Ionizing photon production and escape fractions during cosmic reionization in the TNG50 simulation

In this work, we investigate the dependence of the escape fraction of ionizing photons, f(esc), on various galaxy and host halo properties during the epoch of reionization. We post-process the TNG50 magnetohydrodynamical simulation from the IllustrisTNG project using the three-dimensional multifrequency radiative transfer code CRASH. Our work covers the stellar mass range of 10(6) less than or similar to M-star/M-circle dot less than or similar to 10(8) at redshifts 6 < z < 10. Adopting an unresolved, cloud-scale escape fraction parameter of unity, the average halo escape fraction f(esc) increases with mass from similar to 0.3 at M-star = 10(6) M-circle dot to similar to 0.6 at M-star = 10(7.5) M-circle dot, after which we find hints of a turnover and decreasing escape fractions for even more massive galaxies. However, we demonstrate a strong and non-linear dependence of f(esc) on the adopted subgrid escape fraction, resulting in uncertainties for the absolute value of the escape fraction. In addition, f(esc) has significant scatter at fixed mass, driven by diversity in the ionizing photon rate together with a complex relationship between (stellar) source positions and the underling density distribution. The global emissivity is consistent with observations for reasonable cloud-scale absorption values, and haloes with a stellar mass less than or similar to 10(7.5) M-circle dot contribute the majority of escaping ionizing photons at all redshifts. Incorporating dust reduces f(esc) by a few percent at M-star less than or similar to 10(6.5) M-circle dot, and up to 10 percent for larger haloes. Our multifrequency approach shows that f(esc) depends on photon energy, and is reduced substantially at E > 54.4 eV versus lower energies. This suggests that the impact of high-energy photons from binary stars is reduced when accounting for an energy-dependent escape fraction.