$, but we take $P_k$ to be the non-linear spectrum (see text).
• Figure 4: Matter power spectrum in MICE-GC at several comoving outputs compared to the latest available numerical fits, the revised Halofit (Takahashi et al. 2012) and the Coyote Emulator (Heitmann et al. 2013). The shaded regions show the $1-\sigma$ and $3-\sigma$ error bars given the box-size and the binning (see text for details). The overall matching with the emulator is at the level of $2\%$ (but systematically below), with a jump of $\lesssim 3\%$ at $k\sim 0.3-0.4\,{\it h}^{-1}{\rm Mpc}$. In turn, the revised Halofit over-predicts the power on small scales by $5\%$ to $8\%$ (see text for a more detailed discussion). Vertical arrows show BAO peaks for reference.
• Figure 5: Top Panel: Ratio of the initial Eisenstein & Hu linear power spectrum used in MICE-IR to the one used in MICE-GC from CAMB, both for the same (MICE) cosmology. Bottom Panel: suppression of nonlinear structure formation due to particle mass resolution seen through the ratio of the power spectrum measured in MICE-IR ($m_p \sim 2.3\times 10^{11} \,h^{-1}\,{\rm M_{\odot}}$) to the one in MICE-GC ($m_p \sim 3\times10^{10}\,h^{-1}\,{\rm M_{\odot}}$) at $z=0,0.5$ and $1$. MICE-IR measurements were corrected assuming a Poisson shot-noise and the slight difference in initial spectra was divided out. If we do not correct for shot-noise we find these ratios to be almost independent of redshift and to resemble the $z=0$ case shown by the blue line.