When the first stars emitted light around 100 million years after the Big Bang, their photons began to reionize the Universe bringing it out of the Dark Ages. How long this reionization period lasted, and importantly the main sources providing the photons for reionization are widely debated. Candidates for these sources include dwarf galaxies, massive galaxies, active galactic nuclei (AGN), accretion shocks, stellar mass black holes, globular clusters, and dark matter annihilation and decay.
Ideally one would want to observe these objects directly at the epoch of reionization and simply measure the Lyman Continuum (LyC) radiation that escapes and ionizes the intergalactic medium. However this is near impossible because the photons are absorbed by intervening neutral Hydrogen. Hence it becomes necessary to turn to numerical simulations to gain insight into how many ionizing photons can emerge from these sources and to see whether reionization epoch galaxies that emit LyC, called LyC leakers, share observational signatures with low redshift analogues that can more easily be observed.
As photons escape by moving through channels cleared out by supernovae and/or AGN, their success at escaping a galaxy in simulations depends on resolving the interstellar medium. In Katz et al. 2020, MNRAS, 498, 164 we report on our work using DiRAC to run a suite of high-resolution Adaptive Mesh Refinement cosmological radiation hydrodynamics resimulations (the Aspen suite) of a massive galaxy (MDM = 1011.8 Msun) at redshift 6. The simulations include a seven species non-equilibrium chemistry (including H2) coupled to radiative transfer. We reach resolutions of 1000 Msun for the stars, 4 X 104 Msun/h for the dark matter and 9.1pc/h for the gas. We post-processed the simulations with a photoionization code to predict emission luminosities for IR lines and nebular continuum for ~1000 galaxies up to redshift 9.1.
One way that observations try to identify LyC leakers is through observing an excess of [OIII] to [OII]. But our simulations found that such an excess is not a sufficient condition for high escape fractions as only ~11% of our simulated galaxies with an excess of [OIII] to [OII] had LyC escape fractions greater than 10%. We found that viewing angle (figure 1) as well as ISM physics (figure 2) could cause galaxies with similar [OIII] to [OII] ratios to have very different photon escape fractions.
Therefore, in search of a better diagnostic to identify LyC leaker analogues at low redshift, we used machine learning and found that [CII]158mm and [OIII]5007A are the most predictive indicators. As photon escape fractions appear to be regulated by feedback, it is not surprising that lines that probe neutral ([CII]158mm) and ionized ([OIII]5007A) gas can pick out LyC leakers. But since those two lines are unlikely to be observed by the same instrument as one is in the optical whereas the other is in the infrared, we instead tested the power of replacing [OIII]5007A with the infrared line [OIII]88mm. We applied our new diagnostics on a sample of local dwarf galaxies to identify LyC leakers and then on epoch of reionization galaxies. The most promising candidate to be actively contributing to the reionization of the Universe was found to be MACS1149_JD1 at z = 9.1.
These simulations were carried out on DiAL in Leicester as part of DiRAC project dp016.