ENHANCED STAR FORMATION OF LESS MASSIVE GALAXIES IN A PROTOCLUSTER AT z = 2.5

Since the last decade, the question of a positive correlation between SFR and stellar mass in star-forming galaxies (SFGs), which is called the main sequence (MS) of SFGs, has been a hot topic in the field of galaxy evolution (e.g., Daddi et al. 2007; Elbaz et al. 2007; Noeske et al. 2007). The tight correlation provides perspectives of how the SFGs evolve over cosmic time: they spend most of their lifetime on the sequence and evolve along the MS. However, a small fraction of them show starburst activities and deviate upward from the MS (Rodighiero et al. 2011).

During the course of hierarchical structure formation, galaxy evolution is expected to proceed in different ways in different environments. It is suggested that such environmental effects are more preferentially seen in satellite galaxies rather than in central galaxies at z < 1 (e.g., Peng et al. 2012; Kovač et al. 2014). Some environment-dependent processes, such as galaxy interactions as well as merging gas inflows and outflows, can alter the star formation activity in galaxies by either boosting it or truncating it. Understanding the physical mechanisms of these processes is of vital importance to reveal the origin of early-type galaxies and the strong environmental dependence of galaxy properties seen in the present-day universe.

With this motivation, we are conducting a systematic project called MAHALO-Subaru (MApping HAlpha and Lines of Oxygen with Subaru; Kodama et al. 2013) and mapping star formation activities over a wide range of environments as well as across cosmic times, in particular at 1.5 ≲ z ≲ 2.5, where clusters of galaxies are just vigorously assembling and forming. This project has shown that integrated SFR per dynamical mass in cluster core increases dramatically with redshift up to z ∼ 2.5 (Shimakawa et al. 2014). However, Koyama et al. (2013) show that the location of the MS of SFGs in a protocluster (PKS1138–262) at z ∼ 2 is not different from that in the general field at similar redshifts, although the distribution of galaxies along the MS is skewed to higher SFRs and stellar masses in high-density regions probably due to biased, more advanced galaxy formation there. Because our analyses have so far been limited to relatively massive galaxies (≳10^9.5 M_⊙), we want to extend our study to an even less massive regime.

For this purpose, we target a protocluster around the USS 1558-003 radio galaxy at z = 2.53. Hayashi et al. 2012 (hereafter H12) have already reported a narrowband Hα emission-line survey as a part of MAHALO-Subaru. This previous observation identified as many as 68 HAEs associated with the protocluster within a 27 arcmin² field of view (FOV). It shows a linear filamentary structure that hosts three dense groups of HAEs. This richness and high density make it a unique protocluster target at z > 2 for us to investigate in the early environmental effects of the galaxy-formation phase.

To access less massive galaxies in the cluster, we conducted very deep follow-up observations: one is three times deeper narrowband Hα imaging with the Subaru Telescope and another is deep Hubble Space Telescope (HST)/WFC3 imaging at near-infrared. Based on these new unique imaging data sets, in this letter we report the first intriguing discovery of the nature of the less massive SFGs (≲10^9.5 M_⊙) in the protocluster USS 1558-003 at z = 2.53, which has become accessible only with these deep observations.

Magnitudes are presented in the AB system (Oke & Gunn 1983); the cosmological parameters H₀ = 70 km s⁻¹ Mpc⁻¹, Ω_m = 0.3, and Ω_Λ = 0.7, along with Chabrier (2003) initial mass function, are adopted throughout the letter.

Because many of the imaging data used here are already published in H12, we mainly describe the new additional data obtained after H12. Data in J, K_s, and NB2315 are updated by further deep observations with MOIRCS/Subaru (Ichikawa et al. 2006; Suzuki et al. 2008) at a single pointing whose FOV corresponds to the field called "F2" in H12 (S15A-047, PI: T. Kodama). During the observations from 2015 April 30 to May 6, the weather was fine and the sky condition was photometric. Most of the frames were taken under the seeing condition of better than 06. Combined with the H12 data, the total integration times sum up to 3.18, 3.45, and 9.72 hr in J, K_s, and NB2315 (Table 1), respectively. All of the J, K_s, and NB2315 data are re-reduced in a standard manner using the data reduction package for MOIRCS (MCSRED ver.20150619⁵ by I. Tanaka). The details of the reduction will be shown in our forthcoming main paper.

Table 1.  Optical and Near-infrared Images

Filter	Instrument/Telescope	Integration Time	Limiting mag.a	PSF	Proposal ID
		(minutes)	(5σ)	(arcsec)
B	Suprime-Cam/Subaru	80	27.51	0.70	S10B-028
r'	Suprime-Cam/Subaru	90	27.24	0.63	S10B-028
z'	Suprime-Cam/Subaru	55	26.03	0.66	S10B-028
J	MOIRCS/Subaru	191	24.85	0.55	S10B-028, S15A-047
H160	WFC3/HST	87	27.46	0.21	GO-13291
H	MOIRCS/Subaru	45	23.78	0.47	S10B-028
Ks	MOIRCS/Subaru	207	24.49	0.60	S10B-028, S15A-047
NB2315	MOIRCS/Subaru	583	23.90	0.52	S10B-028, S15A-047

Notes. FWHMs of point-spread function (PSF) in all the Subaru images are matched to 067 finally, except for the B-band image which has a FWHM of 070.

^aLimiting magnitudes are measured with a 12 diameter aperture, except for the 04 aperture for the WFC3/HST image.

Download table as:  ASCIITypeset image

Observations with WFC3/HST in F160W were conducted on 2014 July 5 and 9 (GO-13291, PI: M. Hayashi). Three pointings with WFC3 were required to cover the structures found in H12 (see Figure 1). Because it took two orbits for the observations in each pointing, the integration times were 5224 s in total (see Table 1). The reduction was carried out in a standard manner with pipeline. Using the task multidrizzle, the pixel scale was changed to 0.06 arcsec per pixel (Koekemoer et al. 2011).

Zoom In Zoom Out Reset image size

To summarize, the data set that we use consists of six broadband data (B, r', z', J, H, and K_s), the narrowband (NB2315) data with Subaru, and the H₁₆₀ data with HST/WFC3. The J, K_s, and NB2315 images are deeper by 0.67, 0.84 and 0.89 mag. than those in H12 (Table 1).

The photometric catalog is updated by taking the same procedures on the latest data set as in H12. However, because the HST image has a much better point-spread function (PSF) than that of the Subaru images (see Table 1), source detection and photometry on the HST image are conducted independently. The HST and Subaru photometric catalogs are then combined by matching the detected objects within a 05 radius circle.

A procedure to select emission-line galaxies is basically the same as in H12, but we redo the selection based on the updated catalog. First, we extract galaxies with more than 3σ excess in K_s − NB2315 color (the left panel of Figure 2). Note that a correction of 0.1 mag. in K_s − NB2315 color is required as a color term in estimating a continuum level underneath the Hα emission line because there is a difference of 0.163 μm in the effective wavelengths between K_s and NB2315 filters. We also apply the criterion of K_s − NB2315 > 0.50. These criteria allow us to select 171 NB2315 emitters with line fluxes larger than 1.1 × 10⁻¹⁷ erg s⁻¹ cm⁻² and equivalent widths in the observed frame larger than 66 Å. The limiting line flux corresponds to L(Hα + [N ii]) = 5.8 × 10⁴¹ erg s⁻¹, which turns out to be 2.4 times deeper than that in H12. It can be converted to a dust-free SFR of 2.2 M_⊙ yr⁻¹ (Kennicutt 1998), where the contribution of [N ii] to the line flux measured with NB, i.e., [N ii]/(Hα + [N ii]), is assumed to be 0.25 (Sobral et al. 2013).

Zoom In Zoom Out Reset image size

To eliminate lower-z contaminant lines from the NB2315 emitter sample, we use the same color selection criteria as in H12 on the color–color diagram, $r^{\prime} -J$ versus J − K_s (the middle panel of Figure 2). However, some emitters are not securely detected in r', J, or K_s, and we notice that almost all such galaxies are very faint in J and K_s. For many cases, we are thus not able to classify them according to color selection. We regard them as "possible Hα emitters." This may be justified by our Suprime-Cam/Subaru observation with NB428 narrowband filter targeting Lyα emitters at the same redshift, which shows that most of the Lyα emitters are not detected or very faint (<3σ) in J and/or K_s.

We also found three contaminant [O iii] emitters at z ∼ 3.6 in the follow-up spectroscopy (Shimakawa et al. 2014, 2015b). This suggests a necessity of applying an additional color selection to distinguish [O iii] and Hβ at z ∼ 3.6 from Hα at z ∼ 2.5. We thus utilize the $r^{\prime} -{H}_{160}$ versus ${H}_{160}-{K}_{s}$ diagram as well for galaxies whose HST H₁₆₀ data are available. The deep HST H₁₆₀ photometries correspond to a bluer side of the Balmer/4000 Å break for galaxies at z ∼ 3.6. We set the boundary at ($r^{\prime} -{H}_{160}$) = 0.324 (${H}_{160}-{K}_{s}$) + 1.2 based on the colors of the spectroscopically confirmed HAEs to distinguish the two populations (the right panel of Figure 2). The slope of the criterion is parallel to a reddening vector estimated from the dust extinction curve of Calzetti et al. (2000). We note that the color tracks modeled using the Bruzual & Charlot (2003) stellar population synthesis code also support the rHK_s selection criterion. The rHK_s color selection identifies four emitters as galaxies at z ∼ 3.6. Furthermore, B-band magnitudes of [O iii] emitters at z ∼ 3.6 are more sensitive to attenuation by neutral hydrogen in the intergalactic medium than HAEs at z ∼ 2.5 (Madau 1995). Non-detection or faint magnitude in B imply that the galaxies are more likely to be located at z ≳ 3. Thus, among the HAEs without spectroscopic confirmation and possible HAEs, 15 emitters without a detection at more than 2σ in B or with $B-r^{\prime} \gt 1.39$ are excluded from the samples. Note that all of the confirmed HAEs except for one galaxy have $B-r^{\prime} \lt 1.39$. We also note that 98% of the HAEs are detected in r'-band at 2σ level, which suggests that there is little contamination of [O ii] emission at z ∼ 5.2.

Consequently, by applying these color selections, we select 91 HAEs and nine possible HAEs in total (see Figures 1 and 2). Hereafter, we treat both HAEs and possible HAEs as SFGs at z ∼ 2.5. Compared with the sample of H12, the number of HAEs increases from 68 to 100, due to deeper NB2315 and K_s images available in this study.

We estimate stellar masses of the HAEs by fitting a library of evolutionary stellar population synthesis models (GALAXEV, Bruzual & Charlot 2003) to spectral energy distribution (SED) with the six broadbands. The SEDs are hardly affected by nebular emission lines because none of the strong lines but [O ii] enter the broadbands. In modeling, we fix the redshift to the spectroscopic one if it is available. Otherwise we use the redshift of the radio galaxy (z = 2.53). We assume exponentially declining and constant star formation histories. Acceptable ages are chosen from 50 Myr to the age of the universe at that redshift. Stellar metallicities of Z_⊙ or 0.4 Z_⊙ are applied, which are consistent with gas-phase metallicities of the HAEs measured by Shimakawa et al. (2015a). A dust extinction curve by Calzetti et al. (2000) is assumed and E(B − V) ranges from 0.0 to 3.0. Even if galaxies are faint in J or K_s, thanks to the deep H₁₆₀ and optical bands, a model SED is determined for most of the HAEs. However, 10 HAEs are not fitted by any of the model SEDs. In those cases, we estimate the stellar masses based on their K_s magnitudes corrected for the mass-to-luminosity ratio (M_⋆/${L}_{{K}_{s}}$) measured from their J − K_s colors (H12; Kodama et al. 1998). We note that for the galaxies with SEDs available, the stellar masses derived from K_s and J − K_s are consistent with those determined by the SED fitting.

Next, we estimate SFRs of the HAEs using the Hα luminosities derived from the narrowband (NB2315) imaging (see H12 for more details). The contribution of [N ii] is removed from the line flux 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. (2013). If the spectroscopic redshift is available, we correct the Hα flux for the transmission of the NB filter at the wavelength of the line. We then correct for dust extinction using the calibration by Koyama et al. (2015), which estimates A(Hα) from the observed SFR(UV)/SFR(Hα) ratio and the stellar mass, where the σ_rms of A(Hα) in the calibration is 0.282 mag (Koyama et al. 2015). We use r'-band magnitudes (i.e., rest-frame 1795 Å) to estimate the rest-frame UV luminosity densities. The intrinsic Hα luminosities thus derived are converted to SFRs using the Kennicutt (1998) calibration.

Figure 3 shows a positive correlation between SFR and stellar mass for the 100 HAEs, confirming the existence of the MS of SFGs. The shaded gray region in Figure 3 shows our limit in SFR. This limiting SFR indicates that our data are deep enough for us to discuss the MS for SFGs fully down to stellar masses of ∼10¹⁰ M_⊙, and we can still reach the upper half of the MS galaxies down to ∼10⁹ M_⊙. At the lowest mass regime of ≲10⁹ M_⊙, we would no longer be able to access the MS galaxies if the sequence is extrapolated to the low mass, and we could see only the SFGs with enhanced star formation activities above the MS.

Zoom In Zoom Out Reset image size

Until now, there have been a number of previous studies that investigated the MS of SFGs at z ≳ 2 (e.g., Dunne et al. 2009; Karim et al. 2011; Reddy et al. 2012). Figure 3 also compares the MSs derived from the previous studies. Our HAEs with stellar masses of ≳10¹⁰ M_⊙ seem to be located right on the previously measured MSs in the literature (Speagle et al. 2014; Whitaker et al. 2014; Shivaei et al. 2015). As these previous studies mainly look at galaxies in the general fields, the agreement suggests that SFGs in the protocluster at z ∼ 2.5 share the same MS as field SFGs at similar redshifts, which is consistent with the previous studies (Koyama et al. 2013; Cooke et al. 2014).

On the other hand, at lower stellar masses (<10^9.3 M_⊙) in the protocluster, there are several galaxies that are significantly up-scattered above the MSs. If such small-mass galaxies all follow the same extrapolated MS, they would all be located below our detection limit and we would not see any of them, contrary to what we actually do see. Although we cannot discuss the exact locations of the MS at <10^9.3 M_⊙ due to incompleteness, we argue that there are at least some HAEs which are significantly deviated upward from the MS. Those more than 10 up-scattered HAEs at <10^9.3 M_⊙ have exceptionally large specific SFRs (sSFR = SFR/M_⋆) above 10⁻⁸ yr⁻¹ as shown by the light solid line in Figure 3. This indicates that their inferred ages (timescales of star formation) are smaller than 10⁸ years and that they are young starbursting galaxies just being formed. Note that this result is not affected by dust corrections because the amount of dust correction is progressively lower for less massive galaxies (Garn & Best 2010; Koyama et al. 2015). In fact, the inferred A(Hα) for almost all of the galaxies with <10⁹ M_⊙ is smaller than 0.2 mag. Even if we use the rest-frame UV luminosities to derive SFRs of the HAEs instead of Hα luminosities, we also find the existence of HAEs above the MS at the faint end.

Our results suggest that while the majority of massive galaxies are already settled in a secular evolution phase and are thus found on the MS, some less massive galaxies are in a starburst phase and are significantly up-scattered from the MS. This may be consistent with the downsizing scenario of mass-dependent galaxy evolution (e.g., Cowie et al. 1996; Bundy et al. 2006; Muzzin et al. 2013) or because they are located in a dense protocluster, they may be experiencing some influences from the surrounding environment such as galaxy-galaxy interactions. We do not know, however, if this trend is seen only in high-density regions or whether it is a common feature of less massive SFGs (traced by Hα), irrespective of environment.

The SED fitting described in Section 4 indicates that the youngest age of <10⁸ year is preferred for galaxies with stellar masses less than 10⁹ M_⊙, which is again consistent with the less massive HAEs having sSFR of >10⁻⁸ yr⁻¹. This also supports our interpretation that they are young, starbursting galaxies during the vigorous formation/assembly epoch of a rich galaxy cluster.

Cooke et al. (2014) show the lack of galaxies with stellar mass less than ∼10¹⁰ M_⊙ in a protocluster at z = 2.49, and argue that it is possibly due to either a large dust extinction of less massive galaxies or the earlier formation of massive galaxies. However, our results show that there are SFGs on the MS down to stellar mass of 10^9.3 M_⊙ and that even at lower mass bin there are SFGs with SFRs comparable to those of more massive galaxies with 10¹⁰ M_⊙, which are not in agreement with the results by Cooke et al. (2014). Compared with the USS1558, the protocluster discussed in Cooke et al. (2014) is not very rich, although it shows some overdensity in contrast to the general fields. Therefore, the discrepant result between this letter and Cooke et al. (2014) could be due to the intrinsic diversity of the properties of protoclusters at z ∼ 2.5. However, to address this issue, it is essential for us to investigate a much larger sample of protoclusters.

The existence of the less massive HAEs with <10^9.3 M_⊙ up-scattered above the MS may imply that a scatter around the MS increases at lower stellar masses. Diversity of star formation history in the early phase of galaxy evolution and/or sensitivity to the fluctuation of starburst activity at short timescales in individual H ii regions could cause the increased scatter. Another possible remaining issue is a metallicity dependence of the Hα luminosity (e.g., Bicker & Fritze-v. Alvensleben 2005; Dopita et al. 2006). A lower stellar metallicity would result in a higher stellar temperature, thus the larger number of ionizing photons. Therefore, the SFRs for less massive galaxies can be overestimated due to the metallicity effect if they follow the mass–metallicity relation (e.g., Shimakawa et al. 2015a). These are areas of research for future papers.

We thank the anonymous referee for providing constructive comments. M.H., R.S., and T.S. acknowledge support from the Japan Society for the Promotion of Science (JSPS) through the JSPS Research Fellowship for Young Scientists. T.K. acknowledges the financial support in part by a Grant-in-Aid for the Scientific Research (Nos. 21340045 and 24244015) by the Japanese Ministry of Education, Culture, Sports, and Science. This letter is based on data collected at Subaru Telescope, which is operated by the National Astronomical Observatory of Japan, as well as by observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space Telescope Science Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS 5-26555. These HST observations are associated with programs GO-13291.

Facilities: Subaru - Subaru Telescope, HST - .