The importance of the way in which supernova energy is distributed around young stellar populations in simulations of galaxies

Abstract

Supernova (SN) feedback plays a crucial role in simulations of galaxy formation. Because blast waves from individual SNe occur on scales that remain unresolved in modern cosmological simulations, SN feedback must be implemented as a subgrid model. Differences in the manner in which SN energy is coupled to the local interstellar medium and in which excessive radiative losses are prevented have resulted in a zoo of models used by different groups. However, the importance of the selection of resolution elements around young stellar particles for SN feedback has largely been overlooked. In this work, we examine various selection methods using the smoothed particle hydrodynamics code swift. We run a suite of isolated disc galaxy simulations of a Milky Way-mass galaxy and small cosmological volumes, all with the thermal stochastic SN feedback model used in the eagle simulations. We complement the original mass-weighted neighbour selection with a novel algorithm guaranteeing that the SN energy distribution is as close to isotropic as possible. Additionally, we consider algorithms where the energy is injected into the closest, least dense, or most dense neighbour. We show that different neighbour-selection strategies cause significant variations in star formation rates, gas densities, wind mass-loading factors, and galaxy morphology. The isotropic method results in more efficient feedback than the conventional mass-weighted selection. We conclude that the manner in which the feedback energy is distributed among the resolution elements surrounding a feedback event is as important as changing the amount of energy by factors of a few.

Spin-driven jet feedback in idealized simulations of galaxy groups and clusters

“…The gas is allowed to cool radiatively based on the cooling function from Ploeckinger & Schaye ( 2020 ), although it is not allowed to cool down to the molecular phase. Instead, we use an entropy floor (see Nobels et al 2022 for details). Star formation is modelled based on the Kennicutt-Schmidt law (Kennicutt 1998 ) using the gas pressure (Schaye & Dalla Vecchia 2008 ).…”

“…The initial conditions for our set-up are discussed in detail in Nobels et al ( 2022 ). Here we present a summary of the main features of the set-up.…”

“…In this paper, we present results from a study of self-consistent, spin-driven jet feedback in idealized galaxy group and cluster setups simulated with SWIFT (see Nobels et al 2022 for details of the set-up). Our highest resolution simulations have a mass resolution of m g = 10 5 M , which is 20 times better than the only other similar SPH simulation of this kind (involving gas cooling, self-consistent jet feedback, and star formation) that has been performed (Barai et al 2016 ).…”

The impact of stochastic modeling on the predictive power of galaxy formation simulations

“…Supernova energy is injected thermally into the closest gas particle to the star particle at injection time, following Chaikin et al (2022), and now include a distribution in delay times between the birth of star particles and their injection of supernova energy to sample the stellar lifetimes, a notable change between the new model and EAGLE (which used a fixed delay time for all supernovae). The coupling efficiency of stellar feedback to the ISM is modulated through a number of free parameters, discussed in Section 2.3.…”

“…This can primarily be attributed to two main (connected) reasons: the much poorer sampling of black holes (at z = 0 there are of order thousands, compared to the There is significant variation between clones, shown in the lower panel, but this is consistent with the high level of variability of the BHARD on short timescales within an individual simulation. In the background, we show observational data from Aird et al (2015), spanning the range z = 0.0− 5.0. millions of star particles that have effectively sampled the SFRD), and the high cadence variability of the BHARD within an individual clone simulation (see also McAlpine et al 2017;Nobels et al 2022, for further work on this in the EAGLE model). Even between these 16 clones, we are not able to extract a smooth and stable median BHARD, suggesting an extreme level of variability.…”