This week’s entry is entirely dedicated to galaxy- and lensing-related observational results, with surveys ranging from DES to BOSS to DECaLS (a DESI-related survey). Enjoy!
#1 2010.05294: DES Y1 results: Splitting growth and geometry to test ΛCDM by Jessie Muir et al. (DES collaboration)
If one fixes the parameters of ΛCDM and assumes General Relativity (GR) to be the underlying theory of gravity, then the evolution of geometry (expansion rate and consequently distances) and growth of structure is uniquely fixed. Therefore, inconsistencies between geometry and growth could indicate a breakdown of ΛCDM+GR and the need for a new model (besides the more mundane explanation of systematics). In the case of weak lensing experiments such as the Dark Energy Survey (DES), a possible inconsistency could also have some bearing in the mild but growing tensions between these experiments and Planck as far as σ8 and Ωm are concerned. The idea behind this week’s paper by the DES collaboration is then to focus on a subset of cosmological parameters and split them in two versions: one associated to the background expansion (and hence used to compute comoving, angular, and luminosity distances), and one associated to growth (and hence used to compute growth rate and growth factor). Let’s then check if these two split quantities agree within each other. If they do, it is a sign that, so far, nothing is terribly wrong with the assumption of ΛCDM+GR. Moreover, such a split can teach us about whether DES is more sensitive to geometric or growth information, and which of the two is mostly responsible for driving tensions with Planck.
In this week’ paper led by Jessie Muir, this geometry-growth split is performed focusing on the matter density parameter, Ωm, which thus is split into Ωm_geo and Ωm_grow. One of the key plots is the one I’ve copied below, which shows the posterior for the difference between the two split parameters. Fortunately, everything seems nice and consistent, as Ωm_geo-Ωm_grow is consistent with 0 within ~1.5σ (depending on the exact datasets used). The hint for Ωm_geo-Ωm_grow>0 is, I would say, insignificant. For the record it turns out that DES results are more driven by geometry than growth (something which some might find surprising), and that the tension between DES and Planck seems to be driven by growth (although this question is hard to answer). The idea behind this paper has been used in a few previous works, including 1410.5832 by Ruiz and Huterer (the latter having been Muir’s PhD advisor). Its philosophy also reminded me of a very cool paper by Pavel Motloch I covered in my Week 17 entry, which looked at 8 (!) different ways the baryon energy density Ωbh2 entered into the computation of CMB anisotropies, fortunately finding no evidence for disagreement between the split quantities.
#2 2010.04182: Evidence for galaxy assembly bias in BOSS CMASS redshift-space galaxy correlation function by Sihan Yuan et al.
In my Week 25 entry, I gave a heuristic but otherwise rather accurate description of how to think about galaxy bias, the statistical relation between the clustering of galaxies and the clustering of the underlying matter field. In the simple picture I gave, besides complications due to the presence of massive neutrinos, galaxy bias would essentially only depend on two quantities: the mass of the halos within which the galaxies are residing, and of course their redshift. The halo occupation distribution model (HOD) more accurately quantifies the galaxy-halo connection, and is essentially making the assumption that halo mass is the property that most strongly correlates with the abundance and clustering of both halos, as well as the galaxies which reside in halos. But this clearly can’t be the end of the story. At fixed halo mass, it is impossible to believe that halo clustering will not depend, at least to some extent, on so-called secondary properties, related to the halo assembly history. This is somewhat the halo analogous of the nature vs nurture problem. If I give you two tins (proto-halos with equal mass), who then grow in two different environments (different assembly histories), you wouldn’t expect these twins to behave similarly after many years, right? So why should halos? This dependence on secondary properties other than halo mass is encapsuled in the concept of halo assembly bias. Now the thing gets a bit messier here. Besides halo assembly bias, there is an additional related issue known as galaxy assembly bias. I’m going to quote verbatim from this week’s paper here: “at fixed halo mass, the galaxy properties or number of galaxies within dark matter halos may depend on secondary halo properties that themselves show a halo assembly bias signature”. This is the cleanest definition of galaxy assembly bias, which can lead to important errors in studying both the galaxy-halo connection and galaxy formation if ignored.
In this week’s paper, Yuan and collaborators search for evidence of galaxy assembly bias in current data. They use the BOSS CMASS 2-point correlation function (2PCF), which is essentially the Fourier transform of the power spectrum from the same galaxy sample. They first fit the 2PCF ignoring galaxy assembly bias, and then add two additional parameters (called A and A_e) related to galaxy assembly bias. More precisely, these two parameters capture the impact of halo concentration (a measure of how centrally-peaked the halo density profile is) and halo environment (appropriately defined) on galaxy assembly bias. They find that the addition of these two parameters substantially improves the fit to the data, with the improvement in fit large enough to warrant the addition of the two parameters from a statistical point of view (as quantified through rough but otherwise useful Bayesian model comparison tools). I believe this is the first robust evidence for galaxy assembly bias in data, and paves the way towards perhaps more physically motivated galaxy bias prescriptions.
Besides being important for the construction of mock catalogs, a careful treatment of galaxy assembly bias could have something to do with the so-called “lensing is low” tension, something which we do not often hear about but is no less important than other cosmological tensions (and might or might not have something to do with the σ8-Ωm tension). As well discussed in 1611.08606, the “lensing is low” tension is essentially the statement that fits to galaxy-galaxy clustering predict a galaxy-galaxy lensing signal which is ~20-40% larger than observed. The tension is therefore between galaxy-galaxy clustering and galaxy-galaxy lensing, and crucially hinges upon an assumed HOD model for the galaxy-halo connection. As this week’s paper suggests, the “lensing is low” problem might simply be the sign of an incomplete HOD model.
#3 2010.04698: Cross-Correlation of Planck CMB Lensing with DESI-Like LRGs by Ellie Kitanidis and Martin White
The Dark Energy Spectroscopic Instrument (DESI) is an upcoming Stage IV dark energy experiment which will deliver the largest ever 3D map of the Universe and correspondingly the largest spectroscopic sample of, among others, galaxies and quasars. Throughout the community, DESI is considered to be the heir to BOSS/eBOSS. The less heard of DECam Legacy Survey (DECaLS) is a deep, wide-field survey which provides the optical imaging used to conduct targeting for ~2/3 of the DESI footprint. As such, DECaLS data can be used to produce DESI-like data, using the color cuts specified in Eq.(1). In this week’s paper, Kitanidis and White use the DR8 DECaLS data, collected between August 2014 and March 2019. Through the previously mentioned color cuts, they construct a DESI-like luminous red galaxy (LRG) sample. LRGs are galaxies which are very luminous and intrinsically red due to high stellar mass and lack of recent star formation activity. They are expected to reside in very massive halos, and hence cluster strongly (i.e. they are expected to have a high bias). As such, LRGs are excellent tracers of the large-scale structure, and in fact have been used to deliver a large chunk of the clustering science of BOSS/eBOSS.
In this week’s paper, Kitanidis and White cross-correlate their DECaLS galaxy sample with Planck CMB lensing convergence maps. For why this cross-correlation is interesting, you can read more in my Week 14 entry, where I discussed the ACT lensing-BOSS cross-correlation. In essence, it boils down to three things: 1) pin down the growth of matter with time, 2) break degeneracies between galaxy physics and cosmology (such as between galaxy bias, b1, and σ8), and 3) in general reduce systematics. Kitanidis and White measure both the galaxy-galaxy and convergence-galaxy correlations (named gg and kg respectively), and obtain the respective galaxy bias parameters, which they then compare to expectations from the DESI Final Design Report, finding everything to be overall rather consistent. They also test two different modelings of the signals in terms of either linear bias+HALOFIT emulator on top, or 1-loop convolution Lagrangian effective field theory. They find that, given the sensitivity of the data, the two deliver similar constraints, although the perturbation theory modeling is slightly more accurate. The only minor twist is that the kg correlation is found to be slightly lower than expected given the measured gg correlation. This is something we have been observing for a long time now (with Elena Giusarma we also found something similar and 1802.08694 and discussed it in more detail there). There are essentially two widely accepted explanations: 1) systematics contamination (most likely by tSZ) in the Planck lensing maps - again, see my Week 14 entry which discusses precisely this problem in light of a new tSZ cleaning method applied to ACT lensing maps or 2) a failure of the input assumptions, which might or might not have something to do with the “lensing is low” problem I discussed above (though the problem with “lensing is low” is with galaxy-galaxy lensing rather than galaxy-CMB lensing, but it is entirely possible both that the two problems might be related, and that there is some relation to the σ8-Ωm tension).