LIMFAST
5.4 Comparison to Previous Work
Below, we highlight the origin of the main differences between the LIMFAST results and those from the literature. Our goal here is not to detail the specific individual calculations but to identify the parameters that generally play a major role on the results in intensity mapping modeling and that are relevant to LIMFAST.
5.4.1 Comparing Star Formation Results
A significant difference between LIMFAST and other works observed in the previous section is the higher star formation rate density of LIMFAST at redshiftsπ§ β³ 7. This difference mostly arises from assuming the atomic cooling threshold (104K neutral gas) for hosting star formation in halos, which results in evolving minimum halo
mass values from log(π/πβ) βΌ7 atπ§βΌ 20 to log(π/πβ) βΌ8 atπ§βΌ5β6. At the lowest redshifts here explored, the minimum halo mass considered has little effect on the star formation rate because the latter is dominated by massive halos of mass log(π/πβ) β³ 10β12. However, at the highest redshifts most halos have small masses and, therefore, the resulting star formation is very sensitive to the minimum halo mass adopted. Some of the other works consider minimum halo masses above the atomic cooling threshold because in these small systems feedback effects may suppress star formation (e.g., Yue et al. 2014; Yue et al. 2018).
We show how changing the minimum halo mass parameter in LIMFAST affects the star formation rate density in Figure 5.8. The solid black line is the fiducial star formation rate density shown in Figure 5.2, which considers all halos above the atomic cooling threshold. The black dotted and dashed curves represent the star formation assuming minimum halos masses of log(π/πβ) =8 and the contribution from halos in the range of masses 8 β€ log(π/πβ) < 10, respectively. The brown lines show the results from the THESAN 1 simulations by Kannan et al.
2022a (their figure 12), where the brown solid line is the total and the dashed one illustrates the contribution from halos of mass log(π/πβ) = 10. Furthermore, the dotted brown line denotes the contribution from halos in the mass range 8 β€ log(π/πβ) < 9, showing that these mass range drives the total values at the highest redshifts. The other colored lines represent the same lines as in Figure 5.2 for comparison. Overall, the value adopted for the minimum halo mass hosting star formation at high redshift has a strong impact in the star formation rate and, in turn, it affects the level of emission when this is proportional to the star formation of the halos. However, visible differences are still present between LIMFAST and THESAN, especially regarding the contribution from small halos. Although the upper mass limit in the two respective ranges is different, the impact from halos of mass log(π/πβ) β€8 is notable as visible by the difference between the solid and dotted black lines at π§ β³ 8. We attribute the differences between the two codes to the different star formation prescriptions used in each case.
5.4.2 Comparing Line Emission and Power Spectra Results
A fundamental distinction between the LIMFAST and literature results for line emission in Figure 5.6 is that we have presented the intrinsic emission instead of the attenuated one. The effect of dust and gas attenuation can be treated on-the-fly or post-processed, analytically or numerically, and can be linked to the properties of the simulated gas or simply accounted for in the photoionization calculations
Figure 5.8: Contributions to the cosmic star formation rate density. The different line types consider different minimum halo masses and halo mass ranges able to host star formation. The other colored lines represent the results from other works as in Figure 5.2.
as a property of the nebular gas. Because of these different approaches, we leave the implementation of attenuation effects to future work, but it is still important to compare the evolution of the line emission. The left panels in Figure 5.6 for LyπΌ from star formation show good agreement in the shape of the evolution between the LIMFAST and the Comaschi & Ferrara (2016) results, arising from the fact that both calculations assume the atomic cooling threshold as the limiting halo mass to host star formation. Generally, other works show a steeper decrease towards high redshift because of the impact of a lower number of small halos as discussed above.
Similarly, the slope of the LIMFAST evolution for other lines matches well that from the THESAN 1 simulations. There is no such a match for the THESAN 2 case because these simulations do not resolve the small halos that drive the signal at high redshift, resulting in a steeper evolution (Kannan et al. 2022a). Finally, once again we emphasize the differences in amplitude arising from the usage or not of
local relations, and from different ionization parameter values on the oxygen lines, a resource that can be exploited to retrieve information about the physical properties of the interstellar medium (Silva et al. 2017).
In some cases, very large differences exist between some works, especially for the Pullen et al. (2014) and Heneka et al. (2017) results for the intergalactic and background LyπΌcompared to LIMFAST and Comaschi & Ferrara (2016). We have tested that using a minimum halo mass for star formation of log(π/πβ) = 10 produces LyπΌbackground signals closer to those of Pullen et al. (2014) and Heneka et al. 2017 (not shown), and we ascribe further differences between Heneka et al.
(2017) and LIMFAST to the respective modeling and simulations. Finally, we note that Comaschi & Ferrara (2016) pointed to an inaccurate treatment of the ionization history of the IGM in Pullen et al. (2014), which may contribute to the observed signals.
As mentioned before, the differences observed for line emission roughly increase by their squared value when comparing the power spectra. The exact value also depends on the specific clustering bias parameter used in different works, but we find no extraordinary behaviors when comparing these results. Perhaps most significant are the differences between theory and simulations in the Silva et al. (2013) work, highlighting the model sensitivity, and the strong decrease of between three and five orders of magnitude for the star formation LyπΌ signal (both at π§ = 7 andπ§ = 10;
see Figure 5.7 and also Figure 5.10) between Heneka et al. (2017) and Heneka &
Cooray (2021) with the same simulations and modeling.