The stochastic nature of star formation and photon propagation in high-redshift galaxies can result in sizable galaxy-to-galaxy scatter in their properties. Ignoring this scatter by assuming mean quantities can bias estimates of their emissivity and corresponding observables.
We constructed a flexible, semi-empirical model, sampling scatter around the following mean relations: (i) the conditional halo mass function (CHMF); (ii) the stellar-to-halo mass relation (SHMR); (iii) the galaxy star formation main sequence (SFMS); (iv) the fundamental metallicity relation (FMR); (v) the conditional intrinsic luminosity; and (vi) the photon escape fraction.
In our fiducial model, ignoring scatter in these galaxy properties overestimates the duration of the Epoch of Reionization (EoR), delaying its completion by $ z 1--2. We quantified the relative importance of each of the above sources of scatter in determining the ionizing, soft-band X-ray, and Lyman Werner (LW) emissivities as a function of scale and redshift. We find that scatter around the SFMS is important for all bands, especially at the highest redshifts where the emissivity is dominated by the faintest, most "bursty" galaxies. Ignoring this scatter would underestimate the mean emissivity and its standard deviation computed over 5 cMpc regions by factors of up to sim 2--10 at $5 z 15$. The scatter around the X-ray luminosity to star formation rate and metallicity relation is important for determining X-ray emissivity, accounting for roughly half of its mean and standard deviation. The importance of scatter in the ionizing escape fraction depends on its functional form, while scatter around the SHMR contributes at the level of sim 10--20<!PCT!>. Other sources of scatter have a negligible contribution to the emissivities. Although scatter does flatten the UV luminosity functions, shifting the bright end by 1--2 magnitudes, the level of scatter in our fiducial model is insufficient to fully explain recent estimates from JWST photometry (consistent with previous studies).
We conclude that models of the EoR should account for the burstiness of star formation, while models for the cosmic 21cm signal should additionally account for scatter in intrinsic X-ray production.