Abstract
We derive cosmological constraints on the matter density, , and the amplitude of fluctuations, , using , a catalog of 1800 galaxy clusters we identified in the Sloan Digital Sky Survey-DR13 spectroscopic data set using our GalWeight technique to determine cluster membership. By analyzing a subsample of 756 clusters in a redshift range of 0.045 ≤ z ≤ 0.125 and virial masses of M ≥ 0.8 × 1014 with mean redshift of z = 0.085, we obtain (systematic) and (systematic), with a cluster normalization relation of . There are several unique aspects to our approach: we use the largest spectroscopic data set currently available, and we assign membership using the GalWeight technique, which we have shown to be very effective at simultaneously maximizing the number of bona fide cluster members while minimizing the number of contaminating interlopers. Moreover, rather than employing scaling relations, we calculate cluster masses individually using the virial mass estimator. Since is a low-redshift cluster catalog we do not need to make any assumptions about evolution either in cosmological parameters or in the properties of the clusters themselves. Our constraints on and are consistent and very competitive with those obtained from non-cluster abundance cosmological probes such as cosmic microwave background, baryonic acoustic oscillation (BAO), and supernovae. The joint analysis of our cluster data with Planck18+BAO+Pantheon gives and .
Export citation and abstract BibTeX RIS
1. Introduction
In the current picture of structure formation, galaxy clusters arise from rare high peaks of the initial density fluctuation field. These peaks grow in a hierarchical fashion through the dissipationless mechanism of gravitational instability with more massive halos growing via continued accretion and merging of low-mass halos (White & Frenk 1991; Kauffmann et al. 1999, 2003). Galaxy clusters are the most massive virialized systems in the universe and are uniquely powerful cosmological probes. The cluster mass function (CMF), or the abundance of galaxy clusters, is particularly sensitive to the matter density of the universe and , the rms mass fluctuation on the scale of 8 Mpc at z = 0 (e.g., Wang & Steinhardt 1998; Battye & Weller 2003; Dahle 2006; Wen et al. 2010).
Cosmological analyses have been performed using samples of galaxy clusters constructed from galaxy surveys (e.g., Rozo et al. 2010; Kirby et al. 2019; DES Collaboration et al. 2020), X-ray emission (e.g., Vikhlinin et al. 2009; Mantz et al. 2015), and thermal Sunyaev–Zel'dovich (SZ) signals (e.g., Bocquet et al. 2019; Zubeldia & Challinor 2019). These cluster abundance studies showed that varies from ∼0.2 to 0.4 and varies from ∼0.6 to 1.0. The discrepancies or tensions among these various studies is basically dependent on the accuracy of cluster mass estimation. Cluster mass can be calculated from cluster dynamics using, for example, the virial mass estimator (e.g., Binney & Tremaine 1987), the weak gravitational lensing (Wilson et al. 1996; Holhjem et al. 2009), and the application of the Jeans equation for the gas density calculated from the X-ray analysis of the galaxy cluster (Sarazin 1988). It can also be estimated from other observables, the so-called mass proxies, which scale tightly with cluster mass, such as X-ray luminosity (e.g., Pratt et al. 2009), optical luminosity, or richness (e.g., Yee & Ellingson 2003; Simet et al. 2017), and the velocity dispersion of member galaxies (e.g., Biviano et al. 2006; Bocquet et al. 2015). Generally, most of these methods introduce large systematic uncertainties, which limits the accuracy of estimating cluster masses (e.g., Wojtak & Łokas 2007; Mantz et al. 2016).
Cosmological analyses of galaxy cluster abundance introduce a degeneracy between and . Large ongoing and upcoming wide and deep-field imaging and spectroscopic surveys at different redshifts, such as DES (Abbott et al. 2018a), eROSITA (Merloni et al. 2012), Vera C. Rubin Observatory (LSST Science Collaboration et al. 2009), and Nancy Grace Roman Telescope (Akeson et al. 2019), will simultaneously increase the precision of measuring the cosmological parameters and break the degeneracy between them. This is because evolves slowly while evolves strongly with redshift. Also, these galaxy surveys at different redshifts are significant to study the evolution of the CMF, which is critical to measuring structure growth, and therefore can be used to constrain properties of dark energy (e.g., Haiman et al. 2001; Mantz et al. 2008). Introducing advanced methods is essential to analyzing these surveys. One of these methods is the GalWeight technique (Abdullah et al. 2018, hereafter Abdullah+18), which can be applied to the available and upcoming spectroscopic databases of eBOSS (Raichoor et al. 2017), DESI (Levi et al. 2019), and Euclid (Euclid Collaboration et al. 2019) to construct cluster catalogs. These catalogs provide an unlimited data source for a wide range of astrophysical and cosmological applications.
In addition, there are independent cosmological probes to constraining the cosmological parameters that can be applied alongside or in combination with galaxy cluster abundance. The anisotropies in the cosmic microwave background (CMB) are an independent probe of cosmological parameters (e.g., Hinshaw et al. 2013; Planck Collaboration et al. 2016). The likelihoods of the – confidence levels (CLs) introduced by the CMF and CMB are almost orthogonal to each other, which means combining these measurements will eliminate the degeneracy between and and shrink the uncertainties. Other independent cosmological probes that are used to constrain and include cosmic shear, galaxy–galaxy lensing, and angular clustering (e.g., Abbott et al. 2018b; van Uitert et al. 2018). The likelihoods of the – CLs introduced by these probes are almost parallel to those introduced by the CMF. Moreover, the two cosmological probes of baryon acoustic oscillations (BAO, e.g., Eisenstein et al. 2005) and supernovae (SNe, e.g., Perlmutter et al. 1999) can be used to constrain only (independent of ).
In this paper, we aim to derive the CMF and the cosmological parameters and using a subsample of 756 clusters () obtained from the cluster catalog as we discuss below in detail. The (Abdullah et al. 2020, hereafter Abdullah+20) catalog was derived from the Sloan Digital Sky Survey-Data Release 13 spectroscopic data set (hereafter SDSS-DR13,5 Albareti et al. 2017). The clusters were first identified by looking for the Finger-of-God effect (see Jackson 1972; Kaiser 1987; Abdullah et al. 2013). The cluster membership was constructed by applying our own GalWeight technique, which was specifically designed to simultaneously maximize the number of bona fide cluster members while minimizing the number of contaminating interlopers (Abdullah+18). In Abdullah+18, we applied our GalWeight technique to MDPL2 and Bolshoi N-body simulations and showed that it was >98% accurate in correctly assigning cluster membership. The catalog is at low-redshift for which the effects of cluster evolution and cosmology are minimal. Finally, the cluster masses were calculated individually from the dynamics of the member galaxies via the virial theorem (e.g., Limber & Mathews 1960; Abdullah et al. 2011), and corrected for the surface pressure term (e.g., The & White 1986; Carlberg et al. 1997). A huge advantage of our approach relative to mass proxy methods is that it returns an estimate of the total cluster mass (dark matter and baryons) without making any assumptions about the internal complicated physical processes associated with the baryons (gas and galaxies). The publicly available 6 contains1800 clusters at redshift z ≤ 0.2, which is one of the largest available samples that used a high-quality spectroscopic data set.
The paper is organized as follows. In Section 2, we describe in more detail how we created the cluster catalog. In Section 3, we investigate the volume and mass incompleteness of to obtain a mass-complete local subsample of 756 clusters () used to constrain and . In Section 4, we compare our complete sample with theoretical models to constrain the cosmological parameters and . We investigate how systematics affect the recovered cosmological constraints and compare our results with recent results constrained from some cosmological probes and summarize our conclusions in Section 5. Throughout the paper we adopt ΛCDM with Ωm = 1 − ΩΛ, and H0 = 100 h km s−1 Mpc−1.
2. The Cluster Catalog
In this section, we summarize how we created the cluster catalog. Full details may be found in Abdullah+20. Using photometric and spectroscopic databases from SDSS-DR13, we extracted data for 704,200 galaxies. These galaxies satisfied the following set of criteria: spectroscopic detection, photometric and spectroscopic classification as a galaxy (by the automatic pipeline), spectroscopic redshift between 0.001 and 0.2 (with a redshift completeness >0.7, Yang et al. 2007; Tempel et al. 2014), r-band magnitude (reddening-corrected) <18, and the flag SpecObj.zWarning is zero indicating a well-measured redshift.
Galaxy clusters were identified by the well-known Finger-of-God effect (Jackson 1972; Kaiser 1987; Abdullah et al. 2013). The Finger-of-God effect causes a distortion of line-of-sight velocities of galaxies in the redshift-phase space due to the cluster potential well. As described in Abdullah+20, we calculated the membership of each cluster as follows. We first calculated the galaxy number density within a cylinder of radius 0.5 Mpc and height 3000 km s−1 centered on a galaxy. Second, we sorted all galaxies descending from highest to lowest number densities with the condition that the cylinder has at least eight galaxies. Third, starting with the galaxy with the highest number density, we applied the binary tree algorithm (e.g., Serra et al. 2011) to accurately determine a cluster center (αc, δc, zc) and a phase-space diagram. Fourthly, we applied the GalWeight technique (Abdullah+18) to galaxies in the phase-space diagram out to a maximum projected radius of 10 Mpc and a maximum line-of-sight velocity range of ±3000 km s−1 to identify cluster membership. In Abdullah+18, we showed that the cumulative completeness of the FOG algorithm, which we tested using the Bolshoi simulation (Klypin et al. 2016) was approximately 100% for clusters with masses , and ∼85% for clusters with masses .
The virial mass of each cluster was estimated by applying the virial theorem to the cluster members, under the assumption that the mass distribution follows the galaxy distribution (e.g., Giuricin et al. 1982; Merritt 1988). The estimated mass was corrected for the surface pressure term which, otherwise, would overestimate the fiducial cluster mass (e.g., The & White 1986; Binney & Tremaine 1987; Carlberg et al. 1997). The cluster virial mass was calculated at the viral radius within which the cluster is in hydrostatic equilibrium. The virial radius is approximately equal to the radius at which the density , where ρc is the critical density of the universe and Δ200 = 200 (e.g., Carlberg et al. 1997; Klypin et al. 2016). Abdullah+20 showed that the cluster mass estimates returned by the virial theorem after utilizing the GalWeight technique (Abdullah+18) performed very well in comparison to most other mass estimation techniques described in Old et al. (2015). In particular, our procedure was applied to two mock catalogs (HOD2 and SAM2) recalled from Old et al. (2015). We found that the rms differences of the recovered mass by GalWeight relative to the fiducial cluster mass were 0.24 and 0.32 for HOD2 and SAM2, respectively. Also, the intrinsic scatter in the recovered mass was ∼0.23 dex for both catalogs. Moreover, the uncertainty of the virial mass estimator is calculated using the limiting fractional uncertainty (Bahcall & Tremaine 1981).
The scatter and bias in the recovered mass using the virial mass estimator are caused by some factors, including (i) the assumption of hydrostatic equilibrium, projection effect, and possible velocity anisotropies in galaxy orbits, and the assumption that halo mass follows light (or stellar mass); (ii) the presence of substructure and/or nearby structure such as a cluster, or supercluster, to which the cluster belongs, or filament (e.g., Merritt 1988; Fadda et al. 1996); (iii) the presence of interlopers in the cluster frame due to the triple-value problem, for which there are some foreground and background interlopers that appear to be part of the cluster body because of the distortion of phase space (Tonry & Davis 1981; Abdullah et al. 2013); and (iv) the identification of cluster center (e.g., Girardi et al. 1998; Zhang et al. 2019).
The 1800 clusters range in redshift between 0.01–0.2 and in mass between . The catalog contains a large number of cluster parameters including sky position, redshift, membership, velocity dispersion, and mass at overdensities . The 34,471 member galaxies were identified within the radius at which the density is 200 times the critical density of the universe. The galaxy catalog provided the coordinates of each galaxy and the ID of the cluster that the galaxy belongs to. The catalog was made publicly available at the following website https://mohamed-elhashash-94.webself.net/galwcat/.
3. Cluster Mass Function
The catalog is not complete in either volume or mass. In Section 3.1, we analyze to develop an appropriate selection function of our sample, which is used to correct for the volume incompleteness. Also, in Section 3.2, we compute the CMF derived from and compare it with the CMF calculated from the MDPL27 simulation (described in the next paragraph) to obtain a mass-complete subsample () used to constrain the cosmological parameters and .
The MDPL2 is an N-body simulation of 38403 particles in a box of comoving length 1 h−1 Gpc, mass resolution 1.51 × 109 h−1 M⊙, and gravitational softening length 5 h−1 kpc (physical) at low redshifts from the suite of MultiDark simulations (see Table 1 in Klypin et al. 2016). It was run using the L-GADGET-2 code, a version of the publicly available cosmological code GADGET-2 (Springel 2005). It assumes a flat ΛCDM cosmology, with cosmological parameters ΩΛ = 0.693, Ωm = 0.307, Ωb = 0.048, n = 0.967, σ8 = 0.823, and h = 0.678 (Planck Collaboration et al. 2014). Halos and subhalos have been identified with ROCKSTAR (Behroozi et al. 2013a) and merger trees constructed with CONSISTENT TREES (Behroozi et al. 2013b). The catalogs are split into 126 snapshots between redshifts z = 17 and z = 0. We downloaded the snapshot (hlist_0.91520.list8 ) with z ∼ 0.09, which is consistent with the mean redshift of the sample.
3.1. Completeness
The catalog is incomplete in the distribution of clusters with respect to comoving distance (redshift), and in the distribution of clusters with respect to mass. In this section, we discuss such incompleteness and how to make corrections.
The completeness in comoving volume (redshift) of the catalog can be investigated by calculating the abundance of clusters predicted by a theoretical model and comparing it with the abundance of clusters. We adopt the functional form of Tinker et al. (2008; hereafter Tinker08) to calculate the halo mass function (HMF;9 see Section 4.1 for more details) and consequently the predicted abundance of clusters.
The integrated abundance of clusters as a function of redshift for the sample, , is presented in the upper left panel of Figure 1. Note that N (<z) is calculated for the clusters with redshift z ≥ 0.04 to remove the effect of nearby regions where the cosmic variance has a large effect due to the small volume. The plot shows that the catalog is matched with the prediction of Tinker08 for z ≲ 0.09. Also, the fractional error of N (<z) relative to the expectation of Tinker08, , for each model and the expected Poisson noise (gray shaded area) are presented in the lower left panel. The plot shows that the scatter relative to each model is nearly constant (around zero) for z ≲ 0.09 before it blows up after this redshift limit. This indicates that is approximately complete in volume for z ≲ 0.09 (or equivalently comoving distance of D ≲ 265 Mpc for the ΛCMD universe with = 0.3). We call this volume-complete subsample .
Similarly, the integrated abundance of clusters as a function of cluster mass, N (>M), is presented in the upper right panel of Figure 1 in comparison to five Tinker08 models and the scatter is presented in the lower right panel. The plot shows that the data is matched with the models of = [0.20, 0.305, 0.40] with = 0.825 better than the models of = 0.305 and = [0.725, 0.925]. Even though it is not an easy task to specifically determine the mass threshold at which the catalog is complete, the three matched models indicate that is approximately complete for log(M) ≳ 13.9 . We discuss the systematics of adapting this mass threshold on our analysis in Section 5.1. The large scatter at the high-mass end is due to the small number of massive clusters, while the large scatter at the low mass end comes from the incompleteness of .
In order to correct for the incompleteness in volume of each cluster should be weighed by , where is the selection function at a distance D. Figure 2 introduces the normalized number density , defined as the cluster number density normalized by the average number density calculated for clusters within comoving distance D < 265 Mpc, for all clusters and for five mass bins as described in Table 1. The distribution of points in Figure 2 can be described by an exponential function that represents the selection function . It has the form
Download figure:
Standard image High-resolution imageTable 1. The Cluster Average Number Density for Different Mass Bins
Mass | Number of | Average | Color |
---|---|---|---|
Bin | Clusters | Number Density | |
() | Mpc−3) | ||
13.6–15.2 | 1800 | 5.6 | Black |
13.6–13.8 | 527 | 2.2 | Blue |
13.8–14.0 | 461 | 1.5 | Green |
14.0–14.2 | 411 | 1.0 | Red |
14.2–14.5 | 326 | 0.7 | Cyan |
14.5–15.2 | 75 | 0.2 | Magenta |
Note. Columns: (1) the mass bin in units of (); (2) the number of clusters in each mass bin; (3) the average number density calculated for clusters within comoving distance D < 265 Mpc in each mass bin; (4) the color of the number density profile as shown in the right panel of Figure 1.
Download table as: ASCIITypeset image
The parameters a, b, and γ are determined by applying the chi-squared algorithm using the Curve Fitting MatLab Toolbox. The best-fit values of these parameters are a = 1.07 ± 0.12, b = 293.4 ± 20.7 Mpc, and γ = 2.97 ± 0.90 with an rms error of 0.15. Note that the normalization a is greater than unity because of the scatter and the effect of the cosmic variance. However, we apply the selection function with the condition that .
We should be cautious when using at large distances. This is because (D ≳ 500) Mpc drops to ≳0.01 as demonstrated in Figure 2, which means that a distant cluster would be weighted at least 100 times as a nearby cluster. This will overestimate or overcorrect the number of clusters at large distances, and consequently the estimated CMF will be noisy. Thus, in order to avoid the overcorrection and the noisiness of CMF we restrict our sample to a maximum comoving distance of D ≤ 365 (or z ≤ 0.125) for which .
It is well-known that the cluster number density of a given mass decreases with redshift for a 100% complete sample because of the HMF evolution effect. Thus, the CMF should be scaled or corrected by an evolution function, . For a sample with a broad range of redshifts, the only way to take the evolution into account is to calculate this function. However, the disadvantage of this approach is that the correction is model dependent: the measured HMF (i.e., CMF) is a convolution of the true HMF and theoretical estimate of . However, for a sample with a narrow range of redshifts (as in our case) we show in Appendix A that the evolution effect is less than 3% for clusters in the redshift range of 0.045 ≤ z ≤ 0.125. In Appendix B, we discuss the effect of adopting this redshift interval in our results.
3.2. Estimating the Mass Function
In this section, we compute the CMF, dn(M)/dlog(M), and its corresponding cumulative mass function, n (>M), which are estimated for a ΛCDM cosmology with = 0.3 and ΩΛ = 0.7. The CMF is defined as the number density of clusters per logarithmic cluster mass interval. Also, the cumulative CMF is defined as the number density of clusters more massive than a given mass M.
Mathematically, the CMF, weighted by the selection function , is given by
where Di is the comoving distance of a cluster i, and V is the comoving volume, which is given by
where Ωsky = 41,253 deg2 is the area of the sky, Ωsur ≃ 10,000 deg2 is the area covered by GalWCat19, and D1 and D2 are the minimum and maximum comoving distances of the cluster sample.
Figure 3 introduces the cumulative CMF computed from . The black line is the CMF computed from the MDPL2 simulation (for the snapshot hlist_0.91520.list at z ∼ 0.09 or D ∼ 265; Klypin et al. 2016). The blue points introduce the CMF for without the correction of , since this sample is already complete in volume (see Section 3.1 and Figure 1). The red points represent our CMF corrected by for D ≤ 365 Mpc (z ∼ 0.125). Comparing the CMF estimated by the subsample with that derived from the MDPL2 simulation indicates that the sample is approximately complete in mass for , while it drops lower than the CMF of MDPL2 at the low-mass end. Also, our CMF, corrected by , is in good agreement with the CMF derived from with a scatter of 0.026 dex. The mass completeness of is discussed in Section 3.1 and Figure 1. In Appendix B, we show that the results of deriving the cosmological parameters from are consistent with that derived from . This indicates that weighting each cluster in our sample by introduced in Section 3.1 and Equation (1) is sufficient to correct for the volume incompleteness of .
Download figure:
Standard image High-resolution imageTherefore, our final subsample, corrected by S(D) is restricted by log(M) ≥ 13.9 and 0.045 ≤ z ≤ 0.125. The number of clusters of this subsample is 756, which represents ∼42% of the sample. We use this subsample to constrain and and call it a fiducial sample.
4. Implications for Cosmological Models
In Section 4.1, we discuss the prediction of HMF from the theoretical framework. In Section 4 we derive the constrains on the cosmological parameters and , and discuss the degeneracy between these two parameters.
4.1. Prediction of HMF
The number of dark matter halos per unit mass per unit comoving volume of the universe, HMF, is given by
here ρ0 is the mean density of the universe, σ is the rms mass variance on a scale of radius R that contains mass , and f(σ) represents the functional form that defines a particular HMF fit.
Assuming a Gaussian distribution of mass fluctuation, Press & Schechter (1974) used a linear theory to derive the first theoretical model (hereafter PS) of HMF. While fairly successful in matching the results of N-body simulations, the PS formalism tends to predict too many low-mass clusters and too few high-mass clusters. More recently proposed theoretical models provide better approximations to the output from N-body simulations (e.g., Jenkins et al. 2001; Sheth et al. 2001; Warren et al. 2006; Tinker & Wetzel 2010; Bhattacharya et al. 2011; Behroozi et al. 2013a).
In this paper, we adopt the functional form proposed by Tinker et al. (2008) (hereafter Tinker08) as our form of the HMF. This approach assumes universality of the HMF across the cosmological parameter space considered in this work, and uses a fitting function that was calibrated against N-body simulations. The Tinker08 model is formally accurate to better than 5% for the cosmologies close to the ΛCDM cosmology and for the mass and redshift range of interest in our study (e.g., Vikhlinin et al. 2009). Although the formula has been calibrated using dissipationless N-body simulations (i.e., without the effect of baryons), hydrodynamic simulations suggest that these have negligible impact for clusters with masses as high as those considered here (e.g., Rudd et al. 2008; Velliscig et al. 2014; Bocquet et al. 2016). Finally, note that the Tinker08 model is defined in spherical apertures enclosing overdensities similar to the mass we derive for the observed clusters.
where , , , c = 1.19, and , and σ2 is the mass variance defined as
P(k) is the current linear matter power spectrum (at z = 0) as a function of wavenumber k, is the Fourier transform of the real-space top-hat window function of radius R, and g(z) = σ8(z)/σ8(0) is the growth factor of linear perturbations at scales of 8 Mpc, normalized to unity at z = 0.
The current linear power spectrum P(k) is defined as P(k) = B kn T2(k), where T(k) is the transfer function, B is the normalization constant and n is the spectral index. Usually the normalization B is calculated from the cosmological parameter (e.g., Reiprich & Böhringer 2002; Murray et al. 2013). The function kn imprints the primordial power spectrum during the epoch of inflation. The transfer function T(k) quantifies how this primordial form is evolved with time to the current linear power spectrum on different scales. The transfer function T(k) is calculated using the public Code for Anisotropies in the Microwave Background (CAMB;10 Lewis et al. 2000). The quantities and are the main cosmological parameters that define the HMF. The other parameters do not strongly affect the HMF and thus we fix them during the calculation of the HMF as described below (e.g., Reiprich & Böhringer 2002; Bahcall et al. 2003; Wen et al. 2010).
4.2. Constraining and
The HMF is calculated using the publicly available HMFcalc11 code (Murray et al. 2013). The code provides about 20 fitting functions that can be used to calculate the HMF. In this paper, in order to constrain and , we use Tinker08 (Equation (5)) as discussed above. We calculate the HMF by allowing to range between [0.1, 0.6] and between [0.6, 1.2], both in steps of 0.005. We keep the following cosmological parameters fixed: the CMB temperature , baryonic density Ωb = 0.0486, and spectral index n = 0.967 (Planck Collaboration et al. 2014), at redshift z = 0.089 (the mean redshift of ).
In order to determine the best-fit mass function and constrain and we use a standard χ2 procedure
where the likelihood, , of data (CMF) given a model (HMF) is
yo and ym are the data and model cumulative mass functions at a given mass and σ is the statistical uncertainty of the data.
Using the fiducial SelFMC sample of 756 clusters with and 0.045 ≤ z ≤ 0.125, the best-fit parameters for the minimum value of χ2 are and for Tinker08 at redshift z = 0.085. In Section 5.1 we discuss the systematics of cluster mass uncertainty, mass threshold, and selection function.
The banana shape in Figure 4 shows the well-known degeneracy between and . The relationship between and is often expressed as
The parameters α, β, and δ are determined by applying the χ2 algorithm using the Curve Fitting MatLab. The best-fit values of these parameters are α = 0.425 ± 0.006 and β = −0.550 ± 0.007 with an rms error of 0.005 for the Tinker08 model.
Download figure:
Standard image High-resolution imageWe now ask the question—how do and contribute individually to the HMF? In other words, why do cluster abundance studies introduce a degeneracy between and ? The degeneracy occurs because a low abundance of massive clusters could be caused either by a small amount of matter in the universe (a low value of ) or small fluctuations in the density field (a low value of ). Similarly, a high abundance of massive clusters could be caused either by a large amount of matter in the universe (a high value of ) or large fluctuations in the density field (a high value of ). Therefore, it is possible to obtain the same abundance of massive clusters by fixing one parameter and varying the other one. Figure 5 introduces two sets of HMFs calculated by Tinker08. The first set is shown on the left panel for five different values of = [0.1 0.2 0.3 0.4 0.5], while fixing = 0.8. The second set is shown on the right panel for five different values of = [0.6 0.7 0.8 0.9 1.0] while fixing = 0.3. As expected, increasing the matter density of the universe increases the number of clusters of all masses. But increasing the rms mass fluctuation increases the number of high-mass clusters more dramatically than the number of low-mass clusters. In other words, is very sensitive to the high-mass end of the HMF.
Download figure:
Standard image High-resolution image5. Discussion and Conclusion
In this section, we investigate how systematics affect the recovered cosmological constraints from our analysis (Section 5.1). We compare our constraints on the cosmological parameters and with those obtained from cluster abundance studies (Section 5.2). We also compare our constraints with those obtained from other cosmological probes, which we refer to as non-cluster cosmological probes (Section 5.3).
5.1. Systematics
In constraining and in Section 4.2, we only account for the statistical uncertainty of the estimated cumulative CMF using the fiducial sample. In this section, we discuss the systematics due to mass uncertainty, mass threshold, and parameterization of the selection function.
5.1.1. Mass Uncertainty
The first uncertainty comes from the difficulty of calculating cluster masses accurately. Generally, masses that are estimated using scaling relations, such as luminosity, richness, temperature, and dispersion velocity–mass relations, introduce large scatter and consequently large systematic uncertainties (e.g., Mantz et al. 2016; Mulroy et al. 2019). Masses that are computed by dynamical estimators are subject to systematic uncertainties (e.g., Wojtak & Łokas 2007; Rozo et al. 2010; Old et al. 2018). However, using the virial theorem, corrected for the surface pressure term, provides a relatively unbiased estimation of cluster masses (e.g., Rines et al. 2010; Ruel et al. 2014), particularly when using a sophisticated interloper rejection technique such as GalWeight (Abdullah+18). Also, the virial mass estimator calculates the total cluster mass including baryonic (gas and galaxies) and dark matter regardless of the internal complex physical processes associated with the baryonic component in clusters. However, the virial mass estimator still introduces scatter in estimating cluster masses (see Section 2). Abdullah+20 showed that the application of the virial mass estimator on two mock catalogs (HOD2 and SAM2) recalled from Old et al. (2015) returned intrinsic scatter of ∼0.23 dex in the recovered mass relative to the fiducial cluster mass. Also, the catalog introduced the fractional uncertainty (see Section 2) of each cluster mass.
Assuming a normal distribution, we investigate the systematics of the mass uncertainty by generating ∼8000 estimate for each cluster mass using both the fractional uncertainty for each cluster and the intrinsic scatter for the entire sample. In other words, we reanalyze ∼8000 times and refit for and for each time. The left panel of Figure 6 introduces the effect of cluster mass uncertainty on the constraints on and . Using the fractional uncertainty, we obtain = 0.305 ± 0.014 and = 0.816 ± 0.021, where the red ellipse represents 68% CL for the distribution of the reestimated 8000 pairs of and . Using the intrinsic scatter (blue ellipse), we find = 0.309 ± 0.014 and = 0.815 ± 0.022. Both results indicate that the cluster mass uncertainty (fractional or intrinsic) does not affect our constraints on and using .
Download figure:
Standard image High-resolution image5.1.2. Mass Threshold
The second systematic uncertainty comes from the difficulty of accurately determining the mass threshold at which the sample is mass complete. As discussed in Section 3.1 and Figure 1 the catalog is approximately complete around []. However, the mass threshold at which the sample is mass-complete is not accurately specified. Therefore, we investigate the effect of varying the mass threshold between 13.8 and 14.0 [] in steps of 0.05 dex on the recovered cosmological constraints from our analysis. For each mass threshold we calculate the χ2 likelihood and then we obtain the joint 68% CL of all χ2 distributions as shown in the middle panel of Figure 6. The plot shows that the best-fit values of and deviate very slightly from the results of the fiducial sample with and .
5.1.3. Selection Function Parameterization
The constraints on and are affected by parameterization of the selection function. Our selection function depends on three parameters a, b, and γ. The normalization a is already fixed to unity. Assuming a normal distribution, the systematic of the selection function is investigated by generating ∼8000 pairs of b and γ, using the uncertainty in b and γ (see Section 3.1). For each pair we estimate the best-fit values of and . Figure 6 shows the 68% CL for the systematic of the selection function. This analysis rotates the error ellipses slightly compared to our fiducial analysis, but does not affect our results. We obtain = 0.313 ± 0.035 and = 0.809 ± 0.012, which is consistent with our result of the fiducial sample.
5.2. Comparison with External Data from Cluster Abundance
The left panel of Figure 7 introduces the 68% CL derived from in comparison to the results obtained from other cluster abundance studies. Samples of galaxy clusters constructed from galaxy surveys include optical photometric (e.g., Kirby et al. 2019), X-ray (e.g., Mantz et al. 2015), and SZ (e.g., Zubeldia & Challinor 2019) catalogs as listed in Table 2. The figure shows that the CLs of all cluster abundance studies introduce a degeneracy between and as we discussed in Section 4.2. Also, the CL derived from overlaps the CLs obtained from all other results as shown in the figure. Regardless of this overlapping, the right panel of Figure 7 shows that the constraints on and from cluster abundance studies are in tension with each other, even for the studies that use the same type of cluster sample. Specifically, the X-ray independent studies listed in Table 2 introduce different values of and , which vary from ∼0.22 to 0.40 and 0.71 to 0.89, respectively. Also, the independent studies that use SZ-cluster samples show that and vary from ∼0.25 to 0.31 and 0.77 to 0.98, respectively.
Download figure:
Standard image High-resolution imageTable 2. Comparison of Constraints on Cosmological Parameters and Derived from Clusters Abundances (CMF) and from Other Cosmological Probes, Including Cosmic Shear, Galaxy–Galaxy Lensing, Angular Clustering, BAO, Supernovae, and CMB
Sample | Mass Estimation | a | b | Reference | ||
---|---|---|---|---|---|---|
Spectroscopically selected catalogs—cluster abundance | ||||||
Virial theorem | 0.817 | 0.032 | This work | |||
Optical photometrically selected catalogs—cluster abundance | ||||||
MaxBCG | Richness-mass +WL | 0.779 | 0.108 | Rozo et al. (2010) | ||
RedMaPPer | Richness-mass +WL | 0.778 | 0.325 | Costanzi et al. (2019) | ||
RedMaPPer | Richness-mass +X-ray | 0.776 | 0.212 | Kirby et al. (2019) | ||
X-ray catalogs—cluster abundance | ||||||
REFLEX | Luminosity-mass | 0.758 | 0.148 | Schuecker et al. (2003) | ||
Chandra-ROSAT | Luminosity-mass | 0.757 | 0.191 | Vikhlinin et al. (2009) | ||
ROSAT (WtG)c | Luminosity-mass | 0.773 | 0.176 | Mantz et al. (2015) | ||
ROSAT - HIFLUGCS | Luminosity-mass | 0.760 | 0.327 | Schellenberger & Reiprich (2017) | ||
XMM-XXL | Temperature-mass | 0.832 | 0.289 | Pacaud et al. (2018) | ||
SZ catalogs—cluster abundance | ||||||
ACT, [BBN+H0+ACTcl(B12)] | SZ-mass | 0.799 | 0.214 | Hasselfield et al. (2013) | ||
ACT, [BBN+H0+ACTcl(Dyn)] | SZ-mass | 0.977 | 0.207 | Hasselfield et al. (2013) | ||
SPT | SZ-mass | 0.776 | 0.129 | Bocquet et al. (2019) | ||
HECS-SZ | SZ-mass | ⋯ | ⋯ | 0.751 | ⋯ | Ntampaka et al. (2019) |
Planck18 | SZ-mass | 0.783 | 0.138 | Zubeldia & Challinor (2019) | ||
Other cosmological probes | ||||||
DES-Y1 | CS+GGL+ACd | 0.778 | 0.143 | Abbott et al. (2018b) | ||
KiDS+GAMA | CS+GGL+AC | 0.804 | 0.032 | van Uitert et al. (2018) | ||
Pantheon | SNe | ⋯ | ⋯ | ⋯ | Scolnic et al. (2018) | |
6dF+DR7+BOSSe | BAO | ⋯ | ⋯ | ⋯ | Alam et al. (2017) | |
WMAP9 | CMB only | 0.792 | 0.112 | Hinshaw et al. (2013) | ||
Planck18 | CMB only | 0.832 | 0.000 | Planck Collaboration et al. (2018) |
Notes.
aThe cluster normalization condition parameter, S8, is defined as as used in the literature. b is the scatter of and obtained from each method listed the table relative to that obtained from Planck18 (Planck Collaboration et al. 2018). cMantz et al. (2015) used the combination of luminosity, temperature, gas mass, and lensing mass to estimate cluster masses, which was referred to as Weighting the Giant (WtG). dCC = cosmic shear, GGL = galaxy–galaxy lensing, AC = angular clustering. e6dF = Six Degree Field Galaxy Survey (Beutler et al. 2011), DR7 = SDSS data release 7 (Ross et al. 2015), BOSS = Baryon Oscillation Spectroscopic Survey (Alam et al. 2017).Download table as: ASCIITypeset image
The question now is why are the cosmological constraints derived from many of the cluster abundance techniques in tension with each other? All cluster samples constructed from photometric surveys or detected by the SZ effect do not return an estimate of each cluster's mass directly. For such samples the cluster mass has to be inferred indirectly from other observables, which scale tightly with cluster mass. Among these mass proxies are X-ray luminosity, temperature, the product of X-ray temperature and gas mass (e.g., Vikhlinin et al. 2009; Mantz et al. 2016), richness (e.g., Yee & Ellingson 2003; Simet et al. 2017), and SZ signal (e.g., Bocquet et al. 2019). To estimate cluster masses for the clusters in these samples it is necessary to follow-up a subset of clusters and calculate their masses using, e.g., weak lensing or X-ray observations. Then, an observable–mass relation can be calibrated for these subsamples. Finally, the mass of each cluster in the sample can be estimated from this scaling relation. However, this reliance on observable–mass proxies introduces significant systematic uncertainties, which is the dominant source of error (e.g., Henry et al. 2009; Mantz et al. 2015) for the reasons explained in the next paragraph.
First, the masses obtained for the follow-up subsample of clusters are often biased. For example, it is known that X-ray mass estimates are typically biased low and so a mass bias factor, (1-β), needs to be introduced and calibrated. Second, the size of the subsample used for calibration is usually small (tens of clusters), which introduces large uncertainties in both the slope and the normalization of the scaling relation. Third, many cluster catalogs span a large redshift range, so evolution (due to both the evolution of the universe and the physical processes of baryons in clusters) in the the scaling relations used to estimate the masses needs to be carefully handled, introducing another source of uncertainty. All of the aforementioned assumptions can introduce large uncertainties in the estimates of cluster mass and consequently the constraints on cosmological parameters. For instance, is specifically very sensitive to the high-mass end of the CMF and any offset of cluster true masses leads to biased estimation of . Other observational systematics that introduce additional uncertainties are photometric redshift errors and cluster miscentering.
By using the cluster catalog and deriving cluster masses using the virial theorem, we were able to avoid most of the complexities described above. First, we were able to identify clusters, assign membership, and determine cluster centers and redshifts with high accuracy from the high-quality SDSS spectroscopic data set. Second, cluster membership was determined by the GalWeight technique, which has been shown to be ∼98% accurate in assigning cluster membership (Abdullah+18). Third, a mass for each cluster was determined directly using the virial theorem. Therefore, we were able to recover a total (dark plus baryonic) mass for each cluster and circumvent having to make any assumptions about the complicated physical processes associated with the baryons. It has been suggested that cluster masses estimated via the virial theorem are overestimated by 20%. But we note that we have applied a correction for the surface pressure term, which we believe decreases this bias, especially when applied in combination with our GalWeight membership technique (Abdullah+18). Abdullah+20 showed that the virial mass estimator performed well in comparison to the other mass estimators described in Old et al. (2015), and resulted in a relatively low bias and scatter when applied to two semianalytical simulations (see Figure 3 in Abdullah+20). Fourth, since is a low-redshift cluster catalog it eliminates the need to make any assumptions about evolution in clusters themselves and evolution in cosmological parameters. Finally, because of the large size of the we are able to determine the CMF well and consequently constrain the cosmological parameters and with high precision.
5.3. Comparison with External Data from Noncluster Cosmological Probes
Cosmological parameters can be estimated from different cosmological probes rather than cluster abundance studies. We use measurements of primary CMB anisotropies from both WMAP (9 yr data; Hinshaw et al. 2013) and Planck satellites focused on the TT+lowTEB data combination from the 2018 analyses (Planck Collaboration et al. 2018). We also use angular diameter distances as probed by BAO including the 6dF Galaxy Survey (Beutler et al. 2011), the SDSS Data Release 7 (Ross et al. 2015), and the BOSS Data Release 12 (Alam et al. 2017). Furthermore, we use measurements of luminosity distances from Type Ia supernovae from the Pantheon sample (Scolnic et al. 2018). Finally, we use the measurements from a joint analysis of three cosmological probes: cosmic shear, galaxy–galaxy lensing, and angular clustering, including the results of the Kilo Degree Survey and the Galaxies And Mass Assembly survey (KiDS+GAMA; van Uitert et al. 2018) and the first year of the Dark Energy Survey (DES Y1; Abbott et al. 2018b; see Table 2). The left panel of Figure 8 introduces the 68% CL derived from in comparison to the those obtained from the aforementioned cosmological probes. As shown, the CL derived from overlaps the CLs obtained from all non-cluster abundance probes.
Download figure:
Standard image High-resolution imageWe define the scatter
to compare the constraints on and obtained from all cosmological probes, which are listed in Table 2 with that obtained from Planck18 (Planck Collaboration et al. 2018). Note that the constraints on and derived from most of the cluster abundance studies independently introduce a relatively large scatter compared to the CMB experiment of Planck18. However, our constraints on and are very comparable and competitive with Planck18 with a minimum value of Δpl = 0.018. Moreover, our constraint on is in excellent agreement with the results of the BAO and Pantheon, separately. This remarkable consistency demonstrates that our derived cluster catalog at low redshift and calculating cluster masses, using a spectroscopic database of galaxy surveys, is essential to obtain robust cosmological parameters. These results also emphasize the need to construct accurate cluster catalogs at high redshifts using the ongoing and upcoming galaxy surveys and to perform similar analyses as introduced in this work.
As discussed above there is a degeneracy between and derived from the CMF at low redshift. We combine our 68% CL with those obtained from Planck18+BAO+Pantheon, to eliminate the degeneracy of the our likelihood and to remarkably shrink the uncertainties of the cosmological parameters. The joint analysis gives and .
5.4. Conclusion
In this paper, we derived the CMF and the cosmological parameters and using a mass-complete subsample of 756 clusters () obtained from the cluster catalog, which was constructed from the SDSS-DR13 spectroscopic data set. The advantages of using these catalogs are (i) we were able to identify clusters, assign membership, and determine cluster centers and redshifts with high accuracy from the high-quality SDSS spectroscopic data set; (ii) cluster membership was determined by the GalWeight technique, which has been shown to be ∼98% accurate in assigning cluster membership (Abdullah+18); (iii) the cluster masses were calculated individually using the virial theorem and corrected for the surface pressure term; (iv) is a low-redshift cluster catalog that eliminates the need to make any assumptions about evolution in clusters themselves and evolution in cosmological parameters; (v) is one of the largest available spectroscopic samples in order to be a fair representation of the cluster population.
Our CMF closely matches predictions from MultiDark Planck N-body simulations (snapshot hlist_0.91520.list12 , with z ∼ 0.09) for log(M) ≳ 13.9 . Assuming a flat ΛCDM cosmology, we used the publicly available HMFcalc13 code (Murray et al. 2013) to estimate HMFs for the Tinker08 model (Equation (5)). Then, using a standard χ2 procedure, we compared our cumulative mass function to HMFs to determine the best-fit mass function and constrain and . We measured and to be (systematic) and (systematic), with a cluster normalization relation of .
The cosmological constraints we derived are very competitive with those recently derived using both cluster abundance studies and other cosmological probes. In particular, our constraints on and are consistent with Planck18+BAO+Pantheon constraints. This remarkable consistency highlights the potential of using and its subsample , which are derived from the SDSS-DR13 spectroscopic data set utilizing the application of GalWeight to produce precision constraints on cosmological parameters. The joint analysis of our cluster data with Planck18+BAO+Pantheon gives and .
We would like to thank Steven Murray for making his HMFcalc calculator publicly available and also for his guidance in running it. We would also like to thank Jeremy Tinker, Brian Siana, and Benjamin Forrest for their useful comments and help as well as Shadab Alam for providing us with the chain of BOSS-DR12 BAO data. Finally, we appreciate the comments and suggestions of the reviewer, which improved this paper. This work is supported by the National Science Foundation through grant AST-1517863, by HST program number GO-15294, and by grant number 80NSSC17K0019 issued through the NASA Astrophysics Data Analysis Program (ADAP). Support for program number GO-15294 was provided by NASA through a grant from the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Incorporated, under NASA contract NAS5-26555.
Appendix A: Evolution
In this section, we discuss the evolution effect for a sample of clusters with a narrow redshift range between z1 and z2 with an average of . The HMF depends on the mass and redshift and is given by . We test the effect of evolution assuming an analytical model for the evolution of HMF and cosmological model with reasonable parameters. We then take the integral and compare the results with at z = 0.085.
Figure A1 shows the evolution of the cluster number density expected by Tinker08 for cosmological parameters =0.305 and = 0.825. In the left panel, we plot the HMF times M/ρc, where ρc is the critical density of the universe, to clarify the differences between the models at different redshifts. The right panel shows the scatter of models relative to the expectation at z = 0.085 (black line). As expected, the evolution of clusters with z < 0.085 is less than unity relative to that at z = 0.085 and the evolution of clusters with z > 0.085 is larger than unity relative to that at z = 0.085. The two dashed lines show the expectation in the redshift intervals of 0.0 ≤ z ≤ 0.125 (brown) and 0.045 ≤ z ≤ 0.125 (red). The plot indicates that the evolution is >15% for 0.0 ≤ z ≤ 0.125 for massive clusters, while it drops to <3% for 0.045 ≤ z ≤ 0.125.
Note that we do not neglect the effects of evolution. In other words, we do not assume that the HMF at z1 is (nearly) the same as at z2 (admittedly, there is a 10%–20% difference in the most massive M). Because we use ratios of these quantities, most of the cosmological parameters (e.g., ) are canceled for a sensible range (e.g., = 0.75–0.85). We also test other HMF approximations such as Despali HMF (Despali et al. 2016) and obtain the same conclusion. Therefore, we restrict our data (observed clusters) to 0.045 ≤ z ≤ 0.125 for which the evolution effect of the number density of clusters is minimal.
Appendix B: Redshift Threshold
In this section we investigate the choice of the redshift interval and the application on the selection function of our results of the fiducial analysis as shown in Figure B1. In the left panel, we fix the upper redshift threshold to 0.125 and decrease the lower redshift threshold from 0.075 to 0.045. The plots indicate that decreasing the lower redshift threshold does not affect our result for the fiducial sample (black ellipse). It also demonstrates that the evolution effect is unremarkable in this small redshift interval. The left panel also introduces the 68% CL of the sample (dashed brown ellipse), which gives (5% less than the fiducial value) and (1% greater than the fiducial value). The consistency between the results of and demonstrates that applying the selection function for z ≤ 0.125 does not affect the results of the fiducial analysis and is sufficient to correct for the volume incompleteness of . In the right panel, we fix the lower redshift threshold to 0.045 and increase the upper redshift threshold from 0.125 to 0.16. The plots indicate that increasing the upper redshift threshold significantly affects our constraints on and because applying the selection function to higher redshift (>0.125) affects the shape of the CMF by increasing the scatter and noise and overcorrecting the number of clusters at high redshifts.
Download figure:
Standard image High-resolution imageDownload figure:
Standard image High-resolution imageFootnotes
- 5
- 6
- 7
- 8
- 9
We use CMF for mass functions derived from observations and HMF for mass functions computed by theoretical models.
- 10
- 11
- 12
- 13