Cosmology Large Scale Structure



COSMICATLAS Project                                                                                




CATalogs for                                                      Publications

                 LArge-Scale                                             Presentations

           Structure                                            Authors








The analysis of cosmological large-scale structure experiments such as DESI, JPAS, eBOSS and Euclid demands exact models of galaxy clustering and precise estimates  of covariance matrices. Given the known practical restrictions at the time of generating thousand of N-body simulations (and mock catalogs) requested for such objective, the LSS group at IAC is currently developing novel algorithms envisaged to generate mock catalogs at low computational cost, and at the same time, suitable for precise asessments of the covariance matrices of cosmological observables (clustering, weak lensing, abundance). 




COSMICATLAS aims at generating a pipeline capable to produce an arbitrary number of mock catalogs with the precision demanded from the forthcoming galaxy surveys (Euclid, DESI). The core of the project is a novel technique, B AM  ( Bias Assignment Method ), designed to learn the bias of dark matter tracers from a single reference N-body simulation, and populate dark matter fields (produced with fast gravity solvers) to obtain tracer catalogs with percent precision (with respect to the reference) in the one-, two- and three-point statistics. BAM is a non-parametric approach to the concept of halo-bias mapping technique, and represents a generalization of previous methods such as  PATCHY , a parametric method succesfully applied to generate mock catalogs for the BOSS collaboraton.

The non-parametric nature of the BAM approach allows for the full capture of the complexity of halo-bias, while its learning character ensures the precision in the spatial distribution of tracers. The main components f the BAM pipeline can be summarized as follows:

Reference simulation: BAM uses the information from a single N-body simulation. In particular we aim at using pair-fixed simulations such as the UNITSIM . This information is represented by a tracer-(e.g. halo) catalog and the initial conditions of the simulation.

Gravity Solver Fast approaches to the cosmological distribution of dark matter particles, such as FastPM or Augmented-Lagrangian Perturbation Theory .

Learning approach: BAM can be understood as a first-order machine-learning approach to the subject of clustering of dark matter tracers. However,and more significantly, the BAM approach

has a clear physical input, traceable from beginning-to-end, allowing to draw clearer physical interpretations from its outputs.

Halo BiasBAM explores different local and non-local (short- and large- range) dependencies on

the halo-bias, providing a physical framework in which subjects such as assembly bias, can be detailed analyzed.




COSMICATLAS.exe: A C++ code with more than 70000 lines to be made public available at a late stage of the project.






Halo Bias.  Dark matter haloes are biased tracers of the dark matter distribution. The way halos traces the dark matter depends on the halo mass, assembly time, and on the type of cosmic-web envoronments they live in. Knots, filaments, sheets and voids, host halos with different clustering properties.















The non-parametric approach of BAM replicates the spatial distribution of dark matter halos to percent precision..








The method, and the different tests performed in order to validate its applicability to the current problem of generating mock catalogs, have been described in the following papers: 

The cosmic web connection to the dark matter halo distribution through gravity (F.-S. Kitaura, A.

Balaguera-Antolínez et al. 2020).

The bias of dark matter halos: assessing the accuracy of mapping techniques (M. Pellejero-Ibañez,

A. Balaguera-Antolínez, F.-S. Kitaura, et al. 2020)

One simulation to have them all: performance of the Bias Assignment Method against N-body simulations

( A. Balaguera-Antolínez, F.S. Kitaura et al. 2019 ).

BAM: Bias Assigment Method to generate mock catalogs (A. Balaguera-Antolínez, F. S. Kitaura, et al, 2018).





Mock Innsbruck: the connection between galaxies and dark matter halos (March 2020).

See presentation here .

Euclid General meetings (Helsinki 2019). See presentation  here .



Andrés Balaguera-Antolínez (Severo-Ochoa fellow IAC)

Francisco-Shu Kitaura (RyC IAC & Universidad de la Laguna)




* Halo MOCK catalogs

 BAM is currently generating mock catalogs to participate in mock chalenges in JPAS and DESI.  Usinfg the UNIT simulation, wBAM is replicating the halo distribution from redhsift z=1.2 to redhsift z=0 with percent accuracy (Balaguera-Antolínez & Kitaura, in preparation).


The left panel shows slices of 20 Mpc/h from the UNITs simulation compared with two   BAM  reconstructions. The two and tree point statistics (right pànel) of the BAM reconstructions are within 1% and 10% respectively, with respect to the reference.






* Applications to Hydro-simulations

The learning approach of BAM is currently being applied to reproduce main baryon properties from Htydro-simulation (Sinigaglia et al., in preparation).




COLLABORATORS (alphabetic order)

Raúl Angulo (Donostia International Physics Centre DIPC)

Metin Ata (Kavli IPMU)

Claudio Dalla Vecchia (Instituto de Astrofísica de Canarias)

Chia-Hsun (Albert) Huang (Kavli Institute for Particle Astrophysics and Cosmology, Stanford University)

Martha Lippich (Max-Planck Institute for Extraterrestrial Physics)

Kentaro Nagamine (Osaka University)

Marcos Pellejero-Ibañez (Donostia International Physics Centre DIPC)

Ariel Sánchez (Max-Planck Institute for Extraterrestrial Physics)

Francesco Sinigaglia (University of Padova)

Cheng-Zhao (Laboratory of Astrophysics, Ecole Polytechnique Federale de Lausanne)