Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://www.stecf.org/goods/spectroscopy/CDFS_Mastercat/croom01.pdf Äàòà èçìåíåíèÿ: Thu Sep 16 11:25:44 2004 Äàòà èíäåêñèðîâàíèÿ: Tue Oct 2 01:00:24 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ï ð ï

Mon. Not. R. Astron. Soc. 328, 150­158 (2001)

A small-area faint KX redshift survey for QSOs in the ESO Imaging Survey Chandra Deep Field South
Scott M. Croom,1,2P S. J. Warren1 and K. Glazebrook
1 2

2,3

Astrophysics Group, Imperial College of Science, Technology and Medicine, Blackett Laboratory, Prince Consort Road, London SW7 2BW The Anglo-Australian Observatory, PO Box 296, Epping, NSW 2121, Australia 3 Department of Physics & Astronomy, Johns Hopkins University, 3400 North Charles Street, Baltimore, MD 21218-2686, USA

Accepted 2001 July 20. Received 2001 July 19; in original form 2001 January 21

A B STRA CT

In this paper we present preliminary spectroscopic results from a small-area faint K-excess (KX) survey, and compare KX selection against UVX selection. The aim of the KX method is to produce complete samples of QSOs that are flux-limited in the K band, in order to minimize any selection bias in samples of QSOs from the effects of reddening and extinction. Using the photometric catalogue of the ESO Imaging Survey Chandra Deep Field South (48 arcmin2) we have identified compact objects with J 2 K colours redder than the stellar sequence that are brighter than K ¼ 19:5. We have obtained spectra of 33 candidates, using the LDSS11 spectrograph on the Anglo-Australian Telescope (AAT). Amongst the 11 bluer candidates, with V 2 J , 3, three are confirmed as QSOs. Identification of the 22 redder candidates with V 2 J > 3 is substantially incomplete, but so far no reddened QSOs have been found. Near-infrared spectroscopy will be more effective in identifying some of these targets. Only two UVX ÏU 2 B , 20:2÷ sources brighter than K ¼ 19:5 are found that are not also KX selected. These are both identified as galactic stars. Thus KX selection appears to select all UVX QSOs. The surface density of QSOs in the blue subsample ÏV 2 J , 3÷ at K < 19:5 is 3251316 deg22 . Because identification of the red subsample ÏV 2 J > 3÷ is 2177 substantially incomplete, the 2s upper limit on the density of reddened QSOs is large, ,1150 deg22. As anticipated, at these faint magnitudes the KX sample includes several compact galaxies. Of the 14 with measured redshifts, there are roughly equal numbers of early- and late-type objects. Nearly all the early-type galaxies are found in a single structure at z ¼ 0:66. Key words: surveys ­ galaxies: active ­ quasars: general ­ infrared: galaxies.

1

I NTR O DU C T IO N

The most commonly used method for finding QSOs utilizes the fact the QSOs are bluer than most stars at ultraviolet wavelengths. This ultraviolet excess led to the UVX technique (e.g. Schmidt & Green 1983), which in its simplest form uses the blue U 2 B colours of QSOs to differentiate them from stars. This and related methods have been utilized in most of the largest QSO surveys to date [e.g. Durham/Anglo-Australian Telescope (AAT), Boyle et al. 1990; 2dF, Croom et al. 2001]. Warren, Hewett & Foltz (2000) have recently proposed the KX technique for producing K-band fluxlimited samples of QSOs: because of the power-law nature of their spectra, QSOs are not only bluer than stars at UV wavelengths, but also redder than stars at near-infrared wavelengths. In considering the KX technique it is important to distinguish between the two
P

E-mail: scroom@aaoepp.aao.gov.au

effects of reddening and extinction (e.g. Binney & Merrifield 1998). KX QSO samples could be useful in overcoming selection biases from both effects. First, the KX method is insensitive to reddening. That is, the reddening vector in the V 2 J versus J 2 K plane does not drive QSOs towards the stellar locus. So the KX method could detect QSOs that have been missed in a survey based on UV ­ optical colours because they are unusually red, even though they are brighter than the flux limit. Secondly, reddened QSOs suffer greater extinction at short wavelengths and so can fall below the optical flux limit, but could be detected in the K band. More generally, because red QSOs are under-represented in surveys at short wavelengths, and similarly blue QSOs are underrepresented in surveys at long wavelengths, it is clear that surveys at two widely separated wavelengths are necessary to characterize fully the optical ­ infrared emission from QSOs. There is currently considerable interest in quantifying the numbers of red QSOs. Red QSOs have been discovered in radio-selected
q 2001 RAS


A small-area faint KX redshift survey
samples (Webster et al. 1995), and highly absorbed QSOs have been invoked to explain the spectrum of the X-ray background radiation (e.g. Madau, Ghisellini & Fabian 1994). The unified model of active galactic nuclei (AGN) (e.g. Antonucci 1993) postulates a dusty torus surrounding the central black hole engine. At high inclination angles, when observing through the torus, the gas associated with the torus will absorb soft X-rays, hardening the intrinsic QSO spectrum such that it can be reconciled with the harder X-ray background. The dust also preferentially extinguishes blue optical/UV flux, reddening the object at optical wavelengths. A small number of these `type 2 QSOs' have been discovered (e.g. Almaini et al. 1995; Boyle et al. 1998) in deep X-ray samples, showing narrow emission lines in the optical, but broad lines in the near-infrared (IR). One way to reduce the bias against selecting such objects is to construct an IR-selected sample (alternatives would be to use hard X-ray or deep radio surveys). As well as intrinsic absorption, QSOs may also suffer extinction and reddening by dusty objects that lie along their line of sight. For example, a dusty damped Lyman a absorber (DLA) will have the effect of making the background QSO both fainter and redder (Fall & Pei 1993). A bias will then be introduced in the distribution of DLAs detected from optically selected QSO samples. This bias is apparent as a lack of high H I column Ïlog N ½H i . 21÷, high´ metallicity DLAs (Boisse et al. 1998). A second example is that of gravitational lenses. QSOs behind a dusty lens are less likely to be detected, potentially biasing low the number of gravitational lenses found. This will have a significant impact on the cosmological interpretation of the number of gravitationally lensed QSOs. This paper reports on a test of the KX method. Candidate QSOs are stellar sources with J 2 K colours redder than the stellar sequence in a V 2 J versus J 2 K two-colour diagram. Because of the small areas covered so far by surveys in the near-infrared, to produce a QSO sample of significant size we decided to go deep, rather than wide. The advantage of this strategy is that we can use multi-object spectroscopy for candidate confirmation. The main disadvantage is that the list of candidates will be contaminated by compact galaxies, which is not a significant problem at bright magnitudes. This is made worse by the fact that faint QSOs may also be resolved, so that the position of the star/galaxy discrimination boundary must be relaxed, in order for the sample to be complete. Section 2 describes the photometric data, and the candidate selection procedure employed. Section 3 presents the spectroscopic observations that were carried out with LDSS11 at the Anglo-Australian Telescope. We discuss our results in Section 4. 2 P H O T O MET R IC D A T A AN D CAN DID A TE SELECTION The photometric data used was that of the ESO Imaging Survey Chandra Deep Field South (EIS ­ CDFS; formally the EIS ­ AXAF deep field) (Rengelink et al. 1998). The EIS ­ CDFS deep field covers an area of 48 arcmin2, and was observed in seven bands UBVRIJK at the New Technology Telescope (NTT) using the SUSI2 and SOFI instruments. The I band coverage of the field is incomplete. Rengelink et al. produced catalogues of detections in each band using SEXTRACT OR (Bertin & Arnouts 1996), which generates positions, fluxes, magnitudes and a number of other image parameters including a star ­ galaxy discrimination parameter. The 2s magnitude limits are U . 25:2, B . 27:0, V . 26:5, R . 26:3, I . 25:6, J . 23:6 and K . 21:6. The data are publicly available from http://www.eso.org/eis.
q 2001 RAS, MNRAS 328, 150­158

151

Beginning with the K catalogue, we selected all objects brighter than K ¼ 19:5, and then picked out the relevant photometry from the V and J catalogues. Unfortunately at these faint magnitudes QSOs may appear non-stellar. Therefore the selection of the star ­ galaxy discrimination boundary involves balancing the competing requirements of completeness of the QSO sample, and contamination by galaxies. We have included candidates that are

Figure 1. The stellarity index for all K-band sources from the SEXTRACTOR catalogue of Rengelink et al. (1998). The flux limit of the sample, K ¼ 19:5 mag, is marked with the solid vertical line. The stellarity limit of 0.55 is shown by the dotted line.

Figure 2. Colour ­ colour plot in ÏV 2 J ÷ versus ÏJ 2 K ÷ for K < 19:5 objects with stellarity >0.55 (points). The crosses denote objects with stellarity >0.90. The selection limits for our KX sample are shown by the solid line, while the blue and red candidate subsamples are separated by the dotted line. The two UVX candidates that are not KX-selected are circled. In the top right-hand corner we plot the reddening vectors for absorbers at z ¼ 1 and 3 for EÏB 2 V ÷ ¼ 0:1 using the Large Magellanic Cloud extinction curve.


152

S. M. Croom, S. J. Warren and K. Glazebrook
limits. As illustrated in Fig. 2, all but one of the objects lying in the stellar locus of the two-colour diagram have stellarity >0.9, so our applied cut is quite conservative. We can anticipate, therefore, that our candidate list will include several galaxies. The KX and UVX

marginally non-stellar, applying a cut on the K-band SEXTRACT OR stellarity parameter >0.55, where stellarity ¼ 1 denotes a star and stellarity ¼ 0 denotes a galaxy. Fig. 1 shows the stellarity index plotted as a function of K magnitude together with our selection

Table 1. KX-selected objects and spectroscopic identifications for the blue subsample, ÏV 2 J < 3:0÷. The magnitudes are listed in the UBVRJK bands and S/GK denotes the SEXTRACTOR stellarity parameter in the K band. Objects with an ID of `-' have not been observed. Those with a `?' ID could not be identified from their spectra. An (e) or (a) after a `gal' ID denotes emission- or absorption-dominated spectra respectively. The observed spectral features, in both absorption and emission, are also listed. Name KX1 KX2 KX3 KX4 KX5 KX6 KX7 KX8 KX9 KX10 KX11 KX12 KX13 RA (J2000) 3 3 3 3 3 3 3 3 3 3 3 3 3 32 32 32 32 32 32 32 32 32 32 32 32 32 08.70 09.48 16.53 30.02 30.11 31.51 33.06 33.58 38.83 39.13 41.80 44.24 45.15 Dec. (J2000) 227 227 227 227 227 227 227 227 227 227 227 227 227 47 48 44 45 45 46 46 46 47 46 46 47 47 34.5 06.9 49.2 30.2 24.0 23.4 09.1 23.8 32.5 02.1 19.7 33.8 24.2 U 18.90 22.58 23.13 20.54 22.78 ­ 23.92 22.61 21.19 20.58 24.01 22.46 22.58 B 19.23 21.67 23.84 21.53 23.52 24.17 24.10 23.05 22.11 21.55 23.98 23.28 22.99 V 18.64 20.69 23.16 21.33 23.02 23.58 22.80 21.98 21.50 21.21 22.36 22.66 21.92 R 18.44 20.37 22.73 21.03 22.50 22.80 21.85 21.38 20.80 20.77 21.40 21.80 20.90 J 17.07 19.01 20.76 20.22 20.50 21.42 20.33 20.12 19.95 20.07 19.82 20.70 19.37 K 15.92 18.09 19.30 18.43 18.67 19.04 18.79 18.84 18.38 18.42 18.13 19.07 17.64 S/G
K

z 0.544 0.000 0.975 1.221 0.000 0.296 0.000 0.276 0.458 1.221 0.333 0.000 0.437

ID QSO ­ gal(e) QSO ? gal(a) ? gal(e) gal(e) QSO gal(e) ­ gal(e)

spectral features [O II ],[Ne
II I

0.98 0.98 0.71 0.98 0.61 0.80 0.86 0.56 0.94 0.88 0.82 0.92 0.95

],Hd,Hg,Hb,[O

II I

]

[O II ] Mg II ,[O II ] H,K,Hd,G H,K,Hd,G,Hg,Ha,[N II ] [O II ],H,K,Hd,G,Hg,Hb,[O II I ] Mg II ,[O II ] [O II ],H,K,G,Hg,Hb,Ha,[N II ] [O II ],H,K,Hd,Hb,Ha,[N II ]

Table 2. KX-selected objects and spectroscopic identifications for the red subsample, ÏV 2 J . 3:0÷. The format is the same as that in Table 1. Name KX14 KX15 KX16 KX17 KX18 KX19 KX20 KX21 KX22 KX23 KX24 KX25 KX26 KX27 KX28 KX29 KX30 KX31 KX32 KX33 KX34 KX35 KX36 KX37 KX38 KX39 KX40 KX41 KX42 KX43 KX44 KX45 KX46 KX47 KX48 KX49 KX50 KX51 KX52 KX53 KX54 RA (J2000) 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 06.48 06.61 08.70 10.57 11.26 11.65 12.24 12.50 13.05 14.48 14.82 15.56 16.57 17.23 17.80 18.49 18.90 19.27 19.90 19.96 20.51 24.42 26.71 28.49 28.79 29.27 29.99 30.07 30.89 31.26 32.09 33.89 35.80 37.35 38.30 38.83 38.83 39.27 40.71 41.45 47.38 Dec. (J2000) 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 47 46 45 46 45 45 45 47 46 46 44 45 47 46 47 45 47 46 47 48 47 46 46 47 46 47 47 47 46 45 44 46 47 47 44 46 44 47 47 47 46 28.8 23.1 02.0 29.1 33.7 54.4 30.3 29.3 38.2 24.8 33.4 36.5 27.4 49.4 15.1 56.2 34.6 32.4 21.4 31.2 32.5 24.6 59.3 04.0 20.7 07.8 57.4 27.1 22.1 33.1 52.1 00.6 59.1 29.6 55.4 49.2 49.3 58.6 31.2 17.4 27.3 U 23.53 ­ 25.45 ­ ­ ­ 24.77 ­ ­ ­ 25.73 ­ ­ ­ ­ ­ ­ ­ ­ ­ 25.10 ­ ­ 24.06 ­ 25.17 ­ 24.50 24.02 ­ 24.98 24.78 ­ ­ ­ ­ ­ 25.39 23.98 ­ ­ B 23.95 26.02 25.64 ­ 26.66 25.94 24.92 25.61 ­ 25.26 ­ 26.25 25.50 26.89 25.57 25.35 27.11 25.67 26.47 25.53 25.24 26.08 26.33 24.85 24.32 25.11 25.63 24.65 24.74 ­ 25.48 ­ 25.09 24.81 26.08 25.70 25.65 25.54 24.76 24.93 26.21 V 22.89 25.35 23.63 ­ 24.54 25.02 22.60 24.53 ­ 23.94 25.38 24.85 23.51 24.50 23.23 23.35 24.33 23.32 23.98 24.07 23.30 24.33 24.49 23.23 23.24 23.12 24.09 24.90 23.96 ­ 24.15 25.21 22.78 22.37 23.68 23.67 23.81 23.04 23.30 22.84 24.61 R 22.32 23.90 22.38 ­ 23.74 23.34 21.54 23.31 24.64 22.66 23.84 23.47 22.33 23.13 22.06 22.05 23.13 22.11 22.73 22.82 22.37 23.28 23.14 21.84 22.08 21.63 22.96 24.61 23.10 24.33 22.85 24.71 21.37 20.99 22.28 22.12 22.40 21.68 22.05 21.53 23.45 J 19.67 21.13 19.71 21.49 20.39 19.97 19.32 20.42 20.85 20.36 21.15 20.49 19.77 20.53 19.41 19.43 20.75 19.47 20.12 20.30 20.19 20.37 20.72 19.58 19.71 19.26 21.02 21.29 20.30 21.39 20.34 21.45 19.10 18.61 20.15 20.12 20.31 19.40 19.96 19.40 21.46 K 18.04 19.11 18.16 19.45 18.84 18.47 17.89 18.84 19.31 18.89 19.39 18.66 18.29 18.99 18.00 17.76 18.93 17.86 18.47 18.66 18.29 18.90 19.08 17.87 17.95 17.55 18.96 19.16 18.66 19.42 18.47 19.37 17.33 16.73 18.43 18.38 18.34 17.60 18.16 17.61 19.49 S/GK 0.75 0.76 0.98 0.78 0.96 0.80 0.89 0.93 0.95 0.98 0.90 0.80 0.77 0.84 0.65 0.78 0.90 0.88 0.96 0.96 0.81 0.62 0.90 0.86 0.92 0.91 0.79 0.97 0.95 0.80 0.89 0.92 0.85 0.58 0.70 0.89 0.93 0.86 0.89 0.83 0.97 z 0.000 0.000 0.000 0.000 0.333 0.000 0.000 0.000 0.212 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.668 0.000 0.668 0.000 0.000 0.000 0.000 0.000 0.000 0.665 0.669 0.666 0.000 0.000 0.000 0.667 0.000 0.000 ID ­ ­ ­ ? gal(e) ­ ? ? gal(a) ? ­ ? ­ ? ­ ­ ? ? ­ ­ ­ ? ? gal(a) ? gal(a) ­ ­ ? ­ ­ ? gal(a) gal(a) gal(a) ­ ­ ­ gal(e) ? ­ spectral features

Hb,[O

II I

],Ha

H,K,Hd,G,Hb,[O

II I

]

[O II ],H,K,G,Mgb H,K,Hd,G,Hg,Hb,Mgb

H,K,Hd,G,Hg [O II ],H,K,G,Hg H,K,G,Hg

[O II ],H,K,G,Hb,[O

II I

]

q 2001 RAS, MNRAS 328, 150­158


A small-area faint KX redshift survey
methods have this problem in common at faint magnitudes. The KX colour selection was carried out in the ÏV 2 J ÷ versus ÏJ 2 K ÷ plane, as shown in Fig. 2. Compact objects that have ÏJ 2 K ÷ . 1:35 or ÏJ 2 K ÷ . 0:36ÏV 2 J ÷ 1 0:18 were selected for follow-up spectroscopy. This selection is almost identical to that used by Warren et al. (2000). We make a comparison to the UVX selection by finding the U 2 B colours for all our compact K , 19:5 sources. Of those sources with positive detections in the U and B bands, only two are found to be selected as UVX ÏU 2 B , 20:2÷ and not KX. A further 18 sources are selected as both UVX and KX. The non-KX UVX sources have similar colours in both the

153

optical and IR; these are typical of blue (e.g. A/B) stars. They lie at the blue end of the stellar locus in the ÏV 2 J ÷ versus ÏJ 2 K ÷ plane (the circled points in Fig. 2). We divided our KX candidate list into two subsamples, either side of ÏV 2 J ÷ ¼ 3. Table 1 lists the 13 objects in the bluer of the two subsamples, which lie in the same region of the colour ­ colour plane as typical blue optically selected QSOs. These candidates have average R ¼ 21:3. Table 2 lists the 41 objects in the red subsample, containing candidate highly reddened QSOs. These candidates have average R ¼ 22:7. This subsample is expected to be contaminated by compact distant early-type galaxies.

Figure 3. Spectra of the identified objects in the blue subsample. The wavelengths of spectral features are labelled on each spectrum. We have not corrected for telluric absorption bands, and indicate these by a circled cross. q 2001 RAS, MNRAS 328, 150­158


154
3

S. M. Croom, S. J. Warren and K. Glazebrook
wavelength. The telescope was nodded back and forth every 30 s by 3 arcsec, keeping the object in the slit. At the same time, the charge was shuffled up and down the CCD, such that a single readout contained two sets of spectra, one for each position of the object in the slit. By simply subtracting one set of spectra from the other, near-perfect (i.e. Poisson limited) sky removal is obtained, as the sky flux passes through the same part of the slit, and is incident on the same pixel in the CCD as the object pffiffiffi Although the flux. subtraction process increases the sky noise by 2, compared with the usual method, the greatly reduced systematic errors allow reliable spectroscopy of much fainter objects. During the first night of observations we obtained 5 á 2000-s exposures on mask 1, in seeing of ,2 arcsec. On the second night, 7 á 2000-s exposures were obtained on mask 1 in 1­1:5 arcsec seeing and 3 á 2000-s exposures were obtained on mask 2. Copper ­ Helium ­ Neon arcs were taken at intervals throughout the night to wavelength-calibrate the spectra. The CCD reduction, spectral extraction and calibration was carried out within the IRA F package, using the tasks within IMRED.CCD RED and IMRED. SP ECRED . After bias subtracting the images and interpolating over bad pixels and columns, the images

S PECTR O SCOPIC OBSER VA T ION S

Spectroscopic observations were carried out with the Low Dispersion Survey Spectrograph (LDSS11) on the nights of 1999 December 11­12. LDSS11 uses the MITLL3 CCD, along with a volume phase holographic grism to give very high efficiency ° (,25 per cent at 7000 A). Two masks were observed, each containing ,45 slits of length 6 arcsec and width 1.5 arcsec. Faint objects, where possible, were included in both masks. Priority was given to the blue subsample, resulting in 11 of the 13 objects being allocated slits, while a smaller fraction (22 out of 41) of the red subsample was allocated slits. We were also able to allocate a slit to one of the two non-KX UVX candidates. To fill in the remaining space, in positions where there were no KX objects, we randomly selected K < 19 galaxies (stellarity <0.1). The nod-and-shuffle technique was used to optimize sky subtraction (Glazebrook & Bland-Hawthorn 2001). The principle of the method is to record quasi-simultaneously the sky local to the object through exactly the same light path as the object flux, so that the sky can be subtracted. This is in contrast to the usual method, which is to subtract a smooth function fit to the sky counts at each

Figure 4. Spectra of the identified objects in the red subsample. q 2001 RAS, MNRAS 328, 150­158


A small-area faint KX redshift survey
were combined in groups of 2 or 3, averaging and using clipping algorithms to remove cosmic ray events. Combining in small groups, rather than over a whole night, ensured no smearing out of the spectra arising from small shifts (typically a fraction of a pixel) caused by flexure in the spectrograph. Each combined frame was split in half, separating the spectra taken at the two nod-and-shuffle positions. One of these two halves was then subtracted from the other to produce the sky subtracted image containing both positive and negative spectra for each object. The routine APEXTRAC T was then used to trace and extract the positive and negative spectra. A number of the spectra have extremely low signal-to-noise ratio (S/N) and were therefore impossible to trace. In these cases the trace used was one interpolated from surrounding well-traced spectra. Wavelength calibration was performed using the IDEN TI FY and DIS P C O R routines. Finally, multiple spectra of the same object were combined together. Classification and redshift determination were undertaken by eye using a modified version of the REDS HI FT program by Karl Glazebrook. A high fraction of the blue subsample could be identified (9 of 11 observed), while the redder subsample had a much lower identification rate (only 8 of 22), as expected given their much fainter R-band fluxes. The positions, magnitudes, identifications, and redshifts are given in Tables 1 and 2 for the blue and red subsamples respectively, and the spectra are shown in Figs 3 and 4. The two UVX-selected objects are listed in Table 3. We obtained a spectrum of UVX1 and it was found to be an A star, with clear Balmer absorption lines. The K < 19 galaxy sample is listed in Table 4 and the galaxy spectra are shown in Fig. 5. The KX objects are plotted in the ÏV 2 J ÷ versus ÏJ 2 K ÷ plane in Fig. 6. We identify three QSOs; one at z ¼ 0:544 and a pair of QSOs at z ¼ 1:221. All three objects would also be selected in a UVX sample, having colours ÏU 2 B÷ ¼ 20:33, 20.99 and 20.97 for KX1, KX4 and KX10 respectively. KX1 also shows narrow Hb

155

superimposed on a much broader underlying component. The spectra of the z ¼ 1:221 QSO pair and their flux ratio are shown in Fig. 7. This pair has an angular separation of 125 arcsec. The ratio of the two spectra is flat over the entire spectroscopic range (Fig. 7c). They also have similar fluxes in all the optical and IR bands. However, the large angular separation makes this pair an unlikely candidate for a gravitational lens. The remaining identifications are all galaxies, with roughly equal numbers of early- and late-type spectra (by early and late we mean spectra that are dominated by absorption and emission lines respectively). The early-type objects are on average redder. The redshift distribution of the identified galaxies and QSOs is shown in Fig. 8. While the narrow emission-line galaxies are distributed from z . 0:15 to 1.0, 5 out of 7 of the early-type galaxies lie at z . 0:66, suggesting the presence of a galaxy cluster at that redshift.

4 4.1

D ISCUSS I ON Blue candidates, V 2 J , 3

We can draw some preliminary conclusions on the basis of the spectroscopy completed so far. Only two sources brighter than K ¼ 19:5 were found to be UVX and not KX. The UV, optical and IR colours of these two objects suggest that they are Galactic stars and spectroscopy has confirmed this in one case. The three confirmed QSOs also show a UV excess. Thus, within our sample, KX selection appears to contain all the QSOs that would have been selected via the UVX method, as we find no UVX QSOs that are not also KX. Of the 13 KX candidates with V 2 J , 3; 11 were observed, and nine have been identified. The effective area of this sample is then 48 arcmin2 multiplied by 9/13. (We obtain a very similar effective area if we consider the identified fraction of

Table 3. The two UVX non-KX sources with K < 19:5 in our survey area. Name UVX1 UVX2 RA (J2000) 3 32 30.58 3 32 44.08 Dec.(J2000) 227 44 59.8 227 48 24.4 U 19.75 19.50 B 20.01 19.93 V 19.56 19.37 R 19.20 18.92 J 18.76 18.51 K 18.48 18.15 S/G
K

0.98 0.98

Table 4. K < 19 galaxies observed in addition to our KX survey. The format is the same as that in Table 1. The `?' after the [O III ] for KG3 indicates that the features associated with the [O II I ] line are questionable owing to the peculiar line ratios of the two [O II I ] lines. The `(abs)' after an Mg II indicates that this line is seen in absorption. Name KG1 KG2 KG3 KG4 KG5 KG6 KG7 KG8 KG9 KG10 KG11 KG12 KG13 KG14 KG15 KG16 RA (J2000) 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 32 10.62 10.81 11.45 11.55 14.87 21.79 22.02 22.63 28.04 34.16 39.92 42.59 42.85 45.66 45.69 46.60 Dec. (J2000) 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 227 47 46 46 47 47 44 46 44 46 47 47 43 46 45 44 47 02.0 28.0 50.2 13.2 38.5 42.4 56.2 26.0 39.5 12.4 15.3 47.4 05.9 54.9 05.9 08.9 U 24.58 ­ 24.39 22.84 25.16 21.82 22.78 22.06 22.26 21.85 22.43 22.76 22.05 21.65 22.20 22.15 B 24.77 ­ 24.28 23.33 25.46 22.41 23.91 22.55 22.18 22.28 23.34 23.58 22.60 22.30 22.93 22.51 V 23.43 ­ 22.57 22.39 23.79 21.40 22.64 21.82 20.68 21.30 22.74 22.25 21.89 21.18 22.15 21.46 R 22.21 24.57 21.86 21.56 22.80 20.96 21.49 20.94 19.86 20.80 22.21 21.68 20.88 20.50 21.12 20.85 J 20.19 20.18 19.86 19.64 20.43 19.90 19.02 19.02 17.94 19.74 20.54 21.51 19.36 19.53 19.91 19.98 K 18.42 18.34 18.54 17.74 18.90 18.97 17.37 17.39 16.23 18.38 18.91 18.82 17.65 18.00 18.33 18.90 S/GK 0.07 0.03 0.43 0.03 0.45 0.01 0.04 0.03 0.08 0.10 0.01 0.00 0.04 0.01 0.02 0.00 z 0.000 0.000 0.577 0.000 0.000 0.295 0.670 0.738 0.247 0.227 1.097 0.000 0.000 0.337 0.000 0.128 ID spectral features

gal(e) gal(e) gal(a) gal(e) gal(a) gal(e) gal(e) gal(e) gal(e)

[O II ],H,K,Hg,[O

III

]?

[O II ],H,K,Hd,Hb,[O III ],Ha H,K,Hd,G,Hg Mg II (abs),[O II ],H,K,Hd,G,Hg Hb,Mgb,NaD,Ha,[N II ] Hb,[O II I ],Ha,[S II ] Mg II (abs),[O II ],H,K,Hd [O II ],H,K,Hd,G,Hg,Hb,[O III ], 1.3cm NaD,Ha,[N II ],[S II ] NaD,Ha,[S II ]

q 2001 RAS, MNRAS 328, 150­158


156

S. M. Croom, S. J. Warren and K. Glazebrook
Five of the KX candidates are narrow emission-line galaxies. These could be obscured AGN, where the broad-line region is enshrouded in dust. However, a more likely explanation is that they are simply star-forming galaxies. The current spectra are not of sufficient quality to differentiate between these two possibilities. However, galaxy evolution models (e.g. see Eisenhardt et al. 1999) suggest that late-type galaxies should lie in this region of colour ­ colour space. Comparison with the deep Chandra X-ray imaging data soon to be available in this field can answer this question. The QSO pair at z ¼ 1:22 is unexpected given the small size of the sample, and the large angular separation (.2 arcmin) is significantly larger than any confirmed gravitationally lensed pair.

objects with a stricter stellarity cut . 0.8). This gives us a lower limit to the surface density of QSOs brighter than K ¼ 19:5 of 3251316 deg22 , the lower limit allowing for the possibility that the 2177 red sample contains one or more QSOs. We can calculate the number of QSOs expected in this field, based on previous results from blue selected surveys. Taking the result of Boyle, Jones & Shanks (1991), who find a density of 76.8 QSOs per square degree at z , 2:9 and B , 22, an average (B 2 K ÷ ¼ 3:0 and a faint end number counts slope (extrapolated) of 0.3, we expect to find ,1.4 QSOs in our 48 arcmin2 field. Given the small numbers and the uncertainties in this calculation we conclude that our data are consistent with this.

Figure 5. Spectra of the identified objects in the galaxy sample. q 2001 RAS, MNRAS 328, 150­158


A small-area faint KX redshift survey

157

Figure denotes shading galaxies

8. The redshift distribution of our KX sample. Heavy shading the QSOs, light shading the narrow emission-line galaxies and no the absorption-line galaxies. Note that most of the early-type are at z . 0:66.

Figure 6. Colour ­ colour plot in ÏV 2 J ÷ versus ÏJ 2 K ÷ for our KXselected sample. Different symbols denote the different object IDs: QSOs (filled circles); emission-line galaxies (filled squares); absorption-line galaxies (open squares); non-IDs (open triangles); unobserved objects (crosses).

eight have been identified, all as galaxies. The effective area of the search for red QSOs is therefore 9.4 arcmin2. (This could be considered an underestimate, because QSOs are easier to identify than early-type galaxies ­ which appear to dominate the red sample ­ so the 14 objects with unidentified spectra are most likely to be galaxies.) Our spectroscopy of red candidates therefore gives only a weak upper limit on the number of red QSOs of 1150 deg22 (2s). To conclude, our preliminary spectroscopic results confirm that UVX selection is contained by KX selection, and our small sample provides a measurement of the surface density of QSOs brighter than K ¼ 19:5. However our results do not yet provide useful statistics on the number of red QSOs. We plan to obtain nearinfrared spectroscopy, combined with further optical spectroscopy, and an analysis of the Chandra image, in pursuit of this goal. As anticipated (and as also encountered with UVX surveys), contamination by galaxies is a problem for faint optical/nearinfrared surveys for QSOs. With forthcoming wide-field near-IR surveys to moderate depths K lim . 17, where contamination from galaxies is much less pronounced, the KX method is an obvious choice to construct unbiased samples of QSOs. A C K N O W L E DGMENT S

Figure 7. Spectra of the QSO pair found in our KX sample (a) KX4 and (b) KX10 with emission lines denoted by the dashed lines. Panel (c) shows the flux ratio KX10/KX4 boxcar-smoothed with a window of 5 pixels.

The discovery of a galaxy cluster at z ¼ 0:66 may provide a foreground lens for this source, however a lens of ,2 arcmin separation would typically require a cluster with a velocity dispersion ,1500 km s21 (assuming a simple singular isothermal sphere potential), which is exceptionally large. Again, the question of this unlikely event will be resolved with the availability of the deep Chandra data, which will easily detect hot gas from the cluster if it is sufficiently massive to lens a background QSO strongly. 4.2 Red candidates, V 2 J > 3 Of the 41 KX candidates with V 2 J > 3; 22 were observed, and
q 2001 RAS, MNRAS 328, 150­158

This paper was prepared using the facilities of the STARLINK node at the Imperial College of Science, Technology and Medicine. We thank the ESO Imaging Survey team for making their data publicly available. SMC acknowledges the support of PPARC. REFER E NC ES
Almaini O., Boyle B. J., Griffiths R. E., Shanks T., Stewart G. C., Georgantopoulos I., 1995, MNRAS, 277, L31 Antonucci R., 1993, ARA&A, 31, 473 Bertin E., Arnouts S., 1996, A&AS, 117, 393 Binney J., Merrifield M., 1998, Galactic Astronomy. Princeton Univ. Press, Princeton NJ ´ Boisse P., Le Brun V., Bergeron J., Deharveng J.-M., 1998, A&A, 333, 841 Boyle B. J., Fong R., Shanks T., Peterson B. A., 1990, MNRAS, 243, 1 Boyle B. J., Jones L. R., Shanks T., 1991, MNRAS, 251, 482 Boyle B. J., Almaini O., Georgantopoulos I., Blair A. J., Stewart G. C., Griffiths R. E., Shanks T., Gunn K. F., 1998, MNRAS, 297, L53


158

S. M. Croom, S. J. Warren and K. Glazebrook
Rengelink R. et al., 1998, preprint (astro-ph/9812190) Schmidt M., Green R. F., 1983, ApJ, 269, 352 Warren S. J., Hewett P. C., Foltz C. B., 2000, MNRAS, 312, 827 Webster R. L., Francis P. J., Peterson B. A., Drinkwater M. J., Masci F. J., 1995, Nat, 375, 469

Croom S. M., Smith R. J., Boyle B. J., Shanks T., Loaring N. S., Miller L., 2001, MNRAS, 322, L29 Eisenhardt P., Elston R., Stanford S. A., Dickinson M., Spinrad H., Stern D., Dey A., 1999, in Guiderdoni B., et al., eds, Proc. Xth Recontres de Blois, The Birth of Galaxies. Editions Frontieres, Gif-sur-Yvette Fall S. M., Pei Y. C., 1993, ApJ, 402, 479 Glazebrook K., Bland-Hawthorn J., 2001, PASP, 113, 197 Madau P., Ghisellini G., Fabian A. C., 1994, MNRAS, 270, L17

A This paper has been typeset from a TEX/L TEX file prepared by the author.

q 2001 RAS, MNRAS 328, 150­158