FLAMINGO is a project of the Virgo consortium for cosmological supercomputer simulations. The acronym stands for Full-hydro Large-scale structure simulations with All-sky Mapping for the Interpretation of Next Generation Observations.

Observational cosmology based on measurements of the growth of large-scale structure is increasingly limited by the accuracy of theoretical predictions. The models that are compared with the data are nearly always based on dark matter only (DMO) simulations, though some allow for marginalization over expected baryonic effects associated with galaxy formation. However, baryonic effects become increasingly important as observations move to smaller scales, and may be considerably more complex than is assumed in the corrections applied to DMO simulations. Hydrodynamical simulations can in principle help resolve this issue, but they tend to use volumes that are too small to study large-scale structure, they often do not reproduce the relevant observables, and/or they do not include model variations that cover the relevant parameter space. The FLAMINGO project aims to address these shortcomings.

Key features:

- Three resolutions: high/m8 (baryonic particle mass m
_{g}= 1.3 x 10^{8}M_{☉}), intermediate/m9 (m_{g}= 1.1 x 10^{9}M_{☉}), and low/m10 (m_{g}= 8.6 x 10^{9}M_{☉}) - Flagship simulations: 2.8 Gpc box size at m9 resolution (L2p8_m9) and 1 Gpc at m8 resolution (L1_m8)
- Up to 3 x 10
^{11}particles (2 x 5040^{3}+ 2800^{3}) - Subgrid feedback is calibrated to the z=0 galaxy stellar mass function and low-z cluster gas fractions using Gaussian process emulation
- Massive neutrinos are modeled using particles with the 'δf' method that was designed to reduce shot noise
- Twelve L1_m9 variations: eight simulations varying the calibration data and four varying the cosmology
- Full-sky lightcone particle data and HEALPix maps for up to 8 different observer locations
- Run with the Swift code and SPHENIX smoothed particle hydrodynamics implementation
- 3-fluid initial conditions with separate transfer functions for CDM, baryons and neutrinos, perturbing particle masses rather than positions to suppress discreteness noise

The FLAMINGO simulations and their properties are listed below.

The first table below contain the hydrodynamical simulations in the set (Table 2 from Schaye et al. 2023). The first four lines list the simulations that use the fiducial galaxy formation model and assume the fiducial cosmology, but use different volumes and resolutions. The remaining lines list the model variations, which all use a 1 Gpc box and intermediate resolution. The columns list the simulation identifier (where m8, m9 and m10 indicate log_{10} of the mean baryonic particle mass and correspond to high, intermediate, and low resolution, respectively; absence of this part implies m9 resolution); the number of standard deviations by which the observed stellar masses are shifted before calibration, Δm_{*}; the number of standard deviations by which the observed cluster gas fractions are shifted before calibration, Δf_{gas}; the AGN feedback implementation (thermal or jets); the comoving box side length, L; the number of baryonic particles, N_{b} (which equals the number of CDM particles, N_{CDM}; the number of neutrino particles, N_{ν}; the initial mean baryonic particle mass, m_{g}; the mean CDM particle mass, m_{CDM}; the Plummer-equivalent comoving gravitational softening length, ε_{com}; the maximum proper gravitational softening length, ε_{prop}; and the assumed cosmology which is specified table at the bottom of this page.

Identifier | Δm_{*}(σ) | Δf_{gas}(σ) | AGN | L (cGpc) | N_{b} | N_{ν} | m_{g}(M _{☉}) | m_{CDM}(M _{☉}) | ε_{com}(ckpc) | ε_{prop}(pkpc) | Cosmology |
---|---|---|---|---|---|---|---|---|---|---|---|

L1_m8 | 0 | 0 | thermal | 1 | 3600^{3} | 2000^{3} | 1.34 x 10^{8} | 7.06 x 10^{8} | 11.2 | 2.85 | D3A |

L1_m9 | 0 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

L1_m10 | 0 | 0 | thermal | 1 | 900^{3} | 500^{3} | 8.56 x 10^{9} | 4.52 x 10^{10} | 44.6 | 11.40 | D3A |

L2p8_m9 | 0 | 0 | thermal | 2.8 | 5040^{3} | 2800^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

fgas+2σ | 0 | +2 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

fgas-2σ | 0 | -2 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

fgas-4σ | 0 | -4 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

fgas-8σ | 0 | -8 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

M*-1σ | -1 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

M*-1σ_fgas-4σ | -1 | -4 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

Jet | 0 | 0 | jets | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

Jet_fgas-4σ | 0 | -4 | jets | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | D3A |

Planck | 0 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.72 x 10^{9} | 22.3 | 5.70 | Planck |

PlanckNu0p24Var | 0 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.06 x 10^{9} | 5.67 x 10^{9} | 22.3 | 5.70 | PlanckNu0p24Var |

PlanckNu0p24Fix | 0 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.62 x 10^{9} | 22.3 | 5.70 | PlanckNu0p24Fix |

LS8 | 0 | 0 | thermal | 1 | 1800^{3} | 1000^{3} | 1.07 x 10^{9} | 5.65 x 10^{9} | 22.3 | 5.70 | LS8 |

The second table contains the gravity-only FLAMINGO simulations (Table 3 from Schaye et al. 2023). Simulation L1_m9_ip_DMO is identical to L1_m9_DMO except that the phases in the initial conditions were inverted. Note that there are no hydrodynamical counterparts for L5p6_m10_DMO, PlanckNu0p12Var_DMO, L1_m9_ip_DMO and L11p2_m11_DMO.

Identifier | L (cGpc) | N_{CDM} | N_{ν} | m_{CDM}(M _{☉}) | ε_{com}(ckpc) | ε_{prop}(pkpc) | Cosmology |
---|---|---|---|---|---|---|---|

L1_m8_DMO | 1 | 3600^{3} | 2000^{3} | 8.40 x 10^{8} | 11.2 | 2.85 | D3A |

L1_m9_DMO | 1 | 1800^{3} | 1000^{3} | 6.72 x 10^{9} | 22.3 | 5.70 | D3A |

L1_m10_DMO | 1 | 900^{3} | 500^{3} | 5.38 x 10^{10} | 44.6 | 11.40 | D3A |

L1p8_m9_DMO | 2.8 | 5040^{3} | 2800^{3} | 6.72 x 10^{9} | 22.3 | 5.70 | D3A |

L5p6_m10_DMO | 5.6 | 5040^{3} | 2800^{3} | 5.38 x 10^{10} | 44.6 | 11.40 | D3A |

Planck_DMO | 1 | 1800^{3} | 1000^{3} | 6.78 x 10^{9} | 22.3 | 5.70 | Planck |

PlanckNu0p12Var_DMO | 1 | 1800^{3} | 1000^{3} | 6.74 x 10^{9} | 22.3 | 5.70 | PlanckNu0p12Var |

PlanckNu0p24Var_DMO | 1 | 1800^{3} | 1000^{3} | 6.73 x 10^{9} | 22.3 | 5.70 | PlanckNu0p24Var |

PlanckNu0p24Fix_DMO | 1 | 1800^{3} | 1000^{3} | 6.68 x 10^{9} | 22.3 | 5.70 | PlanckNu0p24Fix |

LS8_DMO | 1 | 1800^{3} | 1000^{3} | 6.72 x 10^{9} | 22.3 | 5.70 | LS8 |

L1_m9_ip_DMO | 1 | 1800^{3} | 1000^{3} | 6.72 x 10^{9} | 22.3 | 5.70 | D3A |

L11p2_m11_DMO | 11.2 | 5040^{3} | 2800^{3} | 4.30 x 10^{11} | 89.2 | 22.80 | D3A |

Finally, the table below lists the values of the cosmological parameters used in different simulations (Table 4 from Schaye et al. 2023). The columns list the prefix used to indicate the cosmology in the simulation name (note that for brevity the prefix 'D3A' that indicates the fiducial cosmology is omitted from the simulation identifiers); the dimensionless Hubble constant, h; the total matter density parameter, Ω_{m}; the dark energy density parameter, Ω_{Λ}; the baryonic matter density parameter, Ω_{b}; the sum of the particle masses of the neutrino species, ∑ m_{ν}c^{2}; the amplitude of the primordial matter power spectrum, A_{s}; the power-law index of the primordial matter power spectrum, n_{s}; the amplitude of the initial power spectrum parametrized as the r.m.s. mass density fluctuation in spheres of radius 8 h^{-1} Mpc extrapolated to z=0 using linear theory, σ_{8}; the amplitude of the initial power spectrum parametrized as S_{8} ≡ σ_{8}(Ω_{m}/0.3)^{1/2}; the neutrino matter density parameter, Ω_{ν} ≅ ∑ m_{ν}c^{2}/(93.14 h^{2} eV). Note that the values of the Hubble and density parameters are given at z=0. The values of the parameters that are listed in the last three columns have been computed from the other parameters.

Name | h | Ω_{m} | Ω_{Λ} | Ω_{b} | ∑ m_{ν}c^{2} | A_{s} | n_{s} | σ_{8} | S_{8} | Ω_{ν} |
---|---|---|---|---|---|---|---|---|---|---|

D3A | 0.681 | 0.306 | 0.694 | 0.0486 | 0.06 eV | 2.099 x 10^{-9} | 0.967 | 0.807 | 0.815 | 1.39 x 10^{-3} |

Planck | 0.673 | 0.316 | 0.684 | 0.0494 | 0.06 eV | 2.101 x 10^{-9} | 0.966 | 0.812 | 0.833 | 1.42 x 10^{-3} |

PlanckNu0p12Var | 0.673 | 0.316 | 0.684 | 0.0496 | 0.12 eV | 2.113 x 10^{-9} | 0.967 | 0.800 | 0.821 | 2.85 x 10^{-3} |

PlanckNu0p24Var | 0.662 | 0.328 | 0.672 | 0.0510 | 0.24 eV | 2.109 x 10^{-9} | 0.968 | 0.772 | 0.807 | 5.87 x 10^{-3} |

PlanckNu0p24Fix | 0.673 | 0.316 | 0.684 | 0.0494 | 0.24 eV | 2.101 x 10^{-9} | 0.966 | 0.769 | 0.789 | 5.69 x 10^{-3} |

LS8 | 0.682 | 0.305 | 0.695 | 0.0473 | 0.06 eV | 1.836 x 10^{-9} | 0.965 | 0.760 | 0.766 | 1.39 x 10^{-3} |

The FLAMINGO simulations were run on the Memory Intensive DiRAC facility managed by the Institute for Computational Cosmology, Durham University.