A model for cosmological simulations of galaxy formation physics: multi-epoch validation

Abstract We present a multi-epoch analysis of the galaxy populations formed within the cosmological hydrodynamical simulations presented in Vogelsberger et al. These simulations explore the performance of a recently implemented feedback model which includes primordial and metal line radiative cooling with self-shielding corrections; stellar evolution with associated mass-loss and chemical enrichment; feedback by stellar winds; black hole seeding, growth and merging; and active galactic nuclei (AGN) quasar- and radio-mode heating with a phenomenological prescription for AGN electro-magnetic feedback. We illustrate the impact of the model parameter choices on the resulting simulated galaxy population properties at high and intermediate redshifts. We demonstrate that our scheme is capable of producing galaxy populations that broadly reproduce the shape of the observed galaxy stellar mass function extending from redshift z = 0 to z = 3. We also characterize the evolving galactic B-band luminosity function, stellar mass to halo mass ratio, star formation main sequence, Tully–Fisher relation and gas-phase mass–metallicity relation and confront them against recent observational estimates. This detailed comparison allows us to validate elements of our feedback model, while also identifying areas of tension (e.g., the shape and normalization of the mass–metallicity relation and normalization of the star formation main sequence) that will be addressed in future work. methods: numerical, galaxies: evolution, galaxies: formation, cosmology: theory INTRODUCTION Cosmological simulations are among the most powerful tools available for studying the non-linear regime of cosmic structure formation. While dark matter only simulations have built a solid foundation for our understanding of the origin of haloes via gravitational collapse (e.g. Springel et al. 2005; Fosalba et al. 2008; Boylan-Kolchin et al. 2009; Teyssier et al. 2009; Klypin, Trujillo-Gomez & Primack 2011), applying their findings to our understanding of the observable Universe (i.e. luminous galaxies) requires modelling of baryonic physics as well. Although semi-analytic (e.g. White & Frenk 1991; Kauffmann et al. 1999; Hatton et al. 2003; Kang et al. 2005; Somerville et al. 2008; Guo et al. 2011, 2012) and halo occupation distribution (e.g. Vale & Ostriker 2004; Conroy, Wechsler & Kravtsov 2006; Behroozi, Wechsler & Conroy 2012; Moster, Naab & White 2012) models can estimate galaxy properties based on dark matter only simulations, the most direct and self-consistent way to explore the evolution of observable galaxies theoretically is by including baryons in the simulations (e.g. Katz, Hernquist & Weinberg 1992; Katz, Weinberg & Hernquist 1996; Weinberg, Hernquist & Katz 1997; Murali et al. 2002; Springel & Hernquist 2003b; Kereš et al. 2005; Ocvirk, Pichon & Teyssier 2008; Crain et al. 2009; Croft et al. 2009; Oppenheimer et al. 2010; Schaye et al. 2010; Vogelsberger et al. 2012). The main challenge for any large-scale galaxy formation model is accurately handling the baryonic physics and including proper forms of feedback to regulate star formation. Galaxy formation simulations lacking strong feedback substantially overproduce stars, leading to galaxies with too high baryon fractions (e.g. White & Frenk 1991; Balogh et al. 2001; Scannapieco et al. 2012). This problem is most pronounced for the highest and lowest mass systems, where star formation is known to be relatively inefficient (e.g. Behroozi et al. 2012, and references therein). The problem can be remedied by introducing sources of feedback which either eject gas from galaxies or heat it to prevent continued accretion from the halo. Two commonly employed strong feedback mechanisms are star formation (Dekel & Silk 1986; Thacker & Couchman 2000; Kawata & Gibson 2003; Springel & Hernquist 2003a; Stinson et al. 2006; Dalla Vecchia & Schaye 2008; Scannapieco et al. 2008; Okamoto et al. 2010; Stinson et al. 2013) and black hole (BH) growth (Di Matteo, Springel & Hernquist 2005; Kawata & Gibson 2005; Springel, Di Matteo & Hernquist 2005a; Thacker, Scannapieco & Couchman 2006; Sijacki et al. 2007; Okamoto, Nemmen & Bower 2008; Booth & Schaye 2009; Kurosawa & Proga 2009; Debuhr, Quataert & Ma 2011; Dubois et al. 2012). Winds driven by star formation are a consequence of energy and/or momentum injection from newly formed stellar populations into the interstellar medium (ISM). Observations indicate that star-forming galaxies show signs of outflowing material (e.g. Heckman et al. 2000; Rupke, Veilleux & Sanders 2002, 2005), and that the velocity of the outflowing wind material may scale with the mass of the galaxy (Martin 2005). Including winds in large-scale cosmological simulations requires sub-grid models because the detailed ISM structure remains unresolved (Hopkins, Quataert & Murray 2012b; Hopkins et al. 2013a,b). Several types of explicit wind models have been developed including hydrodynamically decoupled winds (Springel & Hernquist 2003a), injecting thermal energy while shutting down cooling (Stinson et al. 2006, 2013), imposing a minimum temperature threshold for supernovae energy injection (Dalla Vecchia & Schaye 2012) or adding blast particles to launch a Sedov–Taylor blast wave (Dubois & Teyssier 2008). These wind prescriptions vary in their formulations, but all of them eject material from galaxies based on the local star formation rate (SFR) and they have all been shown capable of regulating the growth of low mass galaxies in large-scale simulations (e.g. Ocvirk et al. 2008; Davé, Oppenheimer & Finlator 2011a; Kannan et al. 2013). Feedback from active galactic nuclei (AGN) is the result of energy and/or momentum injection that occurs as gas accretes on to the galaxy's central supermassive BH. This is thought to be responsible for the rapidly moving outflows that can be inferred from UV and X-ray observations of galaxies that host AGN (Chartas et al. 2002; Pounds et al. 2003; Reeves, O'Brien & Ward 2003). It has been shown that AGN feedback is critical for shutting down star formation (e.g. Springel et al. 2005a; Springel, Di Matteo & Hernquist 2005b; Croton et al. 2006; Hopkins et al. 2006; Sijacki et al. 2007; Hopkins et al. 2008b; Booth & Schaye 2009; McCarthy et al. 2010) and setting up self-regulated BH growth (Di Matteo et al. 2005; Hopkins et al. 2006, 2007a,b, 2008a; Di Matteo et al. 2008; Younger et al. 2008; Sijacki, Springel & Haehnelt 2009). Over the last few years, there have been many studies of galaxy formation using large-scale hydrodynamical simulations. For example, Schaye et al. (2010) presented a suite of smoothed particle hydrodynamics (SPH) simulations (the ‘OWLS’ project) to explore the impact of various physical effects, like stellar and AGN feedback, on the resulting galaxy population. The simulations were used to examine, among other things, the evolution of the cosmic SFR (Schaye et al. 2010), observational signatures of (...truncated)