A STARBURSTING PROTO-CLUSTER IN MAKING ASSOCIATED WITH A RADIO GALAXY AT z = 2.53 DISCOVERED BY Hα IMAGING

Source detection is performed on the original NB2315 image with an FWHM of PSF of 053 rather than the PSF matched image, using SExtractor (ver. 2.5.0; Bertin & Arnouts 1996), and photometry on all the images is conducted by the double-image mode of SExtractor. Color indices are defined with aperture magnitudes mag_aper, which are measured with a 15 diameter aperture, while the total magnitudes are defined with mag_auto magnitudes which are measured with an elliptical aperture (see also Kron 1980). Magnitude errors are estimated from 1σ sky noise at each object position taking into account slightly different exposure times and sensitivities, which are thus corrected for the difference in depth over the entire FoV. The regions near the edges of each image where the exposure time is shorter than 75% of the total in each combined image are masked. If an object is not detected in a broadband filter, a 2σ limiting magnitude is used to estimate the upper/lower limits in color indices or the upper limits in magnitudes. We make sure that the stellar colors are consistent with those of stellar templates given in Gunn & Stryker (1983) in order to check the zero points of magnitudes. From such comparison, the zero-point magnitudes in B, r', z', and J are corrected by ∼0.15 mag at most so that the stellar colors are in good agreement with those of the stellar atlas. Moreover, magnitudes are corrected for the Galactic absorption by the following magnitudes: A(B) = 0.62, A(r') = 0.41, A(z') = 0.23, A(J) = 0.14, A(H) = 0.09, A(K_s) = 0.06, and A(NB912) = 0.05, which are derived from the extinction law of Cardelli et al. (1989) on an assumption of R_V = 3.1 and E(B − V) = 0.155 based on Schlegel et al. (1998).

As a result, 1035 objects are detected in the NB2315 image at more than the 5σ level (i.e., 23.01 mag in the F1+F2 region). Among them, 754 galaxies are distinguished from 281 stars based on their B − z' and z' − K_s colors. This method of separating galaxies from stars is devised by Daddi et al. (2004) (see also Kong et al. 2006).

The detection completeness is investigated as follows:⁶ First, artificial objects with given magnitudes and Gaussian profiles with FWHM of 053 are distributed on the NB2315 image. Source detection is conducted in the same manner as described above, and then the fraction of detected artificial objects is calculated. The detection completeness is found to be more than 74 (83)% down to 23.0 (22.0) mag in NB2315.

Galaxies with a nebular emission line that happens to enter the NB2315 narrow-band filter should be observed to be much brighter in NB2315 magnitude compared with the K_s broadband magnitude, which samples primarily the underlying continuum flux density. We thus apply the following criteria to select such galaxies with an emission line at ∼2.313 μm:

$$\begin{equation} K_s-{\rm NB} > -2.5\log \left(1-\sqrt { f^2_{3\sigma,K_s}+f^2_{3\sigma,NB}}/f_{{\rm NB}} \right)+0.1, \end{equation} \tag{ 1 }$$

$$\begin{equation} K_s-{\rm NB} > 0.35, \end{equation} \tag{ 2 }$$

where f_3σ is the 3σ sky noise flux in each band and f_NB is the NB2315-band flux density (Figure 2). The continuum flux density is corrected for a color term by 0.1 mag as given in the rightmost side of Equation (1), because the effective wavelength of the K_s filter is different by 0.163 μm from that of the NB2315 filter (Figure 1). We note that the K_s–NB2315 colors of Coleman et al. (1980) templates redshifted to z = 2.53 are actually distributed around 0.1 mag. The second criterion of K_s–NB > 0.35 corresponds to the observed equivalent width larger than 79 Å (22 Å in the rest frame if a galaxy is located at z = 2.53) after being corrected for the color term of 0.1 mag. We select galaxies with color excesses in K_s–NB2315 greater than 3σ photometric errors. This means that an emission line with a flux larger than 2.6 × 10⁻¹⁷ erg s⁻¹ cm⁻² can be firmly detected. If the color excess is due to an Hα+[N ii] emission line pair at z = 2.53, the limiting line flux corresponds to L(Hα) = 1.1 × 10⁴² erg s⁻¹ and a dust-free SFR of 8.6 M_☉ yr⁻¹ (Kennicutt 1998), where the ratio of [N ii] to Hα([N ii]/Hα) is assumed to be 0.22 (Sobral et al. 2012). We also find 12 galaxies whose line fluxes are larger than this limit, but color excesses are smaller than the 3σ level due to their location near the edge of the images where the depth is shallower. These galaxies are not included in our emitter sample. As a result, we select 102 NB2315 emitters out of 754 galaxies over the ∼4 × 7 arcmin² area (Figure 2).

Zoom In Zoom Out Reset image size

Although we aim to select HAEs at z = 2.53, it is possible that some of the emission lines detected with the NB2315 filter are some other lines at different redshifts such as Hβ or [O iii] at z ∼ 3.6, [O ii] at z = 5.2, or Paschen series lines at z < 1.0. However, it is unlikely that the bulk of the NB2315 emitters originate from Hβ, [O iii], or [O ii] lines at z > 3.5, because the color excesses of the emitters are quite large as shown in Figure 2. Note that the large color excesses we see in K_s–NB2315 may correspond to unrealistically large emission line fluxes if they are all located at such high redshifts. Indeed, all but seven NB2315 emitters are detected in the B band, and 64 out of 68 HAEs classified below also have the detection. If a nebular line detected by the NB2315 filter would be either an [O iii] or an [O ii] line, the line luminosity of [O iii] (or [O ii]) is larger than 2.0(4.7) × 10⁴² erg s⁻¹, respectively. The luminosity functions of [O iii] at z = 0.84 (Ly et al. 2007) and [O ii] at z = 1.47 (Sobral et al. 2012) imply that no more than one [O iii] or [O ii] emitter with such a high luminosity can be included in our NB2315 emitter sample with limited field coverage, if any. Therefore, we believe that there is little contamination from the lines at z > 3.5 in our NB2315 emitter sample. Moreover, it should be noted that our NB2315 survey targets the plausible proto-cluster at z = 2.53 associated with the radio galaxy, and we must be preferentially detecting HAEs at z = 2.53 rather than other high-z emitters in the general field. However, as reported by Geach et al. (2008), who conducted H₂S1 narrow-band (λ_c = 2.121 μm) imaging with WFCAM on UKIRT to search for HAEs at z = 2.23 in the COSMOS field, other lower-z emission lines such as Paα and Paβ at z < 1 can be significant contaminations. In order to discriminate Hα from other lower-z contaminant lines, we set the following color selection criteria on the color–color diagram, r' − J versus J − K_s:

$$\begin{eqnarray} \begin{array}{lcl}J-K_s > 0.7(r^{\prime }-J)+0.05 & {\rm for} & r^{\prime }-J < 1.9 \\ J-K_s > 1.38 & {\rm for} & r^{\prime }-J > 1.9. \end{array} \end{eqnarray} \tag{ 3 }$$

This classification with r'JK_s colors is an analog of the BzK color selection of galaxies at 1.4 ≲ z ≲ 2.5 devised by Daddi et al. (2004), but it is optimized to identify galaxies at z ≳ 2.5. Figure 3 shows the color tracks of model spectra with various star formation histories and fixed formation redshift of z_f = 5 based on Bruzual & Charlot (2003), and four spectral energy distribution (SED) templates in Coleman et al. (1980), which are redshifted from z = 0.0 to z = 4.0. This figure suggests that the criteria enable us not only to select DRGs as galaxies with red colors (J − K_s(vega)>2.3), but also to isolate star-forming galaxies at z ≳ 2.5 from those at lower redshifts. Although no dust extinction is taken into account for the color tracks, the arrow in the figure shows the reddening vector of E(B − V) = 0.2 based on the Calzetti et al. (2000) extinction curve. This indicates that our selection criteria are relatively free from the dust reddening effect, as the vector is almost parallel to the slanted boundary line. Indeed, a group of emission line galaxies that are likely to be Hα emitters at z = 2.53 seems to be located separately from the bulk of galaxies at lower redshifts as shown in the right panel of Figure 3. Thus, our selection criteria work well to select Hα emitters at z = 2.53.

Zoom In Zoom Out Reset image size

We attempt to apply other color selection methods such as the original Bz'K_s selection (Daddi et al. 2004) and JHK_s selection (Kajisawa et al. 2006a; Kodama et al. 2007). However, the Bz'K_s selection is effective for selecting galaxies at 1.4 ≲ z ≲ 2.5, while we aim to pick out HAEs at z ∼ 2.5 among the NB2315 emitters. Since they are located just at the upper edge of the redshift range of the Bz'K selection, the selection of HAEs may well be quite incomplete. On the other hand, it is found that the JHK_s selection does not work well due to the shallowness of H-band data and the incompleteness in the selection of blue star-forming galaxies. Therefore, we judge that the classification with r'JK_s colors works best among them to identify HAEs at z = 2.53. Instead of the color selection, one may consider the use of photometric redshifts to discriminate among different emission lines at different redshifts. We also derive photometric redshifts of our NB2315-detected galaxies using six-band photometry of B, r', z', J, H, K_s and the EAZY code (Brammer et al. 2008). However, we find that it is difficult to clearly discriminate HAEs from other possible lines, especially those at z > 1.0, based on the photometric redshifts. In the end, we decide to use the r'JK_s diagram to identify HAEs. Consequently, we select 68 HAEs in total (Figure 3).

DRG is a class of galaxy populations at 2 ≲ z ≲ 4 with red colors of J − K_s > 1.38 or J − K_s(vega)>2.3 (Franx et al. 2003) which consist of dusty starburst galaxies and quiescent galaxies. To select such red galaxies in this region, the DRG color selection is applied to the K_s-selected catalog, which is made with the same procedures as those described in Section 3.1, except that source detection is performed on the K_s image. The K_s-selected catalog is used only for the selection of DRGs, since such a catalog enables us to select DRGs more completely down to fainter magnitudes than the NB2315-limiting magnitude. The catalog includes 1340 objects brighter than 5σ limiting magnitude in K_s. Among them, 1002 objects are classified as galaxies, while the remaining 338 objects are classified as stars. Consequently, 42 DRGs are selected from the K_s-selected catalog. It is expected that red galaxies associated with the proto-cluster at z = 2.53 tend to be seen as DRGs, although DRGs in general can be located in a much wider redshift range of 2 ≲ z ≲ 4.

Figure 4 shows the number densities of DRGs and K_s-selected galaxies which are corrected for detection incompleteness estimated in the same manner as in Section 3.1. For comparison with other studies, we also plot the results of the general blank field survey, the MOIRCS Deep Survey (MODS; Kajisawa et al. 2006b, 2011) where the deep and wide surveys have FoVs of 28.2 and 103.3 arcmin², respectively. The number densities of DRGs as well as K_s-selected galaxies in the USS1558 proto-cluster do not show strong overdensities compared to those in the MODS, although some excesses by a factor of a few are seen at several bins. Kodama et al. (2007) found that this region shows an overdensity of bright DRGs compared to the GOODS-S field (Giavalisco et al. 2004). Figure 4 indeed shows that there is an excess of bright DRGs with K_s ≲ 21.5 in the USS 1558-003 proto-cluster that are almost absent in the general blank field. The difference in number densities of DRGs between the proto-cluster and the general field is clearer in the cumulative densities, showing an excess at ∼2σ level in the bin with K_s < 21.5. Moreover, the spatial distribution of DRGs is not homogeneous but is rather clustered around the radio galaxy (see Section 4.1). The fact that there is no significant excess of faint DRGs in the proto-cluster region may suggest that most of the faint member galaxies in this proto-cluster have bluer colors, probably due to ongoing star formation activities, and that passive red galaxies would be gradually emerging later on. Although we do not yet have any spectroscopic confirmation of membership of the DRGs except for the radio galaxy itself, it is likely that the region around the USS 1558-003 radio galaxy is a proto-cluster at z = 2.53.

Zoom In Zoom Out Reset image size

Figure 5 shows the spatial distribution of our 68 HAE candidates associated with the USS 1558-003 radio galaxy at z = 2.53 located at (0,0) in the figure. The 42 DRGs are also plotted with red filled circles. We find that there are three outstanding regions where HAEs and/or DRGs are strongly clustered, which we hereafter call clump-1, clump-2, and clump-3. In fact, as summarized in Table 2, the number densities of HAEs and DRGs in these three clumps are all higher than those of averaged values across the observed fields by factors of two to five for HAEs and two to three for DRGs. Clump-1 is in the vicinity of the radio galaxy, and it contains both HAEs and DRGs around the radio galaxy. If we assume that the central dominant radio galaxy grows to a cD galaxy in the future, this region may correspond to the central part of the cluster. The fact that DRGs are also clustered in this clump may also suggest that this part of the proto-cluster is the oldest. Clump-2 is the most conspicuous, densest association of the HAEs and DRGs, and it is located at ∼32 (about 1.5 Mpc in physical scale) away from the radio galaxy to the southwest. Clump-3 is a smaller group of HAEs located between clump-1 and clump-2.

Zoom In Zoom Out Reset image size

It is interesting to note that the HAEs are more strongly clustered toward the southwest clumps (clump-2 and clump-3) rather than in the immediate surrounding region around the radio galaxy (clump-1). They constitute a part of the large-scale structure hosting the radio galaxy and clump-1, and they would all merge together in the near future to form a more massive single cluster around the radio galaxy. It is obvious that this proto-cluster region is not yet relaxed and is just in the process of galaxy assembly from the surrounding regions. We note that such distribution of the Hα emitters is similar to that of DRGs reported by Kodama et al. (2007) based on NTT/SOFI imaging data. In fact, we also confirm that the DRGs tend to be located in and along the structures traced by the Hα emitters, based on our deeper MOIRCS data (Figure 5). Furthermore, we find that these DRGs also meet the BzK criteria which select galaxies primarily at 1.4 ≲ z ≲ 2.5, suggesting that a significant number of these DRGs are likely to be physically associated with the proto-cluster hosting the radio galaxy at z = 2.53.

The three clumps in the 1558-003 proto-cluster host a large number of HAEs. The surface number density of HAEs in the three clumps is ∼37 ± 13 times larger than those in the redshift slice at z = 2.2 in the GOODS-North field (Hα emission survey with NB209 narrow-band filter; Tadaki et al. 2011). Similarly, we also find that it seems that the surface number density is ∼14 ± 3 and 17 ± 4 times larger than those in the redshift slices of z = 2.2 and 2.53 in the SXDS field (Hα emission survey with NB209 and NB2315; K.-I. Tadaki et al., in preparation).

Tanaka et al. (2011) conducted an HAE survey in the field around the 4C 23.56 radio galaxy at z = 2.48 with a CO narrow-band filter on MOIRCS; they found 11 HAE candidates to flux down to ∼7.5 × 10⁻¹⁷ erg s⁻¹ cm⁻¹ and rest-frame EW > 50 Å over a 23.6 arcmin² area, which is similar to our survey area. It is found that HAEs are distributed on the east side of the radio galaxy 4C 23.56, and there is a clump of HAEs ∼2 Mpc (in comoving scale) away from the radio galaxy. Such an offset distribution of HAEs from the radio galaxy is similar to that in our USS 1558-003 field, although our Hα emitters are more strongly clustered than the 4C 23.56 field. However, note that the EW cut and the limiting flux used for HAE selection are slightly different in our survey than in Tanaka et al. (2011). Our survey enables us to sample HAEs with fainter line fluxes and smaller EWs. If we apply the same EW cut and the limiting flux as in Tanaka et al. (2011) to our sample in the USS 1558-003 field, the number of HAEs reduces to 27, but it is still considerably larger than that in the 4C 23.56 field. The discovery of clumps of HAEs in these proto-clusters around the radio galaxy at z ∼ 2.5 clearly indicates that we are witnessing the process of mass assembly of clusters at their early stage when galaxies are vigorously assembling to form dense cluster cores while they are actively forming stars.

Moreover, Hatch et al. (2011) reported the studies for two proto-clusters around the radio galaxies MRC 1138-262 at z = 2.16 and 4C+10.48 at z = 2.35. These studies suggested that proto-clusters at z ≳ 2 tend to show a statistical excess of HAEs compared to the general fields at similar redshifts. However, they found that Hα emitters in the 4C+10.48 region are not strongly clustered, while those in the MRC 1138-262 region are clustered (see Kurk et al. 2004b). Moreover, both proto-clusters show inhomogeneous spatial distribution of HAEs around each radio galaxy, and most of the Hα emitters are distributed only on one side of the radio galaxies, similar to 4C 23.56 (Tanaka et al. 2011) and USS 1558-003 (this study).

Another interesting result we find is that the radio galaxy shows an extremely extended Hα emission spatially as shown in Figure 6. In the NB2315 image, we can note that the structure of the Hα emission is stretched in the northeast–southwest direction. The size of the Hα emission is ∼45, which corresponds to ∼36 kpc in physical scale. We discuss the extended Hα emission of radio galaxy itself in Section 5.2.

Zoom In Zoom Out Reset image size

The color–magnitude diagram is a powerful tool for investigating the properties of galaxies in clusters or proto-clusters. It is well known that galaxy clusters at low redshifts are dominated by quiescent galaxies which make up the red sequence on the color–magnitude diagram. The tight sequence of red galaxies is one of the prominent features seen in galaxy clusters. However, when and how do such red quiescent galaxies form and evolve? The answer to this question still remains unclear, although there is some evidence suggesting that the blue star-forming galaxies in the early universe evolve and become red quiescent galaxies in high density regions during the redshift interval of 2–3 (e.g., Kajisawa et al. 2006a; Kodama et al. 2007; Kriek et al. 2008; Doherty et al. 2010; Gobat et al. 2011). To address this, it is essential to investigate further the color–magnitude diagram for galaxy cluster and proto-cluster at high redshifts when the clusters are vigorously evolving.

Figure 7 shows a color–magnitude diagram of J − K_s versus K_s for the galaxies in the observed field toward the proto-cluster USS 1558-003. The HAEs associated with the proto-cluster are plotted by open circles. As presented by Kodama et al. (2007), we may be able to recognize a sequence of DRGs at J − K_s ∼ 1.5, which corresponds to the color of quiescent galaxies with formation redshift of 3 ≲ z_f < 4 (Kodama et al. 1998), as indicated by the tickmarks in the figure. It is not surprising that the majority of the HAEs are located on the bluer side of the diagram. However, it is intriguing that some HAEs have very red colors satisfying the DRG criterion and constitute mainly the faint end of the red sequence. We also note that there is a significant color scatter among the red HAEs. Similarly, red HAEs and dusty star-forming galaxies have also been recognized by recent observations of clusters at lower redshifts (Geach et al. 2006; Koyama et al. 2010, 2011). They tend to be dusty star-forming galaxies and are considered to be in the transition phase from active star-forming galaxies to passive quiescent galaxies, probably under the influence of some environmental effects, because these are preferentially found in the medium density regions of galaxy clusters at intermediate redshifts (z ≲ 0.8) when and where we see a sharp transition in the distribution of galaxy colors from blue to red (Tanaka et al. 2005; Koyama et al. 2008). Therefore, it is suggested that the faint end of the red sequence is just being built by those transitional galaxies recognized as the red HAEs.

Zoom In Zoom Out Reset image size

For a galaxy at z = 2.53, its K_s-band luminosity is still a good proxy for the stellar mass. However, the mass-to-luminosity ratio (M_⋆/$L_{K_s}$) is dependent on SED, and it should be corrected to get a more precise stellar mass from the K_s-band luminosity. We use a simple method to estimate stellar masses of the HAEs based on their K_s-band magnitudes and J − K_s colors. The relationship between J − K_s colors and M_⋆/$L_{K_s}$ is approximately estimated using the stellar population synthesis model of Kodama et al. (1998, 1999), where we constructed a sequence of models with varying bulge-to-disk ratios. We note here that because of the SED degeneracy among age, metallicity, and dust extinction, the relationship between J − K_s and M_⋆/$L_{K_s}$ is relatively insensitive to the detailed modeling of stellar populations except for the effect of initial mass function (IMF) variation. We here assume the Salpeter (1955) IMF, and the stellar masses in the models are scaled accordingly. From this experiment, we establish the following relationship between stellar mass (M_⋆), K_s magnitude, and J − K_s color:

$$\begin{equation} \log _{10}(M_\star /10^{11}M_\odot) = -0.4(K_s-K_{11}) + \Delta \log _{10} M, \end{equation} \tag{ 4 }$$

where K₁₁ is the K_s-band total magnitude corresponding to the stellar mass of 10¹¹ M_☉ for a passively evolving galaxy formed at z = 5 and observed at z = 2.53, which is estimated to 22.41. The mass-to-luminosity ratio is corrected for depending on the J − K_s color by

$$\begin{equation} \Delta \log _{10} M = 0.12\hbox{--}1.84\cdot \exp (-1.35\cdot (J - K_s)). \end{equation} \tag{ 5 }$$

The isostellar mass curves of 1 × 10¹¹ M_☉ and 1 × 10¹⁰ M_☉ are shown in Figure 7. All the HAEs but the radio galaxy have stellar masses of ≲ 10¹¹ M_☉. Among them, the bluest HAEs with J − K_s < 0.7 are less massive galaxies with stellar masses of ≲ 10¹⁰ M_☉. This gives a dichotomy in the distribution of star-forming galaxies on the stellar mass–color plane, separated at ∼10¹⁰ M_☉, below which the emitters are the bluest and hence the youngest or the most actively star forming.

It should also be noted that the brightest HAE is the radio galaxy, and it is located at the brightest end of the red sequence. Note that similar results are also found in other proto-clusters at z = 2–2.5 (Tanaka et al. 2011; Hatch et al. 2011). Its stellar mass is estimated to be ∼10¹² M_☉, which is typical for HzRGs (e.g., Rocca-Volmerange et al. 2004; Seymour et al. 2007; Hatch et al. 2009). Since the radio galaxy hosts an AGN, it is not straightforward to quantify its SFR from the emission line strength, as it is affected by photoionization by AGNs (Section 5.2).

Hα luminosity is a good indicator of SFR in the individual galaxy because the Hα emission line is originated from the photoionization by young O and B stars; thus, it probes the instantaneous star formation activity on a timescale of <20 Myr. The advantage of using the Hα line to estimate SFR is that it is well calibrated and widely used in studies of galaxies in both the local and high-z universe. Moreover, it is relatively insensitive to dust extinction compared to [O ii] and UV luminosities.

Hα luminosity is estimated from the emission line flux entering into the narrow-band NB2315, which in most cases is the combination of Hα and [N ii] lines. Since the emission lines do not enter the broadband filter for the HAEs at z = 2.53, as shown in Figure 1, the flux densities in the narrow-band and the broadband filters are expressed as $f_{{\rm NB}}=f_{\rm continuum}+F({\rm H\alpha + [N\,{\scriptstyle II}]})/\Delta _{{\rm NB}}$ and f_BB = f_continuum; thus, the emission line flux is

$$\begin{equation} F({\rm H\alpha + [N\,{\scriptstyle II}]}) = (f_{{\rm NB}}-f_{{\rm BB}})\cdot \Delta _{{\rm NB}}, \end{equation} \tag{ 6 }$$

where Δ_NB = 271 Å is FWHMs of the filters and f_continuum is the flux density of the continuum level of a spectrum. It should be noted that the fluxes are derived under the assumption that the emission lines are shifted into the center of the narrow-band filter. Since the response curve of the filter is not perfectly top hat (Figure 1), the fluxes can be underestimated if the emission lines are not located in the center of the filter. In that case, the SFR derived from the flux can be a lower limit. The contribution of [N ii] is removed from the line flux, F(Hα+[N ii]) by assuming the relation between the ratio of [N ii]/Hα and the rest-frame equivalent width of EW₀(Hα+[N ii]) given by Sobral et al. (2012).

We derive SFR from Hα flux using the Kennicutt (1998) relation, where Hα flux is corrected for dust extinction using the SFR-dependent calibration given by Garn et al. (2010). The dust extinction correction is important even for the Hα line, which is located in the rest-frame optical wavelength (λ6563). It is often assumed that the amount of dust extinction, A(Hα), is unity in many papers. However, we adopt the Garn et al. (2010) relation to more realistically estimate the intrinsic Hα luminosity. The specific SFR is also calculated by dividing the SFR by the stellar mass estimated in Section 4.2.

Figure 8 shows SFRs of the HAEs as a function of stellar mass, where dust-uncorrected and dust-corrected SFRs (SFR_corr) are shown in separate panels. There is a correlation that more massive galaxies tend to have higher SFRs, but the dependence is weak because there is significant scatter in SFR for a given stellar mass. A large fraction of the HAEs have high SFR_corr larger than 100 M_☉ yr⁻¹, meaning that many star-forming galaxies in this proto-cluster are in the starburst phase. The object with the stellar mass of ∼10¹² M_☉ is the radio galaxy USS 1558-003, and it apparently shows a considerably high SFR. However, this value must be wrong because its Hα line intensity is severely contaminated by the AGN component. In the right panel of Figure 8, we plot the main sequence of star-forming galaxies at z ∼ 2 (Daddi et al. 2007). In Rodighiero et al. (2011), the galaxies whose SFRs are more than four times larger than those on the main sequence for a given stellar were defined as starburst galaxies. Following this definition, a large fraction of galaxies with <10¹⁰ M_☉ are starburst galaxies, while the massive galaxies are more or less located around the main sequence. This is somewhat expected considering the fact that less massive galaxies show the bluest colors on the color–magnitude diagram (Figure 7). In contrast, none of the massive galaxies show such blue colors. It should be noted, however, that the red HAEs with stellar masses larger than 10¹⁰ M_☉ are subject to a large amount of dust attenuation; thus, they are actually in the dusty starburst phase, as we will discuss later in Section 5.1. Figure 9 shows specific SFRs of the HAEs plotted as a function of stellar mass. This indicates that specific SFR and stellar mass are anti-correlated, as seen in many other studies. Massive galaxies have lower SFRs for a given stellar mass compared to less massive galaxies, indicating that star-forming activities have been somewhat truncated and weakened in massive galaxies.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In Figures 8 and 9, HAEs in the three clumps are marked with open circles, squares, and triangles. It appears that the HAEs in clump-2 (i.e., the southwest clump ∼1.5 Mpc away from the radio galaxy) tend to have higher SFRs than those in other regions. To investigate more quantitatively any environmental dependence in the properties of HAEs, we show in Figure 10 the differential histograms of SFRs, specific SFRs, and stellar masses of the HAEs divided by the regions in the left panels, and their cumulative and normalized number counts in the right panels.

Zoom In Zoom Out Reset image size

There is no large difference in the distribution of all these physical quantities among different regions in general, except for the weak tendency that SFRs in clump-2 may be slightly higher. In fact, the median SFR of HAE in clump-2 is 152 M_☉ yr⁻¹ which is compared to 97 M_☉ yr⁻¹ in the clump-1, 99 M_☉ yr⁻¹ in the clump-3, and 106 M_☉ yr⁻¹ in the other regions. However, the Kolmogorov–Smirnov test that is applied on the top right panel suggests that the difference does not show high enough statistical significance. The probability of accepting the hypothesis that the SFRs in these clumps have different distributions is less than 2σ at most. Therefore, we may conclude that star formation activity at this high redshift (∼2.5) is high everywhere regardless of environment, not only in the low density outskirts but also in the dense clumps of the proto-cluster. Because we are targeting a proto-cluster and its surrounding region, it is yet unknown whether such absence of environmental dependence at z ∼ 2.5 is universal from one proto-cluster to another, as well as in the general low density fields which are not in the vicinity of proto-clusters.

Furthermore, it is interesting to know how much star formation activity and stellar mass are confined in this proto-cluster region as a whole, as most of the structures and the HAEs are likely to assemble to form a single richer cluster at later times. The integrated stellar mass and SFR of the HAEs in each clump and those in the entire region are shown in Table 3. The uncertainties in the integrated stellar mass and SFR originated from photometric errors are small enough to not greatly influence the discussions below. Also, in this analysis we exclude the most massive radio galaxy in clump-1 (∼1 × 10¹² M_☉), since its Hα line flux is severely contaminated by the AGN component. The integrated stellar mass of DRGs associated with the proto-cluster is derived by subtracting the component of DRGs outside the cluster but projected along the line of slight. Here we use the number count of DRGs in the MODS general field survey (Kajisawa et al. 2006b, 2011). We note that the total stellar masses estimated here should be taken as the lower limits because we miss less massive galaxies fainter than our observational limits. All clumps but clump-3 have a total stellar mass of ∼10¹² M_☉, while clump-3 has a factor of ∼10 lower mass of ∼10¹¹ M_☉. The integrated specific SFR of clump-1 is slightly lower than those of clump-2 and clump-3, by factors of 1.6 and 3.9, respectively, which might suggest that the star formation activity per unit of stellar mass is slightly lower in the closest vicinity of the radio galaxy which may grow to the center of the cluster in the future and that the age of the system is slightly older than the other clumps. However, the statistics are again too poor for us to say anything conclusive yet.

As described in Section 4.2, there are some HAEs with colors redder than J − K_s = 1.38. Figure 11 shows the distribution of the HAEs and DRGs, where the red HAEs with J − K_s > 1.38 are marked by red open circles. It seems that the red HAEs are clustered around the radio galaxy (clump-1) and in the southwest clumps (clump-2 and clump-3). Such concentration of the red emitters in high density regions is qualitatively different from what is seen in the lower redshift cluster at z ∼ 0.8 where red HAEs prefer medium density regions, as mentioned in Section 4.2. This may indicate that, in this proto-cluster at z ∼ 2.5, the transition of galaxies is occurring in the densest environment rather than in the outskirts. We note that most of the red HAEs are less massive than 10¹¹ M_☉ as shown in Figure 7. If the dusty starburst phase does not last so long, these red HAEs would stay on the relatively faint end of the red sequence even after they quench star formation. If the DRGs without Hα emission lines with stellar masses of a few × 10¹¹ M_☉ are really proto-cluster members, the active phase of these massive quiescent galaxies should have been seen at z ≳ 3. Some of their progenitors would be like the populations of LBGs or LAEs, which are in fact often identified in overdense regions (proto-clusters) at z ≳ 3 (e.g., Steidel et al. 1998; Miley et al. 2004; Venemans et al. 2005, 2007; Ouchi et al. 2005; Kuiper et al. 2011).

Zoom In Zoom Out Reset image size

On the other hand, as another possibility, the red HAEs can be quiescent galaxies with AGN activity in the galaxy core, rather than dusty starburst galaxies. Even in that case, the fact that the red HAEs are clustered in the high density region is very interesting. This is because the considered situation suggests that AGN activity is enhanced in the core regions of the clumps; thus, AGNs must significantly influence the evolution of high-z galaxies in the high density region. Perhaps AGN activity has something to do with the process of quenching the star formation of cluster galaxies, i.e., the so-called AGN feedback. Although it is difficult to determine which of the hypotheses are true between dust starbursts and quiescent galaxies hosting an AGN with only the data now available, in any case, there is no doubt that galaxies are activated in the densest regions such as clumps at z ∼ 2.5.

In lower-z clusters (z ≲ 1.0), the fraction of star-forming galaxies to all cluster members decreases moving closer to cluster centers or as the number density of galaxies increases (e.g., Kodama et al. 2004; Koyama et al. 2010, 2011). In this USS 1558-003 proto-cluster, however, it is difficult to completely sample red quiescent galaxies since DRGs have broad redshift distribution such as 2 ≲ z ≲ 4 and also contain both populations of quiescent galaxies and dusty starbursts. This makes it hard for us to quantify the fraction of star-forming galaxies in this proto-cluster. However, even if we assume that all the DRGs are quiescent galaxies associated with the proto-cluster at z = 2.53, a large fraction of proto-cluster members are star formers, and there are far fewer red quiescent galaxies.

As shown in Figure 6, the radio galaxy USS 1558-003 has quite an extended Hα emission with an apparent scale of ∼45 or a physical scale of ∼36 kpc. Although the spatially extended nebular line emissions (Lyα and Hα) have already been recognized around this radio galaxy (Villar-Martín et al. 2007; Humphrey et al. 2008), we here reveal for the first time the two-dimensional distribution of Hα emission. Such extended nebular emission is a remarkable feature commonly seen in HzRGs (e.g., Heckman et al. 1991; Humphrey et al. 2007). In particular, it is well known that they show extended Lyα emission aligned with radio structure (alignment effect). Our extended Hα emission is tilted with a position angle (P.A.) of 50° (Figure 6). This direction is approximately aligned to but slightly different from the radio axis which has the P.A. of 75° (Pentericci et al. 2000). Based on the integral field spectroscopy, Villar-Martín et al. (2007) found that the extent of the Lyα emission is ∼90 × 75 and is misaligned by ∼30°± 5° relative to the radio structures. Therefore, the direction of the stretch of the Hα emission is more or less similar to that of Lyα emission. However, the size of the Hα emission is ∼45 × 30, which is much smaller than the Lyα emission. It is possible that this smaller size of Hα emission is due to the fact that the surface brightness of the outer region is less than the detection limit. To check the possibility, we compare the distribution of Lyα flux density given in Villar-Martín et al. (2007) with the NB2315–K_s image and find that the Hα emission can be detected at more than 2σ if the Lyα/Hα ratio is constant over the whole region where the Lyα emission is detected. This means the ratio of Lyα/Hα is not constant but is larger in the outer region of the radio galaxy. Such a smaller size of Hα emission compared to Lyα emission is also seen in the radio galaxy MRC 1138-262 at z = 2.16 (Kurk et al. 2002). Kurk et al. (2002) concluded that such fairly extended Lyα emission is due to scattered radiation. Since we see a similar nature in the USS 1558-003 radio galaxy, the existence of abundant scattering material around the radio galaxy trapped in deep potential wells and large Lyα emission probably due to resonant scattering seem to be common characteristics of HzRGs.

Furthermore, we find that the radio galaxy also has extended emission in the B and r' band, where there seems to be a contribution of Lyα and He ii emissions to some extend in the B and r' band, respectively. However, we do not see any extended emission in the K_s band. Such extended emission in the rest-frame UV wavelengths and the compactness of the old stellar component may suggest that star-forming activity is also currently ongoing in the extended regions. In that case, such extended star formation is likely triggered by some physical processes associated with the radio activity. The jet-induced star formation is one plausible option (e.g., Bicknell et al. 2000; Hatch et al. 2011).

Many studies have been conducted to investigate the physical origin of nebular emission around the radio galaxy. Possible mechanisms include the photoionization by AGNs, young stars, and X-ray emission from shock heated gas, and collisional ionization by jet-induced shock (e.g., Miley & De Breuck 2008 and references therein). Although the ionization mechanism is not yet completely understood, recent studies suggest that AGNs could be the dominant source to ionize the surrounding nebular gas (e.g., Humphrey et al. 2008; Miley & De Breuck 2008). However, for radio galaxies at z ∼ 1, Best et al. (2000) found that the cause of nebular emission is dependent on the size of radio morphology based on the diagnostic with line ratios of nebular emission. Small sources with a radio structure smaller than ≲ 150 kpc tend to have emission originated from shock ionization, while larger sources prefer the mechanism of photoionization by AGNs. Since the radio observation of USS 1558-003 shows that the size of the radio galaxy is 92 (74 kpc; Pentericci et al. 2000), the shock ionization would be preferred. It is therefore interesting to investigate the origin of Hα emission of this radio galaxy in greater detail.

For this purpose, a diagnostic with nebular emission in rest-frame optical wavelength is useful. Humphrey et al. (2008) conducted optical and NIR spectroscopies in which a slit was placed on the radio galaxy along the radio axis and investigated the ratios of detected emission lines. They detected many emission lines in the rest-frame wavelength range of λ = 1216–6585 Å and found that the radio galaxy has a broad Hα emission line with a velocity width of 12,000 km s⁻¹, and that the amount of dust is negligibly small (A_v ∼ 0). Consequently, they concluded that the line ratios were best explained by AGN photoionization. However, they also found that the sources with high A_v do not show evidence for jet–gas interactions, while the sources undergoing strong jet–gas interactions have low A_v. The small amount of dust in this radio galaxy implies that the surrounding gas is undergoing interaction with the jet-induced shock. If this is indeed the case, the result is consistent with the proposed relation between the ionization mechanism and the radio size. It is suggested, therefore, that the shock induced star formation is occurring to some extent as well as AGN photoionization; thus, both mechanisms are probably contributing to the large extent of the Hα emission. It is difficult, however, to quantify the relative contribution of the two mechanisms, and it is beyond the scope of this paper. Integral field spectroscopy at NIR is essential to resolving the two-dimensional structure of the emission line regions and to fully understanding the physical origins of nebular emission.