Subaru Deep Survey I. Near-Infrared Observations

PASJ: Publ. Astron. Soc. Japan 53, 25–36, 2001 February 25 c 2001. Astronomical Society of Japan.
Subaru Deep Survey I. Near-Infrared Observations Toshinori M AIHARA Department of Astronomy, Kyoto University, Kitashirakawa, Sakyo-ku, Kyoto 606-8502 maihara@kusastro.kyoto-u.ac.jp Fumihide I WAMURO, Hirohisa TANABE, Tomoyuki TAGUCHI, Ryuji H ATA Department of Physics, Kyoto University, Kitashirakawa, Sakyo-ku, Kyoto 606-8502 Shin OYA Communications Research Laboratory, Koganei, Tokyo 184-8975 Nobunari K ASHIKAWA, Masanori I YE, Satoshi M IYAZAKI, Hiroshi K AROJI, Michitoshi YOSHIDA Tomonori TOTANI Theoretical Astrophysics Division, National Astronomical Observatory, Mitaka, Tokyo 181-8588 Yuzuru YOSHII Institute of Astronomy, School of Science, The University of Tokyo, Mitaka, Tokyo 181-0015 Sadanori O KAMURA, Kazuhiro S HIMASAKU, Yoshihiko S AITO Department of Astronomy, The University of Tokyo, Hongo, Bunkyo-ku, Tokyo 113-0033 Hiroyasu A NDO, Miwa G OTO, Masahiko H AYASHI, Norio K AIFU, Naoto KOBAYASHI, George KOSUGI, Kentaro M OTOHARA, Tetsuo N ISHIMURA, Jun’ichi N OUMARU, Ryusuke O GASAWARA, Toshiyuki S ASAKI, Kazuhiro S EKIGUCHI, Tadafumi TAKATA, Hiroshi T ERADA, Takuya YAMASHITA, Tomonori U SUDA Subaru Telescope, National Astronomical Observatory, 650 N. A’ohoku Place, Hilo, HI 96720, USA and Alan T. TOKUNAGA Institute for Astronomy, University of Hawaii, 2680 Woodlawn Dr., Honolulu, HI 96822, USA (Received 2000 May 15; accepted 2000 September 1) Abstract Deep near-infrared images of a blank 2′ × 2′ section of sky near the galactic north pole taken by Subaru Telescope are presented. The total integration times of the J and K ′ bands were 12.1 hr and 9.7 hr, resulting in 5 σ limiting magnitudes of 25.1 and 23.5 mag, respectively. The numbers of sources within these limiting magnitudes found with an automated detection procedure are 385 in the J band and 350 in K ′ . Based on photometric measurements of these sources, we present number count vs. magnitude relations, color vs. magnitude diagrams, size vs. color relationships, etc. The slope of the galaxy number count plotted against the AB magnitude scale is about 0.23 in the 22 to 26 AB magnitude range of both bands. The spatial number density of galaxies as well as the slopes in the faint-end region given by the Subaru Deep Field (SDF) survey are consistent with those given by HST–NICMOS surveys, as expressed on the AB magnitude diagram. Several sources having very large J − K ′ color have been found, including a few K ′ objects without detection at J . In addition, a number of faint galactic stars were also detected, most of which are assigned to M-subdwarfs, together with a few brown dwarf candidates. Key words: cosmology: early universe — cosmology: observations — galaxies: evolution — infrared: galaxies — infrared: stars — stars: low-mass, brown dwarfs 1. Introduction Extremely deep imaging of blank fields is a vital method for delineating the nature of the early Universe and gaining general knowledge about the physical conditions at such an early epoch. In this context, the purpose of deep surveys is not only to search for bright, peculiar objects at very high redshift, but also to learn about the overall nature of the early Universe. The optical Hubble Deep Field (HDF) images taken by Hubble Space Telescope have presented views different from the present-day universe. They show that faint, irregular, and smaller galaxy populations seem to have been much more abundant, perhaps inherent to the early Universe. However, since the faint, high-redshift objects seen in the optical HDF images are deemed to represent rest-frame UV emissions, the information is predominantly related to UV-luminous sites, presumably associated with current star formation, rather than the fundamental structure of stellar components in galaxies. On the other hand, a near-infrared deep survey is expected to convey information about the basic galactic structure, or in other words, information related to the fundamental mass distribution. Near-infrared observations may also be crucial to probe galaxies in the most distant region, because the effect of intergalactic reddening; if it occurs, becomes smaller at longer wavelengths. There are already a number of near-infrared surveys Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Optical and Infrared Astronomy Division, National Astronomical Observatory, Mitaka, Tokyo 181-8588 26 T. Maihara et al. [Vol. 53, Table 1. Specifications and performance of CISCO. Item Description Field of view 2′ × 2′ ′′ Pixel scale 0. 116 = pix−1 ′ Filter z , J , H , K , K, NBF2.04, NBF2.12, NBF2.25 zJ , J H , K Grism Limiting magnitude 22.6 mag (K ′ band, 1 hr, S/N = 5) 2. Observations 2.1. Near-Infrared Camera: CISCO A near-infrared imaging and spectroscopic instrument called CISCO (Cooled Infrared Spectrograph/Camera for OHS) was used from 1999 early April to mid-June at the Cassegrain focus of the Subaru Telescope. A detailed description of CISCO has been reported by Motohara et al. (1998). The major specifications are listed in table 1. The filters used in the present deep survey were J (1.16– 1.32 µm) and K ′ (1.96–2.30 µm) bands. A fixed exposure time was employed throughout the observations, namely, 40 s in the J band and 20 s in the K ′ . The exposure times are shorter than the saturation level (about 30%) of the detector readout system, but we adopted them throughout the present survey to secure uniformity in the dithered multiple exposure strategy which we employed (see the following subsection). 2.2. Field Selection Several years ago we chose two regions suitable for a very deep survey, one each in the northern and southern hemi- Fig. 1. Survey area of SDF overlaid on the Digitized Sky Survey (DSS) map. The near-infrared survey is performed in the 2′ ×2′ region marked by a dashed square box. The surrounding region, which we call flanking fields, is also shown enclosed by the solid-line square box. spheres, and designated them as the Subaru Deep Field (SDF). The selected area in the present near-infrared survey is one of the pre-determined blank sky regions near the north galactic pole (NGP), as shown in figure 1. The center coordinates of the observed area of SDF-NGP are RA(2000) = 13h 24m 21.s 38 and DEC(2000) = +27◦ 29′ 23.′′ 0. In selecting the SDF survey region, we placed a requirement that a reasonably bright star be located close to the area to serve as a reference star for possible adaptive-optics-based observations in the near future. The principles and criteria for selecting the SDF regions are described below: i) We chose an independent deep-survey region because the nature (appearance) of the Universe may have different characteristics from one direction to another. ii) The spatial resolution achieved by the Subaru–CISCO combination has proved to be nearly 0.′′ 3 or even better under the best conditions, and about 0.′′ 45 on average, which potentially offers an excellent opportunity to probe remote faint galaxies in the high-redshift Universe. We therefore hope to obtain the deepest survey data in the near-infrared, especially in the 2 µm region, by taking sufficiently long observations. iii) The HDF is at higher airmass at Mauna Kea than that of the SDF. This factor is significant in the latter half of nights from April to June, when the present survey was performed. Note that, although the limiting magnitude of HST–NICMOS is extremely high compared with any ground-based near-infrared imaging, the sensitivity in the K or K ′ band is higher in the ground-based ones due to larger telescope aperture. The sensitivity of the optical bands by Subaru Telescope in terms of the total magnitude is expected to be nearly comparable, or even higher (R band, for example), despite the lower spatial resolution. iv) We should have a reference star nearby for AO-based observations. With the AO system, the spatial resolution is much higher than that Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 (Gardner et al. 1993; Bershady et al. 1998; Yan et al. 1998; Thompson et al. 1999), with spatial coverage, wavelength bands, and limiting magnitudes differing from survey to survey. The observed wavelengths in the near-infrared may correspond to rest-frame wavelengths within a much broader spectral span, if objects with redshifts of 5 or even larger are included. As for the deepest near-infrared survey, NICMOS images of a part of the HDF region have provided deep source counts in both the J and H bands (Thompson et al. 1999). The claimed limiting magnitudes are between 27.5 and 28 AB magnitudes at the 80% completeness level. It is interesting to note that the source count vs. magnitude diagrams of both these bands show an appreciably lower number density of galaxies than previous ground-based results. Since the sensitivity of the NICMOS imager in the K or K ′ band is limited by the thermal radiation of the telescope, extremely deep surveys using ground-based telescopes are important, especially in the K ′ band. Here, we present the K ′ band image as well as that at J , both currently the deepest images taken by a ground-based telescope. In this report, we concentrate on the details of observations using the newly commissioned 8.2 m Subaru Telescope atop Mauna Kea and on data analysis, and then present number count diagrams of galaxies in two near-infrared bands, J (1.25 µm) and K ′ (2.13 µm). We also show the detection and identification of stellar components found in the SDF survey region. The cosmological constants are assumed to be H0 = 65 km s−1 Mpc−1 and q0 = 0.1 throughout this paper. No. 01] Subaru Deep Survey I. Near-Infrared Observations 27 Table 2. Observing log. Band K′ K′ K′ K′ K′ K′ J J J J J K′ K′ K′ J K′ J J K′ K′ Exposure time (s) 20 20 20 20 20 20 40 40 40 40 40 20 20 20 40 20 40 40 20 20 Seeing FWHM(′′ ) 0.35–0.80 0.40–0.70 0.35–0.60 0.70–1.10 0.60–1.20 0.80–1.20 0.45–0.65 0.35–0.50 0.40–0.65 0.45–0.65 0.35–0.50 0.35–0.55 0.25–0.60 0.20–0.30 0.35–0.80 0.35–0.55 0.45–0.75 0.35–0.40 0.50–1.20 0.40–0.65 Fig. 2. J band SDF image. of HST–NICMOS in the H and K bands, and the sensitivity is comparable if the exposure time is the same. v) In selecting SDF, we posed additional requirements, namely, low galactic H I column density, no nearby bright stars and galaxy (except for the reference star for AO), and no known nearby cluster of galaxies. Number of frames 96 × 3 96 × 2 96 × 2.75 96 × 0.5 96 × 4 96 × 2 48 × 2.5 48 × 2.75 48 × 6.5 48 × 2 48 × 2 96 × 1 96 × 6.5 96 × 3.5 48 × 3 96 × 2 48 × 2 48 × 2 96 × 2 96 × 1 Total exposure time (s) 5760 3840 5280 960 7680 3840 4800 5280 12480 3840 3840 1920 12480 6720 5760 3840 3840 3840 3840 1920 Comment test observation flanking field 1–4 flanking field 5–6 flanking field 7–8 flanking field 8 Fig. 3. K ′ band SDF image. 2.3. Method of Dithering Observations We employed the 8-position dithering method in both bands. In the K ′ band case, 12 consecutive frames were taken at each position before offsetting the telescope towards the next position. The order of 8 positions was chosen so as to make a diamond pattern with a diameter of 8′′ –12′′ . Thus, we took 96 frames of 20 s exposures in a single set of dithering observations, resulting in a total on-source integration time of 1920 s. Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Date (UT) [99/2/27] 99/4/ 3 99/4/ 4 99/4/ 9 99/4/25 99/4/25 99/4/29 99/4/30 99/5/ 2 99/5/ 6 99/5/ 7 99/5/ 8 99/5/11 99/5/27 99/6/ 6 99/6/ 7 99/6/ 7 99/6/ 9 99/6/20 99/6/22 28 T. Maihara et al. [Vol. 53, Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Fig. 4. Two-color image composed from the J and K ′ band images. The image sizes are normalized to 0.′′ 45 in FWHM. The entire process of taking 96 exposures was organized by an “abstract command” (kind of a macro command) issued by the observer workstation, which dispatched individual commands to the telescope control system, such as the auto-guiding system, the infrared camera controller, and the telescope itself. Normally, we made observations in one of the two photometric bands, J or K ′ , in one night throughout the SDF project. When the seeing conditions were fairly good, we preferentially implemented J band observations in the fashion described above. The same integration time of 1920 s in a set of dithering observations was secured, but by adopting a single exposure time of 40 s with 6 consecutive exposures at one position instead of 12 in the K ′ band case, in view of the lower sky background level of the J band. The positional accuracy of each offset movement relative to the origin of coordinates, when the direction is put back, has proved to be on the order of 0.′′ 05 under normal conditions, though sometime it is as large as 0.′′ 2, as measured by the center-of-gravity fluctuation of a bright object on the recorded frames. We determined offset values of each frame by the measured coordinates of a particular star in the frames, not by referencing the recorded FITS data. No. 01] Subaru Deep Survey I. Near-Infrared Observations 16 22 17 23 18 19 24 20 21 25 22 22 23 24 25 23 24 25 50 0 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 Fig. 5. Multiple simulations of photometric measurements for mock objects embedded in survey images. An enlarged diagram with numerical fractions in the 0.5 mag cells is shown in the upper-right corner to show how the measured total magnitude is distributed against the model sources. The lower panel shows diagrams of the calculated completeness in both the J band (filled triangles) and the K ′ band (filled circles). 2.4. Log of Observations Observations of the SDF survey started on 1999 April 3 and ended on 1999 June 22, using available nights for CISCO observations during the telescope commissioning phase. Since the SDF survey was supported by the Subaru Project Office as a priority project during this period, we devoted almost all nights with observing conditions better than average to the SDF work. The observing log is presented in table 2. Note that the nominal date in the table corresponds to the starting UT time of each night. During this period, the best seeing data were recorded on May 26; a couple of frames of a field star with 20 s exposures had PSFs of ∼ 0.′′ 20 at FWHM. Although the individual frames have different image resolutions, mostly attributed to the dayto-day variations in seeing conditions, we just accumulated all of the acquired frames without considering the difference in seeing in the present analyses. Thus, the present near-infrared SDF survey has ended up with the target dedicated time of 12.1 hr in the J band and 9.7 hr in the K ′ band. 3. Data Reduction and Source Detection Using all of the frames taken in a set of automatic dithering observations, we first created a sky frame out of 96 dithered exposures by the standard median sky method. In this procedure, each frame is first divided by the flat frame and is also corrected by a bias frame. The flat frame is produced from a large number of sky frames by a method similar to the coadding method used for creating the standard median sky, but using all of the frames obtained in the long-term SDF observations. In generating the sky frame, the contribution of in- dividual stars was maximally reduced by masking all of the discernible stars. Then, we co-added all of the sky-subtracted frames in which the offsets of the shift-and-add procedure had been determined at a sub-pixel level from the bright sources in each frame. Because the pixel values with scatters exceeding 3 σ were rejected, possible cosmic ray events or rare pixel events due to unstable behavior were excluded. The reduced J and K ′ band images are shown in figures 2 and 3, respectively; and the color composite created from these two bands is presented in figure 4. Throughout these images, the center position is 13h 24m 21.s 38 in RA and +27◦ 29′ 28.′′ 3 in Dec at the epoch of 2000.0, as determined by the astrometric measurement based on the position of a bright reference star located in the northeastern corner of the SDF field, which is just trimmed out in these figures. The unvignetted area presented in figure 2 is 118′′ × 114′′ with a pixel scale of 0.′′ 116 per pixel. The stellar image sizes of figures 2 and 3 are 0.′′ 45 in FWHM for the J band, and 0.′′ 35 for the K ′ band. In producing the combined color image of figure 4, we applied a smoothing filter to the K ′ image to match the image size in J of 0.′′ 45 at FWHM. For source detection and photometry on the reduced SDF frames, we employed a routine called SExtractor, developed by Bertin and Arnouts (1996). Before applying it, images of both bands were smoothed out by a Gaussian filter that made the image resolution 0.′′ 55 at FWHM, because we have learned that filtered images give optimal source detection capability with less spurious source detection by iterative trials of SExtractor. We define the detection threshold as the 1.50 σ level of the surface brightness fluctuation of the sky, which corresponds to 25.59 mag arcsec−2 in the J band and to 24.10 mag arcsec−2 in the K ′ band. If an assemblage of at least 18 pixels, which are connected to each other, has an excess signal over the thresholds, we regarded it as positive detection. We have thus defined the isophotal magnitude by integrating the signal within the region that exceeds the threshold level. In addition, we have listed the total magnitude as well as the aperture magnitude in our primary catalog. The total numbers of sources detected and cataloged through the above procedures are 911 and 939 in the J and K ′ bands, respectively. Note that the catalog may contain spurious sources due to the effect of noise. Regardless of the reliability of each source, we have accomplished photometric calibration of the data by a reference star in the same frame whose brightness has been determined by measuring standard stars: FS 23 and FS 27, selected from the UKIRT faint star catalog (Casali, Hawarden 1992: JCMT–UKIRT Newsletter 4, 33). The resultant calibrated catalog will be reported elsewhere when optical data are obtained. To obtain a diagram of number count versus total magnitude, we have to assess the spurious detection of sources due to the effect of statistical noise and also the error of source photometry. For noise evaluation, we first created a reference frame (artificial blank sky) using SDF raw frames, but with all traces of detected objects removed. To ensure removal of much fainter objects from the reference frame, we also referred to the co-added object frames for identification. The final blank-sky frames in both bands were thus obtained by co-adding all the frames without adjusting the dithering offset. Then, by applying SExtractor to the sky frame, we evaluated the rate of spu- Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 26 100 29 30 T. Maihara et al. [Vol. 53, 1000 100 100 10 10 1 1 0.1 0.1 15 16 17 18 19 20 21 22 23 24 25 26 27 Fig. 6. Number count vs. magnitude diagram for the J band. The number counts corrected for completeness are plotted with filled symbols, while the raw counts are open symbols. The circles, triangles, and squares denote total, isophotal, and aperture magnitudes, respectively. rious source detection, and thus obtained the correction factor for the raw result. It was found that spurious source detection tends to affect the source count discussed below, but only in the S/N < 5 range. In the next step, a number of artificial objects (mock galaxies) with a wide range of brightness as well as source size were embedded in the real image frames. The brightness distribution of the objects in the faint region are assumed to be represented by a slope of ∼ 0.23 (derived from the raw data with S/N > 5). In the course of this calculation, the SExtractor software was applied multiple times to establish a relationship between the input photometric brightnesses and the measured ones. The results of simulations are shown in figure 5, where each dot represents the measured magnitude for an input object of a given magnitude. The relation is expressed by a matrix-type operator. This method has been employed in galaxy count studies (e.g., Smail et al. 1995; Minezaki et al. 1998). Based on the matrix, we can evaluate errors in photometry, and also correct the raw number count data using the completeness curves, as shown in the lower panel of figure 5. The source counts against magnitude, with and without a correction for completeness, are plotted in figures 6 and 7, where data points of the total, isophotal, and aperture magnitudes are presented for a comparison. As can be seen from these figures, the correction becomes significant at magnitudes larger than 25.5 mag in J and 24 mag in K ′ . If we define a S/N of 5 for definite detection, the limiting magnitudes are 25.1 mag in J and 23.5 mag in K ′ , with the number of sources being 385 and 350, respectively. The magnitudes corresponding to detection completeness of 50% are 24.4 mag in K ′ and 25.8 mag in J . 14 15 16 17 18 19 20 21 22 23 24 25 26 Fig. 7. Same as figure 6, but for the K ′ band. 4. Results and Discussion 4.1. Corrected Number Counts The corrected number counts in both the J and K ′ bands are tabulated in table 3. In figures 8 and 9, we plot them for J band sources and for K ′ band sources, respectively. Also plotted are those of other surveys taken from the literature. In the table and figures, our data refer to the total magnitude, while some of the other data points are defined by the aperture magnitude. The source sizes in the faintest magnitude range are very small and, in general, are smaller than the aperture adopted in most aperture photometry. This means that the aperture magnitude is virtually the same as the total magnitude, since the aperture is normally taken to be larger than the seeing size. In this first report of the SDF survey program, we shall concentrate mostly on those sources having relatively high S/N-ratios; i.e., S/N of ∼ 5 or more. We will prepare a separate paper in which the galaxy counts, colors, and morphologies at the faint end will be examined. Here, it should be noted that the contribution of stars to the number counts is estimated by a simulation, and that the point sources (14 objects in the SDF field have been identified) have been excluded. The actual procedure for the simulation, by which a criterion is established to identify stellar objects, will be described in a later section. Some quasi-stellar objects (QSOs) may also be included in the identified stellar objects. However, their contribution to the number counts should be smaller than that of stars in view of the result found by Huchra and Burg (1992), who reported that the fraction of all types of Seyfert galaxies at the absolute magnitude limit of −20.0 is about 1.3% in the CfA redshift survey. Brighter objects with a stellar appearance are expected to be less abundant than Seyfert galaxies. In figure 10, we plot the galaxy number counts against the AB magnitude scale to compare the slope as well as the absolute number density of different photometric bands obtained Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 1000 No. 01] Subaru Deep Survey I. Near-Infrared Observations 31 Table 3. Corrected number counts. Magnitude J band Error 0.000e+00 0.000e+00 0.000e+00 2.500e+02 1.942e+03 2.725e+03 3.354e+03 2.776e+03 5.044e+03 5.162e+03 6.705e+03 7.789e+03 9.255e+03 1.070e+04 1.224e+04 1.333e+04 1.524e+04 1.905e+04 2.949e+04 5.623e+04 1.227e+05 Bershady et al. 1998 Saracco et al. 1999 Thompson et al. 1999 This work 1000 100 15 Jenkins & Reid 1991 Gardner et al. 1993 Glazebrook et al. 1994 McLeod et al. 1995 Moustakas et al. 1997 Bershady et al. 1998 Minezaki et al. 1998 Saracco et al. 1999 This work 1000 16 17 18 19 20 21 22 23 24 25 26 27 Fig. 8. Galaxy number count vs. magnitude of the J band. The data and authors of previous surveys are shown in the panel. so far. Here the HST–NICMOS J and H band data are taken from Yan et al. (1998) and Thompson et al. (1999). In these figures, previously published data of galaxy number counts are also plotted. The spatial number density of galaxies in the SDF appears to be a good match with that given by Yan et al. (1998), who presented the H band number count obtained with HST–NICMOS operating in the so-called parallel mode. It is also consistent with Thompson et al. (1999) at the faint end (at H = 23.5–26 AB mag), but deviates greatly at magni- 100 14 Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Count 0.000e+00 0.000e+00 0.000e+00 3.231e+01 1.949e+03 3.851e+03 5.769e+03 3.996e+03 1.314e+04 1.368e+04 2.296e+04 3.099e+04 4.357e+04 5.835e+04 7.499e+04 8.759e+04 1.131e+05 1.615e+05 2.160e+05 2.624e+05 3.451e+05 16.25 16.75 17.25 17.75 18.25 18.75 19.25 19.75 20.25 20.75 21.25 21.75 22.25 22.75 23.25 23.75 24.25 24.75 25.25 25.75 26.25 K ′ band Count Error 1.918e+03 1.931e+03 1.358e+02 5.118e+02 5.635e+03 3.293e+03 5.755e+03 3.338e+03 7.721e+03 3.869e+03 9.653e+03 4.326e+03 1.346e+04 5.100e+03 1.915e+04 6.111e+03 2.885e+04 7.513e+03 4.757e+04 9.647e+03 5.145e+04 1.006e+04 7.073e+04 1.187e+04 8.419e+04 1.307e+04 1.066e+05 1.488e+04 1.487e+05 1.873e+04 1.962e+05 3.129e+04 2.548e+05 5.648e+04 3.421e+05 1.031e+05 4.793e+05 2.214e+05 — — — — 15 16 17 18 19 20 21 22 23 24 25 26 Fig. 9. Same as figure 8 but for the K ′ band. tudes brighter than 22.5 mag, possibly due to lower statistics. In contrast, the absolute number densities given by some of the past ground-based surveys differ substantially from that of the present SDF survey in the magnitude range ≥ 23 AB mag. It would still be necessary to increase the survey areas as well as to obtain higher S/N ratios to examine the possible structural inhomogeneity of the Universe. Nevertheless, the SDF area is so far the largest among deep near-infrared surveys, and therefore should be a more accurate representation of the global distribution of galaxies in the Universe. Another important result 32 T. Maihara et al. 17 18 19 20 21 22 23 24 25 26 Mobasher et al. 1986 Jenkins & Reid 1991 Gardner et al. 1993 Glazebrook et al. 1994 McLeod et al. 1995 Moustakas et al. 1997 Bershady et al. 1998 Minezaki et al. 1998 Saracco et al. 1999 This work 27 28 10 15 20 25 Fig. 10. AB magnitude plots of J , H , and K ′ band data. Previous near-infrared surveys are also plotted for a comparison. Fig. 11. Integrated flux of SDF sources (filled circles) in each magnitude bin. The fluxes of other survey data are also shown. derived from figures 8, 9, and 10 is that the near-infrared color J −K ′ is almost constant, at least in the AB mag range from 22 to 25.5 mag, and the median color is ∼ 1.4, as described next. which assumes that all galaxies are small in size. Since the completeness correction may be larger than this, if there are galaxies with larger spatial extents, and thus with lower surface brightnesses, the above estimate may be an underestimate of the K ′ band EBL. However, as can be seen in figure 11, because the bulk of EBL comes predominantly from relatively bright galaxies at K ′ ∼ 15–20, we consider that the uncertainty in the counts at the faintest magnitudes in the SDF does not significantly change the above estimate of the EBL. 4.2. Extragalactic Background Light (EBL) As for the slope of the galaxy count in the K ′ band, we have to be careful in interpreting it because we have assumed a slope of 0.23 in the model source number count in the simulation process. This affects the result at the faint end, although the slope is iteratively corrected and has converged at 0.23. It should also be noted that this slope is derived by applying the correction factor of completeness, in which we have assumed, for simplicity, that these faint galaxies are point sources, since apparent sizes of the detected faint sources are sufficiently small. Figure 11 shows the contribution to the EBL in the K ′ band as a function of the apparent magnitude. This figure shows that we have already resolved the EBL consisting of discrete galaxies in this band. The optical galaxy counts in the HDF have also shown such a signature (see Pozzetti et al. 1998). The present result of the SDF in the near-infrared band gives clear evidence that the bulk of EBL is contributed by an integration of fairly bright discrete sources, consistent with a similar diagram shown by Pozzetti et al. (1998), who compiled available K band data. We can now evaluate the surface brightness as EBL by integrating individual sources over a wide magnitude range of 10 ≤ K ′ ≤ 25. The estimate using our data as well as published count data is ∼ 5.1 × 10−20 erg cm−2 s−1 Hz−1 sr−1 in the K ′ band. The major contribution to EBL is made by fairly bright galaxies of about K ′ ∼ 15–20 mag. Note that the contribution of sources fainter than K ′ = 25 mag to the surface brightness is small, since the slope is no more than 0.23. Extrapolation both into the faint and bright end adds at most ∼ 5% of the above value of the EBL flux. It should be noted that these estimates are based on a completeness correction, 4.3. Near-Infrared Color and Morphology of Faint Galaxies Figure 12 shows the J − K ′ color vs. K ′ magnitude diagram of near-infrared sources. Here, we have omitted sources assigned as stellar objects. Since the photometric aperture of the total magnitude in the J band is not always the same as for the K ′ band, it is necessary to define a color with the same aperture of the two bands. Therefore, we have calculated colors of smaller sources in figure 12 in terms of the photometric magnitudes defined by 10 pixels, i.e., a 1.′′ 16 diameter. Even if an object is picked up only in the J band and is not detected by our criteria in the K ′ band, a K ′ photometric brightness is artificially given by measuring the encircled intensity with the same aperture as that defined in the J band, thus providing the J − K ′ color of the sources. The median color is shown by the filled squares in figure 12, while the thin lines represent color–magnitude curves for four categories of galaxy (E/S0, Sbc, Scd, and Irr), all of which were drawn by applying only the K-correction (Coleman et al. 1980; Yoshii, Takahara 1988). Pozzetti et al. (1996) presented a study of the pure luminosity evolution (PLE) model to examine faint galaxy count data from the U to the K band in which luminosity evolution as well as mild spectral evolution are incorporated; they have shown that simple PLE models are in general considered to be as baseline models of faint galaxy counts. However, as they also noted, the discrepancy between the PLE Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Jenkins & Reid 1991 Gardner et al. 1993 McLeod et al. 1995 Moustakas et al. 1997 Bershady et al. 1998 Minezaki et al. 1998 Saracco et al. 1999 Yan et al. 1998 Thompson et al. 1999 This work 1000 100 16 [Vol. 53, No. 01] Subaru Deep Survey I. Near-Infrared Observations 7 6 5 4 3 2 1 E/S0 Sbc Scd Irr -1 -2 -3 16 17 18 19 20 21 22 23 24 25 26 27 Fig. 12. Color–magnitude relation for the SDF sources. Objects detected in both the J and K ′ bands are plotted by open circles. Objects detected either in the J or K ′ band are represented by “×” or “+” marks, respectively. The thin solid lines are the color change vs. magnitude for representative types of galaxy assuming a simple no-evolution model based on SED models of Yoshii and Takahara (1988) and Coleman, Wu, and Weedman (1980). The solid squares show the median color with the standard deviation. Note that objects detected only in the J or K ′ band have been photometrically measured with the same aperture sizes to determine the K ′ or J band magnitudes, respectively, as explained in the text. models and the K band data is more significant than in the optical bands. One may notice that, as shown in figure 12, the median color of sources is fairly constant up to about K ′ = 22 mag, and then becomes slightly redder as the brightness gets fainter, although the standard deviation is large. As for the color of faint galaxies, Saracco et al. (1999) argued that the median J − Ks color of galaxies becomes redder from 1.1 to 1.5 up to Ks = 19 mag, and then tends to be somewhat bluer in the fainter magnitude range. Such a trend is not necessarily inconsistent with figure 12, but it is noteworthy that the color is, by and large, constant at least in the magnitude range where the selection effect is still small (S/N > 5 for both bands), but with significant statistics. In order to interpret the observed galaxy counts as well as the color–magniture relation, it is necessary to introduce galaxy evolution models in which different formation epochs (zF ) for different galaxy categories are presumed. Such quantitative analyses for the present J − K ′ color vs. luminosity relation will be discussed in our forthcoming paper. In figure 12, sources with J − K ′ colors redder than 2.5 are discerned, where 4 objects out of 9 are not detected in the J band. These large J − K ′ color objects in the faint region at about K ′ = 23 mag are likely to be remote galaxies, as judged from the apparent spatial extent as well as the derived “stellarity” index determined by SExtractor. Some of the reddest objects in the SDF survey images are tabulated in table 3 and also shown in figure 13. The listed objects are relatively bright in K ′ (< 22.5 mag) and have J − K ′ colors equal to or larger than 2.8. The object on the left appears to be a merger system, and the third one from the left may represent an interacting system, although it is possible that they could just appear as close neighbors. Extremely red objects with unprecedentedly large J − K ′ colors located in the faint-magnitude domain are currently given special attention in connection with galaxy formation in the earliest epoch. Dickinson et al. (2000) found an unusual infrared object in the HDF North field detected only in the H and Ks bands with no detectable fluxes shorter than the J band. It has the H − K color of about 1.2 with the J − H color limit of nearly 3. They discuss three possible interpretations, which are: i) a dusty z > 2 galaxy, ii) an old elliptical at z > 3, or iii) a z > 10 Lyman break galaxy. Similar objects were reported by Yahata et al. (2000), who list possible extremely high redshift galaxies in the HDF South NICMOS field. Nine objects appear to have a break between 1 and 2 µm. Redshifts of the sources have been derived from U to K band data by a method of photometric redshift determination that spans from z = 7.66 to 15.45. Since we have so far not acquired any data in the optical bands for the extreme SDF objects, we cannot infer photometric redshifts on a firm basis. Nevertheless, infrared color as large as J − K ′ ≥ 2.8 suggests that these objects belong to the same population of the HDF-South NICMOS objects having a redshift z ≥ 10. It is however necessary to obtain optical band photometric information, as well as spectroscopic data with highly sensitive instruments, such as the OHS, for further examining these objects. We can determine whether they are moderately redshifted (z ∼ 2–4) ellipticals or unprecedentedly high-redshift galaxies. 4.4. Faint Stellar Populations in the Galaxy In order to extract stars from the table of SDF sources detected at S/N > 5, we have developed a procedure of star– galaxy separation based on a detectivity test using mock stars. It is similar to the completeness test of the source count in conjunction with the SExtractor. The criteria for identifying stars from SDF sources are basically expressed by the following two conditions. The first one is that the FWHMs are smaller than 0.′′ 47 and 0.′′ 34 for bright sources in the J and K ′ bands, respectively. In addition, based on a simulation, it is found that the limiting sizes should be corrected in the fainter source region, by adding terms: i.e., 10(m−me )/2.5 , where m is the magnitude of the source, and me = 26 and 25 mag for the J and K ′ bands, respectively. The second condition is that the stellarity index which is given by SExtractor, is larger than 0.8. By adopting these criteria, 14 objects are classified as stars on a fairly firm basis. Naturally, possible stellar objects with stellarity indexes larger than 0.8 could be excluded due to the first criterion. See Nakajima et al. (2000) for the detailed procedure of galaxy– star separation we have developed. The completeness of the identification of stars has also been estimated with this simulation, which is about 60% for sources brighter than 24 mag in the J band, or 23 mag in the K ′ band. We plot the SDF objects in the FWHM vs. J − K ′ color diagram, as shown in figure 14, where objects satisfying the above criterion are marked by open star symbols. It is interesting to note that several relatively bluer (J − K ′ ∼ 0.6) stellar objects appear to concentrate in the Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 0 33 34 T. Maihara et al. 2 3 4 Fig. 13. Examples of images of the four reddest objects found in the SDF. J band images are shown in the upper row, and K ′ band images are in the lower row. 2.5 2 1.5 1 0.5 0 -1 0 1 2 3 4 Fig. 14. FWHM vs. J − K ′ color diagram. FWHM values are adopted from K ′ band images. The open stars represent objects identified as stars on the basis of the proclaimed criteria (see text). The open circles are objects detected in both bands, while the “×” and “+” marks represent objects selected only in the J band and K ′ band, respectively. lower left: a well-confined region on the diagram. These are presumably extreme M subdwarf (ESD) stars, as classified by Leggett, Allard, and Hauschildt (1998). The extreme M subdwarfs are supposed to be members of the galactic halo and have very low metallicity. They are listed in the tables of lowmass stars presented by Leggett, Allard, and Hauschildt (1998), and are characterized by effective temperatures of about 3000 K and by masses of 0.09–0.15 M⊙ . From their photometric data it is seen that the brightness range of ESD stars is from K = 11 to 15 mag and that the J − K ′ color is 0.65 ± 0.15. In view of these, some 10 stars in the confined region of figure 14 are most probably extreme M subdwarfs. Since their absolute magnitudes span from MJ = 9.5 to 10, we should have reached about 8 kpc to capture ESD stars by the present observations with K ′ band-limiting magnitudes of ∼ 24.5 mag. The estimated volume density due to these stars is 2.5×10−4 M⊙ pc−3 . Even if the density of stars is extended uniformly to 30 kpc, the total mass is slightly lower than 1011 M⊙ , so that it could not be the majority of galactic dark matter. Another group of stellar components, significantly redder than M subdwarfs, is noticed in figure 14 in the J − K ′ color range from 1 to 1.5. It is likely that they correspond to the L-type dwarfs defined by Kirkpatrick et al. (1999), who have found very red stellar objects with J − Ks colors from 1.3 to 2.1 (practically the same color as the J − K ′ color). They have claimed that at least one third of L-type stars show lithium absorption, and that they are definitely in the category of brown dwarfs. On the other hand, note that objects with J −K ′ color of ∼ 0 and small FWHM values are plotted in figure 14. In fact, they had a fairly large stellarity index (> 0.8) in the J band, but were dropped from identification due to faintness in the K ′ band. The color is consistent with a T-type brown dwarf, GL 229B (Nakajima et al. 1995; Kirkpatrick et al. 1999). However, it is necessary to prove the nature of these L-type and T-type candidates through future multi-wavelength photometric and spectroscopic observations. Related discussions on the stellar members in SDF will be presented in Nakajima et al. (2000). Finally, it is worth noting that the SDF sources have a connection to old, cool white dwarfs, which have recently drawn attention because they might account for most of the hidden baryonic mass of the Galaxy. Hodgkin et al. (2000) reported a cool white dwarf showing an extraordinary spectrum affected by the collision-induced absorption by hydrogen molecules, i.e., very red in the optical region, but extremely blue at wavelengths longer than 1 µm, with a J − K ∼ −1.4. In addition, Harris et al. (2000) identified LHS 3250 as a very cool white dwarf with a J −K of −0.86. A couple of old white dwarf candidates were also found with HDF frames taken at a 2-year interval, which are believed to be halo members, as inferred from proper motion data (Ibata et al. 1999). These candidates were spectroscopically shown to be white dwarfs (Ibata et al. 2000). Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 1 [Vol. 53, No. 01] Subaru Deep Survey I. Near-Infrared Observations 35 Table 4. K ′ magnitude and color of extremely red objects. ID No.* 1 2 3 4 Position† (2000.0) 13 24m 22.s 38 +27◦ 29′ 49.′′ 5 13h 24m 22.s 39 +27◦ 29′ 01.′′ 9 13h 24m 21.s 16 +27◦ 29′ 01.′′ 9 13h 24m 22.s 84 +27◦ 30′ 08.′′ 4 h K′ 20.91 (±0.05) 22.03 (±0.09) 21.99 (±0.05) 22.31 (±0.14) J − K′ 2.97 (±0.14) 3.65 (±0.40) 2.81 (±0.20) 4.12 (±1.04) * ID numbers assigned to objects are from left to right in figure 13. Astrometry of these objects was made by the coordinates of an HST guide star found in the flanking field. The estimated accuracy is ± 0.′′ 15. † 5. Summary A deep near-infrared survey with the newly commissioned Subaru Telescope has been reported. We present two color data of J and K ′ bands as deep images as well as diagrams of the galaxy number count vs. magnitude, of color vs. magnitude, and of size vs. color. i) It is found that slopes of galaxy number count plotted against the AB mag scale in both the J and K ′ bands are about 0.23 in the 22 to 26 AB mag range, which remain the same up to the faint end without a significant change. ii) From this result, we argue that the integrated surface brightness of faint SDF sources does not make an appreciable contribution to the extragalactic background light (EBL); in other words, the total EBL contributed by galaxies up to the faint end has almost been completely resolved. However, measurements of diffuse EBL at this wavelength, if performed with sufficient precision, may pose a crucial cosmological issue regarding light sources other than individual galaxies. iii) The color–magnitude diagram shows a fairly constant J − K ′ color (∼ 1.5) up to about K ′ ∼ 23 mag, although a scattered distribution, especially toward redder color, is noticed. vi) Intriguing objects with extremely red J − K ′ colors are found in the SDF region. The listed objects are relatively bright (S/N > 5), and thus are safely classified as galaxies. The possibility of very high redshift objects, that is, candidates of Lyman break galaxies, are briefly discussed. v) A certain number of stellar sources have been identified, most of which are supposed to be M subdwarfs having colors of J − K ′ ∼ 0.6. These are considered to be located at a median distance of about 2 kpc and are expected to provide samples for further studies of stars of this class for the purpose of examining a luminosity function as well as a contribution to the Galaxy mass. Brown dwarfs of the T-type may also be included in the detected source list. This should be confirmed by spectroscopic measurements. Finally, it should be noted that the present near-infrared SDF survey work is the first step of the Subaru Deep Survey project planned to perform using other facility instruments of the Subaru Telescope. For instance, follow-up observations of optical deep imaging with an optical spectrograph/camera, called FOCAS (Kashikawa et al. 2000), will provide essential data to determine photometric redshifts of detected nearinfrared sources. Objects as faint as H = 21.5 or J = 22 mag are well within the feasible range of spectroscopic observations with the OH-airglow suppression spectrograph (OHS), which has a unique capability of obtaining 1.1 to 1.8 µm spectra of 16 objects within a 3′ field simultaneously, giving crucial information about the SDF sources. We deeply appreciate the devoted support of the Subaru Telescope staff for this project. We also thank the engineering staffs of Mitsubishi Electric Co. and Fujitsu Co. for technical assistance during the observations. The authors would like to acknowledge helpful discussion about stellar components with Tadashi Nakajima and Takashi Tsuji. References Bershady, M. A., Lowenthal, J. D., & Koo, D. C. 1998, ApJ, 505, 50 Bertin, E., & Arnouts, S. 1996, A&AS, 117, 393 Coleman, G. D., Wu, C.-C., & Weedman, D. W. 1980, ApJS, 43, 393 Dickinson, M., Hanley, C., Elston, R., Eisenhardt, P. R., Stanford, S. A., Adelberger, K. L., Shapley, A., Steidel, C. C., et al. 2000, ApJ, 531, 624 Gardner, J. P., Cowie, L. L., & Wainscoat, R. J. 1993, ApJ, 415, L9 Glazebrook, K., Peacock, J. A., Collins, C. A., & Miller, L. 1994, MNRAS, 266, 65 Harris, H. C., Dahn, C. C., Vrba, F. J., Henden, A. A., Liebert, J., Schmidt, G. D., & Reid, I. N. 2000, ApJ, 524, 1000 Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 These findings support the idea that very old white dwarfs could be responsible for a substantial portion, but not all, of the galactic dark matter. In this connection, we can list only a few stellar objects (only one below J = 25.0) having a stellarity index of 0.8 or larger, and with very blue color. This scarcity is consistent with the rough estimate by Hodgkin et al. (2000), who gave ∼ 7 × 10−3 pc−3 as the number density of white dwarfs with a typical mass of 0.5 M⊙ in the solar neighborhood. If the absolute J magnitude MJ , of white dwarfs is assumed to be the same as WD 0346+246 of Hodgkin et al. (2000) with the detection limit being mJ = 24.75, the viewable radial distance would be about 750 pc. The survey volume of the 2′ × 2′ SDF field becomes ∼ 50 pc3 , which is about half of the volume of 109 pc3 surveyed by Hodgkin et al. In order to put a definite constraint on the number density of the white-dwarf population, it is necessary to increase the survey region and also to perform an optical survey. Since the magnitudes in the optical bands are roughly the same as in the J band (∼ 25 mag), such dwarf stars will be seen as optical images, by which the peculiar photometric color could be determined. 36 T. Maihara et al. Moustakas, L. A., Davis, M., Graham, J. R., Silk, J., Peterson, B. A., & Yoshii, Y. 1997, ApJ, 475, 445 Nakajima, T., Iwamuro, F., Maihara, T., Motohara, K., Tsuji, T., Tamura, M., Kashikawa, N., & Iye, M. 2000, AJ, 120, 2488 Nakajima, T., Oppenheimer, B. R., Kulkarni, S. R., Golimowski, D. A., Matthews, K., & Durrance, S. T. 1995, Nature, 378, 463 Pozzetti, L., Bruzual, A. G., & Zamorani, G. 1996, MNRAS, 281, 953 Pozzetti, L., Madau, P., Zamorani, G., Ferguson, H. C., & Bruzual, G. 1998, MNRAS, 298, 1133 Saracco, P., D’Odorico, S., Moorwood, A., Buzzoni, A., Cuby, J.-G., & Lidman, C. 1999, A&A, 349, 751 Smail, I., Hogg, D. W., Yan, L., & Cohen, J. G. 1995, ApJ, 449, L105 Thompson, R. I., Storrie-Lombardi, L. J., Weymann, R. J., Rieke, M. J., Schneider, G., Stobie, E., & Lytle, D. 1999, AJ, 117, 17 Yahata, N., Lanzetta, K. M., Chen, H.-W., Fernández-Soto, A., Pascarelle, S. M., Yahil, A., & Puetter, R. C. 2000, ApJ, 538, 493 Yan, L., McCarthy, P. J., Storrie-Lombardi, L. J., & Weymann, R. J. 1998, ApJ, 503, L19 Yoshii, Y., & Takahara, F. 1988, ApJ, 326, 1 Downloaded from https://academic.oup.com/pasj/article/53/1/25/1552097 by guest on 01 July 2022 Hodgkin, S. T., Oppenheimer, B.R, Hambly, N. C., Jameson, R. F., Smartt, S. J., & Steele, I. A., 2000, Nature, 403, 57 Huchra, J., & Burg, R. 1992, ApJ, 393, 90 Ibata, R. A., Irwin, M., Bienaymé O., Scholz, R., & Guibert, J. 2000, ApJ, 532, L41 Ibata, R. A., Richer, H. B., Gilliland, R. L., & Scott, D. 1999, ApJ, 524, L95 Jenkins, C. R., & Reid, I. N. 1991, AJ, 101, 1595 Kashikawa, N., Inata, M., Iye, M., Kawabata, K., Okita, K., Kosugi, G., Ohyama, Y., Sasaki, T., et al. 2000, Proc. SPIE, 4008, 104 Kirkpatrick, J. D., Reid, I. N., Liebert, J., Cutri, R. M., Nelson, B., Beichman, C. A., Dahn, C. C., Monet, D. G., Gizis, J. E., & Skrutskie, M. F. 1999, ApJ, 519, 802 Leggett, S. K., Allard, F., & Hauschildt, P. H. 1998, ApJ, 509, 836 McLeod, B. A., Bernstein, G. M., Rieke, K. J., Tollestrup, E. V., & Fazio, G. G. 1995, ApJS, 96, 117 Minezaki, T., Kobayashi, Y., Yoshii, Y., & Peterson, B. A. 1998, ApJ, 494, 111 Mobasher, B., Ellis, R. S., & Sharples, R. M. 1986, MNRAS, 223, 11 Motohara, K., Maihara, T., Iwamuro, F., Oya, S., Imanishi, M., Terada, H., Goto, M., Iwai, J., et al. 1998, Proc. SPIE, 3354, 659