[Paper Review] A Research Note on the Implementation of Star Formation and Stellar Feedback in Semi-Analytic Models

Fabio Fontanot, G. De Lucia+3 more|arXiv (Cornell University)|Jan 17, 2013

Inertial Sensor and Navigation3 citations

TL;DR

This study investigates how different star formation and supernova feedback prescriptions affect galaxy properties in semi-analytic models (SAMs), using identical dark matter halo merger trees from N-body simulations. Despite consistent overall trends—such as increased hot gas fractions and reduced cold gas—key differences emerge in stellar and gas content, especially at Milky Way-like masses, with the Durham-like model showing significantly lower star formation rates and stellar masses compared to Munich and morgana models.

ABSTRACT

We study the impact of star formation and stellar feedback prescriptions on galaxy properties predicted by means of "stripped-down" versions of independently developed semi-analytic models (SAMs). These include cooling, star formation, feedback from supernovae (SNe) and simplified prescriptions for galaxy merging, but no chemical evolution, disc instabilities or AGN feedback. We run these versions on identical samples of dark matter (DM) haloes extracted from high-resolution N-body simulations in order to perform both statistical analysis and object-by-object comparisons. We compare our results with previous work based on stripped-down versions of the same SAMs including only gas cooling, and show that all feedback models provide coherent modifications in the distribution of baryons between the various gas phases. In particular, we find that the predicted hot gas fractions are considerably increased by up to a factor of three, while the corresponding cold gas fractions are correspondingly decreased, and a significant amount of mass is ejected from the DM halo. Nonetheless, we also find relevant differences in the predicted properties of model galaxies among the three SAMs: these deviations are more relevant at mass scales comparable to that of our own Galaxy, and are reduced at larger masses, confirming the varying impact of stellar feedback at different mass scales. We also check the effect of enhanced star formation events (i.e. starbursts modes), defined in connection with galaxy mergers. We find that, in general, these episodes have a limited impact in the overall star formation histories of model galaxies, even in massive DM halos where merger-driven star formation has often been considered very important.

Motivation & Objective

• To isolate the impact of star formation and stellar feedback prescriptions on galaxy properties in semi-analytic models (SAMs), independent of other physical processes.
• To assess how different modeling choices for feedback and star formation affect the distribution of baryons across gas phases (cold, hot, ejected) and stellar content.
• To evaluate the role of merger-driven starburst modes in shaping galaxy star formation histories and mass assembly.
• To compare model predictions across three independently developed SAMs (Munich, morgana, Durham-like) under identical dark matter halo evolution.
• To quantify the sensitivity of galaxy properties to feedback model architecture at different mass scales, particularly around Milky Way mass.

Proposed method

• Constructed stripped-down versions of three SAMs (Munich, morgana, Durham-like) including only gas cooling, star formation, SN feedback, and simplified galaxy merging.
• Applied these models to identical merger trees extracted from high-resolution N-body simulations (Millennium and Millennium-II), ensuring consistent dark matter halo evolution.
• Used a two-mode star formation scheme: 'quiescent' mode (standard) and 'starburst' mode triggered by galaxy mergers with infinite vs. zero merging times.
• Tracked redshift evolution of key galaxy properties: cold and hot gas fractions, stellar mass, and ejected gas mass.
• Performed statistical analysis and object-by-object comparisons across models to isolate differences due to feedback and star formation prescriptions.
• Compared results with previous 'cooling-only' versions of the same models to isolate the impact of feedback.

Experimental results

Research questions

• RQ1 How do different star formation and supernova feedback prescriptions affect the distribution of baryons among cold, hot, and ejected gas phases in dark matter haloes?
• RQ2 To what extent do model-specific feedback and star formation schemes lead to divergent predictions for galaxy stellar masses and star formation rates?
• RQ3 What is the contribution of merger-driven starburst modes to the total star formation history in galaxies, particularly in Milky Way-like haloes?
• RQ4 How do the predicted galaxy properties vary across models at different halo mass scales, especially near Milky Way mass?
• RQ5 Does the inclusion of starburst modes reduce scatter in predicted stellar masses across haloes of similar mass?

Key findings

• Star formation and SN feedback significantly increase the hot gas fraction in dark matter haloes by up to a factor of three compared to cooling-only models, while reducing cold gas fractions.
• The Durham-like feedback scheme produces the highest hot gas fractions and the most rapid depletion of cold gas, leading to systematically lower predicted star formation rates and stellar masses.
• The morgana and Munich models show good agreement in average star formation rates at z ≥ 2 for Milky Way-like haloes, but diverge at lower redshifts, with only minor differences in final z=0 stellar masses.
• The starburst mode associated with galaxy mergers contributes at most a factor of two increase in central galaxy mass assembly rate, indicating a limited overall impact on total SFR budgets.
• Including merger-driven starbursts reduces the scatter in predicted stellar masses at fixed redshift across haloes of the same mass, suggesting a stabilizing effect on mass assembly.
• Despite coherent trends in baryon redistribution, each model exhibits a unique redshift evolution pattern for cold, hot, and ejected gas, highlighting the sensitivity of predictions to feedback model architecture.