Southern Cosmology Survey I

Southern Cosmology Survey I: Optical Cluster Detections and Predictions for the Southern Common-Area Millimeter-Wave Experiments


We present first results from the Southern Cosmology Survey, a new multiwavelength survey of the southern sky coordinated with the Atacama Cosmology Telescope (ACT), a recently commissioned ground-based mm-band Cosmic Microwave Background experiment. This article presents a full analysis of archival optical multi-band imaging data covering an 8 square degree region near right ascension 23 hours and declination -55 degrees, obtained by the Blanco 4-m telescope and Mosaic-II camera in late 2005. We describe the pipeline we have developed to process this large data volume, obtain accurate photometric redshifts, and detect optical clusters. Our cluster finding process uses the combination of a matched spatial filter, photometric redshift probability distributions and richness estimation. We present photometric redshifts, richness estimates, luminosities, and masses for 8 new optically-selected clusters with mass greater than at redshifts out to 0.7. We also present estimates for the expected Sunyaev-Zel’dovich effect (SZE) signal from these clusters as specific predictions for upcoming observations by ACT, the South Pole Telescope and Atacama Pathfinder Experiment.

Subject headings:
cosmic microwave background — cosmology: observations — galaxies: distances and redshifts — galaxies: clusters: general — large-scale structure of universe — methods: data analysis

1. Introduction

The new generation of high-angular resolution Cosmic Microwave Background (CMB) ground-based experiments represented by the the Atacama Cosmology Telescope (ACT) (Kosowsky, 2006; Fowler et al., 2007) and the South Pole Telescope (SPT) (Ruhl et al., 2004) are currently targeting their observations in a common area in the southern sky that will ultimately cover several hundreds to thousands of square degrees. These experiments will provide a blind survey of the oldest light in the Universe at wavelengths of  mm and angular scales beyond the resolution limits of the WMAP and Planck satellites. At these arcminute angular scales, temperature fluctuations in the CMB are dominated by secondary effects arising from the formation of large-scale structure in the universe. One of the strongest effects is the imprint left by galaxy clusters though the Sunyaev Zel’dovich effect (SZE) (Sunyaev & Zeldovich, 1980) in which CMB photons suffer inverse Compton scattering by the hot intracluster gas. ACT and SPT are designed to detect the SZE, through its frequency-dependence: these experiments will measure temperature shifts of the CMB radiation corresponding to a decrement below and an increment above the “null” frequency around 220 GHz.

Much can be learned about the Universe from these surveys. First, accurate systematics-free maps will allow measurement of the primary power spectrum of temperature fluctuations at all scales on which they are the dominant contribution. Second, these data sets will result in a complete census of massive clusters to arbitrarily large distances, limited only by a minimum cluster mass set largely by the instrumental sensitivity and expected to be several (Ruhl et al., 2004; Sehgal et al., 2007). Thanks to the relatively clean selection function as well as the redshift independence of the SZE, the cluster sample, especially the evolution of the number density of clusters with redshift, will be quite sensitive to the growth of structure in the Universe offering a potentially powerful probe of dark energy (Carlstrom et al., 2002). Moreover, the SZE data in combination with optical, UV and X-ray observations can teach us a great deal about the detailed physics of cluster atmospheres and galaxy evolution in these dense environments.

Significant observing time and effort has been devoted to the development of techniques and the detection of galaxy clusters using large-area optical catalogs and X-ray observations. Several projects have taken advantage of large-area CCD imaging and have developed automated cluster detection schemes to produce large catalogs of clusters of galaxies (see Koester et al., 2007; Postman et al., 1996, 2001; Gal et al., 2000, 2003, 2009; Gladders & Yee, 2005, for example) which target the relative over-abundance of galaxies over a range of redshifts. Similarly, X-ray surveys such as the ROSAT All Sky Survey (Ebeling et al., 1998; Böhringer et al., 2001; Mullis et al., 2003) produced catalogs with hundreds of galaxy clusters, while pointed X-ray observations have discovered systems up to (Mullis et al., 2005).

With this article we inaugurate the Southern Cosmology Survey (SCS). This project, funded by the National Science Foundation under the Partnership in Research and Education (PIRE) program, is a multiwavelength (radio, mm-band, optical, UV, and X-ray) large area survey specifically coordinated with ACT observations of the southern sky. The goal of the SCS is to maximize the scientific return from the new ground-based CMB experiments and therefore focuses on specific observational studies relevant to this science, such as the selection function of galaxy clusters across wavebands, cluster mass determination, and the establishment a ”gold” sample of clusters for cosmology and galaxy evolution studies. Here we present results from an  deg optical imaging survey of the southern sky that overlaps the common SZ survey region. The purpose of this paper is twofold: (1) to present the details of our data reduction pipeline and analysis software and (2) identify new galaxy clusters, constrain their redshifts and masses, and predict their SZ signals. Photometric redshifts come from the 4-band imaging data, while our mass estimates are inferred from the optical luminosity () and richness () of the clusters, using relations calibrated by the Sloan Digital Sky Survey (SDSS). For the eight massive clusters, out of 38 identified (37 are new sources) in the survey area, we present positions, richness estimates, masses, and predictions for the integrated Compton -distortion of the SZE using empirical power-law relations based on -body simulations. Throughout this paper we assume a flat cosmology with  km s Mpc, and matter density .

2. Dataset and Methodology

Our study is based on the optical multi-band analysis of public data from the Blanco Cosmology Survey12 (BCS). This is a NOAO Large Survey Project that was awarded 45 nights over three years on the Blanco 4-m telescope at the Cerro Tololo InterAmerican Observatory (CTIO). The survey aims to image two 50 square-degree patches of the southern sky in four optical bands (griz) using the  pixel (0.36 deg) MOSAIC II camera in order to attain a sensitivity about an order of magnitude deeper than the SDSS imaging. The targeted areas are centered near declinations of 55 and 52 at right ascensions of 23 hr and 5 hr respectively; each of these patches lies within a larger common region of the southern sky that both ACT and SPT plan to survey. The BCS began in 2005 and has completed three years of data taking. For this paper we have processed and analyzed public data from the first year of the survey using an independent software pipeline developed by us at Rutgers University. The data we present was obtained on 15 nights of observing near the end of November and the beginning of December 2005 and cover an area of  deg in the 23 hr region. In Table 1 we show the observing dates, photometric conditions, lunar illumination and observed bands for the 19 tiles that make up the full extent of the observations analyzed here. In the following we describe the steps followed and tasks performed by the pipeline.

# of Tiles Obs
Date Photometric Lunar Illum
18 Nov 2005 yes 89.4% 1.0 1.0 3.0 3.0
19 Nov 2005 yes 0.0% 4.5 4.5 1.0 1.0
20 Nov 2005 no 0.0% 8.0 8.0 0.0 0.0
22 Nov 2005 no 0.0% 0.0 0.0 0.0 0.0
24 Nov 2005 yes 0.0% 2.0 2.0 0.0 0.0
26 Nov 2005 yes 0.0% 3.0 3.0 1.0 1.0
28 Nov 2005 yes 0.0% 0.0 0.0 2.7 2.7
30 Nov 2005 yes 0.0% 0.0 0.0 0.3 0.3
02 Dec 2005 yes 0.0% 0.5 0.5 0.0 0.0
04 Dec 2005 no 15.3% 0.0 0.0 0.0 0.0
05 Dec 2005 no 24.5% 0.0 0.0 1.3 1.7
06 Dec 2005 yes 35.1% 0.0 0.0 1.3 1.0
08 Dec 2005 yes 57.8% 0.0 0.0 2.3 2.7
10 Dec 2005 yes 78.4% 0.0 0.0 3.0 3.0
11 Dec 2005 yes 86.6% 0.0 0.0 3.0 3.3

Note. – Observing conditions during the 2005 run of the BCS, consisting of only the 19 tiles that were fully observed in all 4 bands in the 23hr region. Lunar illumination is the percentage at midnight local time in the direction toward the center of the region surveyed (R.A. 23 hours, decl. 55.2deg).

Table 12005 Observations in the 23hr Field

2.1. The Rutgers Southern Cosmology Pipeline

The Rutgers Southern Cosmology image analysis pipeline is written in Python with a scalable object-oriented design based on existing public astronomical software that is aimed at processing a large dataset in a repeatable, stable and semi-automated fashion.

The initial standard image processing steps for each observing night are handled by the IRAF/mscred (Valdes, 1998) procedures via the STScI/Pyraf interface. These include: overscan trim, bias correction, CCD cross-talk coefficients corrections as well as dome flat field correction. The pipeline also executes secondary CCD calibration steps on the science images which include the creation of super sky-flats, fringe patterns for and bands and their corresponding correction and removal. Additionally, procedures affecting the cosmetic appearance of the images, such as cosmic ray rejection, removal of saturated star bleed-trails, and generation of bad pixel masks, are automatically performed at this stage. Astrometric re-calibration and WCS plate solution are also handled automatically at this stage on the pre-stack science images using IRAF’s mscred/mscmatch task by matching several hundred sources within each tile with stars from the US Naval Observatory Catalog. We achieve good accurate astrometric solutions (the residual error in matched source positions was typically ) as tested using the overlapping regions between neighboring pointings. Photometric standard star fields were processed together with the normal science images and photometric zero points for each observing night were obtained using a few hundred standards from the Southern Hemisphere Standards Stars Catalog (Smith et al., 2007). Like these authors we use the AB magnitude system.

The survey strategy followed a predetermined observing pattern, which typically consisted of exposures of , , and in the and bands respectively with offsets of  arcmin (within each filter) intended to provide significant overlap between neighboring MOSAIC II tiles and fill in the gaps between CCDs chips. We used the overlapping regions between tiles to adjust the photometric zero points of non-photometric nights using matched sources from adjacent photometric tiles. This ensured a homogeneous photometric calibration across the full survey region with typical variations below  mags.

Image alignment, stacking and combination as well as catalog generation are performed at a secondary stage by the pipeline using association files, which describes a logical group of exposures and filters, created for each tile. Science images were mosaiced, aligned and median combined using SWarp (Bertin, 2006) to a plate scale of /pixel. Source detection and photometry measurements for the science catalogs were performed using SExtractor (Bertin & Arnouts, 1996) in dual-image mode in which sources were identified on the  band images using a detection threshold, while magnitudes were extracted at matching locations from all 4 bands.

Figure 1.— The mean lunar illumination fraction for each tile at time of observation in the 23hr field as a function of the -band magnitude limit for 90% completeness.

As our data set is composed of observations taken over several weeks under varying conditions during the 2005 campaign, we determined, for each tile individually, the -band magnitude limit at which the galaxy detection limit was complete to . To compute this limit we use the fact that the galaxy number counts follow a power-law function, which we fitted in the magnitude range in each tile and extrapolated to obtain the magnitude at which the galaxy number counts dropped by . We took this as the completeness limit for the tile. We found variations of roughly 1 magnitude on the limits among the 19 tiles and, in an attempt to understand this, we investigated a possible correlation with lunar illumination at the time of observation. Figure 1 shows a clear trend between the -band magnitude limit and the lunar illumination. We report a mean limit and we set a conservative magnitude limit of for our full catalog.

2.2. Photometric Redshifts

Figure 2.— Comparison between photometric and spectroscopic redshifts for 5000 galaxies in the SDSS DR6. The upper panel shows the comparison between spectroscopic and photometric redshifts obtained under the maximum likelihood assumption. The lower panel uses a Bayesian assumption with a custom empirical prior on galaxy brightness for the photometric redshifts. Symbols are color coded according to the best determined SED by BPZ. SEDs for E/S0 galaxies tend to dominate at higher redshifts as the SDSS is biased towards early-type galaxies which are the most luminous population at these redshifts.

¿From the multi-band photometry the pipeline computes photometric redshifts and redshift probability distributions for each object using the isophotal magnitudes, as defined by the i-band detection, and the BPZ code (Benítez, 2000). We use a magnitude-based empirical prior (Benitez, private communication) taken from the SDSS and HDF-N spectroscopic redshift distributions, which accounts for the tendency of fainter galaxies to be more likely found at higher redshifts (See Fig 4. from Benítez, 2000, as an example). Because the area covered by the available NOAO imaging does not include any publicly available spectroscopic redshift information for (NASA/IPAC Extragalactic Database, NED), we investigated the accuracy of our photometric redshift estimates using ancillary data. As the ability of BPZ to estimate photometric redshifts at fainter magnitudes () from multi-band photometry has been consistently established in the past using filter sets similar to ours here (Benítez et al., 2004; Mobasher et al., 2004; Cross et al., 2004, for example), we focused on the redshift accuracy at . To this end we extracted photometry from the DR6 SDSS for 5000 randomly-selected bright galaxies ( mag) with reliable spectroscopic redshifts, matching the depth and signal-to-noise ratio of our galaxy sample. We computed photometric redshifts for the SDSS spectroscopic sample in the same way as just described and compared the resulting values to the spectroscopic redshifts. We found, not surprisingly, that simply employing the maximum-likelihood (ML) condition is an ill-suited approach for redshifts below in the absence of a bandpass bluer than 3000Å as it largely over-estimates redshifts and produces an unacceptable number of catastrophic outliers. Recently, Niemack et al. (2008) have demonstrated how the addition of bluer bands using GALEX UV imaging greatly improves ML estimates and reduces the need for priors. On the other hand, Bayesian estimates give results with typical rms errors of and with almost no catastrophic outliers. In Fig. 2 we show the results of our comparison between ML and Bayesian photometric versus spectroscopic redshifts color-coded according to the spectral energy distribution (SED) determined by the BPZ code. It is clear from the figure that at higher redshifts the SDSS population is dominated by early types as these tend to be the most luminous objects. We also note (see fig. 2, lower panel) that on average our photometric redshifts tend to overpredict the true redshifts. The mean bias level for galaxies with E/S0s SEDs is largest around where it is on the order of . In table 2 we show the mean bias and standard deviation for three redshift intervals for as well as the standard defined as . In summary we are able to determine the redshifts for early-type galaxies to an accuracy better than 0.1 across the redshift range of the survey. This is encouraging since early-type galaxies are the predominant population in clusters of galaxies and good photometric redshift determination is essential for successful cluster finding, as we discuss in the next section.

In Figure 3 we show the photometric redshift distribution for all galaxies within our flux completeness limit, , as well as the filter responses for the survey13. Our distribution peaks around , which sets a conservative upper limit to the redshift at which we are able to detect optical clusters.

Figure 3.— The photometric redshift distribution of the galaxy sample used for finding clusters (i.e.: galaxies  mag).
-0.017 0.042 -0.015 0.038
-0.027 0.059 -0.020 0.047
-0.002 0.070 -0.001 0.048

Note. – The mean difference and standard deviation between spectroscopic and recovered photometric redshifts as well as for for all galaxies with SED determined to be E/S0s in three redshift ranges.

Table 2Photometric Redshift Simulations Statistics

2.3. Computing Overdensities and Finding Clusters

One of the main goals of the current SZE experiments is to define a mass-selected sample of galaxy clusters out to large redshifts. At long last this is beginning to happen (Staniszewski et al., 2008; Menanteau & Hughes, 2009), after a number of successful individual detections of the SZE in well-known optical or x-ray clusters (see Birkinshaw et al., 1991; Jones et al., 1993; Herbig et al., 1995; Reese et al., 2000; Gómez et al., 2003; Bonamente et al., 2006; Halverson et al., 2008, and references therein). If we want to understand the systematics of SZE surveys it is crucial to compare with cluster identifications using independent methods. In this section, we describe our effort to select clusters of galaxies from multi-wavelength optical imaging. There are several methodologies and a plethora of papers describing these techniques (Postman et al., 1996; Gladders & Yee, 2005; Koester et al., 2007; Eisenhardt et al., 2008, for example) but they all rely on the same well-known properties of galaxy clusters: a) early-type galaxies are the dominant population, b) cluster galaxies have very similar colors, and display tight color-magnitude relationships across several orders of magnitude in luminosity, and c) the surface number density of cluster galaxies falls off with distance from the center roughly as a power law . We search for clusters using a matched filter approach similar to the one described in Postman et al. (1996) and then define membership and estimate richness of the clusters using the MaxBCG prescription (Koester et al., 2007).

Our cluster finder method folds in the contributions from a cluster spatial profile filter function , a luminosity weight and the BPZ redshift probability distribution from each source to generate likelihood density maps (at pixel positions denoted by ) or a “filtered” galaxy catalog over the area covered by the survey as a function of redshift, namely,


Specifically we use a profile with the form

which is normalized as


and where is the typical cluster core radius and is the cutoff limit for the function. In our analysis we chose  kpc and . We also use a luminosity weight given by


where is the apparent magnitude corresponding to . This function is normalized as


where is the flux limit of the sample (), is the number of background galaxies and is the Schechter (1976) galaxy luminosity function. We use the parameters computed from Brown et al. (2007) for the evolving luminosity function of red galaxies, with a faint-end slope and between . For our estimation of we use the number counts from Yasuda et al. (2001). We generate likelihood density maps with a constant pixel scale of  arcmin at intervals between over the surveyed regions. In Figure 4 we show an example of a likelihood density map centered at on which we superpose outlines of the 19 tiles that define the region studied here.

Figure 4.— The computed likelihood density map image centered at and width over the 23hr field. Bright regions in the image represent denser areas. The red lines represent the area covered by each of the 19 tiles that comprise the area studied.

Cluster candidates are selected from the peaks of the likelihood density maps. In order to define peaks consistently we constructed noise maps by randomizing the positions of the input catalog to produce likelihood density maps following exactly the same procedure as described above. These maps represent the noise floor level above which we desire to detect clusters. We define our initial list of cluster candidates from peaks in the likelihood density maps, where is defined as the median value in the noise maps. Cluster candidates were checked for multiple detection in different likelihood maps. To avoid duplication we considered a system unique if detected in two adjacent redshift maps and with the same center (i.e., within 3).

2.4. Contamination and Completeness

We perfomed simulations to investigate our cluster selection function by estimating the contamination and recovery rates of our cluster finding technique. The lack of distance information in imaging surveys is the principal source of contamination as fluctuations in the projected two-dimensional galaxy distribution as well as random alignments of poor groups may result in false apparent overdensities. We explore this issue following the same methodology as used by Postman et al. (2002) and Gal et al. (2003) which rely on generating Monte Carlo representations of the galaxy sample with an angular two-point correlation function similar to that observed (Infante, 1994). As described in Postman et al. (2002) we implemented the Rayleigh-Lévy (RL) random walk process using Mandelbrot’s (1975) elegant fractal prescription to simulate galaxy positions on the sky, such that galaxy pairs are placed in a randomly chosen direction at distance drawn from the distribution:


where we chose and to match the observed galaxy distribution of our sample. In practice we generate simulated distributions by starting from a randomly selected location within the survey boundaries and generate positions following the RL random walk allowing up to seven galaxies to be drawn around this location. We then select a new center randomly and the process is repeated until we generate the same number of galaxies as in the observed sample. We then process the RL distribution to generate likelihood maps using the same procedure and parameters as for the real data and use these to investigate the rate of false detections as a function of estimated redshift. Since the RL distributions by construction do not explicitly include clusters, we assess the false positive cluster detection fraction by taking the ratio of detections in the simulations per area unit to the observed number of candidates in the real data. We find that at low redshift the false positive fraction is zero (there is virtually no contamination), while at redshifts of and the false positive fraction grows to values of 1% and 19%, respectively. We conclude that false positives are not an important source of spurious detections.

We investigate the selection function for our galaxy sample by simulating galaxy clusters of various richness and shapes at different redshifts and examining their recovery fractions. Specifically we generated clusters with random ellipticities uniformly between using an profile for the galaxy distribution (Lubin & Postman, 1996) with and  Mpc and using the luminosity function for red galaxies from Brown et al. (2007). These clusters are inserted 20 at a time in the observed catalogs at random positions and redshifts, but avoiding the locations where clusters were detected. In total we generate 10,000 simulated clusters with richness values, , of 15, 20, 25, 35, 50, 80 and 120 galaxies uniformly distributed in redshift.14 We process each realization using the same parameters as for the observed data and record the number of clusters recovered as a function of redshift and galaxy richness. In Figure 5 we show the results of this exercise where we plot the recovery fraction as a function of redshift for the seven cluster richnesses simulated. We conclude that for the rich clusters we are always nearly complete () for while for poorer clusters we only detect at best around .

Figure 5.— The cluster recovery fraction as a function of redshift as extracted from the simulations for clusters with and galaxies and a profile with slope and core radius  Mpc.

3. Cluster Properties

One of our main drivers in searching for clusters at optical wavelengths is to correlate them with SZ detections in the new blind SZ surveys. The signal to be detected in the mm-band experiments (i.e., the -distortion due to inverse Compton scattering) is related to the number of hot electrons in the intracluster medium ,and simulations have shown (see Motl et al., 2005; Nagai, 2006; Sehgal et al., 2007; Bhattacharya et al., 2007, for example) that the SZE signal is closely linked to cluster mass. Our analysis of the optical survey has provided positions, redshifts, and fluxes of galaxies, from which we infer the underlying cluster mass using scaling relations established by the SDSS, where cluster masses were determined from weak lensing.

Figure 6.— The composed color image (left panel) and color-magnitude relations (right) for X-ray cluster RXJ2325.65443 from Mullis et al. (2003). Red points represent galaxies classified as E/S0 by BPZ that satisfy the conditions to be cluster members as described in the main text. Black dots are non-member galaxies in a region near the cluster center.

3.1. Defining Cluster Membership

The current state-of-the-art mass tracers for clusters of galaxies using optically observed parameters (Johnston et al., 2007; Reyes et al., 2008) have been extracted from a sample of around optically-selected clusters from the SDSS MaxBCG catalog (Koester et al., 2007). In this paper, we apply these scaling relations to our cluster sample and obtain mass estimates from which we additionally predict SZ distortions. To be fully consistent with previous analyses, we define membership and all other relevant cluster observables following the same method as Reyes et al. (2008) and Koester et al. (2007).

We begin by examining each candidate-cluster peak in the density maps and select the brightest elliptical galaxy in the cluster (BCG), which is taken to be the initial center and redshift of the system. We then use galaxies defined as E or E/S0s (i.e., BPZ SED types 0 and 1 only) within a projected radius of Mpc and redshift interval to obtain a local color-magnitude relation (CMR) for each color combination, , and , and the cluster mean redshift, , for all cluster members, using a median sigma-clipping algorithm. We use these to determine , the number of galaxies within Mpc of the cluster center. Formally, we compute by including those galaxies within a projected 1Mpc from the cluster center that satisfy three conditions: (a) the galaxy must have the SED of an E, E/S0 according to BPZ, (b) it must have the appropriate color to be a cluster member (i.e., colors within of the local CMR for all color combinations) and (c) it must have the right luminosity (i.e., dimmer than the BCG and brighter than , where we use the corresponding absolute magnitude from Brown et al. (2007) redshifted with the elliptical SED template from BPZ). We designated cluster members according to the estimated cluster size , defined as the radius at which the cluster galaxy density is times the mean space density of galaxies in the present Universe. We estimated the scaled radius using the empirical relation from Hansen et al. (2005), Mpc which is derived from the SDSS and we assume it holds beyond for our higher redshift clusters.

In our analysis we use , , and  to scale cluster optical parameters with mass, following Reyes et al. (2008). The cluster richness, , is the number of E/S0 galaxies within with colors and luminosities that satisfy conditions (b) and (c) above. Similarly, is the total rest-frame integrated -band luminosity of all member galaxies included in in units of and is the rest-frame -band luminosity of the BCG.

In order to have reliable estimates of it is necessary to determine the galaxy background contamination and implement an appropriate background subtraction method. The lack of spectroscopic redshifts in our sample only allowed a statistical removal of unrelated field galaxies with similar colors and redshifts that were projected along the line of sight to each cluster. We assumed that the presence of a cluster at some redshift is independent of the field population seen in projection. Therefore we estimate the surface number density of ellipticals in an annulus surrounding the cluster (within ) with and the same colors as the cluster members. We measure this background contribution around the outskirts of each cluster and obtain a corrected value which is used to compute and then corresponding values of and . The magnitude of the correction ranges between %. We will refer to the corrected values hereafter. Moreover as our analysis is based on a magnitude-limited sample it is worth considering the fraction of lower luminosity galaxies that will fall below our magnitude limit () at higher redshifts. As in Menanteau & Hughes (2009), if we make the assumption that the cluster population is like that of the five clusters at in our sample (see Table 5) and evolves passively, then we can compute the fraction of missed for clusters at higher redshifts. We estimate that we are missing 4%, 13%, 29% and 38% of the cluster luminosity at and respectively. Given the uncertainty in this correction factor, we do not include it in our quoted luminosity values for the higher redshift clusters. This means our cluster masses, ), are underestimated by roughly these factors.

Figure 7.— The mass observable parameters for the 38 clusters in our sample. Filled circles represent the eight new clusters above the mass limit . The left panel shows the relation between the mass predictions using vs. as the main parameter; the agreement is generally to within a factor of two, except for the nearest systems. The central and right panel show the mass estimates as a function of the cluster redshift using and respectively, in units of solar mass.

3.2. Recovery of Known Clusters

The area covered by the BCS in the 23hr region is a virtual desert in terms of known clusters and spectroscopic redshifts for galaxies with . We found one catalogued X-ray selected cluster from the 160 deg ROSAT survey (Mullis et al., 2003): RXJ2325.65443 at with an X-ray flux of  erg cm s in the 0.5–2 keV band. Our cluster finding algorithm easily recovered this cluster and produced a photometric redshift estimate of . Figure 6 shows the color composite optical image of the cluster as well as the color magnitude diagrams for cluster members. Using the techniques described below we estimate the mass of RXJ2325.65443 to be which is just below the detectability limit of ACT and therefore will not be included in our SZE predictions. From the - (Evrard et al., 1996) and - (Arnaud & Evrard, 1999) relations we estimate a temperature of keV and an X-ray flux of erg cm s (0.5–2 keV band) which is in rough agreement (factor of 2) with the published value. Considering the entire sample of 38 clusters, RXJ2325.65443 is the closest and, based on our estimated masses from the optical properties and the - and - relations, has the highest predicted X-ray flux of the sample. Still, even this value is below the X-ray detection threshold of the ROSAT All Sky Survey (RASS) and so, as expected, we find that none of our new optical clusters are significant RASS X-ray sources.

3.3. Cluster Mass Estimation

Figure 8.— The gray circles show the total number of clusters with observed in the  deg region analyzed here. The thin and thick lines show the predicted number of clusters in the region for cluster masses greater than and respectively, using the WMAP 5-year best-fit cosmological model and the Jenkins et al. (2001) halo mass function.
1.54 7.77 0.41 0.67
1.64 7.92 0.43 0.66
Table 3Mass-richness power-law function Best fitting parameters

In this section we use the mass-richness relations based on  to weigh our new optical clusters. Both Johnston et al. (2007) and Reyes et al. (2008) found that the luminosity-mass and richness-mass relations were well described by power-law functions and they measured the normalizations and slopes in these relations using minimization. Their values are in broad agreement, but we will use the fits provided by Reyes et al. (2008) since they restrict their fits to clusters with and give results for two redshift bins: and . It is important to note that our clusters go well beyond and that we extrapolate the relation in the last redshift bin for clusters with . We investigated the two fitting functions based on and , (see section 5.2.1 from Reyes et al. 2008 for full details), which are described as:


where is the mass observational equivalent of 15 in units of , is in units of and the  dependence is normalized by its mean value. This is also described by a power-law function for a given value of  and :


The published best-fitting parameters for , and in Eqs. (7) and (8) as well as the values of , for Eqs. (9) and (10) are shown in Table 3. The combination of equations (7), (8) and (9), (10) for and respectively enable us to obtain mass estimates for any cluster with .

The left panel of Fig. 7 compares the masses obtained using and for our 38 optical clusters. The solid line denotes equality between the estimates while the two dashed lines show a factor of 2 range. Most clusters fall within this range, establishing a lower bound on our mass error. The two most significant outliers correspond to nearby clusters (see the right panel of Fig. 7) for which is evidently overpredicting the mass compared to . In one of these cases (which is RXJ2325.65443), the mass inferred by grossly overpredicts (by more than an order of magnitude) the estimated X-ray flux. The natural concentration of luminosities in clusters (i.e., that more luminous galaxies dominate the central regions of clusters) suggests that the mass estimate derived from should be more robust than that from as a function of redshift. For these reasons we use the cluster mass estimate derived from in predicting the SZE signal. Hereafter we refer to this mass as .

Table 4 displays the 8 clusters with in the 8 square-degree sky area covered by our analysis. These clusters also have above the same mass limit. In Figure 8 we show the cumulative number distribution of these clusters as a function of redshift, compared with expectations from simulations (H. Trac, private communication), using the mass function of dark matter haloes from Jenkins et al. (2001) and WMAP5 cosmology (Komatsu et al., 2008). The heavy solid curve shows expectations for a mass limit of , while the light solid curve indicates a mass limit of . This figure demonstrates that the number of clusters we observe is consistent with a mass limit in the range of . In Table 5 we display the properties for the remaining clusters with .

4. Predictions of the Sunyaev-Zel’dovich Effect Signal

Scaling relations between the integrated thermal SZE signal and cluster mass have emerged from current N-body plus hydrodynamic simulations of galaxy clusters. In this section we use these relations and our optically-derived mass estimates, , to predict the SZE signal to be observed by ACT and SPT when these experiments survey this sky region.

The SZE signal consists of small distortions to the CMB spectrum originating from inverse Compton scattering by electrons in the hot plasma of clusters of galaxies (Sunyaev & Zeldovich, 1980). Here we consider the thermal SZ flux , defined as the integrated Compton parameter,


where and are the number density and temperature of hot electrons in the cluster, is electron rest mass, is the speed of light, is the Thompson scattering cross section and is the angular diameter distance. The projected area, , and solid angle, , are related by the angular diameter distance as .

Self-similar scaling relations (Kaiser, 1986) predict that the virialized mass in clusters scales with the gas temperature as where for a flat cosmology. If clusters were isothermal, we would expect their SZE signal to scale like and therefore the self-similar SZ flux-mass scaling relation should have the shape where is the cluster mass fraction. However, clusters are not always isothermal or in hydrostatic equilibrium and physical processes like star formation and feedback will also contribute to deviations from self-similarity.

Prompted by the upcoming SZ surveys, several studies have characterized in detail the scaling relation using simple power-law fits to cosmological N-body simulations. Here we will use the recent fits from Sehgal et al. (2007), who included gas simulations employing small-scale cluster physics such as star formation and feedback into a large cosmological N-body simulation. ¿From their catalog of simulated clusters with , Sehgal et al. (2007) fit the relation


where and are the projected SZ Compton -parameter and mass respectively in a disk of radius . We use the best fit values and for all clusters regardless of redshift, as the redshift dependence of the fits is very weak (see Table 2 from Sehgal et al., 2007). The power-law index that they report is slightly steeper than ones quoted by some previous hydrodynamic simulations (see Hernández-Monteagudo et al., 2006; Nagai, 2006; Motl et al., 2005, for example), which Sehgal et al. (2007) attribute to their more realistic feedback prescription (i.e., including the effects of AGN and supernovae). We use the power-law model from equation 12 and the optical mass estimates for our 8 new massive clusters to predict the integrated signal to be observed by ACT and SPT. Table 4 gives the results in physical units of Mpc and observable ones of arcmin.

[Mpc] [arcmin]
SCSO J233430.2–543647.5 0.35
SCSO J233556.8–560602.3 0.52
SCSO J233425.6–542718.0 0.53
SCSO J232211.0–561847.4 0.61
SCSO J233731.7–560427.9 0.61
SCSO J234012.6–541907.2 0.62
SCSO J234004.9–544444.8 0.66
SCSO J232829.7–544255.4 0.68

Note. – Catalog of the optical clusters with mass estimates from the values. Each cluster’s redshift is the mean photometric redshift computed using the elliptical in the center of the cluster. The ID is based on the position of the BCG.

Table 4Optical Clusters with
SCSO J232540.2–544430.9 0.10
SCSO J232230.9–541608.3 0.12
SCSO J233000.4–543707.7 0.14
SCSO J232419.6–552548.9 0.18
SCSO J233106.9–555119.5 0.19
SCSO J233252.9–561454.1 0.20
SCSO J233215.5–544211.6 0.20
SCSO J233037.1–554338.8 0.20
SCSO J233550.6–552820.4 0.22
SCSO J232200.4–544459.7 0.27
SCSO J233522.6–553237.0 0.29
SCSO J233807.5–560304.9 0.30
SCSO J232956.0–560808.3 0.32
SCSO J232839.5–551353.8 0.32
SCSO J232633.6–550111.5 0.32
SCSO J233753.8–561147.6 0.33
SCSO J232156.4–541428.8 0.33
SCSO J233003.6–541426.7 0.33
SCSO J233231.4–540135.8 0.33
SCSO J233110.6–555213.5 0.39
SCSO J233618.3–555440.3 0.49
SCSO J233706.3–541903.8 0.51
SCSO J233816.9–555331.1 0.52
SCSO J232619.8–552308.8 0.52
SCSO J232215.9–555045.6 0.56
SCSO J232247.6–541110.1 0.57
SCSO J232342.3–551915.1 0.67
SCSO J233403.7–555250.7 0.71
SCSO J233951.1–551331.3 0.73
SCSO J233720.2–562115.1 0.75

Note. – Catalog of the optical clusters with mass estimates from the values. Each cluster’s redshift is the mean photometric redshift computed using the elliptical in the center of the cluster. The ID is based on the position of the BCG.

Table 5Optical Clusters with

5. Results and Conclusions

In this article we have laid out the techniques and methods for our analysis of a large multi-band optical survey with the Blanco telescope and Mosaic-II instrument under the aegis of the Southern Cosmology Survey. We have obtained sub-arcsecond astrometric precision and sufficient photometric accuracy for the estimation of redshifts to . Forty-two optical cluster candidates were identified from an area of the sky covering 8 deg; the richness and integrated galaxy luminosity of the clusters were measured. Based on correlations between these optical observables and cluster mass as established by SDSS cluster surveys (Johnston et al., 2007; Reyes et al., 2008) we provide mass estimates (in addition to positions and redshifts) for 8 new clusters whose inferred masses lie above . These clusters are all likely to be detected by ACT and SPT if these experiments reach their expected sensitivity levels. Although the uncertainties on the estimated mass and inferred SZE signal are large (factors of 2 or so), the accuracy of our cluster positions and redshifts are quite good and typical for 4-band imaging survey data. Moreover, it is worth noting that the relation varies quite a bit from simulation to simulation (differences of a factor of 2 in at fixed ), so our predicted masses are well within the uncertainty range of the latest theoretical predictions. In an effort to reduce the mass errors we have begun to estimate weak lensing masses from the current Blanco data (McInnes et al., 2009). We also note that significant additional areas of the BCS are now publicly available and will be presented in future publications.

The strength of the SCS is its multi-wavelength aspect and large sky area coverage. The 23hr region of the sky analyzed here has now been surveyed in the UV by GALEX and in the X-ray band by XMM-Newton; over the next several months as these data become available, we will incorporate them into the SCS. Correlation analyses of these multi-wavelength data should allow us to reduce the large errors on inferred cluster masses and study the cluster selection biases across wavebands. Secure confirmation of all candidates, as well as additional mass constraints, rest on follow-up optical spectroscopy, which we are pursuing at the Southern African Large Telescope and elsewhere. Finally, once ACT and SPT data become available, the rich optical data set analyzed here will be a valuable source for understanding and quantifying the impact of large scale structure on secondary anisotropies in the CMB.

FM would like to thank Txitxo Benitez for discussions and the use of BPZ priors and templates. JPH would like to thank Kent Patterson for help during the early phases of this project and the Aresty Center for Undergraduate Research at Rutgers for financial support. AK has been supported by NSF grant AST-0546035. LI thanks Centro de Astrofisica Fondap. We thank Armin Rest and Frank Valdes for helpful discussion on the image analysis and we also acknowledge David Spergel for several important insights and suggestions. We are thankful to Wayne Barkhouse, Mark Brodwin, Will High, Yen-Ting Lin and the entire BCS team for the careful planning and execution of the NOAO observations on which this paper is based. Financial support was provided by the NASA LTSA program (grant number NAG5-11714) and the National Science Foundation under the PIRE program (award number OISE-0530095).


  1. affiliation: Rutgers University, Department of Physics & Astronomy, 136 Frelinghuysen Rd, Piscataway, NJ 08854, USA
  2. affiliation: Rutgers University, Department of Physics & Astronomy, 136 Frelinghuysen Rd, Piscataway, NJ 08854, USA
  3. affiliation: ICREA & Institute of Space Sciences(CSIC-IEEC), Campus UAB, Bellaterra, Spain
  4. affiliation: Department of Astrophysical Sciences, Peyton Hall, Princeton University, Princeton, NJ 08544, USA
  5. affiliation: Max Planck Institut fur Astrophysik, Karl Schwarzschild Str.1,D-85741, Garching Munchen, Germany
  6. affiliation: ICREA & Institute of Space Sciences(CSIC-IEEC), Campus UAB, Bellaterra, Spain
  7. affiliation: Department of Astrophysical Sciences, Peyton Hall, Princeton University, Princeton, NJ 08544, USA
  8. affiliation: University of Pittsburgh, Physics & Astronomy Department, 100 Allen Hall, 3941 O’Hara Street, Pittsburgh, PA 15260, USA
  9. affiliation: University of KwaZulu-Natal, Astrophysics & Cosmology Research Unit, School of Mathematical Sciences, Durban, 4041, South Africa.
  10. affiliation: Pontificia Universidad Católica de Chile, Departamento de Astronomía, Santiago, Chile
  11. affiliation: University of Pennsylvania, Physics and Astronomy, 209 South 33rd Street, Philadelphia, PA 19104, USA
  14. These richness values fold in the flux limit of the survey and the membership prescription as described in the next section, so that they are roughly comparable to the values we give for the detected clusters.
  15. is the halo mass enclosed within , defined as a radius of spherical volume within which the mean density is 200 times the critical density.


  1. Arnaud, M., & Evrard, A. E. 1999, MNRAS, 305, 631
  2. Benítez, N. 2000, ApJ, 536, 571
  3. Benítez, N., et al. 2004, ApJS, 150, 1
  4. Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393
  5. Bertin, E., SWarp Resample and Coadd Software,
  6. Bhattacharya, S., Di Matteo, T., & Kosowsky, A. 2007, ArXiv e-prints, 710, arXiv:0710.5574
  7. Birkinshaw, M., Hughes, J. P., & Arnaud, K. A. 1991, ApJ, 379, 466
  8. Bonamente, M., Joy, M. K., LaRoque, S. J., Carlstrom, J. E., Reese, E. D., & Dawson, K. S. 2006, ApJ, 647, 25
  9. Böhringer, H., et al. 2001, A&A, 369, 826
  10. Brown, M. J. I., Dey, A., Jannuzi, B. T., Brand, K., Benson, A. J., Brodwin, M., Croton, D. J., & Eisenhardt, P. R. 2007, ApJ, 654, 858
  11. Carlstrom, J. E., Holder, G. P., & Reese, E. D. 2002, ARA&A, 40, 643
  12. Cross, N. J. G., et al. 2004, AJ, 128, 1990
  13. Ebeling, H., Edge, A. C., Bohringer, H., Allen, S. W., Crawford, C. S., Fabian, A. C., Voges, W., & Huchra, J. P. 1998, MNRAS, 301, 881
  14. Eisenhardt, P. R. M., et al. 2008, ArXiv e-prints, 804, arXiv:0804.4798
  15. Evrard, A. E., Metzler, C. A., & Navarro, J. F. 1996, ApJ, 469, 494
  16. Fowler, J. W., et al. 2007, Appl. Opt., 46, 3444
  17. Gal, R. R., de Carvalho, R. R., Odewahn, S. C., Djorgovski, S. G., & Margoniner, V. E. 2000, AJ, 119, 12
  18. Gal, R. R., de Carvalho, R. R., Lopes, P. A. A., Djorgovski, S. G., Brunner, R. J., Mahabal, A., & Odewahn, S. C. 2003, AJ, 125, 2064
  19. Gal, R. R., Lopes, P. A. A., de Carvalho, R. R., Kohl-Moreira, J. L., Capelato, H. V., & Djorgovski, S. G. 2009, AJ, 137, 2981
  20. Génova-Santos, R., et al. 2005, MNRAS, 363, 79
  21. Genova-Santos, R., et al. 2008, ArXiv e-prints, 804, arXiv:0804.0199
  22. Gladders, M. D., & Yee, H. K. C. 2005, ApJS, 157, 1
  23. Gómez, P. L., et al. 2003, Astronomical Society of the Pacific Conference Series, 301, 495
  24. Halverson, N. W., et al. 2008, ArXiv e-prints, 807, arXiv:0807.4208
  25. Hansen, S. M., McKay, T. A., Wechsler, R. H., Annis, J., Sheldon, E. S., & Kimball, A. 2005, ApJ, 633, 122
  26. Herbig, T., Lawrence, C. R., Readhead, A. C. S., & Gulkis, S. 1995, ApJ, 449, L5
  27. Hernández-Monteagudo, C., Genova-Santos, R., & Atrio-Barandela, F. 2004, ApJ, 613, L89
  28. Hernández-Monteagudo, C., Trac, H., Verde, L., & Jimenez, R. 2006, ApJ, 652, L1
  29. Infante, L. 1994, A&A, 282, 353
  30. Jenkins, A., Frenk, C. S., White, S. D. M., Colberg, J. M., Cole, S., Evrard, A. E., Couchman, H. M. P., & Yoshida, N. 2001, MNRAS, 321, 372
  31. Johnston, D. E., et al. 2007, ArXiv e-prints, 709, arXiv:0709.1159
  32. Jones, M., et al. 1993, Nature, 365, 320
  33. Kaiser, N. 1986, MNRAS, 222, 323
  34. Komatsu, E., et al. 2008, ArXiv e-prints, 803, arXiv:0803.0547
  35. Koester, B. P., et al. 2007, ApJ, 660, 239
  36. Kosowsky, A. 2006, New Astronomy Reviews, 50, 969
  37. Kravtsov, A. V., Berlind, A. A., Wechsler, R. H., Klypin, A. A., Gottlöber, S., Allgood, B., & Primack, J. R. 2004, ApJ, 609, 35
  38. Lubin, L. M., & Postman, M. 1996, AJ, 111, 1795
  39. Mandelbrot, B. 1975, Academie des Sciences Paris Comptes Rendus Serie Sciences Mathematiques, 280, 1551
  40. McInnes, R. N., Menanteau, F., Heavens, A. F., Hughes, J. P., Jimenez, R., Massey, R., Simon, P., & Taylor, A. N. 2009, arXiv:0903.4410
  41. Menanteau, F., & Hughes, J. P. 2009, ApJ, 694, L136
  42. Mobasher, B., et al. 2004, ApJ, 600, L167
  43. Motl, P. M., Hallman, E. J., Burns, J. O., & Norman, M. L. 2005, ApJ, 623, L63
  44. Mullis, C. R., et al. 2003, ApJ, 594, 154
  45. Mullis, C. R., Rosati, P., Lamer, G., Böhringer, H., Schwope, A., Schuecker, P., & Fassbender, R. 2005, ApJ, 623, L85
  46. Nagai, D. 2006, ApJ, 650, 538
  47. Niemack, M. D., Jimenez, R., Verde, L., Menanteau, F., Panter, B., & Spergel, D. 2008, ArXiv e-prints, 803, arXiv:0803.3221
  48. Postman, M., Lubin, L. M., Gunn, J. E., Oke, J. B., Hoessel, J. G., Schneider, D. P., & Christensen, J. A. 1996, AJ, 111, 615
  49. Postman, M., Lauer, T. R., Oegerle, W., & Donahue, M. 2002, ApJ, 579, 93
  50. Postman, M., Lubin, L. M., & Oke, J. B. 2001, AJ, 122, 1125
  51. Reese, E. D., et al. 2000, ApJ, 533, 38
  52. Reyes, R., Mandelbaum, R., Hirata, C., Bahcall, N., & Seljak, U. 2008, MNRAS, 390, 1157
  53. Ruhl, J., et al. 2004, Proc. SPIE, 5498, 11
  54. Schechter, P. 1976, ApJ, 203, 297
  55. Sehgal, N., Trac, H., Huffenberger, K., & Bode, P. 2007, ApJ, 664, 149
  56. Smith, J. A., Allam, S. S., Tucker, D. L., Stute, J. L., Rodgers, C. T., & Stoughton, C. 2007, AJ, submitted
  57. Staniszewski, Z., et al. 2008, arXiv:0810.1578
  58. Sunyaev, R. A., & Zeldovich, Y. B. 1980, ARA&A, 18, 537
  59. Valdes, F. G. 1998, Astronomical Data Analysis Software and Systems VII, 145, 53
  60. Yasuda, N., et al. 2001, AJ, 122, 1104
Comments 0
Request Comment
You are adding the first comment!
How to quickly get a good reply:
  • Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
  • Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
  • Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minumum 40 characters
Add comment
Loading ...
This is a comment super asjknd jkasnjk adsnkj
The feedback must be of minumum 40 characters
The feedback must be of minumum 40 characters

You are asking your first question!
How to quickly get a good answer:
  • Keep your question short and to the point
  • Check for grammar or spelling errors.
  • Phrase it like a question
Test description