Abstract

We present the results of a z≥ 2.9 survey for H i 21-cm and molecular absorption in the hosts of radio quasars using the Giant Metrewave Radio Telescope and the Tidbinbilla 70-m telescope. Although the atomic gas has been searched to limits capable of detecting most known absorption systems, no H i was detected in any of the 10 sources. Previously published searches, which are overwhelmingly at redshifts of z≲ 1, exhibit a 42 per cent detection rate (31 out of 73 sources), whereas the inclusion of our survey yields a 17 per cent detection rate (two out of 12 sources) at z > 2.5. We therefore believe that our high-redshift selection is responsible for our exclusive non-detections, and find that at ultraviolet (UV) luminosities of LUV≳ 1023W Hz−1, 21-cm absorption has never been detected. We also find this to not only apply to our targets, but also those at low redshift exhibiting similar luminosities, giving zero detections out of a total of 16 sources over z= 0.24 to 3.8. This is in contrast to the LUV≲ 1023W Hz−1 sources where there is a near 50 per cent detection rate of 21-cm absorption.

The mix of 21-cm detections and non-detections is currently attributed to orientation effects, where according to unified schemes of active galactic nuclei, 21-cm absorption is more likely to occur in sources designated as radio galaxies (type 2 objects, where the nucleus is viewed through dense obscuring circumnuclear gas) than in quasars (type 1 objects, where we have a direct view to the nucleus). However, due to the exclusively high UV luminosities of our targets it is not clear whether orientation effects alone can wholly account for the distribution, although there exists the possibility that the large luminosities are indicative of a changing demographic of galaxy types. We also find that below luminosities of LUV∼ 1023W Hz−1, both type 1 and type 2 objects have a 50 per cent likelihood of exhibiting 21-cm absorption.

Finally, we do not detect molecular gas in any of the sources. The lack of H i absorption, combined with the results from Paper I, suggests these sources are not conducive to high molecular abundances.

1 INTRODUCTION

Redshifted molecular and atomic absorption lines can provide excellent probes of the contents and nature of the early Universe. In particular, with redshifted radio and microwave lines we can investigate the gaseous content and large-scale structure as well any possible variations in the values of the fundamental constants at large look-back times. However, these are currently rare, with only 67 H i 21-cm absorption systems at z≳ 0.1 known, comprising of 37 associated systems and 30 intervening (Table 1). For molecular absorption in the radio band the situation is considerably worse with only five redshifted OH 18-cm systems currently known (Chengalur, de Bruyn & Narasimha 1999; Kanekar & Chengalur 2002; Kanekar et al. 2003), four of which also exhibit a plethora of molecular absorption lines in the millimetre regime (see Combes & Wiklind 1998).

Table 1

The known redshifted (zabs≳ 0.1) H i 21-cm absorbers (updated from Paper I, Curran et al. 2006). Absorber types are BLRG – broad line radio galaxy; CSO – compact symmetric object; CSS – compact steep spectrum source; DLA – damped Lyman α absorption system (Mg ii– DLA candidate); GPS – gigahertz peaked spectrum source; HFP - high-frequency peaker galaxy; Lens – gravitational lens; OHM – OH megamaser; Red – red quasar; RG – radio galaxy. The number of each type is given, as well as the absorption redshift and column density ranges.

Reference Type No. zabs NH I(cm−2
Associated absorbers 
Mirabel (1989) RG 0.10 and 0.12 2 and 6 × 1018 (Ts/f
van Gorkom et al. (1989) RG 0.06 and 0.10 11 and 8 × 1018 (Ts/f
Uson et al. (1991) RG 3.40 3 × 1018 (Ts/f
Carilli, Perlman & Stocke (1992) Red 0.25 1 × 1019 (Ts/f
Carilli et al. (1998) Red 0.58–0.67 0.8– 8 × 1019 (Ts/f
Moore et al. (1999) Red 2.64 8 × 1018 (Ts/f
Peck, Taylor & Conway (1999) CSO 0.10 3 × 1019 (Ts/f
Peck et al. (2000) CSO 0.25 5 × 1018 (Ts/f
Ishwara-Chandra, Dwarakanath & Anantharamaiah (2003) CSS/Red 1.19 4 × 1019 (Ts/f
Vermeulen et al. (2003) BLRG 0.22 7 × 1017 (Ts/f
… CSS 0.19–0.80 0.1– 2 × 1018 (Ts/f
… GPS 10 0.08–0.65 0.07– 3 × 1019 (Ts/f
… RG 0.24 1 × 1018 (Ts/f
Pihlström et al. (2005) OHM 0.22 6 × 1018 (Ts/f
Curran et al. (2006) RG 0.10 4 × 1019 (Ts/f
Gupta & Saikia (2006a) RG 0.08 6 × 1018 (Ts/f
Gupta et al. (2006) CSS 0.17 5 × 1018 (Ts/f
Orienti, Morganti & Dallacasa (2006)a HFP 0.67 8 × 1019 (Ts/f
Intervening absorbers 
Carilli, Rupen & Yanny (1993) Lens 0.69 1 × 1019 (Ts/f
Lovell et al. (1996) Lens 0.19 ≈2 × 1018 (Ts/f
Chengalur et al. (1999) Lens 0.89 1 × 1019 (Ts/f
Kanekar & Briggs (2003) Lens 0.76 1 × 1019 (Ts/f
Kanekar & Chengalur (2003)b DLA 15 0.09–2.04 0.02– 6 × 1019 (Ts/f
Darling et al. (2004) DLA 0.78 2 × 1019 (Ts/f
Kanekar et al. (2006) DLA 2.35 4 × 1017 (Ts/f
Kanekar, Chengalur & Lane (2007) DLA 3.39 1 × 1018 (Ts/f
Gupta et al. (2007) Mg ii 1.17–1.37 0.4– 2 × 1018 (Ts/f
Curran et al. (2007b) Lens 0.96 2 × 1018 (Ts/f
Curran et al. (2007a) DLA 0.66 4 × 1018 (Ts/f
York et al. (2007) DLA 2.29 2 × 1018 (Ts/f
Zwaan et al. (in preparation) Mg ii ∼0.6 – 
Reference Type No. zabs NH I(cm−2
Associated absorbers 
Mirabel (1989) RG 0.10 and 0.12 2 and 6 × 1018 (Ts/f
van Gorkom et al. (1989) RG 0.06 and 0.10 11 and 8 × 1018 (Ts/f
Uson et al. (1991) RG 3.40 3 × 1018 (Ts/f
Carilli, Perlman & Stocke (1992) Red 0.25 1 × 1019 (Ts/f
Carilli et al. (1998) Red 0.58–0.67 0.8– 8 × 1019 (Ts/f
Moore et al. (1999) Red 2.64 8 × 1018 (Ts/f
Peck, Taylor & Conway (1999) CSO 0.10 3 × 1019 (Ts/f
Peck et al. (2000) CSO 0.25 5 × 1018 (Ts/f
Ishwara-Chandra, Dwarakanath & Anantharamaiah (2003) CSS/Red 1.19 4 × 1019 (Ts/f
Vermeulen et al. (2003) BLRG 0.22 7 × 1017 (Ts/f
… CSS 0.19–0.80 0.1– 2 × 1018 (Ts/f
… GPS 10 0.08–0.65 0.07– 3 × 1019 (Ts/f
… RG 0.24 1 × 1018 (Ts/f
Pihlström et al. (2005) OHM 0.22 6 × 1018 (Ts/f
Curran et al. (2006) RG 0.10 4 × 1019 (Ts/f
Gupta & Saikia (2006a) RG 0.08 6 × 1018 (Ts/f
Gupta et al. (2006) CSS 0.17 5 × 1018 (Ts/f
Orienti, Morganti & Dallacasa (2006)a HFP 0.67 8 × 1019 (Ts/f
Intervening absorbers 
Carilli, Rupen & Yanny (1993) Lens 0.69 1 × 1019 (Ts/f
Lovell et al. (1996) Lens 0.19 ≈2 × 1018 (Ts/f
Chengalur et al. (1999) Lens 0.89 1 × 1019 (Ts/f
Kanekar & Briggs (2003) Lens 0.76 1 × 1019 (Ts/f
Kanekar & Chengalur (2003)b DLA 15 0.09–2.04 0.02– 6 × 1019 (Ts/f
Darling et al. (2004) DLA 0.78 2 × 1019 (Ts/f
Kanekar et al. (2006) DLA 2.35 4 × 1017 (Ts/f
Kanekar, Chengalur & Lane (2007) DLA 3.39 1 × 1018 (Ts/f
Gupta et al. (2007) Mg ii 1.17–1.37 0.4– 2 × 1018 (Ts/f
Curran et al. (2007b) Lens 0.96 2 × 1018 (Ts/f
Curran et al. (2007a) DLA 0.66 4 × 1018 (Ts/f
York et al. (2007) DLA 2.29 2 × 1018 (Ts/f
Zwaan et al. (in preparation) Mg ii ∼0.6 – 

a Also J1407+2827, which is counted as one of the 10 GPSs of Vermeulen et al. (2003).

b See the paper for the full reference list and Curran et al. (2005) for the calculated column densities. Note that since PKS 1413+135 is an associated system, it has been included in the top panel.

From our previous survey for radio absorption lines in the hosts of the sources in the Parkes half-Jansky flat-spectrum sample (PHFS), one H i absorption system was clearly detected (out of five searched) and one OH system tentatively detected (of the 13 searched), both at zabs∼ 0.1 (Paper I). Upon examination of the previous detections, we concluded the following.

  • For the H i 21-cm absorption there is no overwhelming correlation between the line strength and the optical–near-infrared (VK) colour.

  • However, for the OH 18-cm absorption there is a clear relationship, thus suggesting that the reddening of these quasars is due to dust, the amount of which is correlated with the molecular abundance. Furthermore, for the molecular absorbers:

  • all of the absorption lines were found at redshifts were absorption (usually H i) was already known to occur;1

  • in all cases the absorption occurs towards flat spectrum radio sources, suggesting compact radio sources and thus a large effective coverage by the absorber.

Unfortunately, prior to the analysis undertaken in Paper I, not all of the above criteria were fully formulated, and so we have also targeted sources for which only the last criterion is satisfied. In this paper we present the results of a survey for redshifted atomic and molecular absorption within the hosts of sources selected from the Parkes quarter-Jansky flat-spectrum sample (PQFS; Jackson et al. 2002).

Because of the search for coincident molecular absorption, we initially prioritized sources in which the HCO+ 0 → 1 transition would be redshifted into the 12-mm band of the Tidbinbilla 70-m telescope. This gave 70 targets at z > 2.3 out of the 878 PQFS sources, and limiting the sample further to sources with flux densities in excess of 0.5 Jy at ≈0.4 GHz, gave a total of 19 sources at z > 2.9.

2 OBSERVATIONS AND RESULTS

2.1 The redshifted decimetre wave observations

The redshifted decimetre wave observations where performed with the Giant Metrewave Radio Telescope (GMRT)2 in 2004 March, during the first run as described in Paper I. Again, as per Paper I, we searched for absorption within the host at the emission redshift of the quasar. Our target sample therefore consisted (primarily) of objects in the PQFS where either the H i 21-cm or OH 18-cm (2Π3/2J= 3/2) transition was redshifted into the 90-cm (≈327 MHz) band. Because of time constraints during the observations, we prioritized the targets in which B≳ 19, in order to preferentially select sources where the presence of dust, associated with dense gas, would dim the visible and ultraviolet (UV) light. This left 13 out of the original 19 targets (see Table 2).

Table 2

Summary of the search for decimetre absorption lines in the hosts of z≳ 3 PQFS sources. νobs is the observed frequency of the line, σrms is the rms noise reached per Δv channel, Scont is the continuum flux density (uncalibrated for 2215+02, Section 2), τ is the optical depth of the line calculated per channel, where τ= 3σrms/Scont is quoted for these non-detections, N is the resulting column density, where Ts is the spin temperature of the H i 21-cm line, Tx is the excitation temperature of the corresponding OH line and f the respective covering factor. In all cases OH refers to the 2Π3/2J= 3/2 (1667 MHz) transition, with the exception of 1228−113 and 1534+004 for which we observed the 2Π1/2J= 1/2 (4751 MHz) transition. Finally, in light of the results of Paper I, we list the V, B and K magnitudes (where available) with their respective references.

PKS zem Line νobs (MHz) σrms (mJy) Δv (km s−1Scont (Jy) τ N (cm−2z range B V K Ref 
0335−122 3.442 H i 319.77 23 15 0.90 <0.077 <2.1 × 1018 (Ts/f3.428–3.456 21.018 20.110 17.510 1, 2 
0347−211 2.944 H i 360.14 17 13 0.51 <0.10 <2.4 × 1018 (Ts/f2.935–2.953 20.476 – 17.900 
0537−286 3.104 H i 346.10 54 14 1.61 <0.10 <2.6 × 1018 (Ts/f3.092–3.106 19.290 – 16.770 2, 3 
… … … … RFI required removal of z= 3.1064– 3.1088 3.109–3.115 … … …  
0913+003 3.074 H i 348.65 – 13 – RFI dominant 20.998 20.775 – 2, 4 
1026−084 4.276 OH 316.03 53 15 0.65 <0.24 <1.5 × 1015 (Tx/f4.265–4.283 21.070 – – 
1228−113 3.528 H i 313.69 26 15 0.49 <0.16 <4.3 × 1018 (Ts/f3.516–3.532 22.010 – 16.370 5, 3 
… … OH 1049.17 3.7 18 0.64 <0.017 <4.9 × 1013 (Tx/f3.513–3.543 … … …  
1251−407 4.460 OH 305.38 12 15 0.25 <0.14 <5.0 × 1014 (Tx/f4.442–4.478 – 23.7 – 
1351−018 3.707 H i 301.76 – 16 – <0.090 <2.6 × 1018 (Ts/f3.693–3.722 21.030 19.696 17.070 7, 4, 3 
… … OH 354.23 39 13 0.75 <0.052 <1.6 × 1014 (Tx/f3.695–3.719 … …   
1535+004 3.497 H i 315.86 8.6 15 0.37 <0.070 <1.9 × 1018 (Ts/f3.485–3.509 22.500 – 19.54 
… … OH 1056.41 4.0 18 0.47 <0.032 <9.4 × 1013 (Tx/f3.486–3.514 … …   
1630−004 3.424 H i 321.07 – 15 – RFI dominant – 21.81 – 
1937−101 3.787 H i 296.72 9.4 16 0.61 <0.046 <1.3 × 1018 (Ts/f3.773–3.802 18.800 – 13.816 2, 9 
… … OH 348.31 3.5 14 0.51 <0.021 <7.0 × 1013 (Tx/f3.775–3.795 … … …  
2215+02 3.572 H i 310.67 14 15 3.07 <0.014 <3.7 × 1017 (Ts/f3.558–3.585 21.840 20.420 19.340 10 
… … OH 364.69 22 13 – <0.089 <2.8 × 1014 (Tx/f3.561–3.583 … … … … 
2245−059 3.295 H i 330.71 – 14 – RFI dominant 19.523 – – 
PKS zem Line νobs (MHz) σrms (mJy) Δv (km s−1Scont (Jy) τ N (cm−2z range B V K Ref 
0335−122 3.442 H i 319.77 23 15 0.90 <0.077 <2.1 × 1018 (Ts/f3.428–3.456 21.018 20.110 17.510 1, 2 
0347−211 2.944 H i 360.14 17 13 0.51 <0.10 <2.4 × 1018 (Ts/f2.935–2.953 20.476 – 17.900 
0537−286 3.104 H i 346.10 54 14 1.61 <0.10 <2.6 × 1018 (Ts/f3.092–3.106 19.290 – 16.770 2, 3 
… … … … RFI required removal of z= 3.1064– 3.1088 3.109–3.115 … … …  
0913+003 3.074 H i 348.65 – 13 – RFI dominant 20.998 20.775 – 2, 4 
1026−084 4.276 OH 316.03 53 15 0.65 <0.24 <1.5 × 1015 (Tx/f4.265–4.283 21.070 – – 
1228−113 3.528 H i 313.69 26 15 0.49 <0.16 <4.3 × 1018 (Ts/f3.516–3.532 22.010 – 16.370 5, 3 
… … OH 1049.17 3.7 18 0.64 <0.017 <4.9 × 1013 (Tx/f3.513–3.543 … … …  
1251−407 4.460 OH 305.38 12 15 0.25 <0.14 <5.0 × 1014 (Tx/f4.442–4.478 – 23.7 – 
1351−018 3.707 H i 301.76 – 16 – <0.090 <2.6 × 1018 (Ts/f3.693–3.722 21.030 19.696 17.070 7, 4, 3 
… … OH 354.23 39 13 0.75 <0.052 <1.6 × 1014 (Tx/f3.695–3.719 … …   
1535+004 3.497 H i 315.86 8.6 15 0.37 <0.070 <1.9 × 1018 (Ts/f3.485–3.509 22.500 – 19.54 
… … OH 1056.41 4.0 18 0.47 <0.032 <9.4 × 1013 (Tx/f3.486–3.514 … …   
1630−004 3.424 H i 321.07 – 15 – RFI dominant – 21.81 – 
1937−101 3.787 H i 296.72 9.4 16 0.61 <0.046 <1.3 × 1018 (Ts/f3.773–3.802 18.800 – 13.816 2, 9 
… … OH 348.31 3.5 14 0.51 <0.021 <7.0 × 1013 (Tx/f3.775–3.795 … … …  
2215+02 3.572 H i 310.67 14 15 3.07 <0.014 <3.7 × 1017 (Ts/f3.558–3.585 21.840 20.420 19.340 10 
… … OH 364.69 22 13 – <0.089 <2.8 × 1014 (Tx/f3.561–3.583 … … … … 
2245−059 3.295 H i 330.71 – 14 – RFI dominant 19.523 – – 

References: 1 –Ellison, Hall & Lira (2005); 2 – SuperCOSMOS Sky Survey, Hambly et al. (2001); 3 – Francis (private communication); 4 – SDSS DR6, Adelman-McCarthy et al. (2008); 5 –Chun et al. (2006); 6 –Jackson et al. (2002); 7 –Drinkwater et al. (1997); 8 –Winn et al. (2002); 9 –Smith & Heckman (1989); 10 –Francis et al. (2000).

We used all 30 antennas and the 90-cm receiver over a 2-MHz bandwidth (giving a coverage of ≈± 900 km s−1). Over the 128 channels (two polarizations) this gave a channel width of 16 kHz, or a velocity resolution of ≈15 km s−1. For all of the runs we used 3C 48, 3C 147 and 3C 286 for bandpass calibration and used separate phase calibrators for all of the sources, as heavy flagging of the target data could result in poor self-calibration. The flagging and all of the reduction was done by the miriad interferometry reduction package. Synthesized beam sizes were typically ≳10 arcsec for the 90-cm and ≈6 arcsec for the 21-cm observations (OH 2Π1/2J= 1/2 redshifted from 6-cm). As per the other observations (Paper I), channel 117 of the RR polarization and the telescope pairing between antennas E02 and E03 (15 and 16 in aips/miriad convention) were removed. Phase stability on all but the shortest baselines was excellent. In this band radio frequency interference (RFI) was considerably more severe than at 21- and 50-cm (Paper I). Regarding each source observation details are as follows:

  • 0335−122

    was observed for 1.5 h at a centre frequency of 319.77 MHz. While there were no overwhelmingly bad frequencies, after flagging inferior data only 135 good antenna pairs were retained.

  • 0347−211

    was observed for 1.4 h at a centre frequency of 360.14 MHz. This band was relatively RFI free and 300 good antenna pairs were retained.

  • 0537−286

    was observed for 1.4 h at a centre frequency of 346.10 MHz, with 210 good antenna pairs. High-amplitude spikes from 345.7 to 345.9 MHz, required the removal of these frequencies.

  • 0913+003

    was observed for 1.4 h at a centre frequency of 348.65 MHz. A spike was also dominant over 348.1 to 348.3 MHz and the full band before 18:00 ut was swamped with RFI, leaving only 40 min on source over a fragmented band (347.5 to 347.9 MHz). Furthermore, since the bandpass was observed at 17:00 ut, no gain calibration is possible for this source.

  • 1026−084

    was observed for 1.9 h at a centre frequency of 316.03 MHz and 250 good antenna pairs were retained. Again severe RFI was present over 315.3 to 315.6 MHz and frequencies above 316.7 MHz.

  • 1228−113

    was searched for H i in a 1.4 h observation centred on 313.69 MHz. In the 210 good antenna pairs severe RFI remained from 313.0 to 313.3 MHz, leaving little remaining band below these frequencies. Since the OH 2Π1/2J= 1/2 (4751 MHz) transition at the redshift of this source fell into the 20-cm band of the GMRT, this line was searched at a centre frequency of 1049 MHz for 4.4 h. RFI was minimal and 350 good antenna pairs were retained.

  • 1251−407

    was observed for 1.4 h at a centre frequency of 305.38 MHz. This band was fairly clean and 285 good antenna pairings were retained.

  • 1351−018

    was searched for H i in a 1.0 h observation centred on 301.76 MHz. Although there was no overwhelmingly apparent RFI over 300.8 to 302.7 MHz, the reference antenna was affected, and so no reliable image could be produced. The optical depth is therefore derived from the averaged visibilities of the 260 good antenna pairings. OH was also searched at a frequency of 354.23 MHz for 1.4 h. No particular frequency was especially subject to RFI, although flagging of some affected baselines was required, leaving 310 good antenna pairs.

  • 1535+004

    was searched for H i in a 1.4 h observation centred on 315.86 MHz. Very little RFI was present giving excellent data over the 400 good antenna pairs used. Again, since the OH 2Π1/2J= 1/2 (4751 MHz) transition at the redshift of this source fell into the 20-cm band of the GMRT, this transition was searched at a centre frequency of 1056 MHz for 3.1 h. One polarization (LL) was severely affected by RFI, and subsequently removed, while for the remaining polarization RFI was minimal and 370 good antenna pairs were retained.

  • 1630−004

    was observed for 1.0 h at a centre frequency of 321.07 MHz. Unfortunately, RFI totally swamped this band.

  • 1937−101

    was searched for H i in a 1.4 h observation centred on 296.72 MHz. No major RFI was present and 300 good antenna pairs were retained. OH was also searched at a frequency of 348.31 MHz for 1.8 h. RFI was minimal and 370 good antenna pairs were retained.

  • 2215+02

    was searched for H i in a 1.4 h observation centred on 310.67 MHz. Although no particular frequencies where affected by RFI, only 190 antenna pairs proved to be of good quality. OH was also searched at a frequency of 364.69 MHz for 3.54 h. Because of RFI many baseline pairs had to be removed, leaving 210 pairs. Note that the bandpass calibrator, 3C 295, used for this source is unknown to miriad and so this could not be used to correct the gains for 2215+02. The flux density scale for this source (Fig. 1) should therefore not be considered as absolute. Furthermore, unlike the lower frequency H i observations, the phase stability was very poor during this observation, preventing the extraction of a high-quality cube.

  • 2245−059

    was observed for 1.4 h at a centre frequency of 330.71 MHz. This band was so dominated by RFI, that no useful data could be retained.

Figure 1

The GMRT spectra. The ordinate in each spectrum shows the flux density (Jy) and the abscissa the Doppler-corrected frequency (MHz). All spectra have been extracted from the spectral cube, with the exception 1351−018 at 302 MHz and 2215+02 at 365 MHz, where the visibilities are averaged together. For the latter and 2215+02 at 311 MHz the flux scale is not absolute.

Figure 1

The GMRT spectra. The ordinate in each spectrum shows the flux density (Jy) and the abscissa the Doppler-corrected frequency (MHz). All spectra have been extracted from the spectral cube, with the exception 1351−018 at 302 MHz and 2215+02 at 365 MHz, where the visibilities are averaged together. For the latter and 2215+02 at 311 MHz the flux scale is not absolute.

2.2 The redshifted millimetre wave observations

The redshifted millimetre wave observations were performed with the Australia Telescope's Tidbinbilla 70-m telescope,3 over several sessions between 2003 November and 2005 March. Again, the sources were selected according to those in which a strong transition4 fell into the 1-cm (≈22 GHz) band. For the backend we used the das_xxyy_64_2048 configuration (dual polarization with 2048 channels over a 64-MHz band), giving a resolution of ≈0.5 km s−1 over a range of ≈± 500 km s−1: such high resolution was required, since although optically thick, the linewidths in the four known millimetre absorption only span a few km s−1 (Wiklind & Combes 1994,1995,1996b,1998), cf. up to 240 km s−1 for the, optically thin, OH absorption (see fig. 5 of Curran et al. 2007b).

The half power beamwidth (HPBW) of the single dish at these frequencies is ≈50 arcsec and, in order to ensure good baselines, we used position switching with 1 min per position over a four-point position switching cycle, each source being observed for a total of 4 h. In general, the baselines were excellent with no baseline subtraction of the spectra (Fig. 2) being required. Flagging of data was only required when communication with the subreflector had been lost. During the observations, the T*A flux scale was periodically calibrated against a thermal load provided by a noise diode. An additional correction for the variation of antenna gain with elevation was applied to each spectrum off-line. The data were reduced using the dfm,5 graphical interface to the spc6 package. A quotient was formed between the source and reference positions and the resulting spectra were averaged and weighted by Tint/T2sys. Lastly, the data were corrected on to the TMB temperature scale by multiplying by the beam efficiency at 22 GHz (η≈ 0.48; Greenhill et al. 2003).

Figure 2

The Tidbinbilla spectra. The ordinate in each spectrum shows the flux density (Jy) and the abscissa the Doppler-corrected frequency (GHz). Each spectrum has been smoothed to a channel width of 4 km s−1. The results are summarized in Table 3.

Figure 2

The Tidbinbilla spectra. The ordinate in each spectrum shows the flux density (Jy) and the abscissa the Doppler-corrected frequency (GHz). Each spectrum has been smoothed to a channel width of 4 km s−1. The results are summarized in Table 3.

2.3 Results

In Table 2 we summarize our results. The column density is derived from the velocity integrated depth, τ≡−ln[1 − (σ/fS)], where σ is rms noise in the case of our exclusive non-detections and S and f the flux density and covering factor of the background continuum source, respectively. In the optically thin regime (where σ/fS≲ 0.3), which applies to the vast majority of the known H i and all of the OH absorbers, the 1σ column density limit is given by  

1
formula
where for H i 21-cm, X= 1.823 × 1018 (Wolfe & Burbidge 1975) and T is the spin temperature (Ts) and for OH, X= 2.38 × 1014 for the 2Π3/2J= 3/2 (1667 MHz) transition and X= 1.61 × 1014 for the 2Π1/2J= 1/2 (4751 MHz) transition (Henkel, Güsten & Baan 1987), with T being the excitation temperature (Tx).

For OH, the values of X are derived from the total column density obtained from a rotational transition, which is given by  

2
formula
where ν is the rest frequency of the J→ J + 1 transition, gJ+1 and AJ+1→J are the statistical weight and the Einstein A coefficient7 of the transition, respectively, and forumla, with gJ= 2J+ 1, is the partition function.8 Since kTxhν for Tx≈ 10 K and ν≲ 5 GHz, equation (2) can be simplified to the above expression (equation 1).9 For the millimetre-wave column densities, based on the four known systems, the covering factor is expected to be close to unity (Wiklind & Combes 1994,1995,1996b,1998) so, as in the optically thin regime, this can be written outside of the integral. However, for the same excitation temperatures the higher frequencies give kTxEJhν and so the column density cannot be approximated via a linear dependence on the excitation temperature. We therefore adopt the canonical value of 10 K at z= 0.10

3 POSSIBLE EFFECTS IN THE NON-DETECTION OF ATOMIC ABSORPTION

From Table 2, it is seen that we do not detect 21-cm absorption in any of the 10 sources for which good data were obtained. In our earlier study of sources from the PHFS (Paper I), we detected 21-cm absorption in one out of four sources, although the only real difference between the PHFS and PQFS samples is that the PQFS has a lower flux limit (0.25 Jy, cf. 0.5 Jy).11 However, due to our restriction that the sources are at redshifts where the HCO+ 0 → 1 transition is redshifted into the 12-mm band, unlike the previous search which spanned z≈ 0.1–3.5, our targets are all at z≳ 2.9, where published searches have been very rare (Fig. 3): in the figure all of the sources in the hatched region are from our search (Table 2), in addition to the zem= 3.497 quasar 1535+004 from Paper I. However, there is one 21-cm detection in this band (z= 3.4 in 0902+343; Uson, Bagri & Cornwell 1991) and so the high-redshift selection on its own cannot be responsible for our non-detections, all of which have been searched to limits comparable to the detections.

Figure 3

The scaled velocity integrated optical depth of the H i line forumla versus the host redshift for the published searches for associated 21-cm absorption (see Table 1 and Appendix A). The filled symbols represent the 21-cm detections and the unfilled symbols the non-detections, with stars designating quasars and circles galaxies (see Section 3.2.2). The hatched region shows the range of our H i searches (Table 2). The other results are from the references quoted in Table 1 (see Appendix A), with the addition of the two zem > 5 non-detections of Carilli et al. (2007).

Figure 3

The scaled velocity integrated optical depth of the H i line forumla versus the host redshift for the published searches for associated 21-cm absorption (see Table 1 and Appendix A). The filled symbols represent the 21-cm detections and the unfilled symbols the non-detections, with stars designating quasars and circles galaxies (see Section 3.2.2). The hatched region shows the range of our H i searches (Table 2). The other results are from the references quoted in Table 1 (see Appendix A), with the addition of the two zem > 5 non-detections of Carilli et al. (2007).

Since we are searching for associated absorption, although covering factors may be an issue, these will not be subject to the same geometrical effects found for intervening systems: specifically damped Lyman α absorption systems (DLAs), where Curran & Webb (2006) show a strong correlation between low absorber/quasar angular diameter distance ratios and 21-cm detections, while high ratios are correlated with non-detections. This suggests that the absorbers located ‘closer’ to us (in an angular sense) have significantly larger covering factors than those which share a similar angular diameter distance to the background quasar.

Therefore, our lack of 21-cm detections must be due to another effect – either low neutral hydrogen column densities, high spin temperatures or low covering factors (see equation 1), although, as stated, the latter would have to be due to small intrinsic absorption cross-sections, since for both the 21-cm detections and non-detections in associated systems, the absorber/quasar angular diameter distance ratios are close to unity. In the absence of measurements of the neutral hydrogen column densities from the Lyman α line, unlike DLAs, we cannot compare how NH I varies between the detections and non-detections nor speculate on spin temperature effects, although Curran et al. (2005,2007c) have shown that spin temperatures in DLAs may not vary by nearly as much as was previously believed (Kanekar & Chengalur 2003), and that these may generally be below Ts≈ 2000 K (at least up to zabs≈ 3.4). Again, however, for these sources we have a degeneracy of three variables, all of which may be mutually coupled (see Curran et al. 2007c), thus making the relative contribution of each very difficult to ascertain.

3.1 Incident fluxes

3.1.1 21-cm luminosities

Although our sample could be highly heterogeneous, since our only criterion for our own targets is that they are high-redshift PQFS sources (Section 1), there must be some reason why 21-cm absorption is detected in some quasars and radio galaxies, while not being detected in others, particularly at high redshift. Curran & Webb (2006) previously investigated a ‘proximity effect’ in DLAs, where a high 21-cm flux may maintain a high population in the upper hyperfine level, thus having relatively few antiparallel spin atoms available to absorb the 1420-MHz radiation (Wolfe & Burbidge 1975). Although there were a few non-detections and zero detections at incident flux densities of ≳104Jy in this sample, there was no overwhelming trend, with the aforementioned geometrical effects being apparently much more significant with regard to the detection of 21-m absorption.

When we consider our sample, we cannot determine incident fluxes for the non-detections, as we have no knowledge of where the neutral gas would be located relative to the emission region. However, since we are just looking for statistical differences in this sample, we can still investigate this effect through the luminosities. The specific luminosity of the quasar at the rest-frame emission frequency, νem, is Lν= 4π D2QSOSobs/(zem+ 1), where DQSO is the luminosity distance to the quasar,12Sobs is the observed flux density13 and zem+ 1 corrects for the redshifting of the frequency increment. Using this expression, in Fig. 4 we show the derived quasar frame luminosities in relation to the velocity integrated optical depth of the 21-cm absorption. From this, the vertical histograms show that, on the whole, the non-detections have been searched sufficiently deeply, many to a higher sensitivity required for the detections. The horizontal histograms show considerable overlap, and, while the average log Lradio is higher for the non-detections, the difference is small (log Lradio= 27.06 cf. 26.71), and the distributions of luminosities for detections and non-detections are not statistically different (the probability of the null hypothesis for the Kolmogorov–Smirnov test is 33.7 per cent). Therefore, although a high incident 21-cm flux may make a detection less likely through a highly populated upper hyperfine level, this does not appear to be an overwhelming cause of the non-detections.

Figure 4

The scaled velocity integrated optical depth of the H i line versus the quasar frame 21-cm luminosity for the quasars searched for associated H i absorption. Again, the hatched region shows the range of our H i searches. Throughout this paper the filled symbols/hatched histogram represent the 21-cm detections and the unfilled symbols/coloured histogram the non-detections.

Figure 4

The scaled velocity integrated optical depth of the H i line versus the quasar frame 21-cm luminosity for the quasars searched for associated H i absorption. Again, the hatched region shows the range of our H i searches. Throughout this paper the filled symbols/hatched histogram represent the 21-cm detections and the unfilled symbols/coloured histogram the non-detections.

3.1.2 Ultraviolet luminosities

Although a 21-cm proximity effect is not apparent for DLAs or associated absorbers, in the UV such an effect is well known, where a high ionizing flux from the quasar is believed to be responsible for the decrease in the number density of the UV Lyman α lines as zabs approaches zem (Weymann, Carswell & Smith 1981; Bajtlik, Duncan & Ostriker 1988). To excite the hydrogen beyond the realm of 21-cm absorption does not require ionization of the gas (by 912-Å photons), but ‘merely’ excitation above the ground state by a Lyman α (1216 Å) photon, although the lifetime in this state is only ∼10−8s. In any case, since both the ionizing and Lyman α photons are ∼106 times as energetic as the spin–flip transition, if the gas is excited by Lyman α absorption, much of it will also be ionized.

Therefore, in order to determine the ∼1000-Å fluxes, we have exhaustively searched the literature and on-line archives to obtain optical and near-IR photometry of as much of the sample as possible (see Appendix A). In some cases, we use the photometry of the Sloan Digital Sky Survey (SDSS; York et al. 2000),14 taken from Data Release 6 (DR6; Adelman-McCarthy et al. 2008), applying the transformations of Fukugita et al. (1996) to ensure consistency between the bands. From these measurements, we estimate the λ≈ 1216(1 +z) Å flux according to the prescription in Appendix B, and thence the luminosity in the rest frame of the galaxy/quasar.

Since the sources cover a range of redshifts, the interpolations/extrapolations necessary to estimate LUV involve different bands for different sources. We can check the validity of the estimates by comparing the interpolated value of the flux at the nominal 1216-Å wavelength (from BVR) with that extrapolated from the JHK bands. The latter situation is similar to the case of extrapolating from optical bands for low-redshift sources. In our sample, there are only three sources in the high (z > 2.5) redshift group for which we have more than one near-IR band available: J0414+0534, 1937−101 [both of which have The Two Micron All Sky Survey (2MASS)15 photometry] and 2215+020 (Francis, Whiting & Webster 2000) and for these, we can compare the extrapolation of the JH measurements to that obtained from BR. 2215+020 has the same extrapolation for both, 1937−101 has JH overestimating the flux by a factor of ∼5, and J0414+0534 overestimates by ∼50. The analysis of the latter source could be affected by the extreme optical–near-IR colour of VK= 10.26 (Lawrence et al. 1995) [see Section 4], most likely due to the presence of dust in the intervening gravitational lens or host galaxy.16 The other two sources, however, show that the extrapolation from longer wavelengths (rest-frame optical) can give a value of the UV flux in broad agreement with the extrapolation from rest-frame UV. We note that an extrapolation using JH for these sources is roughly equivalent to using BR for sources at 0.6 ≲z≲ 1.0.

Plotting our results, in Fig. 5 we see an equal mix of 21-cm detections and non-detections below the median value of the UV luminosity range [(18.04 + 24.91)/2 = 21.48]. However, at higher luminosities the distribution is dominated by non-detections which become exclusive at LUV≳ 1023W Hz−1, the range occupied by our observed sample (Table 2).17 Investigating the likelihood of such a distribution occurring by chance, defining a partition at the median of the sample gives 52 objects (26 detections and 26 non-detections) in the low-luminosity bin and 33 objects (seven detections and 26 non-detections) in the log LUV > 21.48 bin. For an unbiased sample, i.e. there is an equal likelihood of either a detection or non-detection (as may be expected from orientation effects, see next section), the binomial probability of 26 or more detections out of 52 objects occurring is 55 per cent. However, the probability of 25 non-detections or more out of 33 objects in the other bin is just 0.23 per cent. Moving the partition to log LUV= 23.0, thus covering the range of our targets (Fig. 5), in the low-luminosity bin there are 33 detections out of 69 objects (again a near 50 per cent detection rate) and in the high-luminosity bin, 0 detections out of 16. Again, assuming an unbiased sample, the binomial probability of this latter distribution is 0.0015 per cent, a significance of 4.3σ assuming Gaussian statistics,18 against the probability of a 21-cm absorption detection being unrelated to the UV luminosity. Applying a Kolmogorov–Smirnov test, there is a 99.25 per cent confidence that the UV luminosity distribution of the detections differs from that of the non-detections. Therefore, there is a relationship between the UV luminosity and the likelihood of detecting 21-cm absorption.

Figure 5

As Fig. 4, but with the scaled velocity integrated optical depth of the H i line versus the quasar frame UV luminosity.

Figure 5

As Fig. 4, but with the scaled velocity integrated optical depth of the H i line versus the quasar frame UV luminosity.

3.2 Orientation effects

3.2.1 Unified schemes of AGN

Previously, the mix of 21-cm detections and non-detections has been attributed to unified models of active galactic nuclei (AGN). This is an orientation effect due to the presence of a dense subpc circumnuclear disc or TORUS (Thick Obscuration Required by Unified Schemes; Conway 1999) of gas which obscures the UV/optical light from the AGN: in type 1 objects, the obscuration has its rotation axis directed towards us and the AGN and the centralised broad-line region are viewed directly, whereas in type 2 objects the AGN is hidden, and only the more extended narrow-line region is visible (see Antonucci 1993; Urry & Padovani 1995). It is hypothesized that the 21-cm absorption occurs in this obscuration and so is only visible in type 2 objects, where the gas intercepts the line-of-sight to the AGN (e.g. Jaffe & McNamara 1994; Conway & Blanco 1995).

For instance, of four 21-cm detections in a sample of 23 zem < 0.7 radio galaxies (Morganti et al. 2001), two are narrow-line radio galaxies, indicating type 2 objects, and the other two are weak-line radio galaxies. Therefore, superficially at least, this is consistent with the 21-cm absorption occurring in an intervening torus or disc (Morganti et al. 2001). However, the fact that there are significant redshift (velocity) offsets between the 21-cm and optical lines, leads the authors to conclude that some of the H i responsible for the absorption is located farther out than the central sub-pc occupied by the dense obscuration. The presence of significant gas motions, due to infall, outflows, jet interactions as well as general galactic rotation19 is also suggested by the large velocity offsets found in a sample of 19 detections by Vermeulen et al. (2003)[see Fig. 6].

Figure 6

The 21-cm absorption strength versus the velocity offset of the 21-cm absorption from the host. Here we have plotted each individually resolved absorption component at its line strength and velocity offset for each of the detections (Table 1). The solid symbols represent blueshifted (approaching) H i and the hollow symbols redshifted (receding) H i. In associated systems these offsets represent peculiar motions rather than the Hubble flow. Furthermore, there are often velocity differences between different optical lines and so these velocities cannot be reliably converted to distance offsets to yield the incident flux on the absorbing gas.

Figure 6

The 21-cm absorption strength versus the velocity offset of the 21-cm absorption from the host. Here we have plotted each individually resolved absorption component at its line strength and velocity offset for each of the detections (Table 1). The solid symbols represent blueshifted (approaching) H i and the hollow symbols redshifted (receding) H i. In associated systems these offsets represent peculiar motions rather than the Hubble flow. Furthermore, there are often velocity differences between different optical lines and so these velocities cannot be reliably converted to distance offsets to yield the incident flux on the absorbing gas.

Furthermore, Pihlström, Conway & Vermeulen (2003) note that 21-cm absorption is more likely to be detected in radio galaxies than in quasars, suggesting that the quasars are viewed at lower inclinations, thus having the observed radio emission bypass the absorbing gas which obscures the optical/UV light. The orientations for a very limited sample (Pihlström et al. 2003) are quantified by the core prominence, where the galaxies have core emission fractions of fc < 0.03, cf. fc > 0.04 for the quasars, the larger core prominence being attributed to a more direct viewing angle to the AGN. From a larger sample, Gupta & Saikia (2006b) find that the galaxies have a median value of fc= 0.011, compared to fc= 0.028 for the quasars. However, the distribution becomes more mixed with nearly half of the quasars located below the median value for the galaxies (fc∼ 0.01) and with five galaxies at fc≳ 0.03. Additionally, although none of the quasars has been detected in 21-cm absorption, all of these five galaxies exhibit strong (NH I≥ 1.6 × 1018Ts/f cm−2) absorption, thus weakening the argument that core dominance indicates lower inclinations, or at least, the amount of cold neutral gas lying along the sightline to the AGN.

Many of these core fractions have however been estimated20, and overestimates could explain the five 21-cm absorbing galaxies exhibiting a large core prominence. Nevertheless, from their CSS and GPS sample (Gupta & Saikia 2006b) there is a 21-cm detection rate of one out of nine for quasars, cf. 15 out of 23 for galaxies, supporting the existence of a bias. Quantifying this, the binomial probability of 15 or more out of 23 detections occurring in one class, while eight or more non-detections occur in another class is 0.20 per cent, again consistent with quasars being the result of lower inclined obscuring tori.

3.2.2 Quasar–galaxy classifications

The classification schemes used by these authors do not apply to all the sources under consideration in this work. We have therefore applied our own scheme in order to examine the effect of the relative strength of the quasar/AGN nucleus over the host galaxy. The aim is to distinguish sources whose appearance is dominated by the nuclear source (the ‘quasars’) from those whose appearance is dominated by the extended stellar light of the host galaxy (the ‘galaxies’). The latter sources are those with either intrinsically weaker AGN or the type 2 AGN that are obscured from the line-of-sight. The classification was mostly done on a morphological basis, using SuperCOSMOS Sky Survey images and classification (which does a star/galaxy classification based on the source profile), DSS2 images, as well as detailed imaging and spectroscopic analysis (where the spectra are examined for evidence of galaxy light or quasar emission lines) available from the literature. Each source was examined individually to ensure the classifications were consistent. There are some obvious caveats with this process. By looking at the morphology, we are biasing ourselves somewhat against high-redshift galaxies, since we lose the resolution and surface-brightness sensitivity for detecting the galaxy light. However, many of the high-redshift sources have higher resolution imaging or spectroscopic observations available that allow us to determine the importance of the host galaxy in the source's appearance.

In order to compare the proportions of detections and non-detections for galaxies and quasars, we use the following statistic. If we have two measured proportions forumla and forumla, with the total proportion forumla, then  

3
formula
which has a standard normal distribution (mean zero, variance one) under the null hypothesis, being that the two proportions (forumla and forumla) are the same. We examined the proportion of sources that we classified as galaxies (as opposed to quasars), and found that sources with detected absorption had a much higher galaxy proportion (28 out of 34) than sources without absorption (28 out of 56). These fractions are different at 99.67 per cent confidence. Additionally, almost all the high-LUV sources are ‘quasars’– only one out of the 24 sources with LUV > 1022.5W Hz−1 is a galaxy.

3.2.3 Our results in the context of the unified schemes

All of the above discussion on orientation effects applies to the low-redshift sample, where searches have previously been concentrated (Fig. 3).21 Of our own low redshift searches (Paper I), one target out of three was detected in 21-cm absorption, i.e. as per the 33 per cent detection rate of Vermeulen et al. (2003), the rate expected from unified schemes.22 For the two undetected, 1450−338 (z= 0.368) has an estimated UV luminosity of LUV= 4 × 1018W Hz−1 and 2300−189 (z= 0.129) LUV= 1 × 1020W Hz−1, cf. LUV= 5 × 1019W Hz−1 for the detection in 1555−140 (z= 0.097). That is, the UV luminosity of the detection lies between those of the non-detections, although all lie well within the LUV≲ 1023W Hz−1 detection range (Fig. 5). Regarding their orientations, 1450−338 is an apparently dust-reddened quasar (Francis et al. 2000), 2300−189 is a Seyfert 1 galaxy, whereas the strong (NH I= 4.2 × 1019Ts/f cm−2) 21-cm absorber 1555−140 has a reddened type 2 AGN spectrum superimposed on a galaxy (Wilkes et al. 1983), and is itself a large galaxy in the centre of a group. With the one detection occurring in the only type 2 object, our low-luminosity results are consistent with unified schemes, as discussed above.

However, although our high-redshift criterion selects highly luminous sources, note that half (eight out of 16) of those with LUV≳ 1023W Hz−1 have redshifts of zem≤ 0.7323 and, in common with the other LUV≳ 1023W Hz−1 targets, these are all non-detections. This raises the question of whether orientation effects alone can account for the observed differences. That is, why do some of the low-redshift non-detections exhibit low UV luminosities when, like their high-luminosity counterparts, we expect to have a direct view to the UV emitting region in these type 1 objects?

3.3 The significance of known intervening absorbers

Most of our targets also have intervening DLAs and so if the photons emitted from the quasar are ionizing, a significant number of these must have been redshifted to 1216 Å by time they encounter the absorber, providing a continuum for the Lyman α absorption. In fact, DLAs could be common to many of the objects searched for host absorption, although ground-based observations of the Lyman α line are usually restricted to redshifts of z≥ 1.7, where the line is redshifted into optical bands.24 From Fig. 3, we see that only our targets, the two high-z 21-cm detections (Uson, Bagri & Cornwell 1991; Moore, Carilli & Menten 1999, Table 1) and the two targets of Carilli et al. (2007) are of sufficiently high redshift to illuminate such absorbers. The DLAs which intervene our targets (nine in total towards seven sources) are all themselves at redshifts (zabs≥ 1.947), comparable to those of the background quasars (Table 4).

Table 4

The targets with detected intervening DLAs (NH I≥ 2 × 1020cm−2) and sub-DLAs. Note that 0913+003, 1026−084 and 1251−407 have not featured in any of our prior statistics since the H i search in 0913+003 was ruined by RFI and in the latter two we searched for OH only (H i is redshifted out of the band).

QSO zabs log NH I Ref. zem Δz 
0335−122 3.178 20.8 E01 3.442 0.063 
0347−211 1.947 20.3 E01 2.944 0.338 
0537−286 2.974 20.3 E01 3.104 0.033 
0913+003 2.774 20.3 E01 3.074 0.079 
1026−084 3.42 20.1 P01 4.276 0.193 
… 4.05 19.7 P01 … 0.045 
1228−113 2.193 20.6 E01 3.528 0.418 
1251−407 3.533 20.6 E01 4.464 0.205 
.. 3.752 20.3 E01 … 0.150 
QSO zabs log NH I Ref. zem Δz 
0335−122 3.178 20.8 E01 3.442 0.063 
0347−211 1.947 20.3 E01 2.944 0.338 
0537−286 2.974 20.3 E01 3.104 0.033 
0913+003 2.774 20.3 E01 3.074 0.079 
1026−084 3.42 20.1 P01 4.276 0.193 
… 4.05 19.7 P01 … 0.045 
1228−113 2.193 20.6 E01 3.528 0.418 
1251−407 3.533 20.6 E01 4.464 0.205 
.. 3.752 20.3 E01 … 0.150 

Notes. No DLAs have been found towards 1351−018, 1535+004, 1937−101, 2215+02, 2245−059 (Ellison et al. 2001), 1937−101 (Lanzetta et al. 1991).

References: E01 –Ellison et al. (2001); P01 –Péroux et al. (2001).

The redshift of the background source in the rest frame of the absorber is given by Δz=[(zem+ 1)/(zabs+ 1) − 1] and at Δz= 0.33, the 912-Å (ionizing) photon is redshifted to 1216 Å. All but two of the DLAs have Δz < 0.33 (Table 4), meaning that the radiation which is Lyman α absorbed must have been non-ionizing at the source. Two of the DLAs have also been searched, and not detected, in 21-cm absorption; 0335−122 and 0537−286 (Kanekar & Chengalur 2003). The non-detections place spin temperature/covering factor ratios of Ts/f≳ 2000 and ≳700 K, respectively, and since both occult very compact radio sources (table 2 of Curran et al. 2005), f may be close to unity, indicating high spin temperatures in these absorbers (cf. Curran & Webb 2006). Therefore, as shown by these two cases, the non-detection of 21-cm does not necessarily imply a lack of neutral gas close to the quasar.

In the absence of total neutral hydrogen column densities, we cannot estimate the limits for our targets (Table 2), although for NH I≳ 1020cm−2, Ts≳ 103K, which is around the typical upper limit for the detection of 21-cm absorption in DLAs (fig. 5 of Curran et al. 2007c). The spin temperature measures the relative populations of the two possible spin states (Purcell & Field 1956), although excitation to the n= 2 level by Lyman α photons will also raise the spin temperature (Field 1959). Furthermore, Bahcall & Ekers (1969) show that both the 21-cm (cf. Fig. 4) and the Lyman α (cf. Fig. 5) flux can contribute to the spin temperature at absorber–quasar separations of less than a few tens of kpc, i.e. for associated systems.

In addition to many of the background quasars of our sample emitting a large fraction of non-ionizing, high-energy (1216 < λ < 912 Å) photons, there exists a number of proximate damped Lyman α absorption systems (PDLAs), where at Δv≤ 3000 km s−1, the ionization of the gas is expected to be dominated by the quasar, rather than the background UV flux.25 Unlike the detections in the associated sample (Table 1), many of the PDLAs are subject to LUV≳ 1023W Hz−1 (Fig. 7). From the figure we see that few of the PDLAs appear to arise in the host (i.e. at Δv≲ 300 km s−1). Comparing these with the associated sample (Fig. 6), a Kolmogorov–Smirnov test on the velocity offsets gives a probability of only 9.7 × 10−12 that the associated detections and the PDLAs are drawn from the same sample. Therefore, although there are several PDLAs for which LUV≳ 1023W Hz−1 at Δv≲ 300 km s−1, PDLAs are not (generally, at least) associated systems (see also Møller, Warren & Fynbo 1998). That is, although it is feasible that a search for associated absorbers could overlap significantly with a search for PDLAs, we show that this is generally not the case. For the PLDAs, Δz≲ 0.01 and, as per the DLAs detected towards our targets (Table 4), this would suggest again that much of the flux from the quasar is non-ionizing at the source.26

Figure 7

The UV (≈912 Å) luminosity versus the velocity offset of the Lyman α absorption from the background quasar for the SDSS Data Release 5 (DR5) PDLA sample (Prochaska et al. 2008). The quasar redshifts range from zem= 2.308 to 5.185 and the luminosities have been estimated as per the associated sample (Section 3.1.2). The solid symbols represent blueshifted (approaching) H i and the hollow symbols redshifted (receding) H i. The horizontal lines signify the detection cut-off for the associated systems and the vertical line, the fiducial 300 km s−1 offset above which the gas is believed to be unassociated.

Figure 7

The UV (≈912 Å) luminosity versus the velocity offset of the Lyman α absorption from the background quasar for the SDSS Data Release 5 (DR5) PDLA sample (Prochaska et al. 2008). The quasar redshifts range from zem= 2.308 to 5.185 and the luminosities have been estimated as per the associated sample (Section 3.1.2). The solid symbols represent blueshifted (approaching) H i and the hollow symbols redshifted (receding) H i. The horizontal lines signify the detection cut-off for the associated systems and the vertical line, the fiducial 300 km s−1 offset above which the gas is believed to be unassociated.

Since, by definition, the DLAs towards our targets are subject to a high 1216-Å flux, there must be a significant portion of ionizing photons present, which could render the gas less detectable in Lyman α (and 21-cm) absorption, although at these high column densities self-shielding effects (Zheng & Miralda-Escudé 2002) should counteract this somewhat. Note, however, that none of the sightlines towards our targets exhibit further DLAs between those listed in Table 4 and zem, where the ionizing photons will be less redshifted. In Fig. 8 we convert the luminosities to fluxes for the DLAs towards our targets and the PDLA sample. Although, there is a tentative trend for the column density of the neutral gas to decrease with flux for our targets, this is not borne out by the larger PDLA sample. Although this would suffer a selection effect by being limited to the NH I≥ 2 × 1020cm−2 DLA defined cut-off (where self-shielding becomes particularly significant; Zheng & Miralda-Escudé 2002), at higher incident fluxes there appears to be no additional appreciable photoionization. However, PDLAs are considerably less numerous than expected, which Prochaska, Hennawi & Herbert-Fort (2008) attribute to photoevaporation. Another factor which could cover any neutral gas column density–incident UV flux anticorrelation is the possibility that the observed redshift differences between the absorbers and quasars do not provide accurate distances in these cluster environments. This could be the case for the Δv≤ 300 km s−1 PDLAs of Fig. 7.

Figure 8

The neutral hydrogen column density versus the incident UV flux at the DLAs towards our targets (Table 4, unfilled markers) and the Δv≥ 300 km s−1, zabs < zem PDLAs (Prochaska et al. 2008, filled markers). These criteria have been applied in order to select absorbers directly along the line-of-sight and with the redshifts not dominated by peculiar motions. The scatter in the distribution remains if the whole PDLA sample is used.

Figure 8

The neutral hydrogen column density versus the incident UV flux at the DLAs towards our targets (Table 4, unfilled markers) and the Δv≥ 300 km s−1, zabs < zem PDLAs (Prochaska et al. 2008, filled markers). These criteria have been applied in order to select absorbers directly along the line-of-sight and with the redshifts not dominated by peculiar motions. The scatter in the distribution remains if the whole PDLA sample is used.

3.4 The chicken or the egg: ionizing UV flux or orientation effects

3.4.1 Orientation: ionization

Since Lyman α absorption is not detected in the hosts of any of our high-redshift quasar sample (Ellison et al. 2001; Péroux et al. 2001), it is not surprising that 21-cm absorption remains undetected. Naturally, intervening DLAs require sufficient background UV flux against which to detect absorption, and by selecting such high-redshift sources, we are selecting those which are sufficiently luminous in the UV to enable the detection of intervening absorption, but perhaps too bright to host a large column of neutral gas close to the host.

This raises the question of whether it is the UV luminosity ionizing the neutral gas, which would otherwise be there, or the distribution of the gas (a face-on torus), which is responsible for the non-detection of 21-cm absorption in these sources. For the PDLAs there are several cases where LUV≳ 1023W Hz−1 in what could be associated absorption (Δv≤ 300 km s−1) and orientation effects could be consistent with the complete lack of any PDLAs in our own (admittedly, small) target sample (Table 4). This paucity may however be expected without invoking orientation effects, since PDLAs constitute only a small fraction (∼5 per cent) of the DLA sample (Ellison et al. 2002; Prochaska et al. 2008) and those which may be associated (Δv≤ 300 km s−1) are even rarer (≲1 per cent).

In AGN, large-scale (kpc) outflows of ionized gas, directed along the radio jets are rife (particularly in Seyfert galaxies, see table 1.2 of Curran 2000b,27). These are believed to be due to either nuclear gas which is ionized and driven out along the radio jets (e.g. Begelman, Blandford & Rees 1984; Schulz 1988; Colbert et al. 1996,1998) or photoionized ambient galactic gas (Pedlar, Unger & Dyson 1985; Unger et al. 1987; Falcke, Wilson & Simpson 1998). This latter model28 is consistent with the UV radiation ionizing all of the gas, while giving a high UV luminosity in our direction. Furthermore, the paucity of extended ionized structures in type 1 Seyfert galaxies (Pogge 1989), which unified schemes dictate are intrinsically identical to their type 2 counterparts, suggests that these are being viewed at low inclinations and so do not extend beyond the central AGN from our viewpoint. Furthermore, at all redshifts (Fig. 9), all of the log LUV≳ 23 radio sources have been flagged as quasars (by us as well as by Vermeulen et al. 2003; Gupta et al. 2006).29 That is, the sources in which the UV flux appears to be directed towards us are believed to be type 1 objects, in addition to having a very low (possibly zero) likelihood of exhibiting 21-cm absorption.

Figure 9

The UV luminosity–redshift distribution for the sample. The symbols and histograms are as Fig. 4, but now the shapes represent the AGN classifications, with triangles representing type 1 objects and squares type 2s – the legend shows the number of each according to the UV luminosity cut-off (cf. Table 5). The distribution indicates that the extrapolations used for the low-redshift sources (Section 3.1.2) do not appear to introduce an overwhelming bias in comparison to the high-redshift sources (at least for the quasars). Should the extrapolations be overly erroneous, this would result in a systematic shift in the luminosities of the low-redshift sample, although there would still be the subsample of those undetected in 21-cm absorption at higher UV luminosities.

Figure 9

The UV luminosity–redshift distribution for the sample. The symbols and histograms are as Fig. 4, but now the shapes represent the AGN classifications, with triangles representing type 1 objects and squares type 2s – the legend shows the number of each according to the UV luminosity cut-off (cf. Table 5). The distribution indicates that the extrapolations used for the low-redshift sources (Section 3.1.2) do not appear to introduce an overwhelming bias in comparison to the high-redshift sources (at least for the quasars). Should the extrapolations be overly erroneous, this would result in a systematic shift in the luminosities of the low-redshift sample, although there would still be the subsample of those undetected in 21-cm absorption at higher UV luminosities.

Although random orientation effects could give the 50 per cent split of 21-cm detections and non-detections below the median value of log LUV < 21.5 (Section 3.1.2), at higher UV luminosities the number of 21-cm detections drops significantly, especially at log LUV≳ 23. Also, for the low-luminosity sample, there is a detection rate of 1/2 for the galaxies and 1/3 for the quasars (Table 5) and, while this is consistent with previous studies (Pihlström et al. 2003), the fact that the detection rate for the low UV luminosity quasars is significantly greater than for the LUV≳ 1023W Hz−1 quasars (0/16), suggests that these may be different a beast than their low UV luminosity counterparts.

Table 5

Partitioning of the 21-cm detections and non-detections with respect to LUV= 1023W Hz−1.

 LUV≲ 1023W Hz−1 LUV≳ 1023W Hz−1 
Dets Nons Total Dets Nons Total 
Galaxies 27 24 51 – – 
Quasars 12 18 16 16 
Total 33 36 69 16 16 
 LUV≲ 1023W Hz−1 LUV≳ 1023W Hz−1 
Dets Nons Total Dets Nons Total 
Galaxies 27 24 51 – – 
Quasars 12 18 16 16 
Total 33 36 69 16 16 

This is also illustrated in Fig. 9,30 where it is seen that all of zem≳ 1 targets have UV luminosities in excess of LUV∼ 1022W Hz−1, demonstrating that there is a selection effect at play, where at high redshifts we are targeting the most optically bright sources, which are naturally better known at such high-luminosity distances. At zem≲ 1 (luminosity distances ≲6 Gpc), we see the whole range of luminosities. Again, the fact that all of the detections occur at LUV≲ 1023W Hz−1 is clearly evident, but whether the difference in luminosities is an intrinsic property is difficult to ascertain: the low UV luminosity quasars undetected in 21-cm absorption may be at sufficiently low inclinations so that a large column of neutral gas is not intercepted by the radio emission, while being inclined highly enough so that the axis of the jet does not cross our line of sight, i.e. the UV radiation is not directed towards us in these cases. Considering this, these low-luminosity quasars may represent intermediate types (1 to 1.5; Keel 1980; Maiolino & Rieke 1995), which are known to have stronger ionization lines than type 2 objects (Cohen 1983).

3.4.2 Orientation: obscuration

If the orientation is important in the detection rate of H i absorption, one manifestation of it should be a higher extinction of the quasar emission by dust associated with the torus. We can test this via an optical–near-IR colour–colour diagram, using the photometry in Appendix A, and in Fig. 10 we show the VR and RK colours, where available, for the sample. From this there is little apparent difference between the two populations: although there are a few sources which lie off the main distribution in the direction expected from dust reddening, the bulk of the detections and non-detections are found to be concentrated in the same part of the plot. This indicates that the detections are not significantly more dust reddened than the non-detections. While this might indicate that the orientation hypothesis is not supported, it should be noted that if the dusty torus did significantly extinguish the quasar light, the host galaxy starlight would then dominate, lessening the apparent reddening.

Figure 10

The RK colour versus the VR colour for the sample. As per Fig. 4, the filled symbols represent the 21-cm detections and the unfilled symbols the non-detections.

Figure 10

The RK colour versus the VR colour for the sample. As per Fig. 4, the filled symbols represent the 21-cm detections and the unfilled symbols the non-detections.

As a check on the effect of orientation, we have examined the spectral type for each source in the full sample. This entailed an exhaustive literature search to find, ideally, published measurements of emission line fluxes, or alternatively, a published spectrum (Tables 6 and 7). This information was then used to classify the AGN as a type 1 (i.e. showing broad permitted lines), or a type 2, where only narrow lines are present. The result is shown in Fig. 9, where the detection rate for type 1 objects is 10/37, compared to 19/39 for the type 2s, indicating a preference for type 2 objects to exhibit 21-cm absorption.

Table 6

The sources detected in 21-cm absorption, listed by their B1950.0 or J2000.0 name as given in the 21-cm search paper (Table 1). Columns 12 and 13 give the estimated luminosity at 1216 Å (W Hz−1) and our determination of the AGN type.

Source Class zem B (mag) Ref V (mag) Ref R (mag) Ref K (mag) Ref log LUV (W Hz−1Type Ref 
J0025−2602 Gal 0.3220 20.300 30 –  18.084 30 15.674 61 20.100 75 
0108+388 Gal 0.6685 –  –  22.000 65 16.690 71 20.309 43 
J0119+3210 Gal 0.0600 16.271 30 –  14.749 30 12.600 31 20.548 28 
J0141+1353 Gal 0.6210 22.327 30 20.920 27 20.876 30 16.680 27 20.777 64 
J0414+0534 Gal 2.6365 24.100 29 23.800 42 21.270 13.540 42 22.188 42 
J0410+7656 Gal 0.5985 –  –  21.200 70 –  – 43 
J0431+2037 Gal 0.2190 22.174 30 –  19.085 30 14.924 16 18.039 –  
0500+019 Gal 0.5846 22.500 17 21.350 20.682 15.430 24 20.367 35 
0758+143 QSO 1.1946 19.976 30 17.460 57 18.972 30 15.300 60 22.825 73 
J0834+5534 Gal 0.2420 18.921 30 17.390 17.180 46 14.180 61 20.719 
J0901+2901 Gal 0.1940 19.321 30 18.078 18.600 12 15.200 16 21.280 26 
0902+343 Gal 3.3980 –  23.800 23.500 21 19.900 21 22.422 – 20 
J0909+4253 QSO 0.6700 18.960 19.049 18.220 14.860 60 22.699 
J1124+1919 Gal 0.1650 22.082 30 21.448 20.513 30 15.930 16 19.190 –  
12032+1707 Gal 0.2170 18.758 30 –  17.327 30 14.864 61 20.949 11 
J1206+6413 Gal 0.3710 21.847 30 20.790 19.910 55 –  19.908 26 
J1326+3154 Gal 0.3700 21.367 30 19.822 18.882 30 14.940 16 19.638 18 
J1347+1217 QSO 0.1217 16.615 62 16.050 62 15.718 30 13.216 61 21.736 39 
J1357+4354 Gal 0.6460 –  22.708 20.951 30 –  18.620 –  
J1400+6210 Gal 0.4310 22.137 30 20.373 19.530 30 16.130 16 19.459 43 
J1407+2827 Gal 0.0766 16.345 30 14.910 13 14.240 13 11.601 61 20.144 23 
1413+135 QSO 0.2467 21.055 30 20.000 33 18.461 30 14.928 61 19.105 72 
1504+377 Gal 0.6715 –  21.808 20.800 69 16.100 71 20.295 69 
1555−140 Gal 0.0970 18.280 25 16.930 25 16.280 25 13.060 25 19.719 79 
J1815+6127 QSO 0.6010 21.272 30 –  19.122 30 –  20.665 76 
J1816+3457 Gal 0.2448 20.342 30 –  18.459 30 15.525 61 20.034 –  
J1821+3942 Gal 0.7980 19.598 30 –  18.135 30 15.023 61 22.202 77 
J1944+5448 Gal 0.2630 21.732 30 –  18.591 30 15.000 61 18.424 67 
J1945+7055 Gal 0.1010 18.726 30 –  17.199 30 13.369 61 20.067 67 
J2052+3635 Gal 0.3550 22.083 30 –  21.200 30 –  20.648 14 
J2245+3941 Gal 0.0811 17.788 30 16.550 58 15.993 30 12.388 61 19.947 43 
J2255+1313 QSO 0.5430 19.535 30 19.590 19.190 15 –  22.530 26 
J2316+0405 Gal 0.2199 18.595 62 17.440 62 17.220 30 13.991 61 21.081 74 
J2355+4950 Gal 0.2379 21.101 30 –  18.400 51 15.112 61 18.940 43 
Source Class zem B (mag) Ref V (mag) Ref R (mag) Ref K (mag) Ref log LUV (W Hz−1Type Ref 
J0025−2602 Gal 0.3220 20.300 30 –  18.084 30 15.674 61 20.100 75 
0108+388 Gal 0.6685 –  –  22.000 65 16.690 71 20.309 43 
J0119+3210 Gal 0.0600 16.271 30 –  14.749 30 12.600 31 20.548 28 
J0141+1353 Gal 0.6210 22.327 30 20.920 27 20.876 30 16.680 27 20.777 64 
J0414+0534 Gal 2.6365 24.100 29 23.800 42 21.270 13.540 42 22.188 42 
J0410+7656 Gal 0.5985 –  –  21.200 70 –  – 43 
J0431+2037 Gal 0.2190 22.174 30 –  19.085 30 14.924 16 18.039 –  
0500+019 Gal 0.5846 22.500 17 21.350 20.682 15.430 24 20.367 35 
0758+143 QSO 1.1946 19.976 30 17.460 57 18.972 30 15.300 60 22.825 73 
J0834+5534 Gal 0.2420 18.921 30 17.390 17.180 46 14.180 61 20.719 
J0901+2901 Gal 0.1940 19.321 30 18.078 18.600 12 15.200 16 21.280 26 
0902+343 Gal 3.3980 –  23.800 23.500 21 19.900 21 22.422 – 20 
J0909+4253 QSO 0.6700 18.960 19.049 18.220 14.860 60 22.699 
J1124+1919 Gal 0.1650 22.082 30 21.448 20.513 30 15.930 16 19.190 –  
12032+1707 Gal 0.2170 18.758 30 –  17.327 30 14.864 61 20.949 11 
J1206+6413 Gal 0.3710 21.847 30 20.790 19.910 55 –  19.908 26 
J1326+3154 Gal 0.3700 21.367 30 19.822 18.882 30 14.940 16 19.638 18 
J1347+1217 QSO 0.1217 16.615 62 16.050 62 15.718 30 13.216 61 21.736 39 
J1357+4354 Gal 0.6460 –  22.708 20.951 30 –  18.620 –  
J1400+6210 Gal 0.4310 22.137 30 20.373 19.530 30 16.130 16 19.459 43 
J1407+2827 Gal 0.0766 16.345 30 14.910 13 14.240 13 11.601 61 20.144 23 
1413+135 QSO 0.2467 21.055 30 20.000 33 18.461 30 14.928 61 19.105 72 
1504+377 Gal 0.6715 –  21.808 20.800 69 16.100 71 20.295 69 
1555−140 Gal 0.0970 18.280 25 16.930 25 16.280 25 13.060 25 19.719 79 
J1815+6127 QSO 0.6010 21.272 30 –  19.122 30 –  20.665 76 
J1816+3457 Gal 0.2448 20.342 30 –  18.459 30 15.525 61 20.034 –  
J1821+3942 Gal 0.7980 19.598 30 –  18.135 30 15.023 61 22.202 77 
J1944+5448 Gal 0.2630 21.732 30 –  18.591 30 15.000 61 18.424 67 
J1945+7055 Gal 0.1010 18.726 30 –  17.199 30 13.369 61 20.067 67 
J2052+3635 Gal 0.3550 22.083 30 –  21.200 30 –  20.648 14 
J2245+3941 Gal 0.0811 17.788 30 16.550 58 15.993 30 12.388 61 19.947 43 
J2255+1313 QSO 0.5430 19.535 30 19.590 19.190 15 –  22.530 26 
J2316+0405 Gal 0.2199 18.595 62 17.440 62 17.220 30 13.991 61 21.081 74 
J2355+4950 Gal 0.2379 21.101 30 –  18.400 51 15.112 61 18.940 43 

References: 1 – SDSS DR6, Adelman-McCarthy et al. (2008); 2 –Allen, Ward & Hyland (1982); 3 –Angonin-Willaime et al. (1999); 4 –Carballo et al. (1999); 5 –Chavushyan et al. (2001); 6 –Chu, Zhu & Butcher (1986); 7 –Chun et al. (2006); 8 –Cody & Braun (2003); 9 –Cohen & Osterbrock (1981); 10 –Dallacasa, Falomo & Stanghellini (2002); 11 –Darling & Giovanelli (2006); 12 –de Koff et al. (1996); 13 –de Vaucouleurs & Longo (1988); 14 –de Vries et al. (2000); 15 –de Vries et al. (1997); 16 –de Vries et al. (1998); 17 –Drinkwater et al. (1997); 18 –Dunlop et al. (1989); 19 –Eales (1985); 20 –Eales et al. (1993); 21 –Eisenhardt & Dickinson (1992); 22 –Ellison et al. (2005); 23 –Eracleous & Halpern (1994); 24 – Francis (private communication); 25 –Francis et al. (2000); 26 –Gelderman & Whittle (1994); 27 –Glikman et al. (2004); 28 –Gao & Solomon (2004); 29 –Gregg et al. (2002); 30 – SuperCOSMOS Sky Survey, Hambly et al. (2001); 31 –Heckman et al. (1983); 32 –Henstock et al. (1997); 33 –Hewitt & Burbidge (1989); 34 –Hirst, Jackson & Rawlings (2003); 35 –Hook et al. (2003); 36 –Hunstead, Murdoch & Shobbrook (1978); 37 –Hyland & Allen (1982); 38 –Jackson & Browne (1991); 39 –Kim, Veilleux & Sanders (1998); 40 –Kristian, Sandage & Katem (1974); 41 –Lanzetta et al. (1991); 42 –Lawrence et al. (1995); 43 –Lawrence et al. (1996); 44 –Lilly & Longair (1984); 45 –Lipovetsky, Neizvestny & Neizvestnaya (1988); 46 –Marchã et al. (2001); 47 –Marchã et al. (1996); 48 –Marziani et al. (2003); 49 –Marziani et al. (1996); 50 – NASA Extragalactic Data base (NED); 51 –O'Dea, Baum & Morris (1990); 52 –Overzier et al. (2006); 53 –Prestage & Peacock (1983); 54 –Rao, Turnshek & Nestor (2006); 55 –Roche & Eales (2000); 56 –Sandage, Véron & Wyndham (1965); 57 –Sandage (1965); 58 –Sandage (1973); 59 –Schneider et al. (2005); 60 –Simpson & Rawlings (2000); 61 – 2MASS, Skrutskie et al. (2006); 62 –Smith & Heckman (1989); 63 –Snellen et al. (1999); 64 –Spinrad et al. (1985); 65 –Stanghellini et al. (1993); 66 –Stickel, Fried & Kühr (1993); 67 –Srianand & Khare (1993); 68 –Stickel & Kühr (1996); 69 –Stickel & Kühr (1994); 70 –Schmitt & Kinney (1996); 71 –Stickel et al. (1996); 72 –Stocke et al. (1992); 73 –Stockton & Ridgway (2001); 74 –Tadhunter et al. (2002); 75 –Tadhunter et al. (1993); 76 –Vermeulen & Taylor (1995); 77 –Vermeulen et al. (1996); 78 –White et al. (2000); 79 –Wilkes et al. (1983); 80 –Wills & Lynds (1978); 81 –Winn et al. (2002); 82 –Zickgraf et al. (1997).

Table 7

As Table 6 but for the sources not detected in 21-cm absorption.

Source Class zem B (mag) Ref V (mag) Ref R (mag) Ref K (mag) Ref log LUV (W Hz−1Type Ref 
J0003+2129 QSO 0.4520 21.005 30 20.580 10 19.650 10 –  20.971 –  
0131−001 QSO 0.8790 23.340 25 22.500 25 20.780 25 16.780 25 20.221 –  
J0157−1043 QSO 0.6160 17.504 30 –  17.039 30 –  23.380 48 
J0201−1132 QSO 0.6690 16.232 30 –  16.073 30 13.860 37 24.176 75 
J0224+2750 Gal 0.3102 19.502 30 –  18.263 30 15.250 44 21.225 23 
0335−122 QSO 3.4420 21.018 30 20.110 22 20.199 30 17.510 22 23.722 
0347−211 QSO 2.9940 20.476 30 –  20.297 30 17.900  23.722 35 
J0348+3353 Gal 0.2430 20.723 30 –  19.110 15 14.390 16 20.121 23 
J0401+0036 Gal 0.4260 20.200 40 19.010 40 18.532 30 –  20.969 
J0521+1638 QSO 0.7590 19.370 56 18.840 56 18.480 15 15.380 60 22.580 26, 38 
0537−286 QSO 3.0140 19.290 17 –  18.789 30 16.770 24 24.231 79 
J0542+4951 QSO 0.5450 18.450 57 17.800 57 17.210 15 –  22.311 26, 34 
J0556−0241 Gal 0.2350 20.968 30 –  19.533 30 –  20.150 14 
J0609+4804 Gal 0.2769 21.198 30 –  18.767 30 –  19.349 –  
J0655+4100 Gal 0.0216 15.021 30 –  13.996 30 10.357 61 20.648 47 
J0709+7449 Gal 0.2921 19.982 30 –  17.540 19 13.790 53 19.898 50 
J0741+3112 QSO 0.6350 16.517 30 16.100 54 16.322 30 16.100 23.990 
J0815−0308 Gal 0.1980 18.490 62 16.940 62 16.797 30 13.858 61 20.707 –  
J0840+1312 QSO 0.6808 18.370 80 17.940 80 17.622 30 15.280 60 22.947 49 
J0913+5919 QSO 5.1200 –  23.281 24.948 –  22.071 
J0924−2201 Gal 5.2000 –  –  25.850 52 –  21.893 –  
J0927+3902 QSO 0.6948 17.064 30 –  16.486 30 –  23.603 
J0939+8315 Gal 0.6850 –  –  20.140 44 –  – 64 
J0943−0819 Gal 0.2280 19.401 30 –  18.100 70 14.750 16 20.868 14 
J0954+7435 Gal 0.6950 –  –  21.700 70 –  – –  
1026+084 QSO 4.2760 21.070 30 –  19.154 30 –  24.308 82 
J1035+5628 Gal 0.4590 –  21.244 20.200 65 –  19.889 43 
J1120+1420 Gal 0.3620 –  20.935 20.100 30 17.100 16 20.098 –  
J1159+2914 QSO 0.7290 17.489 30 18.113 17.652 30 –  23.955 78 
1228−113 QSO 3.5280 22.010 17 –  19.115 30 16.370 24 23.754 17 
J1252+5634 QSO 0.3210 17.760 56 17.930 56 17.660 15 –  22.949 
J1308−0950 Gal 0.4640 20.767 30 20.500 53 18.439 30 –  20.340 75 
J1313+5458 QSO 0.6130 –  21.735 20.374 30 –  19.581 77 
1351−018 QSO 3.7070 21.030 17 19.696 19.277 30 17.070 24 24.014 59 
J1421+4144 Gal 0.3670 20.496 30 19.330 18.560 19 15.910 44 20.435 50 
J1443+7707 Gal 0.2670 –  –  18.730 15 –  – 26 
1450−338 Gal 0.3680 22.520 25 20.400 25 19.390 25 15.230 25 18.629 17 
J1511+0518 Gal 0.0840 17.993 30 16.200 40 16.634 30 12.081 61 20.353 
1535+004 QSO 3.4970 –  –  –  19.540 24 – –  
J1540+1447 QSO 0.6050 17.480 80 17.000 80 17.240 30 13.640 23.529 66 
J1546+0026 Gal 0.5500 19.730 80 18.900 80 –  16.420 16 22.703 –  
J1623+6624 Gal 0.2030 19.477 30 –  17.430 30 –  20.004 63 
J1642+6856 QSO 0.7510 19.723 30 –  19.219 30 –  22.667 43 
J1658+0741 QSO 0.6210 19.993 30 –  19.598 30 –  22.441 79 
J1823+7938 Gal 0.2240 19.269 30 –  17.415 30 13.866 61 20.385 32 
J1829+4844 QSO 0.6920 16.260 30 –  16.860 15 14.250 60 24.692 26 
J1831+2907 Gal 0.8420 21.917 30 –  20.200 30 –  21.201 68 
J1845+3541 Gal 0.7640 –  –  21.900 65 –  – 77 
1937−101 QSO 3.7870 18.800 30 –  17.188 30 13.816 61 24.910 41 
J2022+6136 Gal 0.2270 19.830 30 –  18.146 30 –  20.334 26 
J2137−2042 Gal 0.6350 20.400 53 –  19.286 30 –  21.808 74 
2149+056 QSO 0.7400 23.700 25 22.050 25 20.850 25 17.170 25 19.582 68 
2215+02 QSO 3.5720 21.840 25 20.420 25 20.140 25 19.340 25 23.613 17 
J2250+1419 QSO 0.2370 16.760 80 16.640 80 17.243 30 –  23.616 
2300−189 Gal 0.1290 18.430 17 –  16.569 30 13.060 61 20.099 36 
J2321+2346 Gal 0.2680 20.315 30 –  18.468 30 14.710 16 20.187 –  
J2344+8226 QSO 0.7350 21.769 30 –  20.220 70 15.850 16 21.165 43 
Source Class zem B (mag) Ref V (mag) Ref R (mag) Ref K (mag) Ref log LUV (W Hz−1Type Ref 
J0003+2129 QSO 0.4520 21.005 30 20.580 10 19.650 10 –  20.971 –  
0131−001 QSO 0.8790 23.340 25 22.500 25 20.780 25 16.780 25 20.221 –  
J0157−1043 QSO 0.6160 17.504 30 –  17.039 30 –  23.380 48 
J0201−1132 QSO 0.6690 16.232 30 –  16.073 30 13.860 37 24.176 75 
J0224+2750 Gal 0.3102 19.502 30 –  18.263 30 15.250 44 21.225 23 
0335−122 QSO 3.4420 21.018 30 20.110 22 20.199 30 17.510 22 23.722 
0347−211 QSO 2.9940 20.476 30 –  20.297 30 17.900  23.722 35 
J0348+3353 Gal 0.2430 20.723 30 –  19.110 15 14.390 16 20.121 23 
J0401+0036 Gal 0.4260 20.200 40 19.010 40 18.532 30 –  20.969 
J0521+1638 QSO 0.7590 19.370 56 18.840 56 18.480 15 15.380 60 22.580 26, 38 
0537−286 QSO 3.0140 19.290 17 –  18.789 30 16.770 24 24.231 79 
J0542+4951 QSO 0.5450 18.450 57 17.800 57 17.210 15 –  22.311 26, 34 
J0556−0241 Gal 0.2350 20.968 30 –  19.533 30 –  20.150 14 
J0609+4804 Gal 0.2769 21.198 30 –  18.767 30 –  19.349 –  
J0655+4100 Gal 0.0216 15.021 30 –  13.996 30 10.357 61 20.648 47 
J0709+7449 Gal 0.2921 19.982 30 –  17.540 19 13.790 53 19.898 50 
J0741+3112 QSO 0.6350 16.517 30 16.100 54 16.322 30 16.100 23.990 
J0815−0308 Gal 0.1980 18.490 62 16.940 62 16.797 30 13.858 61 20.707 –  
J0840+1312 QSO 0.6808 18.370 80 17.940 80 17.622 30 15.280 60 22.947 49 
J0913+5919 QSO 5.1200 –  23.281 24.948 –  22.071 
J0924−2201 Gal 5.2000 –  –  25.850 52 –  21.893 –  
J0927+3902 QSO 0.6948 17.064 30 –  16.486 30 –  23.603 
J0939+8315 Gal 0.6850 –  –  20.140 44 –  – 64 
J0943−0819 Gal 0.2280 19.401 30 –  18.100 70 14.750 16 20.868 14 
J0954+7435 Gal 0.6950 –  –  21.700 70 –  – –  
1026+084 QSO 4.2760 21.070 30 –  19.154 30 –  24.308 82 
J1035+5628 Gal 0.4590 –  21.244 20.200 65 –  19.889 43 
J1120+1420 Gal 0.3620 –  20.935 20.100 30 17.100 16 20.098 –  
J1159+2914 QSO 0.7290 17.489 30 18.113 17.652 30 –  23.955 78 
1228−113 QSO 3.5280 22.010 17 –  19.115 30 16.370 24 23.754 17 
J1252+5634 QSO 0.3210 17.760 56 17.930 56 17.660 15 –  22.949 
J1308−0950 Gal 0.4640 20.767 30 20.500 53 18.439 30 –  20.340 75 
J1313+5458 QSO 0.6130 –  21.735 20.374 30 –  19.581 77 
1351−018 QSO 3.7070 21.030 17 19.696 19.277 30 17.070 24 24.014 59 
J1421+4144 Gal 0.3670 20.496 30 19.330 18.560 19 15.910 44 20.435 50 
J1443+7707 Gal 0.2670 –  –  18.730 15 –  – 26 
1450−338 Gal 0.3680 22.520 25 20.400 25 19.390 25 15.230 25 18.629 17 
J1511+0518 Gal 0.0840 17.993 30 16.200 40 16.634 30 12.081 61 20.353 
1535+004 QSO 3.4970 –  –  –  19.540 24 – –  
J1540+1447 QSO 0.6050 17.480 80 17.000 80 17.240 30 13.640 23.529 66 
J1546+0026 Gal 0.5500 19.730 80 18.900 80 –  16.420 16 22.703 –  
J1623+6624 Gal 0.2030 19.477 30 –  17.430 30 –  20.004 63 
J1642+6856 QSO 0.7510 19.723 30 –  19.219 30 –  22.667 43 
J1658+0741 QSO 0.6210 19.993 30 –  19.598 30 –  22.441 79 
J1823+7938 Gal 0.2240 19.269 30 –  17.415 30 13.866 61 20.385 32 
J1829+4844 QSO 0.6920 16.260 30 –  16.860 15 14.250 60 24.692 26 
J1831+2907 Gal 0.8420 21.917 30 –  20.200 30 –  21.201 68 
J1845+3541 Gal 0.7640 –  –  21.900 65 –  – 77 
1937−101 QSO 3.7870 18.800 30 –  17.188 30 13.816 61 24.910 41 
J2022+6136 Gal 0.2270 19.830 30 –  18.146 30 –  20.334 26 
J2137−2042 Gal 0.6350 20.400 53 –  19.286 30 –  21.808 74 
2149+056 QSO 0.7400 23.700 25 22.050 25 20.850 25 17.170 25 19.582 68 
2215+02 QSO 3.5720 21.840 25 20.420 25 20.140 25 19.340 25 23.613 17 
J2250+1419 QSO 0.2370 16.760 80 16.640 80 17.243 30 –  23.616 
2300−189 Gal 0.1290 18.430 17 –  16.569 30 13.060 61 20.099 36 
J2321+2346 Gal 0.2680 20.315 30 –  18.468 30 14.710 16 20.187 –  
J2344+8226 QSO 0.7350 21.769 30 –  20.220 70 15.850 16 21.165 43 

If we split the sample at LUV= 1023W Hz−1, we see that all 16 of the high-luminosity sources are found to be type 1 objects. Being exclusive non-detections, this is in line with unified schemes. However, for the low-luminosity sources, the detection rates are close to 50 per cent for both types (10/21 for type 1s and 19/37 for type 2s). So while the raw detection rates for type 1s and type 2s appear different (actually only at a 1.95σ level), the dominant cause of the different detection rates between the type 1 and type 2 objects appears to be due to the 16 high-luminosity type 1 objects. Without these, i.e. for LUV < 1023W Hz−1, both AGN types have a ∼50 per cent probability of exhibiting H i absorption. In other words, a type 1 object does not automatically result in a non-detection of H i absorption, nor does a type 2 necessarily result in a detection. This flies in the face of the notion that the strength of the 21-cm absorption is determined by the aspect of the central obscuring torus (Section 3.2) and is strong evidence that the absorption occurs beyond the parsec-scale, possibly in the main disc of the galaxy, the orientation of which appears to have little bearing on the AGN type.

3.4.3 Host galaxy

The third point to consider is thus the quasar host galaxy. The detection of associated absorption simply requires that there is neutral gas at similar velocities to the quasar nucleus. While we have been discussing the prospect of gas associated with the active nucleus (for instance, with the surrounding torus) and whether the nuclear obscuration plays a part, the role of gas in the host galaxy needs to be considered. Quasars, particularly at high redshifts, are observed to reside in galaxies of diverse types: for example, Peng et al. (2006) find that quasar hosts at 1 < z < 4.5 span a range of morphologies consistent with early-type to discy/late-type galaxies. While some early-type galaxies are known from targeted searches to have significant H i content, particularly in the field (e.g. Morganti et al. 2006; Orienti et al. 2007), blind surveys, such as the Arecibo Legacy Fast ALFA (ALFALFA) survey (di Serego Alighieri et al. 2007), show that early-type galaxies in clusters have a much lower neutral gas content. Those quasars with detected H i absorption are then more likely to be found in discy or late-type galaxies.

Our observed trend in the detection rate as a function of UV luminosity could thus be explained by a changing mix of host galaxy types, in the sense that a larger fraction of the more luminous quasars are found in early-type galaxies. While observational constraints limit our knowledge of such a tendency at z > 2, this trend is known to apply at lower redshifts. For instance, Taylor et al. (1996) find that the hosts of all radio-loud quasars studied, as well as the most powerful radio-quiet quasars (all with z < 0.35), have a de Vaucouleurs r1/4 law profile (characteristic of elliptical galaxies), whereas the less powerful radio-quiet quasars exhibited exponential disc profiles. Taylor et al. (1996) suggest that the most luminous quasars reside in elliptical galaxies, regardless of their radio properties. More recent observations of z < 0.3 quasars, using adaptive optics (Guyon, Sanders & Stockton 2006), find that most luminous quasars (with L > 2L*H) have elliptical host galaxies (this includes the majority of the radio-loud quasars). To summarize, in this scenario the most luminous quasars of our sample are in early-type galaxies, which have a lower neutral gas content than later types, and so the chance of detecting significant H i absorption is greatly reduced compared to the less-luminous quasars.

4 POSSIBLE EFFECTS IN THE NON-DETECTION OF MOLECULAR ABSORPTION

In light of the absence of atomic absorption due to the large UV luminosities, possibly due to unfavourable orientations, it is not surprising that molecular absorption remains undetected. However, were H i detected, a detectable molecular abundance would still not necessarily be expected from this sample because of the following.

  • The relationship between molecular fraction and optical–near-IR colour (Paper I) suggests that our sources would simply not be red enough to be detectable in molecular absorption at these sensitivities. The current sample had been selected before the conclusions of Paper I had been fully formulated, and so the range of VK colours most likely lie off to the left in Fig. 11 (the colours range from VK= 1.08–2.63, where available).

  • Furthermore, due to the metallicity and molecular fraction evolution noted in DLAs (Curran et al. 2004a), at z≥ 3 we may expect much lower abundances than present day values. From a search of millimetre lines in DLAs with the Green Bank Telescope, Curran et al. (2004b) reached similar limits as this survey (Table 3). However, upon applying the molecular fraction evolution to the limits, they found that the survey was only sensitive of molecular fractions of close to unity, a value which even the DLAs detected in H2 fall very far short of, although for VK≳ 5.3 such high fractions may be expected (fig. 1 of Curran et al. 2006).

  • Lastly, at such high redshifts the cosmic microwave background will raise excitation temperatures of Tx= 10 K at z= 0 to ≳20 K. This has the effect of decreasing the sensitivities to these ground state transitions – by a factor of 2 for OH 18-cm (2Π3/2J= 3/2) and by up to a factor of 4 for the J= 0 → 1 transitions (Curran et al. 2004b). One solution to this is to search higher transitions redshifted into the 3-mm band (cf. the 12-mm band), although lower flux densities, compounded with the need for much better observing conditions, makes this a poor trade.

Figure 11

The normalized OH line strength forumla versus optical–near-IR colour for the known OH absorbers. These are represented by the filled symbols, with the least-squares fit shown for the four millimetre systems. The correlation for the five OH absorbers is significant at the 2.0σ level, which rises to 3.0σ for the molecular fraction/VK correlation for these sources plus the H2-bearing DLAs (fig. 1 of Curran et al. 2006). The unfilled symbols show the other sources where H i has been detected and OH absorption searched: the three limits shown in fig. 7 of Curran et al. (2006)[blue] (the lower limits on the abscissa designate RK magnitudes) with a further three from Gupta et al. (2006)[blue]. Note that the OH limits have been rescaled according to the method described in Curran et al. (2007b), who find full width at half-maximum FWHMOH= FWHMH I for the five known OH absorbers: combining the otherwise unknown FWHMOH with the optical depth limit, gives a more accurate estimate for the upper limit than quoting the column density limit per each Δv channel. This has the effect of degrading the apparent sensitivity, although some can be recovered since usually FWHMOH > Δv. Therefore, for the non-detections, we scale each of the OH column density limits by forumla, thus giving the limit for a single channel ‘smoothed’ to FWHMOH. Associated absorption is designated by a square and intervening absorption (due to a gravitational lens) by a triangle.

Figure 11

The normalized OH line strength forumla versus optical–near-IR colour for the known OH absorbers. These are represented by the filled symbols, with the least-squares fit shown for the four millimetre systems. The correlation for the five OH absorbers is significant at the 2.0σ level, which rises to 3.0σ for the molecular fraction/VK correlation for these sources plus the H2-bearing DLAs (fig. 1 of Curran et al. 2006). The unfilled symbols show the other sources where H i has been detected and OH absorption searched: the three limits shown in fig. 7 of Curran et al. (2006)[blue] (the lower limits on the abscissa designate RK magnitudes) with a further three from Gupta et al. (2006)[blue]. Note that the OH limits have been rescaled according to the method described in Curran et al. (2007b), who find full width at half-maximum FWHMOH= FWHMH I for the five known OH absorbers: combining the otherwise unknown FWHMOH with the optical depth limit, gives a more accurate estimate for the upper limit than quoting the column density limit per each Δv channel. This has the effect of degrading the apparent sensitivity, although some can be recovered since usually FWHMOH > Δv. Therefore, for the non-detections, we scale each of the OH column density limits by forumla, thus giving the limit for a single channel ‘smoothed’ to FWHMOH. Associated absorption is designated by a square and intervening absorption (due to a gravitational lens) by a triangle.

Table 3

Summary of the search for millimetre absorption lines in the hosts of z≳ 3 PQFS sources. νobs is the observed frequency of the line, σrms is the rms noise reached per 4 km s−1 channel, Scont is the continuum flux density and τ is the optical depth of the line calculated per channel. The column density of each molecule is calculated for f= 1 and an excitation temperature of Tx= 10 K at z= 0. The magnitudes, where available, are given in Table 2. For 0434−188, B= 20.86 (Ellison et al. 2005) and K= 16.21 (Drinkwater et al. 1997) and for 0601−17, B= 20.445 and R= 20.172 (Hambly et al. 2001).

PKS zem Line νobs (GHz) σrms (mJy) Scont (Jy) τ N (cm−2z range 
0335−122 3.442 HCN 0 → 1 19.953 4.5 0.245 <0.055 <9.7 × 1012/f 3.435–3.449 
0434−188 2.702 HCN 0 → 1 23.942 8.7 0.220 <0.12 <1.7 × 1013/f 2.698–2.707 
… … HCO+ 0 → 1 24.092 4.6 0.202 <0.068 <4.3 × 1012/f 2.698–2.707 
0601−17 2.711 HCN 0 → 1 23.884 7.0 0.202 <0.10 <1.4 × 1013/f 2.706–2.716 
… … HCO+ 0 → 1 24.034 6.3 0.207 <0.091 <5.7 × 1012/f 2.705–2.716 
1026−084 4.276 CO 0 → 1 21.848 3.9 0.137 <0.085 <7.8 × 1015/f 4.268–4.284 
PKS zem Line νobs (GHz) σrms (mJy) Scont (Jy) τ N (cm−2z range 
0335−122 3.442 HCN 0 → 1 19.953 4.5 0.245 <0.055 <9.7 × 1012/f 3.435–3.449 
0434−188 2.702 HCN 0 → 1 23.942 8.7 0.220 <0.12 <1.7 × 1013/f 2.698–2.707 
… … HCO+ 0 → 1 24.092 4.6 0.202 <0.068 <4.3 × 1012/f 2.698–2.707 
0601−17 2.711 HCN 0 → 1 23.884 7.0 0.202 <0.10 <1.4 × 1013/f 2.706–2.716 
… … HCO+ 0 → 1 24.034 6.3 0.207 <0.091 <5.7 × 1012/f 2.705–2.716 
1026−084 4.276 CO 0 → 1 21.848 3.9 0.137 <0.085 <7.8 × 1015/f 4.268–4.284 

5 SUMMARY

We have undertaken a survey for H i 21-cm and rotational molecular absorption in the hosts of radio sources at redshifts of z≥ 2.9, and report no detections in the 13 sources for which we have good data. Upon comparing our search criteria with those formulated in Paper I (Curran et al. 2006), we are not surprised that molecules (OH, HCN, HCO+ and CO) were not detected in the 10 separate sources searched: with optical–near-IR colours of VK≤ 2.63 (at least where these are available), the sources do not exhibit the degree of reddening indicative of the dust abundances which would permit molecular fractions of close to unity, the limit to which current radio observations are sensitive.

However, since the comoving density of H i at z∼ 3 is expected to be many times higher than present values (e.g. Péroux et al. 2001), the absence of detections of 21-cm absorption was surprising. We rule out the possibility that the non-detections are due exclusively to high 1420-MHz continuum fluxes maintaining an overpopulated 21-cm upper hyperfine (antiparallel) level. We do find, however, that all of our targets are quasars and the UV continuum luminosities are in excess of LUV∼ 1023W Hz−1 (Tables 6 and 7). In comparison to the previous searches for redshifted 21-cm absorption, the following facts are observed:

  • A mix of 21-cm detections and non-detections at lower redshifts (mostly z≲ 1) is well documented, where the detection rate is higher in galaxies than in quasars. This skew in the distribution is attributed to the possibility that radio galaxies are type 2 sources, whereas quasars are type 1 objects. That is, the more direct view to the active nucleus in a quasar means that the dusty obscuring torus, invoked by unified models, is orientated so that 21-cm absorption does not occur along our sightline. Although our own low-redshift results (Paper I) are consistent with this scenario, we find little evidence for a significantly higher degree of dust reddening by any obscuring gas in the 21-cm detections.

  • Despite the fact that half of the whole LUV≳ 1023W Hz−1 sample are at z≤ 0.73, this is the first time that a UV luminosity bias has been noted. The high UV luminosity again may be consistent with these being type 1 objects, where the bright UV continuum suggests we are seeing the accretion disc directly, unobscured by the circumnuclear torus. However, although random orientation effects can explain the mix of detections and non-detections in the low-luminosity sample, the high-luminosity sample, in which there are no detections, remains unexplained. This suggests that there are additional effects at play.

As well as several of our high-redshift targets having intervening DLAs at similar redshifts to the background quasar, there exists a sample of proximate DLAs, where zabszem, and both of these groups show that large columns of neutral gas can in fact exist close to LUV≳ 1023W Hz−1 QSOs. Since the gas in these intervening absorbers is not associated, and are thus free to have any aspect with respect to the AGN, this does not contradict the possibility that our non-detections are the result of orientation between the associated gas and the line-of-sight to the quasar. Furthermore, no other absorbers have been found closer to our targets (Ellison et al. 2001; Péroux et al. 2001), perhaps suggesting that either this gas is unfavourably orientated or that there is a proximity effect, where the intense UV radiation is photoionizing the associated gas clouds (Bajtlik et al. 1988; Prochaska et al. 2008). This latter scenario could also be responsible for high spin temperatures in the DLAs towards our targets, which are sufficiently remote to host large columns of neutral gas, two of which have been searched and not detected in 21-cm absorption (Kanekar & Chengalur 2003).

Finally, although 21-cm absorption shows a slight preference to be present in galaxies over quasars, by determining the AGN spectroscopic type for each object (a total of 76), we find that below LUV∼ 1023W Hz−1 the presence of 21-cm absorption shows no preference for AGN type. That is, both type 1 and type 2 objects have a 50 per cent likelihood of exhibiting 21-cm absorption and any apparent bias against type 1 objects is due solely to the 16 LUV≳ 1023W Hz−1 objects. This means the following.

  • The UV luminosity, rather than the orientation of the AGN, can determine whether 21-cm absorption can be detected in the host galaxy, where at high luminosities 21-cm is never detected and at low luminosities the odds are even.

  • The H i absorption probably does not occur in the obscuring torus, but in the large-scale galactic disc or is possibly associated with in-falling or out-flowing material (e.g. Jaffe & McNamara 1994; Pihlström et al. 1999; Morganti et al. 2001; Vermeulen et al. 2003). As such, there is as yet no definite explanation why there is only a 50 per cent detection rate in LUV≲ 1023W Hz−1 sources.

  • If our classifications are to be trusted, although type 1 objects are more likely to arise in quasars and type 2s in galaxies, this is not the case for every object.

Whether all UV luminous sources arise in type 1 objects is difficult to ascertain, and the fact that there are low UV luminosity non-detections is somewhat of a puzzle. If, as we believe, the absorption is occurring in the galactic disc, it may be the orientation of this which is responsible for the observed 21-cm optical depths. This would suggest that the galactic disc does not necessarily share a similar orientation to the dense circumnuclear gas on the parsec scale.31 Alternatively, a possible explanation for the high-luminosity non-detections is that this selection biases towards a specific class of host galaxy, i.e. gas-poor, early types, in our targets as well as the LUV≳ 1023W Hz−1 quasars at z≤ 0.73. This may render the orientation argument also invalid in these cases.

The exclusive non-detections of 21-cm absorption for LUV≳ 1023W Hz−1 quasars indicate why we do not see absorption in our PQFS sample. Our selection of targets was biased towards high UV luminosity sources in several ways. First, the requirement of having a measured optical redshift preferentially selects objects that are relatively bright in the optical (i.e. rest-frame UV), where a suitable spectrum can be obtained in a feasible observing time. In the PQFS, 509 of the 878 sources have measured redshifts, so that about 42 per cent of the sample is unavailable for this study. If anything, these missing sources would have luminosities lower than the quasars of our sample and so the detection of 21-cm absorption in any of these would have little bearing on our result.

Secondly, our high-redshift selection clearly biases towards the brightest UV sources. This is despite our B≳ 19 selection, which gives the luminosity ceiling of LUV≲ 3 × 1024W Hz−1 at z∼ 3,32 well above the 1023W Hz−1 fiducial limit. Therefore, in order to detect associated H i at zem≥ 3, sources with UV luminosities of LUV≲ 1023W Hz−1, should therefore be targeted. At luminosity distances of ≥24.5 Gpc, this corresponds to λ≥ 4860 Å flux densities of ≲6 μJy, or V≳ 22. For the one detection of associated H i at zem≥ 3, zem= 3.3968 in the radio galaxy 0902+343 (Uson et al. 1991), the observed flux density is ≈1 μJy, rendering this detectable towards a UV luminosity of LUV= 3 × 1022W Hz−1, which is, not surprisingly, at the upper end of the 21-cm detections.

1
Except in the case of PKS 1830−211 where a gravitational lens of undetermined redshift was previously known (Subrahmanyan et al. 1990). The redshift was finally determined through a 14 GHz wide spectral scan of the 3-mm band (Wiklind & Combes 1996a).
2
The GMRT is run by the National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research.
3
The Australia Telescope is funded by the Commonwealth of Australia for operations as a National Facility managed by CSIRO.
4
That is, transitions of CO, HCO+ and HCN, which are optically thick in the four known redshifted millimetre absorption systems.
7
These are taken from Chandra et al. (1995), Chandra, Maheshwari & Sharma (1996) or derived from the dipole moment (e.g. Rohlfs & Wilson 2000).
8
The energy of each level, EJ, is obtained from the JPL spectral line catalogue (Pickett et al. 1998). An on-line column density calculator based on equation (2) is available at http://www.phys.unsw.edu.au/~sjc/column/
9
Where the values for X are derived using a weighting factor of forumla and AJ+1→J= 7.71 × 10−11s−1 (Carrington & Miller 1967) for the 1667-MHz transition. For the 4751-MHz transition, forumla and AJ+1→J= 7.7597 × 10−10s−1 (http://www.strw.leidenuniv.nl/~moldata/datafiles/oh@hfs.dat).
10
Since we are analysing high-redshift sources, we scale this with the temperature of the cosmic microwave background, TCMB= 2.73 (1 +z), giving Tx= 17–21 K over z= 2.702– 4.276.
11
Also a slightly stricter lower limit to the spectral index (α > −0.4, cf. α > −0.5, where Sν∝να).
12
Throughout this paper we use H0= 75 km s−1Mpc−1, Ωmatter= 0.27 and ΩΛ= 0.73.
13
Since we are searching for absorption at the host redshift, the quasar frame 21-cm flux density is given by the observed value.
14
The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are the American Museum of Natural History, Astrophysical Institute Potsdam, University of Basel, University of Cambridge, Case Western Reserve University, the University of Chicago, Drexel University, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, the Joint Institute for Nuclear Astrophysics, the Kavli Institute for Particle Astrophysics and Cosmology, the Korean Scientist Group, the Chinese Academy of Sciences (LAMOST), Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, Ohio State University, University of Pittsburgh, University of Portsmouth, Princeton University, the United States Naval Observatory and the University of Washington.
15
2MASS is a joint project of the University of Massachusetts and the Infrared Processing and Analysis Center/California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation.
16
Note that on the basis of their detection of H i in the lens in conjunction with a strong limit on the OH column density, Curran et al. (2007b)[and references therein] suggest that the reddening is occurring in the z= 2.64 host galaxy.
17
The mean luminosity of the detections is log LUV= 20.5 ± 1.2, cf. log LUV= 21.8 ± 1.7 for all of the non-detections and log LUV= 20.8 ± 1.1 for the low-luminosity non-detections.
18
Should there be an overwhelming error in the band extrapolations, to the point where the z > 1 luminosities cannot be used in the comparison, for the z < 1 sources alone the result remains significant at 2.9σ.
19
Although gas in the galactic disc may be expected to share the same orientation as the sub-pc obscuration (Curran et al. 1998, cf. Greenhill et al. 2003; see also Curran 2000a).
20
Using assumed spectral indices and at a frequency of 8 GHz, which is an order of magnitude higher than the typical observed redshifted 21-cm frequency.
21
The large gap in redshift space between these and the high-redshift targets (Fig. 3) is due to a lack of (RFI free) coverage over ≈350 to 700 MHz.
22
The detection rate for a given half-opening angle, α, is 1 − cos α, giving a detection rate of 29 per cent for a randomly oriented population with a 90° opening angle.
23
This demonstrates that the segregation of our high-luminosity sample is not due to a systematic underestimate in the low-redshift UV luminosities, where the extrapolation of the spectral energy distributions (SEDs) are generally more extreme than for the high-redshift sources (Section 3.1.2).
24
Additionally, at redshifts of z≳ 4, the increasing line density per unit redshift of the Lyman α forest makes the identification of DLAs very difficult.
25
At 104≳Δv≳ 105km s−1 (Table 4), none of the DLAs intervening our targets is proximate.
26
Since DLAs do not generally occult quasars of sufficient radio flux to be detected in 21-cm absorption at present (≳0.1 Jy; see Curran et al. 2002), it is likely to remain unknown for some time whether these PDLAs are host to cold neutral gas.
28
Where the torus may be a consequence of the weak radiation emitting from the equator of the continuum source, with the cone arising from gas ionized by the strong polar radiation (Pedlar et al. 1998).
29
Although having classified all of the sources for ourselves (Section 3.2.2), not all of the assigned designations agree.
30
The symbols are discussed in the next section.
31
Contradicting low-redshift surveys (Keel 1980; Maiolino & Rieke 1995, and to a certain degree, Curran 2000a), which find that find that intermediate Seyferts of types 1, 1.2 and 1.5 will occur in face-on galaxies while those of types 1.8 and 1.9 will occur in the edge-on cases.
32
The fact that one of our targets lies on the flux limit with B= 19, while also being the highest redshift source observed (z= 3.8), gives the one point above LUV= 3 × 1024W Hz−1 (Fig. 9).

We would like to thank the anonymous reviewers for their helpful and supportive comments. Christian Henkel for the 6-cm OH Einstein A coefficients. Also, many thanks to Jim Lovell for performing all of the Tidbinbilla observations and the GMRT telescope operators for their extensive assistance.

We acknowledge financial support from the Access to Major Research Facilities Programme which is a component of the International Science Linkages Programme established under the Australian Government's innovation statement, Backing Australia's Ability.

This research has made use of the NASA/IPAC Extragalactic Data base (NED) which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration. This research has also made use of NASA's Astrophysics Data System Bibliographic Services.

This work made use of the Frequently Asked Questions of the Statistical Consulting Center for Astronomy, operated at the Department of Statistics, Penn State University, M. G. Akritas Director).

Funding for the SDSS and SDSS-II has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Science Foundation, the US Department of Energy, the National Aeronautics and Space Administration, the Japanese Monbukagakusho, the Max Planck Society and the Higher Education Funding Council for England. The SDSS Web site is http://www.sdss.org/.

REFERENCES

Adelman-McCarthy
J. K.
et al.,
2008
,
ApJS
 ,
175
,
297
Allen
D. A.
Ward
M. J.
Hyland
A. R.
,
1982
,
MNRAS
 ,
199
,
969
Angonin-Willaime
M.-C.
Vanderriest
C.
Courbin
F.
Burud
I.
Magain
P.
Rigaut
F.
,
1999
,
A&A
 ,
347
,
434
Antonucci
R. R. J.
,
1993
,
ARA&A
 ,
31
,
473
Bahcall
J. N.
Ekers
R. D.
,
1969
,
ApJ
 ,
157
,
1055
Bajtlik
S.
Duncan
R. C.
Ostriker
J. P.
,
1988
,
ApJ
 ,
327
,
570
Begelman
M. C.
Blandford
R. D.
Rees
M. J.
,
1984
,
Rev. Mod. Phys.
 ,
56
,
255
Carballo
R.
González-Serrano
J. I.
Benn
C. R.
Sánchez
S. F.
Vigotti
M.
,
1999
,
MNRAS
 ,
306
,
137
Carilli
C. L.
Perlman
E. S.
Stocke
J. T.
,
1992
,
ApJ
 ,
400
,
L13
Carilli
C. L.
Rupen
M. P.
Yanny
B.
,
1993
,
ApJ
 ,
412
,
L59
Carilli
C. L.
Menten
K. M.
Reid
M. J.
Rupen
M. P.
Yun
M. S.
,
1998
,
ApJ
 ,
494
,
175
Carilli
C. L.
Wang
R.
Hoven
M.
Dwarakanath
K.
Chengalur
J.
Wyithe
S.
,
2007
,
AJ
 ,
133
,
2841
Carrington
A.
Miller
T. A.
,
1967
,
Nat
 ,
214
,
998
Chandra
S.
Kegel
W. H.
Roy
R. J. L.
Hertenstein
T.
,
1995
,
A&AS
 ,
114
,
175
Chandra
S.
Maheshwari
V. U.
Sharma
A. K.
,
1996
,
A&AS
 ,
117
,
557
Chavushyan
V.
Mujica
R.
Gorshkov
A. G.
Konnikova
V. K.
Mingaliev
M. G.
Valdéz
J. R.
,
2001
,
Astron. Rep.
 ,
45
,
79
Chengalur
J. N.
De Bruyn
A. G.
Narasimha
D.
,
1999
,
A&A
 ,
343
,
L79
Chu
Y.-Q.
Zhu
X.-F.
Butcher
H.
,
1986
,
Acta Astron. Sin.
 ,
27
,
151
Chun
M. R.
Gharanfoli
S.
Kulkarni
V. P.
Takamiya
M.
,
2006
,
AJ
 ,
131
,
686
Cody
A. M.
Braun
R.
,
2003
,
A&A
 ,
400
,
871
Cohen
R. D.
,
1983
,
ApJ
 ,
273
,
489
Cohen
R. D.
Osterbrock
D. E.
,
1981
,
ApJ
 ,
243
,
81
Colbert
E. J. M.
Baum
S. A.
Gallimore
J. F.
O'Dea
C. P.
Christensen
J. A.
,
1996
,
ApJ
 ,
467
,
551
Colbert
E. J. M.
Baum
S. A.
O'Dea
C. P.
Veilleux
S.
,
1998
,
ApJ
 ,
496
,
786
Combes
F.
Wiklind
T.
,
1998
,
ESO Messenger
 ,
91
,
29
Conway
J.
,
1999
, in
Carilli
C.
Radford
S.
Menton
K.
Langston
G.
, eds, ASP Conf. Ser. Vol. 156,
Highly Redshifted Radio Lines
 .
Astron. Soc. Pac.
, San Francisco, p.
259
Conway
J. E.
Blanco
P. R.
,
1995
,
ApJ
 ,
449
,
L131
Curran
S. J.
,
2000a
,
A&AS
 ,
144
,
271
Curran
S. J.
,
2000b
, PhD thesis, Chalmers Univ. Technology
Curran
S. J.
Webb
J. K.
,
2006
,
MNRAS
 ,
371
,
356
Curran
S. J.
Johansson
L. E. B.
Rydbeck
G.
Booth
R. S.
,
1998
,
A&A
 ,
338
,
863
Curran
S. J.
Webb
J. K.
Murphy
M. T.
Bandiera
R.
Corbelli
E.
Flambaum
V. V.
,
2002
,
Publ. Astron. Soc. Aust.
 ,
19
,
455
Curran
S. J.
Webb
J. K.
Murphy
M. T.
Carswell
R. F.
,
2004a
,
MNRAS
 ,
351
,
L24
Curran
S. J.
Murphy
M. T.
Pihlström
Y. M.
Webb
J. K.
Bolatto
A. D.
Bower
G. C.
,
2004b
,
MNRAS
 ,
352
,
563
Curran
S. J.
Murphy
M. T.
Pihlström
Y. M.
Webb
J. K.
Purcell
C. R.
,
2005
,
MNRAS
 ,
356
,
1509
Curran
S. J.
Whiting
M.
Murphy
M. T.
Webb
J. K.
Longmore
S. N.
Pihlström
Y. M.
Athreya
R.
Blake
C.
,
2006
,
MNRAS
 ,
371
,
431
Curran
S. J.
Tzanavaris
P.
Murphy
M. T.
Webb
J. K.
Pihlström
Y. M.
,
2007a
,
MNRAS
 ,
381
,
L6
Curran
S. J.
Darling
J. K.
Bolatto
A. D.
Whiting
M. T.
Bignell
C.
Webb
J. K.
,
2007b
,
MNRAS
 ,
382
,
L11
Curran
S. J.
Tzanavaris
P.
Pihlström
Y. M.
Webb
J. K.
,
2007c
,
MNRAS
 ,
382
,
1331
Dallacasa
D.
Falomo
R.
Stanghellini
C.
,
2002
,
A&A
 ,
382
,
53
Darling
J.
Giovanelli
R.
,
2006
,
AJ
 ,
132
,
2596
Darling
J.
Giovanelli
R.
Haynes
M. P.
Bower
G. C.
Bolatto
A. D.
,
2004
,
ApJ
 ,
613
,
L101
De Koff
S.
Baum
S. A.
Sparks
W. B.
Biretta
J.
Golombek
D.
Macchetto
F.
McCarthy
P.
Miley
G. K.
,
1996
,
ApJS
 ,
107
,
621
De Vaucouleurs
A.
Longo
G.
,
1988
,
Catalogue of Visual and Infrared Photometry of Galaxies from 0.5 Micrometer to 10 Micrometer (1961–1985)
 .
University of Texas Monographs in Astronomy
, University of Texas, Austin
De Vries
W. H.
et al.,
1997
,
ApJS
 ,
110
,
191
De Vries
W. H.
O'Dea
C. P.
Perlman
E.
Baum
S. A.
Lehnert
M. D.
Stocke
J.
Rector
T.
Elston
R.
,
1998
,
ApJ
 ,
503
,
138
De Vries
W. H.
O'Dea
C. P.
Barthel
P. D.
Thompson
D. J.
,
2000
,
A&AS
 ,
143
,
181
Di Serego Alighieri
S.
et al.,
2007
,
A&A
 ,
474
,
851
Drinkwater
M. J.
et al.,
1997
,
MNRAS
 ,
284
,
85
Dunlop
J. S.
Peacock
J. A.
Savage
A.
Lilly
S. J.
Heasley
J. N.
Simon
A. J. B.
,
1989
,
MNRAS
 ,
238
,
1171
Eales
S. A.
,
1985
,
MNRAS
 ,
213
,
899
Eales
S.
Rawlings
S.
Puxley
P.
Rocca-Volmerange
B.
Kuntz
K.
,
1993
,
Nat
 ,
363
,
140
Eisenhardt
P.
Dickinson
M.
,
1992
,
ApJ
 ,
399
,
L47
Ellison
S. L.
Yan
L.
Hook
I. M.
Pettini
M.
Wall
J. V.
Shaver
P.
,
2001
,
A&A
 ,
379
,
393
Ellison
S. L.
Yan
L.
Hook
I. M.
Pettini
M.
Wall
J. V.
Shaver
P.
,
2002
,
A&A
 ,
383
,
91
Ellison
S. L.
Hall
P. B.
Lira
P.
,
2005
,
AJ
 ,
130
,
1345
Eracleous
M.
Halpern
J. P.
,
1994
,
ApJS
 ,
90
,
1
Falcke
H.
Wilson
A. S.
Simpson
C.
,
1998
,
ApJ
 ,
502
,
199
Field
G. B.
,
1959
,
ApJ
 ,
129
,
536
Francis
P. J.
Whiting
M. T.
Webster
R. L.
,
2000
,
Publ. Astron. Soc. Aust.
 ,
17
,
56
Fukugita
M.
Ichikawa
T.
Gunn
J. E.
Doi
M.
Shimasaku
K.
Schneider
D. P.
,
1996
,
AJ
 ,
111
,
1748
Gao
Y.
Solomon
P. M.
,
2004
,
ApJ
 ,
606
,
271
Gelderman
R.
Whittle
M.
,
1994
,
ApJS
 ,
91
,
491
Glikman
E.
Gregg
M. D.
Lacy
M.
Helfand
D. J.
Becker
R. H.
White
R. L.
,
2004
,
ApJ
 ,
607
,
60
Greenhill
L. J.
et al.,
2003
,
ApJ
 ,
590
,
162
Gregg
M. D.
Lacy
M.
White
R. L.
Glikman
E.
Helfand
D.
Becker
R. H.
Brotherton
M. S.
,
2002
,
ApJ
 ,
564
,
133
Gupta
N.
Saikia
D. J.
,
2006a
,
MNRAS
 ,
370
,
L80
Gupta
N.
Saikia
D. J.
,
2006b
,
MNRAS
 ,
370
,
738
Gupta
N.
Salter
C. J.
Saikia
D. J.
Ghosh
T.
Jeyakumar
S.
,
2006
,
MNRAS
 ,
373
,
972
Gupta
N.
Srianand
R.
Petitjean
P.
Khare
P.
Saikia
D. J.
York
D. G.
,
2007
,
ApJ
 ,
654
,
L111
Guyon
O.
Sanders
D. B.
Stockton
A.
,
2006
,
ApJS
 ,
166
,
89
Hambly
N.
et al.,
2001
,
MNRAS
 ,
326
,
1279
Heckman
T. M.
Lebofsky
M. J.
Rieke
G. H.
Van Breugel
W.
,
1983
,
ApJ
 ,
272
,
400
Henkel
C.
Güsten
R.
Baan
W. A.
,
1987
,
A&A
 ,
185
,
14
Henstock
D. R.
Browne
I. W. A.
Wilkinson
P. N.
McMahon
R. G.
,
1997
,
MNRAS
 ,
290
,
380
Hewitt
A.
Burbidge
G.
,
1989
,
ApJS
 ,
69
,
1
Hirst
P.
Jackson
N.
Rawlings
S.
,
2003
,
MNRAS
 ,
346
,
1009
Hook
I. M.
Shaver
P. A.
Jackson
C. A.
Wall
J. V.
Kellermann
K. I.
,
2003
,
A&A
 ,
399
,
469
Hunstead
R. W.
Murdoch
H. S.
Shobbrook
R. R.
,
1978
,
MNRAS
 ,
185
,
149
Hyland
A. R.
Allen
D. A.
,
1982
,
MNRAS
 ,
199
,
943
Ishwara-Chandra
C. H.
Dwarakanath
K. S.
Anantharamaiah
K. R.
,
2003
,
JA&A
 ,
24
,
37
Jackson
N.
Browne
I. W. A.
,
1991
,
MNRAS
 ,
250
,
414
Jackson
C. A.
Wall
J. V.
Shaver
P. A.
Kellermann
K. I.
Hook
I. M.
Hawkins
M. R. S.
,
2002
,
A&A
 ,
386
,
97
Jaffe
W.
McNamara
B. R.
,
1994
,
ApJ
 ,
434
,
110
Kanekar
N.
Briggs
F. H.
,
2003
,
A&A
 ,
412
,
L29
Kanekar
N.
Chengalur
J. N.
,
2002
,
A&A
 ,
381
,
L73
Kanekar
N.
Chengalur
J. N.
,
2003
,
A&A
 ,
399
,
857
Kanekar
N.
Chengalur
J. N.
De Bruyn
A. G.
Narasimha
D.
,
2003
,
MNRAS
 ,
345
,
L7
Kanekar
N.
Subrahmanyan
R.
Ellison
S. L.
Lane
W.
Chengalur
J. N.
,
2006
,
MNRAS
 ,
370
,
L46
Kanekar
N.
Chengalur
J. N.
Lane
W. M.
,
2007
,
MNRAS
 ,
375
,
1528
Keel
W. C.
,
1980
,
AJ
 ,
85
,
198
Kim
D.-C.
Veilleux
S.
Sanders
D. B.
,
1998
,
ApJ
 ,
508
,
627
Kristian
J.
Sandage
A.
Katem
B.
,
1974
,
ApJ
 ,
191
,
43
Lanzetta
K. M.
Wolfe
A. M.
Turnshek
D. A.
Lu
L.
McMahon
R. G.
Hazard
C.
,
1991
,
ApJS
 ,
77
,
1
Lawrence
C. R.
Elston
R.
Januzzi
B. T.
Turner
E. L.
,
1995
,
AJ
 ,
110
,
2570
Lawrence
C. R.
Zucker
J. R.
Readhead
A. C. S.
Unwin
S. C.
Pearson
T. J.
Xu
W.
,
1996
,
ApJS
 ,
107
,
541
Lilly
S.
Longair
M.
,
1984
,
MNRAS
 ,
211
,
833
Lipovetsky
V. A.
Neizvestny
S. I.
Neizvestnaya
O. M.
,
1988
,
Soobshch. Spets. Astrofiz. Obs.
 ,
55
,
5
Lovell
J. E. J.
et al.,
1996
,
ApJ
 ,
472
,
L5
Maiolino
R.
Rieke
G. H.
,
1995
,
ApJ
 ,
454
,
95
Marchã
M. J. M.
Browne
I. W. A.
Impey
C. D.
Smith
P. S.
,
1996
,
MNRAS
 ,
281
,
425
Marchã
M. J.
Caccianiga
A.
Browne
I. W. A.
Jackson
N.
,
2001
,
MNRAS
 ,
326
,
1455
Marziani
P.
Sulentic
J. W.
Dultzin-Hacyan
D.
Calvani
M.
Moles
M.
,
1996
,
ApJS
 ,
104
,
37
Marziani
P.
Sulentic
J. W.
Zamanov
R.
Calvani
M.
Dultzin-Hacyan
D.
Bachev
R.
Zwitter
T.
,
2003
,
ApJS
 ,
145
,
199
Mirabel
I. F.
,
1989
,
ApJ
 ,
340
,
L13
Møller
P.
Warren
S. J.
Fynbo
J. U.
,
1998
,
A&A
 ,
330
,
19
Moore
C. B.
Carilli
C. L.
Menten
K. M.
,
1999
,
ApJ
 ,
510
,
L87
Morganti
R.
Oosterloo
T. A.
Tadhunter
C. N.
Van Moorsel
G.
Killeen
N.
Wills
K. A.
,
2001
,
MNRAS
 ,
323
,
331
Morganti
R.
et al.,
2006
,
MNRAS
 ,
371
,
157
O'Dea
C. P.
Baum
S. A.
Morris
G. B.
,
1990
,
A&AS
 ,
82
,
261
Orienti
M.
Morganti
R.
Dallacasa
D.
,
2006
,
A&A
 ,
457
,
531
Orienti
M.
et al.,
2007
,
New Astron. Rev.
 ,
51
,
8
Overzier
R. A.
et al.,
2006
,
ApJ
 ,
637
,
58
Peck
A. B.
Taylor
G. B.
Conway
J. E.
,
1999
,
ApJ
 ,
521
,
103
Peck
A. B.
Taylor
G. B.
Fassnacht
C. D.
Readhead
A. C. S.
Vermeulen
R. C.
,
2000
,
ApJ
 ,
534
,
104
Pedlar
A.
Unger
S. W.
Dyson
J. E.
,
1985
,
MNRAS
 ,
214
,
463
Pedlar
A.
Fernandez
B.
Hamilton
N. G.
Redman
M. P.
Dewdney
P. E.
,
1998
,
MNRAS
 ,
300
,
1071
Peng
C. Y.
Impey
C. D.
Rix
H.-W.
Falco
E. E.
Keeton
C. R.
Kochanek
C. S.
Lehár
J.
McLeod
B. A.
,
2006
,
New Astron. Rev.
 ,
50
,
689
Péroux
C.
Storrie-Lombardi
L. J.
McMahon
R. G.
Irwin
M.
Hook
I. M.
,
2001
,
AJ
 ,
121
,
1799
Pickett
H. M.
Poynter
R. L.
Cohen
E. A.
Delitsky
M. L.
Pearson
J. C.
Müller
H. S. P.
,
1998
,
J. Quant. Spectrosc. Radiat. Transfer
 ,
60
,
883
Pihlström
Y. M.
Vermeulen
R. C.
Taylor
G. B.
Conway
J. E.
,
1999
,
ApJ
 ,
525
,
L13
Pihlström
Y. M.
Conway
J. E.
Vermeulen
R. C.
,
2003
,
A&A
 ,
404
,
871
Pihlström
Y. M.
Baan
W. A.
Darling
J.
Klöckner
H.-R.
,
2005
,
ApJ
 ,
618
,
705
Pogge
R. W.
,
1989
,
ApJ
 ,
345
,
730
Prestage
R. M.
Peacock
J. A.
,
1983
,
MNRAS
 ,
204
,
355
Prochaska
J. X.
Hennawi
J. F.
Herbert-Fort
S.
,
2008
,
ApJ
 ,
675
,
1002
Purcell
E. M.
Field
G. B.
,
1956
,
ApJ
 ,
124
,
542
Rao
S.
Turnshek
D.
Nestor
D. B.
,
2006
,
ApJ
 ,
636
,
610
Roche
N.
Eales
S. A.
,
2000
,
MNRAS
 ,
317
,
120
Rohlfs
K.
Wilson
T. L.
,
2000
,
Tools of Radio Astronomy.
 
Springer-Verlag
, Berlin
Sandage
A.
,
1965
,
ApJ
 ,
141
,
1560
Sandage
A.
,
1973
,
ApJ
 ,
183
,
711
Sandage
A.
Véron
P.
Wyndham
J. D.
,
1965
,
ApJ
 ,
142
,
1307
Schmitt
H. R.
Kinney
A. L.
,
1996
,
ApJ
 ,
463
,
498
Schneider
D. P.
et al.,
2005
,
AJ
 ,
130
,
367
Schulz
H.
,
1988
,
A&A
 ,
203
,
233
Simpson
C.
Rawlings
S.
,
2000
,
MNRAS
 ,
317
,
1023
Skrutskie
M. F.
et al.,
2006
,
AJ
 ,
131
,
1163
Smith
E. P.
Heckman
T. M.
,
1989
,
ApJS
 ,
69
,
365
Snellen
I. A. G.
Schilizzi
R. T.
Bremer
M. N.
Miley
G. K.
De Bruyn
A. G.
Röttgering
H. J. A.
,
1999
,
MNRAS
 ,
307
,
149
Spinrad
H.
Marr
J.
Aguilar
L.
Djorgovski
S.
,
1985
,
PASP
 ,
97
,
932
Srianand
R.
Khare
P.
,
1993
,
ApJ
 ,
413
,
486
Stanghellini
C.
O'Dea
C. P.
Baum
S. A.
Laurikainen
E.
,
1993
,
ApJS
 ,
88
,
1
Stickel
M.
Kühr
H.
,
1996
,
A&AS
 ,
115
,
11
Stickel
M.
Kühr
H.
,
1994
,
A&AS
 ,
105
,
67
Stickel
M.
Fried
J. W.
Kühr
H.
,
1993
,
A&AS
 ,
98
,
393
Stickel
M.
Rieke
G. H.
Kühr
H.
Rieke
M. J.
,
1996
,
ApJ
 ,
468
,
556
Stocke
J. T.
Wurtz
R.
Wang
Q.
Elston
R.
Jannuzi
B. T.
,
1992
,
ApJ
 ,
400
,
L17
Stockton
A.
Ridgway
S. E.
,
2001
,
ApJ
 ,
554
,
1012
Subrahmanyan
R.
Narasimha
D.
Pramesh-Rao
A.
Swarup
G.
,
1990
,
MNRAS
 ,
246, 263
Tadhunter
C. N.
Morganti
R.
Di Serego-Alighieri
S.
Fosbury
R. A. E.
Danziger
I. J.
,
1993
,
MNRAS
 ,
263
,
999
Tadhunter
C.
Dickson
R.
Morganti
R.
Robinson
T. G.
Wills
K.
Villar-Martin
M.
Hughes
M.
,
2002
,
MNRAS
 ,
330
,
977
Taylor
G. L.
Dunlop
J. S.
Hughes
D. H.
Robson
E. I.
,
1996
,
MNRAS
 ,
283
,
930
Unger
S. W.
Pedlar
A.
Axon
D. J.
Whittle
M.
Meurs
E. J. A.
Ward
M. J.
,
1987
,
MNRAS
 ,
228
,
671
Urry
C. M.
Padovani
P.
,
1995
,
PASP
 ,
107
,
803
Uson
J. M.
Bagri
D. S.
Cornwell
T. J.
,
1991
,
Phys. Rev. Lett.
 ,
67
,
3328
Van Gorkom
J. H.
Knapp
G. R.
Ekers
R. D.
Ekers
D. D.
Laing
R. A.
Polk
K. S.
,
1989
,
AJ
 ,
97
,
708
Vermeulen
R. C.
Taylor
G. B.
,
1995
,
AJ
 ,
109
,
1983
Vermeulen
R. C.
Taylor
G. B.
Readhead
A. C. S.
Browne
I. W. A.
,
1996
,
AJ
 ,
111
,
1013
Vermeulen
R. C.
et al.,
2003
,
A&A
 ,
404
,
861
Weymann
R. J.
Carswell
R. F.
Smith
M. G.
,
1981
,
ARA&A
 ,
19
,
41
White
R. L.
et al.,
2000
,
ApJS
 ,
126
,
133
Wiklind
T.
Combes
F.
,
1994
,
A&A
 ,
286
,
L9
Wiklind
T.
Combes
F.
,
1995
,
A&A
 ,
299
,
382
Wiklind
T.
Combes
F.
,
1996a
,
Nat
 ,
379
,
139
Wiklind
T.
Combes
F.
,
1996b
,
A&A
 ,
315
,
86
Wiklind
T.
Combes
F.
,
1998
,
ApJ
 ,
500
,
129
Wilkes
B. J.
Wright
A. E.
Jauncey
D. L.
Peterson
B. A.
,
1983
,
Proc. Astron. Soc. Aust.
 ,
5
,
2
Wills
D.
Lynds
R.
,
1978
,
ApJS
 ,
36
,
317
Winn
J. N.
et al.,
2002
,
AJ
 ,
123
,
10
Wolfe
A. M.
Burbidge
G. R.
,
1975
,
ApJ
 ,
200
,
548
York
D. G.
et al.,
2000
,
AJ
 ,
120
,
1579
York
B. A.
Kanekar
N.
Ellison
S. L.
Pettini
M.
,
2007
,
MNRAS
 ,
382
,
L53
Zheng
Z.
Miralda-Escudé
J.
,
2002
,
ApJ
 ,
568
,
L71
Zickgraf
F.-J.
Voges
W.
Krautter
J.
Thiering
I.
Appenzeller
I.
Mujica
R.
Serrano
A.
,
1997
,
A&A
 ,
323
,
L21

Appendices

APPENDIX A

In this section we list all of the sources used in the analysis, which all been obtained from the literature cited in Table 1, complete with the compiled photometry and classification.

APPENDIX B

The UV luminosities (Section 3.1.2) are calculated at a standard rest-frame wavelength for each source, in this case λU= 1216 Å. As before (Section 3.1.1), the luminosity is given by the expression forumla, although rather than using a K-correction term, to correct a given observed passband to the appropriate rest-frame wavelength, we instead interpolate between or extrapolate from the observed photometry to obtain the flux at the observed wavelength λU(z+ 1).

The extrapolation or interpolation is done by fitting a power law (i.e. a linear fit in log λ– log F space). Thus, for two observations at bands 1 and 2, with fluxes F1 and F2, the UV flux is  

formula

Although the observational data we have obtained from the literature is a heterogeneous mix of photometry, we try as much as possible to maintain consistency by using the same bands. The combination that covers the largest number of sources is B and R bands. Because of the relatively low redshift of most of the sources, extrapolation from B band is required. If B is not available, V is used instead, and if one of V or R is not available, K band was used. In one case, J0924−2201, only R was used since the high redshift meant that λU(1 +z) fell into this band.