Explore the first galaxies through physics-based modeling, simulation, and data exploration
All of the Renaissance Simulations are performed in the same comoving volume of (40 Mpc)3. The initial conditions for this volume are generated using MUSIC Hahn & Abel 2011 with second-order Lagrangian perturbations at z=99 using a 5123 root grid resolution. We use the cosmological parameters from the seven-year WMAP ΛCDM+SZ+LENS best fit Komatsu et al. 2011: ΩM=0.266, Ω=0.734, Ωb=0.0449, h=0.71, σ8=0.81, and n=0.963.
It is computationally prohibitive to have the necessary parsec-scale spatial resolution (and accompanying mass resolution), which is required to marginally resolve star-forming molecular clouds, throughout the entire simulation volume. We perform zoom-in simulations on three selected regions, ranging from 220 to 430 comoving Mpc3 with different overdensities, providing a mixture of large-scale environments. We first run a 5123 N-body only simulation to z = 6. We then select an overdense region (“Rarepeak”), a nearly mean density region (“Normal”) and an underdense region (“Void”), which are displayed in Figure 1. The selection of the survey volume and detailed setup of the Rarepeak have been described in Xu et al. (2013), which is centered on two 3x1010Msun halos at z=6 with a survey volume of (3.8 × 5.4 × 6.6) Mpc3. For both the Normal and Void runs, we select comoving volumes of (6.0 × 6.0 × 6.125) Mpc3 as the survey volumes. We then re-initialize all simulations, having the survey volume at the center, with three more static nested grids to have an effective resolution of 40963 and an effective dark matter mass resolution of 2.9×104 Msun inside the highest static nested grid that encompasses the survey volume.
During the course of the simulation, we allow a maximum refinement level of l=12, resulting in a maximal resolution of 19 comoving parsecs. The refinement criteria employed are the same as in Wise et al. (2012b), refining on baryon and dark matter overdensities of four and local Jeans length by at least four cells Truelove et al. 1997 and is restricted to the survey volumes. While the Rarepeak simulation adjusts the survey volume size during the simulation to contain only the highest resolution dark matter particles of the highest static nested grid, matter in the Normal and Void simulations is not fully contained in a large-scale potential well and have significant peculiar velocities, causing some of the high-resolution particles to migrate out of the initial static grid. Thus, we simplify the simulation setup by restricting grid refinement to occur in the initial high-resolution grid instead of its Lagrangian region.
A total of 7 Renaissance Simulations listed on the Investigate page were performed with different treatments for the inclusion of Lyman-Werner UV radiation (10.2-13.6 eV) which photodissociates molecular hydrogen, the primary coolant of primordial gas. In various combinations, they employ 2 different models for local sources of LW radiation, and 2 different models for the Lyman-Werner radiation background:
| Model | Description |
|---|---|
| LS1 | Lyman-Werner radiation from Pop III stars inside the simulated volume, geometrically attenuated. |
| LS2 | Lyman-Werner radiation from Pop III stars and metal-enriched star clusters inside the simulated volume, geometrically attenuated. |
| BG | Lyman-Werner background from early generations of Pop III stars based on the analytic model of Wise & Abel (2005), as updated in Wise et al. (2012). |
| BG1 | Lyman-Werner background is self-consistently computed from Pop III and metal-enriched stellar sources formed in the Normal simulation, as described in Xu et al. (2016). |
We stop the simulations of the Rarepeak, Normal, and Void regions at approximately z=(15, 12, 8), respectively, because of the high computational cost of the radiative transfer once a substantial fraction of the volume becomes ionized. The Renaissance simulations were run on the Blue Waters system at NCSA. Each simulation used approximately eight million core-hours.