statistical reconstruction and simulation of the initial
play

Statistical reconstruction and simulation of the initial conditions - PowerPoint PPT Presentation

Statistical reconstruction and simulation of the initial conditions of our Local Universe (PLUS project) Guilhem Lavaux Institut d'Astrophysique de Paris / CNRS Collaboration: J. Jasche, S. Peirani, PLUS collaboration November 27 th 2014


  1. Statistical reconstruction and simulation of the initial conditions of our Local Universe (PLUS project) Guilhem Lavaux Institut d'Astrophysique de Paris / CNRS Collaboration: J. Jasche, S. Peirani, PLUS collaboration November 27 th 2014 Workshop "Deciphering the Dynamic Universe"

  2. Motivation ● Galaxy catalogues static picture of the Universe ● Large scale structure non-trivial statistical properties ● Initial conditions trivial (and a test is interesting per se) ● Foreground/selection effects must be taken care of Constrained simulations complementary observations

  3. A static picture of Local Universe 2M++, Lavaux & Hudson (2011)

  4. The 2M++ galaxy redshift catalog Lavaux & Hudson (2011, MNRAS) Galaxy distribution Completeness at K=11.5 Completeness at K=12.5

  5. Local Universe simulation (PLUS) Aim: Biggest and most accurate simulation of the Local Universe Current observational base: 2M++ (Lavaux & Hudson 2012) SFI++/Taipan/6dF Larger surveys (2MPZ...) 2M++ dataset BORG SuperVelMod (Jasche & Wandelt 2013) (Lavaux 2014) Large scale Initial conditions (600 Mpc/h, 256 3 phases, extended to 1024 3 ) Fine bias calibration Dark Matter Baryon physics Small scale improvements (10-15 Mpc/h subbox, enforce distances) Local Group + Local Sheet simulation DONE TO DO NEW TECHNOLOGY

  6. The BORG data model Linear response operator (mask, radial selection) « Mean » density of tracers bias ∑ | ^ 2 δ k | − log P (δ , α , ~ ~ n )= A + ∑ α − N i α log ( 1 +δ NL , i [δ])+ k n R i ( 1 +δ NL ,i [δ]) 2 P k i NGP binned data Gaussian prior Poisson probability δ NL [δ] Field derived from Cloud-In-Cell + 2LPT dynamics δ sampled using Hamiltonian Monte Carlo algorithm

  7. Parameter exploration using HMC Assimilate the posterior to a potential energy : Create an Hamiltonian : Evolve system with Hamiltonian dynamics Compute acceptance : Pump and dump energy in the system by varying Generate a MC sample dropping the state of

  8. Parameter exploration using HMC Assimilate the posterior to a potential energy : Create an Hamiltonian : INVOLVE GRADIENT COMPUTATIONS ! INVOLVE GRADIENT COMPUTATIONS ! Evolve system with Hamiltonian dynamics Compute acceptance : Pump and dump energy in the system by varying Generate a MC sample dropping the state of

  9. Power spectrum / Phase convergence 10 -1 10 -2 10 -3 10 -4 10 -5 10 -6 0.1 1

  10. Large scale structure detection and caracterization

  11. Binned distribution of galaxies Sloan Wall

  12. Final dark matter density distribution Sloan Wall

  13. Dark matter density distribution (ARES view) Mpc/h Sloan Wall Mpc/h

  14. Supergalactic plane ? Coma Shapley concentration Perseus-Pisces Pisces-Cetus

  15. Another example: Tully's void Thick 2M++ slice (± 20 Mpc/h), Density field in the galactic plane

  16. Another example: Tully's void Thin HIPASS slice (± 20 Mpc/h), Density field in the galactic plane

  17. BORG/DIVA structure classification Dynamically based structure classifier (DIVA, Lavaux & Wandelt 2010) Only possible when displacement field is available BORG ! MC Sample Displacement +Shear field DIVA classification (Watershed + Eigenvalues) Initial condition voxel marking Lagrangian transport Eulerian classification MC Map Lavaux & Jasche (in prep.), Leclercq et al. (arxiv : 1410.0355)

  18. Another example: Tully's void Thin HIPASS slice (± 20 Mpc/h), Density field in the galactic plane

  19. Structure finding with BORG/DIVA Void probability density in the galactic plane (DIVA criterion), HIPASS slice (± 10 Mpc/h)

  20. Full resimulation

  21. Initial conditions Sloan Wall

  22. Local Universe Simulation OTHER MOVIE 600 Mpc/h, 60 Mpc/h projection, 512 3 dark matter PLUS collaboration

  23. Local Universe Simulation Large scales: BORG (Jasche & Wandelt) 2M++ galaxy catalogue (Lavaux & Hudson 2012) 600 Mpc/h, 200 Mpc/h projection, 1024 3 dark matter PLUS collaboration

  24. Local Universe Simulation Large scales: BORG (Jasche & Wandelt) 2M++ galaxy catalogue (Lavaux & Hudson 2012) 600 Mpc/h, 60 Mpc/h projection, 2048 3 dark matter, z=1 PLUS collaboration

  25. Conclusion Galaxy surveys + Data model (selection, bias, completeness, ...) BORG Final density Initial density Structure prediction and/or Full resimulation classification Preliminary results 512 3 , 1024 3 (2048 3 to come) Available ensemble final density fields Stable large scale structuration Showcases for Tully's void and Sloan Wall PLUS project - redshift space distortion - Local Group dynamics - hydrodynamics

Recommend


More recommend