Merger tree comparison : impact on the semi-analytic model GALFORM
Many previous studies using semi-analytic models have focused predominantly on improving the phenomenological prescriptions of the baryonic physics governing galaxy formation as the vast uncertainties in these processes are critical in understanding the observable Universe. On the other hand, studies dedicated to "the dark universe" imply that technical issues concerning dark matter (DM) halo merger trees can also impose other uncertainties when modeling the observable universe. We examine the effect of using four different merger tree building algorithms, SUBFIND, HBT, ROCKSTAR and VELOCIraptor, on galaxy properties using the semi-analytic model of galaxy formation and evolution GALFORM with fixed model parameters, run on the EAGLE simulation in its dark matter only version. The construction of each Halo Finder causes some to be able to find some haloes that other finders do not and vice versa affecting the total number of haloes that are found. The differences in definitions or limits to distinguish the satellite and central haloes present in the Halo Finders, produce haloes which are found in two different halo catalogs with different central/satellite classification which impacts the way in which the galaxies that reside in these haloes evolve as, by construction in GALFORM, central galaxies retain their hot gas whereas satellite galaxies do not. The results of GALFORM show that the number of central and type 1 satellite galaxies (those that are hosted by a dark matter satellite subhalo) depend directly on the halo finder and its definition for central and satellite subhalo, which results in the ROCKSTAR run having a slight excess of galaxies. In contrast the number of type 2 satellite galaxies, i.e. galaxies that are not hosted by a resolved dark matter subhalo also called orphan galaxies, depends strongly on the tree builder. The number of type 2 satellite galaxies (orphans) resulting from the VELOCIraptor run is up to ∼ 2x higher than HBT, ROCKSTAR and SUBFIND, in agreement with the excess of merged dark matter satellite subhalo progenitors of VELOCIraptor; but these are only a small fraction of the galaxy population. When matching individual GALFORM galaxies in the four finders we find that the hot gas mass MHot Gas, cold gas mass MCold Gas and instantaneous star formation rates are significantly affected by the satellite and central subhalo type. When we have a central subhalo that has a satellite subhalo as its counterpart, the satellite galaxy has MHot Gas= 0 by definition and, consequently, lower MCold Gas and instantaneous SFR for MStellar< 108h −1M than the central that it is matched to, independently of the halo finder algorithm. However, when comparing matched centrals and matched satellites, their average properties agree between the outputs from the different finders. Even though the scatter can be quite high, we find that the distributions of different baryon properties do not vary significantly between the different runs, even including orphan galaxies. Overall, the run with the strongest differences in the stellar mass function is that with HBT which contains ∼ 30% less galaxies, while in the other runs the match is excellent. Another discrepancy is that GALFORM run on VELOCIraptor is accompanied by higher star formation rate density and smaller sizes. However, in all cases the amplitude of these differences is small and insufficient to make an impact on any comparison to observational properties. The reason behind the small impact of the choice of finder lies in the homogenising effect of GALFORM in the merger trees, which ensures a small uncertainty on the model results coming from the dark matter halo finding procedure.