We use the FIRE-2 cosmological simulations to study the formation of a virial temperature, quasi-static gas phase in the circumgalactic medium (CGM) at redshifts 0 < z < 5, and how the formation of this virialized phase affects the evolution of galactic discs. We demonstrate that when the halo mass crosses ∼ 10 12 M , the cooling time of shocked gas in the inner CGM (∼ 0.1R vir , where R vir is the virial radius) exceeds the local free-fall time. The inner CGM then experiences a transition from on average sub-virial temperatures (T T vir ), large pressure fluctuations and supersonic inflow/outflow velocities, to virial temperatures (T ∼ T vir ), uniform pressures and subsonic velocities. This transition occurs when the outer CGM (∼ 0.5R vir ) is already subsonic and has a temperature ∼ T vir , indicating that the longer cooling times at large radii allow the outer CGM to virialize at lower halo masses than the inner CGM. This outsidein CGM virialization scenario is in contrast with inside-out scenarios commonly envisioned based on more idealized simulations. We demonstrate that virialization of the inner CGM coincides with abrupt changes in the properties of the central galaxy and its stellar feedback: the galaxy settles into a stable rotating disc, star formation transitions from 'bursty' to 'steady,' and stellar-driven galaxy-scale outflows are suppressed. Our results hence suggest that CGM virialization is initially associated with the formation of rotation-dominated thin galactic discs, rather than with the quenching of star formation as often assumed.

Understanding the rate at which stars form is central to studies of galaxy formation. Observationally, the star formation rates (SFRs) of galaxies are measured using the luminosity in different frequency bands, often under the assumption of a time-steady SFR in the recent past. We use star formation histories (SFHs) extracted from cosmological simulations of star-forming galaxies from the FIRE project to analyse the time-scales to which the H α and far-ultraviolet (FUV) continuum SFR indicators are sensitive. In these simulations, the SFRs are highly time variable for all galaxies at high redshift, and continue to be bursty to z = 0 in dwarf galaxies. When FIRE SFHs are partitioned into their bursty and time-steady phases, the best-fitting FUV time-scale fluctuates from its ∼10 Myr value when the SFR is time-steady to ≳100 Myr immediately following particularly extreme bursts of star formation during the bursty phase. On the other hand, the best-fitting averaging time-scale for H α is generally insensitive to the SFR variability in the FIRE simulations and remains ∼5 Myr at all times. These time-scales are shorter than the 100 and 10 Myr time-scales sometimes assumed in the literature for FUV and H α, respectively, because while the FUV emission persists for stellar populations older than 100 Myr, the time-dependent luminosities are strongly dominated by younger stars. Our results confirm that the ratio of SFRs inferred using H α versus FUV can be used to probe the burstiness of star formation in galaxies.

We investigate the effects of AGN heating and the ultraviolet background on the low-redshift Lymanα forest column density distribution (CDD) using the Illustris simulation. We show that Illustris reproduces observations at z = 0.1 in the column density range 10 12.5 − 10 13.5 cm −2 , relevant for the "photon underproduction crisis." We attribute this to the inclusion of AGN feedback, which changes the gas distribution so as to mimic the effect of extra photons, as well as the use of the Faucher-Giguère ultraviolet background, which is more ionizing at z = 0.1 than the Haardt & Madau background previously considered. We show that the difference between simulations run with smoothed particle hydrodynamics and simulations using a moving mesh is small in this column density range but can be more significant at larger column densities. We further consider the effect of supernova feedback, Voigt profile fitting and finite resolution, all of which we show to have little influence on the CDD. Finally, we identify a discrepancy between our simulations and observations at column densities 10 14 − 10 16 cm −2 , where Illustris produces too few absorbers, which suggests the AGN feedback model should be further refined. Since the "photon underproduction crisis" primarily affects lower column density systems, we conclude that AGN feedback and standard ionizing background models can resolve the crisis.

Pressure balance plays a central role in models of the interstellar medium (ISM), but whether and how pressure balance is realized in a realistic multiphase ISM is not yet well understood. We address this question using a set of FIRE-2 cosmological zoom-in simulations of Milky Way-mass disk galaxies, in which a multiphase ISM is self-consistently shaped by gravity, cooling, and stellar feedback. We analyze how gravity determines the vertical pressure profile as well as how the total ISM pressure is partitioned between different phases and components (thermal, dispersion/turbulence, and bulk flows). We show that, on average and consistent with previous more idealized simulations, the total ISM pressure balances the weight of the overlying gas. Deviations from vertical pressure balance increase with increasing galactocentric radius and with decreasing averaging scale. The different phases are in rough total pressure equilibrium with one another, but with large deviations from thermal pressure equilibrium owing to kinetic support in the cold and warm phases, which dominate the total pressure near the midplane. Bulk flows (e.g., inflows and fountains) are important at a few disk scale heights, while thermal pressure from hot gas dominates at larger heights. Overall, the total midplane pressure is well-predicted by the weight of the disk gas, and we show that it also scales linearly with the star formation rate surface density (ΣSFR). These results support the notion that the Kennicutt-Schmidt relation arises because ΣSFR and the gas surface density (Σg) are connected via the ISM midplane pressure.

We explore the origin of stellar metallicity gradients in simulated and observed dwarf galaxies. We use FIRE-2 cosmological baryonic zoom-in simulations of 26 isolated galaxies as well as existing observational data for 10 Local Group dwarf galaxies. Our simulated galaxies have stellar masses between 105.5 and 108.6 M⊙. Whilst gas-phase metallicty gradients are generally weak in our simulated galaxies, we find that stellar metallicity gradients are common, with central regions tending to be more metal-rich than the outer parts. The strength of the gradient is correlated with galaxy-wide median stellar age, such that galaxies with younger stellar populations have flatter gradients. Stellar metallicty gradients are set by two competing processes: (1) the steady ‘puffing’ of old, metal-poor stars by feedback-driven potential fluctuations and (2) the accretion of extended, metal-rich gas at late times, which fuels late-time metal-rich star formation. If recent star formation dominates, then extended, metal-rich star formation washes out pre-existing gradients from the ‘puffing’ process. We use published results from ten Local Group dwarf galaxies to show that a similar relationship between age and stellar metallicity-gradient strength exists among real dwarfs. This suggests that observed stellar metallicity gradients may be driven largely by the baryon/feedback cycle rather than by external environmental effects.