The Importance of Subtleties in the Scaling of the 'Terminal Momentum' For Galaxy Formation Simulations

Abstract

In galaxy formation simulations, it is increasingly common to represent supernovae (SNe) at finite resolution (when the Sedov-Taylor phase is unresolved) via hybrid energy-momentum coupling with some 'terminal momentum' $p_{\rm term}$ (depending weakly on ambient density and metallicity) that represents unresolved work from an energy-conserving phase. Numerical implementations can ensure momentum and energy conservation of such methods, but these require that couplings depend on the surrounding gas velocity field (radial velocity $\langle v_{r} \rangle$). This raises the question of whether $p_{\rm term}$ should also be velocity-dependent, which we explore analytically and in simulations. We show that for simple spherical models, the dependence of $p_{\rm term}$ on $\langle v_{r} \rangle$ introduces negligible corrections beyond those already imposed by energy conservation if $\langle v_{r} \rangle \ge 0$. However, for SNe in some net converging flow ($\langle v_{r} \rangle<0$), naively coupling the total momentum when a blastwave reaches the standard cooling/snowplow phase (or some effective cooling time/velocity/temperature criterion) leads to enormous $p_{\rm term}$ and potentially pathological behaviors. We propose an alternative $\Delta$-Momentum formulation which represents the differential SNe effect and show this leads to opposite behavior of $p_{\rm term}$ in this limit. We also consider a more conservative velocity-independent formulation. Testing in numerical simulations, these directly translate to large effects on predicted star formation histories and stellar masses of massive galaxies, explaining differences between some models and motivating further study in idealized simulations.