Cosmological Simulation with Population III Stellar Feedback and Metal Enrichment I: Model Description And Convergence Test

TL;DR

This paper introduces a comprehensive simulation framework for early universe star formation, incorporating Pop III and Pop II feedback, and demonstrates its convergence and computational feasibility across various initial conditions.

Contribution

The paper presents a novel, detailed subgrid model for Pop III and Pop II stellar feedback within the { m arepo} code, enabling accurate and efficient cosmological simulations of early star formation.

Findings

Reproduces Pop II star formation rate density consistent with previous models.

Metal-enriched gas volume filling factor converges to ~1% at z=10.

Total stellar mass at z=10 is insensitive to initial conditions and resolution.

Abstract

We present a new Pop III + Pop II subgrid framework implemented in the moving-mesh code {\sc arepo}, designed to study the impact of Pop III feedback on star formation in the early universe. The framework combines primordial non-equilibrium chemistry, metal-line cooling, IMF-sampled stellar evolution with SN feedback, and approximate Lyman-Werner (LW) and ionizing radiation transport. We run a suite of $1c{\rm Mpc}/h$ box simulations with different initial conditions and resolutions from $z=127$ to $z=10$. The highest gas mass and spatial resolution in the fiducial simulation reach $\sim10\,{\rm M_{\odot}}$ and $\sim4\,{\rm pc}$, respectively. The model successfully reproduces the Pop II star formation rate density (SFRD) consistent with previous theoretical works across all initial conditions, with only minor variation driven by local halo interactions and LW irradiation. We find that the volume filling factor of metal-enriched gas converges to $\sim1\%$ at $z=10$. Convergence is achieved once subhalos with $M_{\rm subhalo}\gtrsim 10^{6.5}\,{\rm M_{\odot}}$ are resolved, and the total stellar mass at $z=10$ is largely insensitive to initial conditions or the resolution considered in this work. A fiducial simulation requires $\sim 10^4$ CPU hours, making the framework computationally tractable for larger box simulations and enabling future large parameter studies of stellar physics or environment effects such as Pop III IMF variations, X-ray radiation, or the streaming velocity at high redshift.