A&A 468, 877-883 (2007)
T. Nagao1,2 - T. Murayama3 - R. Maiolino1,4 - A. Marconi1,5 - N. Kashikawa2 - M. Ajiki3 - T. Hattori6 - C. Ly7 - M. A. Malkan7 - K. Motohara8 - K. Ohta9 - S. S. Sasaki3,10 - Y. Shioya10 - Y. Taniguchi10
1 - INAF - Osservatorio Astrofisico di Arcetri, Largo Enrico Fermi 5, 50125 Firenze, Italy
2 - National Astronomical Observatory of Japan, 2-21-1 Osawa, Mitaka, Tokyo 151-8588, Japan
3 - Astronomical Institute, Graduate School of Science, Tohoku University, Aramaki, Aoba, Sendai 980-8578, Japan
4 - INAF - Osservatorio Astrofisico di Roma, via di Frascati 33, 00040 Monte Porzio Catone, Italy
5 - Dipartimento di Astronomia e Scienza dello Spazio, Università di Firenze, Largo E. Fermi 2, 50125 Firenze, Italy
6 - Subaru Telescope, National Astronomical Observatory of Japan, 650 North A'ohoku Place, Hilo, HI 96720, USA
7 - Department of Astronomy, University of California at Los Angeles, PO Box 951547, Los Angeles, CA 90095-1547, USA
8 - Institute of Astronomy, Graduate School of Science, University of Tokyo, 2-21-1 Osawa, Mitaka, Tokyo 181-0015, Japan
9 - Department of Astronomy, Graduate School of Science, Kyoto University, Kitashirakawa, Sakyo, Kyoto 606-8502, Japan
10 - Department of Physics, Graduate School of Science and Engineering, Ehime University, 2-5 Bunkyo-cho, Matsuyama 790-8577, Japan
Received 3 August 2006 / Accepted 12 January 2007
We report new follow-up spectroscopy of -dropout galaxies with an NB921-band depression found in the Subaru Deep Field. The NB921-depressed -dropout selection method is expected to select galaxies with large equivalent-width Ly emission over a wide redshift range, . Two of four observed targets show a strong emission line with a clear asymmetric profile, identified as Ly emitters at z = 6.11 and 6.00. Their rest-frame equivalent widths are 153 and 114 , which are lower limits on the intrinsic equivalent widths. Through our spectroscopic observations (including previous ones) of NB921-depressed -dropout galaxies, we identified 5 galaxies in total with a rest-frame equivalent width larger than 100 at out of 8 photometric candidates, which suggests that the NB921-depressed -dropout selection method is possibly an efficient way to search for Ly emitters with a large Ly equivalent width, in a wider redshift range than with usual narrow-band excess techniques. By combining these findings with our previous observational results, we infer that the fraction of broad-band selected galaxies having a rest-frame equivalent width larger than 100 is significantly higher at (the cosmic age of 1 Gyr) than that at (2 Gyr), being consistent with the idea that the typical stellar population of galaxies is significantly younger at than that at . The NB921-depressed -dropout galaxies may be interesting candidates for hosts of massive, zero-metallicity Population III stars.
Key words: cosmology: early universe - galaxies: evolution - galaxies: formation - galaxies: individual: J132345.6+271701 - galaxies: individual: J132519.4+271829 - galaxies: starburst
Observational searches for high-z galaxies have progressed a lot in this decade. Thanks to large telescopes such as Subaru, VLTs, and Keck, a few dozen galaxies at 6<z<7 have been spectroscopically confirmed so far (Hu et al. 2002; Kodaira et al. 2003; Cuby et al. 2003; Rhoads et al. 2004; Kurk et al. 2004; Nagao et al. 2004, 2005a; Taniguchi et al. 2005; Stern et al. 2005; Kashikawa et al. 2006). Statistical properties of these high-z galaxies, such as their luminosity function and their correlation function provide information on galaxy evolution, the cosmic star-formation rate, and the re-ionization history of the universe.
Table 1: Photometric properties of NB921-depressed -dropout galaxies in the Subaru Deep Field.
In addition to statistical properties, properties of individual galaxies at high redshift also shed light on galaxy evolution. In particular, the equivalent width () of the Ly emission provides clues on the stellar population of high-z galaxies, for which it is generally difficult to investigate stellar spectral features due to the limited observational sensitivity of currently available instruments. Malhotra & Rhoads (2002) present their theoretical estimates that large Ly equivalent widths ( for a Salpeter initial-mass function (IMF) and for a flatter IMF) suggest a very young (<107 years) stellar population (e.g., Tumlinson et al. 2003). More interestingly, it has been theoretically predicted that galaxies hosting zero-metallicity stars (or Population III stars, hereafter PopIII) should show huge Ly equivalent widths that could reach up to (e.g., Schaerer 2002, 2003; Tumlinson et al. 2003; Scannapieco et al. 2003). Therefore, the frequency distribution of the Ly equivalent width in high-z galaxies and its evolution with redshift are important for understanding the early stages of galaxy evolution (e.g., Shimasaku et al. 2006; Ando et al. 2006).
However, observational study of the Ly equivalent width in galaxies at high redshift is not straightforward. Although Malhotra & Rhoads (2002) report that more than half of their sample of 150 Ly emitters (LAEs) have Ly equivalent width larger than 240 , their analysis is only based on photometric data. Especially for LAEs at (which corresponds to the window between airglow emission at ), the -band magnitude is contaminated by the Ly emission, so is an unreliable measure of continuum flux density. Even for the sample with spectroscopic data, the continuum emission of LAEs selected by using narrow-band magnitude is generally too faint to be detected in their spectra.
We are exploiting a new selection method for identifying LAEs with a large Ly equivalent width, that is, dropout with a "depression'' in the narrow-band filter NB921 (see Sect. 2) (Nagao et al. 2004, 2005a). If galaxies are at 6.0 < z < 6.5, their redshifted Ly emission is expected at . In this case, the narrow-band magnitude at has no contamination from Ly emission and thus only traces the continuum, while the -band magnitude is enhanced by the contribution of the strong Ly emission and thus the narrow-band magnitude is depressed with respect to the -band magnitude. Therefore this selection method finds galaxies with a large Ly equivalent width in a wide redshift range, 6.0 < z < 6.5. Note that this selection method does not select Galactic late-type stars, unlike the usual -dropout selection (see Nagao et al. 2005a, for more details).
|Figure 1: Color-magnitude diagram of vs. ( left) and vs. color-color diagram ( right). The 48 -dropout objects detected in the SDF are plotted. The NB921-depressed -dropout objects are shown by open squares, and the other -dropout objects are shown by filled circles. Error bars denote 1 uncertainty in the observed color or magnitude. In the left panel, 1, 2, and 3 uncertainties of the color are shown by the dotted, dashed, and solid lines, respectively.|
|Open with DEXTER|
In this paper, we report new spectroscopy for a sample of NB921-depressed -dropout galaxies. Throughout this paper, we adopt a cosmology with ( , , ) = (1.0, 0.3, 0.7) and H0 = 70 km s-1 Mpc-1.
We selected our -dropout galaxy sample from the public catalog of the Subaru Deep Field (SDF) imaging survey (Kashikawa et al. 2004), which contains broad-band (B, V, , , and ) and narrow-band photometric data (NB816 and NB921; the central wavelengths and the half-widths of the transmittance are (8150 , 120 ) and (9196 , 132 ), respectively). The adopted criteria to select the -dropout galaxy sample are:
We have already obtained optical spectra of 3 NB921-depressed -dropout galaxies; SDF J132440.6+273607 (#4; z=6.33; Nagao et al. 2004), SDF J132426.5.0+271600, and SDF J132442.5+272423 (#3 and #5; z=6.03 and 6.04; Nagao et al. 2005a). Among the remaining 5 NB921-depressed -dropout galaxies, we observed 4 objects on 26 April 2006 (UT). This spectroscopy was carried out with the Faint Object Camera And Spectrograph (FOCAS; Kashikawa et al. 2002) on the Subaru telescope (Iye et al. 2004) in its multi-object slit mode. The four objects were chosen to optimize the slit mask design (by also accomodationg several other objects observed within the broader context of the SDF project). The 175 lines mm-1 echelle grating and the SDSS filter were used. The resulting wavelength coverage was , with a dispersion of pixel-1. The adopted slit width was 0.83 arcsec, giving a spectral resolution of or at as measured from the widths of atmospheric OH emission lines. The spatial sampling was 0.31 arcsec per resolution element, as we adopted 3 pixel on-chip binning. The seeing was variable during the observation (0.5-1.0 arcsec). The total integration times were 7200 s for SDF J132345.6+271701 (#1) and SDF J132422.0+271742 (#2), and 9000 s for SDF J132519.4+271829 (#6) and SDF J132526.1+271902 (#8). Note that the position angle of the slits was set to the North-South direction for all the target objects. Since no neighboring objects are around the target objects (Fig. 2), the detected signal is believed to be from the targets. We also obtained spectra of the spectrophotometric standard star Feige 34 (Oke 1990) for flux calibration. The obtained data were reduced in the standard manner by using IRAF.
|Figure 2: Thumbnail images of NB921-depressed -dropout galaxies. The square regions around each object in the , , and NB921 images are shown from left to right. The IDs given on the left side of the panels correspond to those in Table 1. The size of panels and the radius of circles are 16 arcsec and 8 arcsec, respectively.|
|Open with DEXTER|
Among the four observed objects, SDF J132345.6+271701 (#1) and SDF J132519.4+271829 (#6) show a strong emission line in their spectra, whose peak wavelengths are at 8634 and 8512 , respectively. In Figs. 3 and 4, the sky-subtracted position-velocity spectrogram, the extracted one-dimensional spectrum, and the typical sky spectrum are shown for these two objects, respectively. The aperture size used for the extraction of the one-dimensional spectrum is 5 binned pixels (1.6 arcsec). No continuum emission is detected for either object. Both emission lines show a clear asymmetric profile, i.e., a sharp decline on the blue side and a prominent tail on the red side, which suggests that the observed emission lines are Ly. Note that if the detected emission line were H, [O III]5007 or H, other rest-frame optical emission line(s) should be seen in the observed wavelength range. If the detected emission line was [O II]3727, it should be resolved as a doublet emission, since the expected wavelength separation of the redshifted [O II] 3726, 3729 is (or 6 pixels), which corresponds to 210 km s-1. Here we assume that the velocity width of the [O II] lines is not very broad; otherwise we could not resolve the [O II] doublet. However our assumption seems valid for normal star-forming galaxies, because the [O II] doublet of some star-forming galaxies in the SDF is indeed resolved in our previous spectroscopic follow-up observations (Shimasaku et al. 2006; Ly et al. 2007). The fact that we see only one strong emission line with an asymmetric profile in each spectrum strongly suggests that we have detected Ly at .
To quantify the asymmetry of the detected emission lines, we calculated two independent parameters; and . The former is the ratio between and , where is the flux at wavelengths longer than the emission-line peak, while is that at shorter wavelengths (Taniguchi et al. 2005; Nagao et al. 2005a; see also Haiman 2002 for a theoretical discussion on the parameter ). The measured ratios are 2.08 0.12 and 2.47 0.30 for SDF J132345.6+271701 (#1) and SDF J132519.4+271829 (#6), respectively. Note that the latter value may be overestimated, since the OH airglow emission on the blue side of the Ly emission may be over-subtracted (Fig. 4). The ratio of both line-detected objects is significantly higher than unity, and also higher than that of most LAEs reported by Taniguchi et al. (2005). This result is consistent with the interpretation that the detected emission lines are Ly. The latter parameter () is the weighted skewness (Shimasaku et al. 2006; Kashikawa et al. 2006), which is larger for objects with higher asymmetries and/or larger emission-line widths. The calculated values are 4.00 0.22 and 8.72 0.28 for SDF J132345.6+271701 (#1) and SDF J132519.4+271829 (#6), respectively. Since the weighted skewness of emission lines arising from possible low-redshift interlopers generally do not exceed 3 (Kashikawa et al. 2006; see also Shimasaku et al. 2006), the derived values of the weighted skewness are also consistent with the interpretation that the detected emission lines are Ly. Taking the derived values of and for the two line-detected objects into account, we conclude that the detected emission lines are Ly and, accordingly, that the redshifts of the two objects are 6.11 and 6.00. The observed Ly fluxes are (4.5 0.1) 10-17 erg s-1 cm-2 and (3.4 0.2) 10-17 erg s-1 cm-2, without any correction for slit loss. The Ly luminosities are thus calculated to be 1.9 1043 erg s-1 and 1.3 1043 erg s-1. The measured emission-line widths in FWHM are and . These values correspond to and , or velocity widths of 330+50-70 km s-1 and 220+80-220 km s-1, respectively, after correction for the instrumental broadening effect. Here the absorption effects are not taken into account.
Adopting these redshifts, the N V1240 emission line would be expected at and for SDF J132345.6+271701 (#1) and SDF J132519.4+271829 (#6), respectively, if these two objects were active galactic nuclei (AGNs). However, no emission-line features are seen at the corresponding wavelengths (see Figs. 3 and 4). The 3 upper limits for the N V emission are 5.7 10-18 erg s-1 cm-2 and 6.9 10-18 erg s-1 cm-2 for these two objects, assuming the same velocity widths as Ly. However, we cannot completely rule out the possibility that one or both of these two objects are AGNs, since some high-z narrow-line radio galaxies show very weak N V compared with Ly (e.g., De Breuck et al. 2000; Nagao et al. 2006; see also Malkan et al. 1996).
|Figure 3: Sky-subtracted optical position-velocity spectrogram ( top) and one-dimensional spectrum ( middle) of SDF J132345.6+271701 (#1) obtained with FOCAS on Subaru. The typical spectrum of the sky emission is also shown in the bottom panel.|
|Open with DEXTER|
As mentioned in Sect. 1, it is sometimes very difficult to estimate the Ly equivalent width because measuring the continuum in the low-S/N spectra is very difficult. However, we can determine the flux density of the continuum emission for those two galaxies from the NB921-band magnitude, since the NB921-band flux does not contain Ly flux, unlike the case of LAEs selected by the NB921 excess (e.g., Kodaira et al. 2003; Taniguchi et al. 2005). The NB921-band magnitudes of the two galaxies are 26.37 mag and 26.38 mag, which correspond to the flux densities of 10-20 erg s-1 cm-2 and 10-20 erg s-1 cm-2 at the wavelengths of the Ly peak, respectively. Then the rest-frame equivalent widths of the Ly emission are calculated to be 153 and 114 for SDF J132345.6+271701 (#1) and SDF J132519.4+271829 (#6), assuming a flat UV continuum. Note that if the detected emission lines were [O II], the rest-frame equivalent widths would be 470 and 349 . These high values are quite rare for [O II] emitters (e.g., Ajiki et al. 2006), which also supports our interpretation that the detected emission lines are Ly.
The obtained spectroscopic properties are summarized in Table 2. In the same table, properties of our previous spectra (Nagao et al. 2004, 2005a) are also given for the reader's convenience. Note that there are no spectral features in the obtained spectra of SDF J132422.0+271742 (#2) and SDF J132526.1+271902 (#8). We will discuss these non-detections briefly in the next section. In this table, we also give the minimum EW0(Ly) that can be detected by the NB921-depressed -dropout method for galaxies with the redshift and the -band magnitude of each NB921-depressed -dropout galaxy. We will discuss this quantity in Sect. 4.1.
Table 2: Spectroscopic properties of NB921-depressed -dropout galaxies in the Subaru Deep Field.
In this observing run we observed four NB921-depressed -dropout galaxies, and found strong emission lines in two of them. Could the two remaining emission-line undetected objects also be strong LAEs at 6.0 < z < 6.5, as expected by the NB921-depressed -dropout selection? Since high-z Ly emission shows a clear asymmetric profile and especially a prominent tail toward the red side of the emission-line peak, the Ly emission of LAEs is expected to be resolved with the current wavelength resolution . This means that the Ly emission should be found, even when the redshifted Ly line falls on an isolated (i.e., unresolved) OH airglow emission line. However, the Ly detection would be difficult when the Ly emission line falls on blended OH lines, such as the lines at seen in the bottom panel of Fig. 3 (which corresponds to a redshift range ). SDF J132422.0+271742 (#2) has and magnitudes, which are the faintest among the NB921-depressed -dropout galaxy sample. Thus the non-detection of the emission line might be simply due to an insufficient integration time. On the contrary, the and magnitudes of SDF J132526.1+271902 (#8) are the brightest among the NB921-depressed -dropouts. Since this object is so bright, it is selected as an NB921-depressed object in spite of its small amount of the NB921 depression, . Therefore the expected Ly equivalent width of this object is so small (which is estimated to be in the rest frame by assuming z=6.0 and a flat UV slope) that the detection of Ly could be more difficult than the other NB921-depressed -dropout galaxies. Taking all of these considerations into account, we cannot reject the possibility that the two Ly non-detected objects are also LAEs at 6.0 < z < 6.5. Deeper spectroscopic observations are necessary to investigate these objects further.
Therefore, the "success rate'' of the NB921-depressed -dropout selection method (i.e., the probability that this selection identify LAEs at 6.0 < z < 6.5 correctly) is between 5/8 (63%) and 8/8 (100%). The rest-frame Ly equivalent width of all of the five Ly-detected objects is quite large, at least , as also expected by the selection method (see Table 2). It should be mentioned here that, as discussed in Nagao et al. (2005a), the NB921 depression is not expected in Galactic late-type stars and thus this selection method is not contaminated by stars, unlike a simple -dropout sample (see, e.g., Stanway et al. 2004).
|Figure 5: Minimum EW0(Ly) that can be selected by the NB921-depressed -dropout method [ (Ly)] as a function of the -band magnitude, estimated by assuming z = 6.0 and a flat UV slope.|
|Open with DEXTER|
Here we discuss the limiting EW(Ly) of the NB921-depressed -dropout selection method. This selection technique does not select only LAEs with a large EW0(Ly) in principle, because the selection criterion on the NB921 depression is based on the sky error as described in Sect. 2. This is different from the usual LAE surveys based on the narrow-band excession, for which a certain excess magnitude is generally used instead of the sky noise. Therefore the NB921-depressed -dropout method would select LAEs with a relatively small EW0(Ly) if the imaging data were deep enough and accordingly the sky noise was very small. In other words, the minimum EW0(Ly) that can be selected by the NB921-depressed -dropout method [hereafter (Ly)] depends on the magnitude of target galaxies, since the color of brighter objects is less affected by the sky error with respect to that of fainter objects. To see this effect quantitatively, we show the dependence of (Ly) on the magnitude for the case of the SDF dataset in Fig. 5, where z=6.0 and a flat UV slope are assumed. It is demonstrated that only LAEs with EW0(Ly) 100 are selected among faint galaxies with , while LAEs with EW0(Ly) a tens of 10 could be also selected among relatively bright galaxies with . Note that the latter case corresponds to the case of SDF J132526.1+271902 (#8). To compare (Ly) and the detected EW0(Ly) for each NB921-depressed -dropout galaxy, we also give (Ly) for each set of the redshift and the -band magnitude of 5 spectroscopically identified NB921-depressed -dropout galaxies in Table 2.
In order to carry out systematic surveys for strong LAEs utilizing the NB921-depressed -dropout method, it would be more appropriate to adopt a certain depression magnitude as a selection criterion rather than using the sky error. The reason for adopting the selection criterion based on the sky noise is that one of our motivations in the photometric selection is to construct a sample of the target objects for our spectroscopic observations to find candidates for PopIII-hosting galaxies (see Nagao et al. 2005b). Due to the limited observing time, we had to focus on convincing candidates, i.e., objects with a significant NB921 depression with respect to the sky error.
The rest-frame Ly equivalent width of the NB921-depressed -dropout galaxies with a Ly detection (given in Table 2) ranges from 94 to 236 . Although these values are lower than the "critical value'' EW0(Ly) = 240 above which it cannot be explained by normal stellar populations (e.g., Malhotra & Rhoads 2002), it should be noted that the values we obtained are not corrected for absorption effects. The amount of the Ly absorption depends on various parameters, including the neutral fraction of the inter-galactic matter (IGM) and the kinematic status of neutral hydrogen within LAEs themselves (e.g., Neufeld 1990; Haiman 2002; Mas-Hesse et al. 2003; Ahn 2004). Here it should be kept in mind that the IGM is not perfectly re-ionized at , which is suggested by recent observations of high-z quasars (Fan et al. 2006) and LAEs (Kashikawa et al. 2006). There have been extensive arguments about whether Ly photons from galaxies at an earlier epoch than the completion of the cosmic re-ionization are observable or not. Some calculations suggest that the Ly photons from such galaxies are strongly suppressed (e.g., Miralda-Escude 1998; Miralda-Escude & Rees 1998; Loeb & Rybicki 1999) and other theoretical works infer that a large fraction of the Ly photons can transmit due to cosmological H II regions (e.g., Madau & Rees 2000; Cen & Haiman 2000). Haiman (2002) investigated various parameter dependences of the Ly transmission fraction for high-z LAEs by taking both the neutral hydrogen within LAEs themselves and the neutral IGM into account. They showed that the ratio of the transmitted Ly flux to the total (intrinsic) Ly flux [F0(Ly)/ (Ly)] is estimated to be 10% for typical LAEs with a star-formation rate (SFR) 10 yr-1 and 50% for LAEs with yr-1.
Following this result, we assume F0(Ly)/ (Ly) 0.5 to correct the absorption effect on the observed Ly flux rather conservatively, in the sense that the actual F0(Ly)/ (Ly) of our spectroscopic sample would be smaller than 0.5 since the SFR is less than 100 yr-1 (Nagao et al. 2004, 2005a). See also, e.g., Santos (2002) and Dijkstra et al. (2006) for the justification of our adopted value of F0(Ly)/ (Ly). Interestingly, by adopting this assumption, the intrinsic EW0(Ly) for three out of the five spectroscopically identified NB921-depressed -dropout galaxies is then expected to exceed the critical value EW0(Ly) = 240 . This result is consistent with the idea that galaxies at contain young stellar populations, whose age may be younger than 107 years as discussed by Malhotra & Rhoads (2002); see also Shimasaku et al. (2006).
The huge intrinsic Ly equivalent width would indicate that the NB921-depressed -dropout galaxies may contain a significant number of PopIII stars. Therefore our NB921-depressed -dropout galaxy sample offers fascinating targets for observational searches for PopIII stars, which will be an important topic in the forthcoming decade (see, e.g., Jimenez & Haiman 2006). We have already searched for an observational signature in one of the NB921-depressed -dropout galaxies, SDF J132440.6+273607 (#4; Nagao et al. 2005b), through He II1640 emission (see, e.g., Tumlinson & Shull 2000; Oh et al. 2001; Schaerer 2002, 2003). Although the He II1640 emission was not detected in this object, similar observations will provide important constraints on theoretical PopIII models (e.g., Scannapieco et al. 2003; Nagao et al. 2005b).
It has been known that narrow-band selected galaxies (e.g., emission-line galaxies) at high redshift tend to have large Ly equivalent widths, which sometimes exceed the critical value of EW0(Ly) = 240 (e.g., Malhotra & Rhoads 2002; Shimasaku et al. 2006). However, some broad-band selected high-z galaxies such as Lyman-break galaxies (LBGs) rarely have large Ly equivalent widths. For instance, at the fraction of LBGs showing EW0(Ly) > 100 is 1% of 1000 galaxies (Shapley et al. 2003). Since our sample is also broad-band selected, it is interesting to compare the fraction of galaxies with a large equivalent width to investigate whether the stellar population of galaxies evolves as a function of redshift. At , our -dropout galaxy sample (48 objects) contains at least 5 galaxies with EW0(Ly) 100 (here we retain SDF J132426.5+271600 (#3) as a large EW object although its EW is slightly below the criterion). However, -band flux of the NB921-depressed -dropout galaxies is strongly boosted by the Ly contamination. After correction for this effect, they may fail to satisfy the -dropout criterion, . Indeed if adopting the NB921 magnitude instead of as a continuum magnitude at the long side of the Lyman break, SDF J132442.5+272423 (#5) should be removed from the SDF -dropout galaxy sample ( ). Therefore we estimate the fraction of -dropout galaxies having EW0(Ly) 100 is 4/48 8%. Note that this estimated fraction is a conservative lower limit, because (1) some Ly non-detected NB921-depressed -dropout galaxies may also have as large an equivalent width as discussed above, and (2) some of the 48 -dropout objects may be emission-line galaxies at lower redshift or Galactic late-type stars, since most of the SDF -dropout objects have not been confirmed by spectroscopic follow-up observations (see, e.g., Stanway et al. 2004).
The derived lower limit of the fraction of -dropout objects having EW0(Ly) 100 is significantly higher than the value at . However, it should be noted that the limiting magnitude of these two samples is different; that is, while the spectroscopic survey of LBGs at by Shapley et al. (2003) reaches down to , our SDF -dropout sample consists of galaxies with . These limiting magnitudes correspond to the absolute UV magnitudes of M1500 = -20.0 and -20.6, respectively, and thus the galaxies in the sample are intrinsically brighter with respect to those in the sample, systematically. Recently, Ando et al. (2006) report that the LBGs with a high Ly equivalent width are rarer in brighter samples than in fainter samples, at . Based on their huge number of a LBG spectroscopic sample, Shapley et al. (2003) also reported that the broad-band magnitude of LBGs with a stronger Ly emission tends to be fainter than that of LBGs with a weaker Ly emission (or with a Ly absorption instead of emission; see Table 3 of Shapley et al. 2003).
Therefore, taking the difference in the limiting magnitude between the samples at and at and the dependence of the equivalent width on the luminosity into account, the difference in the fraction of galaxies with EW0(Ly) 100 between and should be even more significant. Note that the significance may be intrinsically much more if the IGM effect against the Ly transmission is stronger at than at . This result is consistent with a recent study by Shimasaku et al. (2006) that the fraction of LBGs having EW0(Ly) 100 significantly evolves from to . All of these results are consistent with the idea that the typical stellar population of galaxies is significantly younger at than that at . This may also be consistent with recent findings that the slope of the rest-frame UV continuum of some galaxies is bluer at than the typical UV slope at (e.g., Stanway et al. 2004, 2005; Bouwens et al. 2005; Yan et al. 2005).
In earlier papers we identified 8 NB921-depressed -dropout galaxies, which are expected to be strong LAEs at , through the narrow-band and broad-band photometric data of the SDF (Nagao et al. 2004, 2005a). In addition to 3 previously spectroscopically confirmed ones, we found that two other NB921-depressed -dropout galaxies are also LAEs with EW0(Ly) > 100 at z = 6.11 and 6.00, by new optical spectroscopy. This result, combined with our previous spectroscopic runs, means that at least 5 objects among 8 NB921-depressed -dropout galaxies are indeed LAEs having EW0(Ly) 100 at ; these results suggest that the NB921-depressed -dropout selection method is an efficient technique for identifying strong LAEs in a wide redshift range, .
These results also suggest that more than 8% of the -dropout galaxies in the SDF have a large Ly equivalent width of EW0(Ly) 100 . This contrasts with LBGs at , where such strong LAEs are much rarer (1%). This also implies a strong redshift evolution in the Ly equivalent-width distribution from to , consistent with a stellar population of broad-band selected galaxies, which is significantly younger at than at .
This research is based on data collected at the Subaru telescope, which is operated by the National Astronomical Observatory of Japan. We are grateful to the staff of the Subaru telescope and to all members of the Subaru Deep Field project. We also thank the anonymous referee for useful comments. TN and SSS are financially supported by the Japan Society for the Promotion of Science (JSPS) through the JSPS Research Fellowship for Young Scientists.