THREE CANDIDATE CLUSTERS OF GALAXIES AT REDSHIFT ∼1.8: THE "MISSING LINK" BETWEEN PROTOCLUSTERS AND LOCAL CLUSTERS?

The search for clusters of galaxies at z > 1 has proven to be particularly difficult, mainly because of the reduced contrast between cluster members and field galaxies. High-z clusters have been searched for using a variety of techniques, including the so-called red sequence technique (Gladders & Yee 2000), and the use of both X-ray surveys (Rosati et al. 2002, and references therein), and infrared surveys (e.g., Elston et al. 2006). Recently, Eisenhardt et al. (2008) presented a sample of clusters of galaxies from the Spitzer Infrared Array Camera Shallow Survey. Of the 335 cluster candidates in their sample, 12 have confirmed spectroscopic redshift 1 ≲ z ≲ 1.4, and a few (less than 10) have photometric redshift (obtained using photometry in four bands, B_W, R, I, and 3.6 μm) in the range 1.5 ≲ z_phot ≲ 1.7. None of their candidates have either photometric or spectroscopic redshift higher than 1.7. A similar effort is being pursued by the SpARCS project (e.g., Wilson et al. 2009), which found no candidates at 1.4 < z < 2.2.

Radio galaxies (e.g., Hall et al. 2001; Best et al. 2003; Venemans et al. 2007) and quasars (e.g., Haas et al. 2009) have also been extensively used to find high-z clusters (see also Miley & De Breuck 2008, for a review), as they are associated with massive elliptical galaxies and clusters in the local universe. However, it is known that only low-power radio galaxies, most of which possess an "edge-darkened" radio morphology, namely FR I (Fanaroff & Riley 1974), are almost invariably located in clusters of galaxies (e.g., Zirbel 1996; Owen 1996, and references therein) and are also associated with the brightest cluster members (e.g., Best et al. 2007). For z > 0.5, an increasing fraction of the more powerful "edge-brightened" FR IIs are also found in rich environments (e.g., Prestage & Peacock 1988; Hill & Lilly 1991; Best 2000), but the fraction of FR II in rich clusters is still lower than that of FR Is. Furthermore, the host galaxies of FR Is are more similar to normal "inactive" ellipticals than those of FR IIs. Giant Lyα emitters, possibly representing protocluster regions, have been discovered around powerful radio galaxies at z > 2 (e.g., Steidel et al. 2000; Pentericci et al. 2001; Zirm et al. 2005), but the relationship with today's clusters is still a matter of debate, mainly because of the small sample of clusters known between z ∼ 1 and 2, and in particular at 1.5 < z < 2.

Here, we take advantage of the specific properties of the environment of FR Is to set the stage for filling that redshift gap. We utilize our newly selected sample of high-z low-luminosity radio galaxies (Chiaberge et al. 2009) (hereinafter Paper I) as "beacons" for clusters in the unexplored range of redshift. In this Letter, we present first results that returned three very promising cluster candidates, most likely located at z ∼ 1.7–2.0.

We use the following cosmological parameters (H₀ = 71 Km s⁻¹ Mpc⁻¹, Ω_M = 0.27, Ω_vac = 0.73; Hinshaw et al. 2009).

Details of the selection of the high-z (1 < z < 2) FR I sample are given in Paper I. Here, we briefly summarize the relevant steps.

We proceed in four steps, under the basic assumptions that (1) the FR I/FR II break in radio power per unit frequency (usually set at L_1.4 GHz ∼ 4 × 10³² erg s⁻¹ Hz⁻¹; Fanaroff & Riley 1974) does not change with redshift and that (2) the magnitude and color of the hosts of FR Is at 1 < z < 2 are similar to those of FR IIs within the same redshift bin, as in the case of local radio galaxies (e.g., Donzelli et al. 2007). Note that the photometric redshift is not a selection constraint.

• 1.  
  We select FIRST (Becker et al. 1995) radio sources in the COSMOS field (Scoville et al. 2007), whose 1.4 GHz fluxes correspond to the range of fluxes expected for FR Is at 1 < z < 2 (1 < F_1.4 < 13 mJy).
• 2.  
  Sources with FR II radio morphology are excluded.
• 3.  
  Bright (m_I > 22) galaxies are rejected since they are most likely low-z galaxies with faint radio emission (e.g., nearby starbursts).
• 4.  
  U-band dropouts are rejected as they are likely to be at z > 2.5.

We are left with a sample of 37 FR I candidates, all of which are identified in the COSMOS catalog (Capak et al. 2007).

To confirm that the FR I sources reside in a cluster environment at z > 1, spectroscopic redshifts of the target itself and eventually of a significant number of galaxies in the surrounding region of the sky would be ideal. Unfortunately, the z-COSMOS catalog (Lilly et al. 2007) does not include our sources.

In the absence of spectroscopic data, we take advantage of the recently published catalog of photometric redshifts obtained using 30 band photometry by Ilbert et al. (2009) to identify the best candidates. The catalog, which includes objects with I <25 in the deep Subaru area of the COSMOS field (Taniguchi et al. 2007), represents a significant improvement with respect to the list of Mobasher et al. (2007), providing a factor of ∼3–5 higher accuracy. The authors estimate the redshift accuracy to be σ_Δz ∼ 0.04 for galaxies of i ∼ 25 at z < 1.25. For 1.5 < z < 2.5 and i ∼ 25 (similar to our faintest FR I hosts), the accuracy is σ_Δz ∼ 0.19, still well suitable for the purpose of this Letter. Note that the spectral energy distribution of FR Is host galaxies is not dominated by strong emission lines,⁷ contrary to those of all other active galactic nuclei. The galaxy templates used by Ilbert et al. (2009) are therefore adequate for our analysis.

We search the Ilbert et al. (2009) catalog for the 37 sources presented in Paper I and we find all of them, but one. Four have 1.4 < z_phot < 1.7 and 5 have 1.7 ⩽ z_phot < 2.0. Here, we focus on the latter objects, since they are among the highest redshift cluster candidates known to date.

The five objects with 1.7 ⩽ z_phot < 2.0 are objects 03, 05, 22, 226, and 228 in the catalog of Paper I. We refer to these objects as "COSMOS-FR I nn." The Ilbert et al. (2009) catalog gives a "best fit" z_phot and a range of probable redshifts, at the 68% and 99% confidence level (see Table 1, where we also compare these photo-z with the previous estimates by Mobasher et al. 2007).

Two out of the five objects (COSMOS-FR I 03 and COSMOS-FR I 05) are detected in the X-ray band (as point sources) both in the XMM-COSMOS (Hasinger et al. 2007) and in the Chandra-COSMOS data (Elvis et al. 2009).

In this Letter, we focus on three objects, namely COSMOS-FR I 03, COSMOS-FR I 05, and COSMOS-FR I 226, which display significant overdensities of galaxies in the Mpc-scale environment.

In Figure 1 we show RGB images of a region of 180 × 180 square arcsec, corresponding to ∼1.5 × 1.5 Mpc² at a redshift z ∼ 1.8, centered on each of the three radio galaxy hosts. The images are obtained using Spitzer 3.6 μm, Subaru r- and B-band images for the R, G, and B channels, respectively. The white circles indicate objects with 1.6 < z_phot < 2.3 (see Section 4). The "beacon" FR I radio galaxies are at the center of each of the three fields. Note that the host galaxy of COSMOS-FR I 03, which in Paper I we reported as "non-detected" in the Hubble Space Telescope (HST)-COSMOS I-band images (Koekemoer et al. 2007), is in fact detected in the deeper Subaru-COSMOS i-band data.

Zoom In Zoom Out Reset image size

In order to find more support for the presence of clusters of galaxies around our "beacon" radio galaxies, we again search the Ilbert et al. (2009) catalog for sources that lie within a radius of 90'' from the FR I host. We restrict the search to a bin of photometric redshift 1.6 < z_phot < 2.3. The choice of that particular redshift range is motivated by the uncertainties on the estimate of z_phot outlined above. We find 58, 51, and 53 objects in the selected three-dimensional regions around the three FR Is, respectively.

In Figure 1, we mark with white circles the galaxies with 1.6 < z_phot < 2.3. The K-band magnitude of the marked objects is in the range K_s ∼ 20.5–25 (Ilbert et al. 2009). Note that the expected magnitude of a passively evolving L^⋆ galaxy at z ∼ 1.8 is K_s ∼ 21 (e.g., Strazzullo et al. 2006). The "central" radio galaxy host (always located at the center of the fields shown in the figures) has a distinct "red" color i − K_s ≳3 and the K_s magnitude is in the range 20.5–21.9 (Ilbert et al. 2009). In Paper I, we showed that our FR I hosts lie on the K–z relation for more powerful radio galaxy hosts with evolved stellar populations.

Note that a large number of "red" objects are not marked in the figure. A significant number of those objects (e.g., the galaxies ∼30'' NW of COSMOS-FR I 03) are indeed only bright in the infrared and do not make the optical selection cut to be included in the Ilbert et al. (2009) catalog. These red objects are as bright as or slightly fainter than the radio galaxy host in the K band. Their color is i − K_s ≳3, consistent with evolved stellar populations, and their magnitude is in the range K_s ∼ 20.5–22. A few other galaxies, which appear slightly less "red" than the radio galaxy host (e.g., the group in the southeast corner of the field of COSMOS-FR I 05, are simply foreground (z_phot ∼ 0.8–1) galaxies with an old stellar population.

We also perform the same search on six randomly selected fields inside the COSMOS survey area. For three-dimensional regions of the same size as those described above, we find 28–35 objects in the six control fields, with a mean value of 32 ±  2 and a median value of 32. This implies an overdensity factor of ∼1.7 for the Mpc environments of the three selected FR Is, statistically significant at the ∼4σ level, when the three candidate cluster fields are combined. We also checked that doubling the number of control fields does not alter the results.

In Figure 2, we show histograms representing the photometric redshift distribution of the sources within a 90'' radius around our radio galaxies (red histograms), compared with the average distribution of an identical area in the control fields. Clearly, this assumes that the "best" values of z_phot are accurate to better than the width of redshift bins, which therefore is chosen to be sufficiently large (Δz = 0.2, similar to the photo-z accuracy for faint objects at 1.5 < z < 2.5).

Zoom In Zoom Out Reset image size

The figure shows the overdensities of the cluster candidates with respect to the control fields, arguing for the presence of clusters. The space density enhancements appear to be centered between z ∼ 1.8–2.0, thus very close to the FR Is' best fit z_phot (indicated by the solid line in each plot). For the fields of COSMOS FRI 03 and COSMOS FRI 226 the distribution of the "extra" sources is spread across a larger redshift range (∼1.6–2). This is not entirely surprising, because of the uncertainty of the photo-z estimates.

We checked that the overdensities at z_phot ∼ 1.8 are not "artificially" generated by foreground clusters coupled with large photo-z errors at faint magnitudes. In fact, they are unrelated to the presence of any concentration of lower-z objects. On the other hand, in a few control fields we observe overdensities at z_phot ≲ 1, which do not artificially produce any overdensities at higher redshifts.

Finally, we note that the two other sources in the range of photo-z of interest (namely COSMOS-FR I 22 and COSMOS-FR I 228) do not show significant overdensities with respect to the control fields, at least in the range 1.6 < z_phot < 2.3. One possibility is that the photo-z for these objects is significantly offset.

We identified three cluster candidates with z_phot ∼ 1.8, according to the photometric redshift catalog of Ilbert et al. (2009). The possible existence of clusters is supported by the significant overdensity measured on the Mpc scale environment of the three selected radio galaxies. These first results show that the method of using low-luminosity radio galaxies as "beacons" for clusters is returning promising candidates. A systematic study of the whole sample of 37 radio galaxies is under way. That will allow us to determine the typical environment of these high-z sources and firmly establish whether they are more likely to be associated with clusters than FR IIs in the same range of redshifts.

We have submitted observational proposals to spectroscopically measure the redshift of the candidates. If confirmed, these would represent extremely important structures that would contribute to filling the existing gap between z ∼ 1.5 and z ∼ 2 in cluster studies.

A study of the fields with high-accuracy photometry and deep high angular resolution images (with HST) is mandatory to investigate the spatial distribution and the morphology of the cluster galaxies, to derive the photometric properties of a significant number of member candidates, and to assess the existence (or the absence) of the color-magnitude relation at these high redshifts.

In summary, follow-up observations of these fields from radio through X-rays will allow us to tackle some of the most important questions on the origin of clusters of galaxies and cluster members, to investigate the link between protoclusters at z > 2 and well-formed clusters in the nearby universe, and to study the effects of the evolution of the intracluster gas on the radio structures.

We acknowledge the effort of the entire COSMOS team. We thank A. Zirm for insightful comments on the manuscript.