astro-ph0211637/ms.tex
1: \documentclass{aa}
2: %\documentclass[referee]{aa}
3: \usepackage{graphics,epsfig}
4: 
5: \newcommand{\kms}{km~s$^{-1}$}
6: \newcommand{\dV}{\mathrm{d} V}
7: \newcommand{\cm}{cm$^{-2}$}
8: \newcommand{\Dla}{ Damped Lyman-$\alpha$ }
9: \newcommand{\dla}{ damped Lyman-$\alpha$ }
10: \newcommand{\NHI}{N_{\rm HI}}
11: \newcommand{\noi}{\noindent}
12: \newcommand{\lb}{\left(}
13: \newcommand{\DV}{\Delta V}
14: \newcommand{\rb}{\right)}
15: \newcommand{\acmsqr}{atoms~cm$^{-2}$}
16: \newcommand{\lya}{Lyman-$\alpha\ $}
17: 
18: \begin{document}
19: \title{A deep search for 21cm absorption in high redshift damped Lyman-$\alpha$ systems.}
20: \titlerunning{21cm absorption in high $z$ DLAs}
21: \author{Nissim Kanekar \inst{1}\thanks{nissim@ncra.tifr.res.in},
22: Jayaram N Chengalur\inst{2}\thanks{chengalu@ncra.tifr.res.in}}
23: \authorrunning{Kanekar \& Chengalur }
24: \institute{Kapteyn Institute, University of Groningen, Post Bag 800, 9700 AV Groningen
25: \and National Centre for Radio Astrophysics, Post Bag 3, Ganeshkhind, Pune 411 007 }
26: \date{Received mmddyy/ accepted mmddyy}
27: \offprints{Nissim Kanekar}
28: \abstract{We present deep GMRT 21cm absorption spectra of 10 damped Lyman-$\alpha$ 
29: systems (DLAs), of which 8 are at redshifts $z \ga 1.3$. HI absorption was detected 
30: in only one DLA, the $z = 0.5318$ absorber toward PKS~1629+12. This absorber has been
31: identified with a luminous spiral galaxy; the spin temperature limit ($T_{\rm s} \le 310$~K)
32: derived from our observations continues the trend of DLAs associated with bright spirals
33: having low spin temperatures. In seven of the remaining 9 systems, the observations place 
34: strong lower limits on the spin temperature of the HI gas. \\
35: 	We combine this sample with data taken from the literature to study the
36: properties of all known DLAs with 21cm absorption studies. The sample of DLAs which have 
37: been searched for 21cm absorption now consists of 31 systems, with $T_{\rm s}$ estimates 
38: available in 24 cases; of these, 16 are at $z < 2$ and 8 at $z > 2$, with 11 (all at $z < 1$) 
39: having optical identifications. For the latter 11 DLAs, we find that all of the 
40: low $T_{\rm s}$ DLAs have been identified with large, luminous galaxies, while all the DLAs 
41: with high spin temperature ($T_{\rm s} \ga 1000$~K) have been identified either with LSBs 
42: or dwarfs. Further, we find no correlation between impact parameter and spin 
43: temperature; it is thus unlikely  that the high measured $T_{\rm s}$ values for DLAs 
44: arise from lines of sight passing through the outskirts of large disk galaxies. 
45: Instead, the spin temperature of DLAs appears to correlate with the host galaxy type.\\
46: 	The trend (noted earlier by Chengalur \& Kanekar 2000) that low $z$ DLAs 
47: exhibit both high and low $T_{\rm s}$ values while high redshift ($z \ga 3$) DLAs only show high 
48: spin temperatures is present in this expanded data set. Based on this difference in
49: spin temperatures, the Gehan test rules out the hypothesis that DLAs at $z > 2$ and 
50: DLAs at $z < 2$ are drawn  from the same parent population at $\sim 99$\% confidence 
51: level. \\
52: 	Finally, we use the new GMRT spectra along with 2 spectra from the literature
53: to estimate upper limits on the fraction of cold HI, $f_{\rm CNM}$, in DLAs at $z \ga 3$.
54: For local spirals, $f_{\rm CNM} \sim 0.5$; in contrast, we find that $f_{\rm CNM} < 0.3$ 
55: in all 7 high $z$ absorbers, and $f_{\rm CNM} < 0.1$ in 5 of the 7 cases. 
56: \keywords{galaxies: evolution: --
57:           galaxies: formation: --
58:           galaxies: ISM --
59:           cosmology: observations --
60:           radio lines: galaxies}
61: }
62: \maketitle
63: 
64: \section{Introduction}
65: 	
66: 	While molecular gas can be detected in emission at high redshifts 
67: (albeit only in a handful of somewhat unusual systems), it is 
68: as yet a severe challenge to detect neutral atomic gas in emission, 
69: even at redshifts as low as $z \sim 0.2$. As a result, absorption spectra 
70: against bright background sources provide the sole source of information on 
71: neutral atomic gas in the high redshift Universe. An important advantage of using 
72: absorption spectra as a probe of systems lying along the lines of sight to 
73: distant QSOs is that such studies tend to be unbiased, at least to first order --
74: sources which give rise to absorption are presumably typical systems at 
75: that redshift. In contrast, emission studies of a flux-limited, high redshift
76: sample are usually dominated by the brightest (often atypical) sources.
77: 
78: 	The large absorption cross-section of the \lya transition and the
79: dominance of hydrogen in baryonic material makes the \lya line one of the 
80: easiest to detect in QSO spectra. In fact, lines arising from neutral gas with 
81: column densities as low as $10^{13}$~\acmsqr~can be routinely detected with
82: present day telescopes. However, although these lines (the \lya ``forest'') 
83: are by far the most numerous, they contain only a small fraction of the total 
84: neutral gas at high redshifts; most of the neutral gas is, in fact, contained 
85: in relatively rare, high column density ($\NHI \ga 10^{20}$~\acmsqr) systems, 
86: the so-called \dla absorbers (DLAs). As the major repository of neutral gas at 
87: high redshift, these systems are natural candidates for the precursors of today's
88: galaxies. However, despite their importance in the context of galaxy evolution, 
89: the nature of high redshift DLAs is presently quite controversial, with models 
90: ranging from large rapidly rotating disks (\cite{wolfe86}; \cite{pw97}) to merging 
91: sub-galactic blobs (\cite{haehnelt98}) to outflows from dwarf galaxies 
92: (\cite{schaye01}) present in the recent literature.\\
93: 	A successful model for DLAs needs to not only reproduce the 
94: observed $\Omega_{\rm g}(z)$ and column density distribution but also to 
95: describe the observed physical conditions in the absorbing gas. The latter
96: is important for our understanding of DLAs, quite independent of their precise 
97: size and structure. Conversely, knowledge of these conditions is useful in 
98: modelling the further temporal evolution of damped systems. HI 21cm observations (possible 
99: only for those DLAs for which the background quasar is also radio loud) 
100: are particularly useful because they allow one to estimate the average 
101: spin temperature $T_{\rm s}$ of the absorbing gas. DLA spin temperatures have been, 
102: in general, found to be significantly higher than those typically seen 
103: in lines of sight through the Galactic disk, or in those through the
104: disks of nearby galaxies (see e.g. \cite{chengalur2000}, \cite{kanekar2001c} 
105: and references therein). The only exceptions to the above are DLAs which 
106: have been identified as associated with low redshift ($z~<~0.6$) spiral galaxies 
107: -- these systems all have low spin temperatures, comparable to those seen 
108: in the Milky Way and nearby spirals. Unfortunately, the number of systems 
109: for  which deep 21cm observations are available is relatively small; further, 
110: while the present data (\cite{chengalur2000}) suggest that high $z$ DLAs 
111: are different from low $z$ ones, the observations are currently biased toward 
112: systems at $z< 1$, with not too many systems of the sample at high redshift. 
113: Are high redshift DLAs indeed systematically different from low redshift ones ? 
114: Also, how significant is the fact that only DLAs associated with spiral galaxies 
115: have low spin temperatures ? To find answers to these and similar questions,
116: it is crucial to increase the number of systems with 21cm observations, 
117: particularly at high redshifts. We present, in this paper, deep Giant 
118: Meterwave Radio Telescope (GMRT) observations of 10 DLAs, 8 of which are 
119: at $z>1$. Two of these systems, the $z = 2.9084$ DLA towards TXS~2342+342 
120: and the $z = 3.0619$ absorber towards PKS~0336$-$014 have been earlier observed 
121: by Carilli et al. (1996); the present observations are far more sensitive
122: than the earlier ones. The observations themselves are summarised in 
123: Sect.~\ref{sec:obs}, while the results are presented in Sect.~\ref{sec:res} and their 
124: implications for physical conditions in DLAs discussed in Sect.~\ref{sec:dis}.
125: 
126: \section{Observations and Data Analysis}
127: \label{sec:obs}
128: \subsection{The GMRT observations}
129: \label{ssec:gmrt}
130: 
131: The GMRT has five currently operational frequency bands, at 150~MHz, 233~MHz, 327~MHz, 
132: 610~MHz and 1420~MHz.  The present observations utilised the 327, 610 and 1420~MHz bands, 
133: which cover redshift ranges $z  \sim 2.9 - 3.6$, $ z \sim 1.15  - 1.5$ and $ z \sim 
134: 0 - 0.6$ respectively, with reasonable sensitivity (i.e. better than half the sensitivity 
135: at the band centre).  The observations were carried out between January and November 2001, 
136: using the 30-station FX correlator as the backend; this gives a fixed number of 128 
137: channels over a bandwidth which can be varied between 64~kHz and 16~MHz. All our 
138: observations (except for those of PKS~2128$-$123 (bandwidth = 2~MHz) and PKS~1629+12, 
139: which are described below) used a bandwidth of 1~MHz, sub-divided into 128 channels; 
140: this gave a channel resolution of 7.8~kHz. The number of available antennas varied 
141: between 18 and 26, due to various debugging and maintenance activities. One (or more) 
142: of the standard calibrators 3C48, 3C147 or 3C286 was used in all cases to calibrate 
143: the absolute flux scale and the system bandpass; these sources were observed at least 
144: every two hours in all cases, to ensure a good bandpass calibration.  A nearby 
145: standard phase calibrator (see Cols.~4 and 5 of Table~1) was observed every forty 
146: minutes in all cases to determine the instrumental phase, except in the case of 
147: PKS~2128$-$123 which is itself a phase calibrator for the GMRT. Our experience with 
148: the GMRT indicates that the flux calibration is reliable to $\sim 15$\%, in this
149: observing mode. Observational details 
150: are summarised in Table~1, where the sources are arranged in order of increasing 
151: redshift; note that the values listed here for spectral resolution and RMS 
152: noise are before any smoothing. 
153: 
154: The only source towards which we detected 21cm absorption, viz. PKS~1629+12, was initially 
155: observed with a 1 MHz bandwidth on the 12th and 22nd of October 2001, 
156: with 20 antennas and a total on-source time of $\sim 6$~hours. These observations resulted in 
157: the detection of a narrow 21cm absorption feature; the source was hence re-observed on the 17th
158: and 18th of November, 2001 with a 0.25~MHz bandwidth (yielding a resolution of 1.95~kHz,
159: i.e. 0.6~\kms), to try to resolve out the absorption line. The total on-source time
160: of these latter observations was 13~hours, with 25 antennas. In all cases, 3C286 was
161: used for absolute flux and system bandpass calibration while the compact source
162: 1640+123 was used for phase calibration. 
163: 
164: 
165: 
166: \begin{table*}
167: \begin{center}
168: \label{tab:obs}
169: \caption{Observing details}
170: \begin{tabular}{|c|c|c|c|c|c|c|c|c|c|} \hline
171: &&&&&&&&&\\
172:  QSO &        $z_{\rm abs}$ & $\nu_{\rm obs}$ & Phase & Phase Cal. & No. of & Time & Source & Channel & RMS    \\
173:      &                  & MHz   & Cal.  & Flux &antennas& Hrs. & Flux & resolution& mJy         \\
174:      &                  &       &       & Jy   & 	&      & Jy   & \kms & \\
175: &&&&&&&&&\\
176: \hline
177: &&&&&&&&&\\
178: PKS~2128$-$123 & 0.4298 & 993.43 &$^a$       &$^a$ & 20 & 8   & 1.90    & 4.7   & 1.4      \\
179: &&&&&&&&&\\
180: PKS~1629+12    & 0.5318 & 927.28 &1640+123  & 2.5 & 25 & 13  & 2.35    & 0.6   & 2.8      \\
181: &&&&&&&&&\\
182: PKS~0215+015   & 1.3439 & 605.72 & 0204+152 & 5.0 & 26 & 5   & 0.92    & 3.9   & 2.2     \\
183: &&&&&&&&&\\
184: QSO~0957+561A  & 1.3911 & 594.04 & 0834+555 & 7.4 & 26 & 6.5 & 0.59    & 3.9   & 2.4     \\
185: &&&&&&&&&\\
186: PKS~1354+258   & 1.4205 & 586.82 & 3C286    & $^b$& 22 & 5.5 & 0.30$^c$& 4.0   & 2.8      \\
187: &&&&&&&&&\\
188: TXS~2342+342   & 2.9084 & 363.42 & 3C19     & 8.3 & 23 & 8   & 0.31    & 6.4   & 2.8      \\
189: &&&&&&&&&\\
190: PKS~1354$-$107 & 2.966  & 358.15 & 3C283    & 21  & 22 & 9.5 & 0.12    & 6.5   & 2.0      \\
191: &&&&&&&&&\\
192: PKS~0537$-$286 & 2.974  & 357.43 &0521$-$207& 7.2 & 25 & 5   & 1.05    & 6.5   & 2.4     \\
193: &&&&&&&&&\\
194: PKS~0336$-$014 & 3.0619 & 349.69 & 0323+055 & 7.4 & 26 &  5  & 0.94    & 6.7   & 2.1     \\
195: &&&&&&&&&\\
196: PKS~0335$-$122 & 3.178  & 339.97 &0409$-$179& 6.1 & 18 & 5.5 & 0.68    & 6.9   & 1.9     \\
197: &&&&&&&&&\\
198: \hline
199: \end{tabular}
200: \end{center}
201: \vskip 0.1 in
202: ${}$~$^a$~No phase calibrator was used, as PKS~2128$-$123 is itself a phase calibrator 
203: for the GMRT. \\
204: ${}$~$^b$~3C286 was also used as the flux calibrator; we hence do not have an independent 
205: estimate of its flux.\\
206: ${}$~$^c$~PKS~1354+258 is resolved by the GMRT synthesised beam; this is the peak flux.\\
207: \vskip 0.1 in
208: 
209: \end{table*}
210: 
211: \subsection{Data Analysis}
212: \label{ssec:anal}
213: 
214: The data were analysed in classic AIPS using standard procedures. After the initial flagging 
215: and gain and bandpass calibration, continuum images were made for all sources and 
216: then used to self-calibrate the U-V data. This was carried out in an iterative manner 
217: until the quality of the image was found to not improve on further self-calibration. 
218: In general, a few rounds of phase self-calibration were carried out on all fields,
219: followed by one or two rounds of amplitude self-calibration. The continuum emission 
220: was then subtracted from the multi-channel U-V data set, using the AIPS task UVSUB; 
221: any residual continuum was subtracted by fitting a linear baseline to the U-V 
222: visibilities, using the task UVLIN. The continuum subtracted data were then mapped 
223: in all channels and a spectrum extracted at the quasar location from the resulting
224: three dimensional data cube. In the case of PKS~1629+12, which showed 21cm absorption, 
225: spectra were also extracted from other locations in the cube to ensure that the 
226: data were not corrupted by RFI. In general, all the observing frequencies were 
227: found to be reasonably free of RFI, except for sporadic interference on individual 
228: antennas which was edited out. In the case of multi-epoch observations 
229: of a single source, the spectra were shifted to the heliocentric frame outside AIPS 
230: and then averaged together. 
231: 
232: \section{Results}
233: \label{sec:res}
234: 
235: \begin{figure*}[t!]
236: \begin{center}
237: \epsfig{file=montage.eps,height=7truein}
238: \end{center}
239: \caption{ GMRT HI absorption spectra towards the ten sources of our sample, 
240: arranged in order of increasing redshift. 
241: {\bf [A]}~$z = 0.4298$~DLA towards PKS~2128$-$123.
242: {\bf [B]}~$z = 0.5318$~DLA towards PKS~1629+12
243: {\bf [C]}~$z = 1.3449$~DLA towards PKS~0215+015. 
244: {\bf [D]}~$z = 1.3911$~DLA towards QSO~0957+561A. 
245: {\bf [E]}~$z = 1.4205$~ DLA towards PKS~1354+258. 
246: {\bf [F]}~$z = 2.9084$~DLA towards TXS~2342+342. 
247: {\bf [G]}~$z = 2.966$~DLA towards PKS~1354$-$107. 
248: {\bf [H]}~$z = 2.974$~DLA towards PKS~0537$-$286
249: {\bf [I]}~$z = 3.0619$~DLA towards PKS~0336$-$014. 
250: {\bf [J]}~$z = 3.178$~DLA towards PKS~0335$-$122.}
251: \label{fig:spc}
252: \end{figure*}
253: 
254: The final absorption spectra towards the ten sources of our sample are shown 
255: in Fig.~\ref{fig:spc}, arranged in order of increasing redshift; the measured 
256: source fluxes, channel resolutions (in \kms) and RMS noise values are listed in 
257: Cols.~8, 9 and 10 of Table~1. 
258: 
259: Our only detection of HI absorption is in the $z = 0.5318$ DLA towards PKS~1629+12. 
260: Fig.~\ref{fig:1629} shows a more detailed version of the GMRT 0.25 MHz HI spectrum towards 
261: this source; the spectrum has a resolution of $\sim 0.6$~\kms~and an RMS noise of 
262: 2.8~mJy. Highly asymmetric HI absorption can be clearly seen, extending 
263: over a velocity width of $\sim 40$~\kms. The peak absorption occurs at a heliocentric 
264: frequency of 927.287~MHz, i.e. at a heliocentric redshift of $ 0.531787 \pm 0.000002$. 
265: This lies in between (and in reasonable agreement with) the redshifts of the 
266: damped Lyman-$\alpha$ line ($z_{\rm DLA} = 0.532$; \cite{nestor2001}) and the MgI 
267: absorption ($z_{\rm MgI} = 0.5316$; (\cite{aldcroft94}; \cite{barthel90}).
268: (Note that Aldcroft et al. (1994) quote $z = 0.5313$ for the strong MgII and FeII 
269: absorption lines.) The flux of PKS~1629+12 was measured to be 2.35~Jy; the peak optical 
270: depth is thus $\tau_{\rm max} = 0.039$. The equivalent width of the profile is 
271: $\int \tau \dV = 0.494 \pm 0.001$~\kms, obtained by integrating the observed absorption profile.
272: 
273: \section{The spin temperature}
274: \label{sec:Ts}
275: 
276: \begin{table*}
277: \begin{center}
278: \label{tab:ts_obs}
279: \caption{The spin temperature for the 10 DLAs observed here}
280: \begin{tabular}{|c|c|c|c|c|c|c|c|c|} \hline
281: &&&&&&&&\\
282:  QSO &        $z_{\rm abs}$ & $\NHI \times 10^{20}$ & $\Delta V$~$^a$  & RMS$^c$ & $\tau_{\rm max}^d$ 
283:  &  $(1/f)\int \tau {\mathrm d}V^e$ & $ T_{\rm s}$ & Refs.$^f$ \\
284:      &                  & \cm  &   \kms     & mJy &        & \kms    & K     &  \\
285: &&&&&&&&\\
286: \hline
287: &&&&&&&&\\
288: PKS~2128$-$123   & 0.4298 & $0.25\pm0.06$ & 7    & 1.2 & $< 0.0019$ & $< 0.014$ & $ > 980 $ &  1 \\ 
289: &&&&&&&&\\
290: PKS~1629+12      & 0.5318 & 2.8  & 0.6  & 2.8 &  0.039     & $0.494 \pm 0.001$   & 310       &  2 \\
291: &&&&&&&&\\
292: PKS~0215+015     & 1.3439 & 0.8  & 7.8  & 1.6 & $< 0.0052$ & $< 0.043$ & $ > 1020$ &  3 \\
293: &&&&&&&&\\
294: QSO~0957+561A$^b$& 1.3911 & $2.1\pm0.5$ & --   &  -- &    --      & --        & --        &  4 \\
295: &&&&&&&&\\
296: PKS~1354+258$^b$ & 1.4205 & $32\pm2$   &  --  &  -- &    --      & --        & --        &  4 \\
297: &&&&&&&&\\
298: TXS~2342+342     & 2.9084 & $20\pm0.5$   & 15   & 2.0 & $< 0.0192$ & $< 0.306$ & $ > 3585$ &  5 \\ 
299: &&&&&&&&\\
300: PKS~1354$-$107   & 2.966  & 6.0  & 6.5  & 2.0 & $< 0.05$   & $< 0.345$ & $ > 955 $ &  6 \\
301: &&&&&&&&\\
302: PKS~0537$-$286   & 2.974  & 2.0  & 10   & 1.9 & $< 0.0054$ & $< 0.058$ & $ > 1890 $ &  6 \\
303: &&&&&&&&\\
304: PKS~0336$-$014   & 3.0619 & $16\pm1.3$ & 20   & 1.4 & $< 0.0045$ & $< 0.095$ & $ > 9240$ &  7 \\
305: &&&&&&&&\\
306: PKS~0335$-$122   & 3.178  & $6\pm1$  & 15   & 1.2 & $< 0.0053$ & $< 0.084$ & $ > 3920 $ &  6 \\
307: &&&&&&&&\\
308: \hline
309: \end{tabular}
310: \end{center}
311: \vskip 0.1 in
312: ${}^a$$\Delta V$ is the FWHM used to compute the spin temperature in the
313: case of non-detections (see the text for more details). For 1629+12, it is the velocity
314: resolution of the observations.\\
315: ${}^b$The large uncertainty in the covering factor makes it difficult to
316: estimate the spin temperature (see the text for more details).\\
317: ${}^c$This is the RMS acually measured from the smoothed spectra. The spectra
318: were smoothed to a velocity resolution given in the preceeding column, except for
319: the following cases for which the RMS was measured at a slightly better spectral 
320: resolution than $\Delta V$, viz 12.8~\kms~for TXS~2342+342, 13.4~\kms~for
321: PKS~0336$-$014 and  13.8~\kms~for PKS~0335$-$122.\\
322: ${}^d$All limits (in this and other columns) correspond to 3$\sigma$. \\
323: ${}^e$The covering factor $f$ has been set to unity in all cases (see the text for
324: details).\\
325: ${}^f$References for the HI column densities : 1.~\cite{ledoux2002}; 
326: 2.~\cite{nestor2001}; 3.~\cite{lanzetta95}; 4.~\cite{rao2000}; 5.~\cite{white93}; 
327: 6.~\cite{ellison2001}; 7.~\cite{prochaska01}.
328: \vskip 0.1 in
329: \end{table*}
330: 
331: 
332: 
333: \begin{figure}
334: \centering
335: \epsfig{file=1629.eps,height=3.5truein,width=3.5truein}
336: \caption{ High resolution (0.6~\kms) GMRT HI absorption spectrum towards 
337: PKS~1629+12. Asymmetric HI absorption is clearly visible, with the peak 
338: optical depth at $z = 0.531787$.}
339: \label{fig:1629}
340: \end{figure}
341: 
342: \noi In the case of an optically thin, homogenous cloud in thermal equilibrium, its 
343: HI column density, $\NHI$, 21cm optical depth, $\tau$, and spin temperature, $T_{\rm s}$, 
344: are related by the expression (e.g. \cite{rohlfs86})
345: \begin{equation}
346: \label{eqn:tspin}
347: N_{\rm HI} = { 1.823\times10^{18} T_{\rm s} \over f} \int \tau \dV \; ,
348: \end{equation}
349: \noi where $N_{\rm HI} $ is in cm$^{-2}$, $T_{\rm s}$ in K and $\dV$ in km s$^{-1}$. $f$ is 
350: the covering factor of the absorber. Of course, a given line of sight through an 
351: absorber is quite likely to pass through a number of clouds, with different
352: spin temperatures and column densities. In such a situation, the spin temperature 
353: obtained using the above expression is the column density weighted harmonic mean 
354: of the spin temperatures of the individual clouds (provided, of course, that 
355: all the clouds are optically thin).
356: 
357: 	It should be noted that Eq.~(\ref{eqn:tspin}) relates the HI column density 
358: to the integrated optical depth $\int \tau {\mathrm d}V$; this means that, 
359: in the case of a non-detection, one needs to know the shape of the absorption 
360: profile to compute a lower limit on the spin temperature. This shape is, of course, 
361: not a priori known. However, since our limits on the optical depth imply that 
362: the HI line is optically thin, the line can be assumed to have a Gaussian shape; 
363: here, $\int \tau dV = 1.06 \tau_{\rm max} \Delta V$, where $\tau_{\rm max}$ is the peak 
364: optical depth (or the limit on the peak) and $\Delta V$ is the FWHM of the Gaussian. 
365: Further, for a gas cloud in thermal equilibrium, the kinetic temperature $T_{\rm k}$ (with
366: $T_{\rm s} \approx T_{\rm k}$, for a single cloud) and the FWHM of the observed line are 
367: related ($T_{\rm k} \approx 21.855 {\Delta V}^2$, with $T_{\rm k}$ in K and $\Delta V$ in 
368: km~s$^{-1}$); care must hence be taken to derive the limits on $T_{\rm s}$ in a self-consistent 
369: manner, so that the velocity width over which the integral is carried out is 
370: consistent with (at least) thermal broadening of the gas at the derived temperature
371: (any bulk kinematic motion will further broaden the absorption features; we will 
372: assume that no such motions are present). For example, if the $3 \sigma$ lower limit on 
373: the spin temperature is 10000~K, one should use $\Delta V \ga 20$~\kms, while, if the 
374: limit is 1000~K, it is appropriate to use $\Delta V \ga 7$~\kms.  We have, in the 
375: following analysis, used $\Delta V \approx 7$, $15$ and $20$~\kms~for temperatures 
376: $T_{\rm s} \sim 1000$, $5000$ and $10000$~K, respectively, and have smoothed the spectra 
377: to the above resolutions while computing the spin temperature.
378: 
379: The HI column density of a DLA can be directly estimated from the equivalent 
380: width of the Lyman-$\alpha$ profile. The original criterion for a system to be classified
381: as a damped absorber ($N_{\rm HI} \ge 2 \times 10^{20}$~\cm) was an observational one, 
382: linked to the ability to identify damped profiles in medium resolution spectra 
383: from the relatively small telescopes used to carry out the early DLA surveys. 
384: Throughout this paper, however, we will consider all systems for which the 
385: Lyman-$\alpha$ profile shows damping wings as DLAs, regardless of whether or not 
386: their HI column density is greater than the traditionally used threshold. We hence 
387: retain the absorbers towards PKS~0215+015 and PKS~2128$-$123 in our DLA sample, 
388: despite their HI column densities being less than $10^{20}$~\cm.
389: 
390: Given the HI column density, a search for 21cm absorption directly yields the spin 
391: temperature of the absorbing gas (or limits on the temperature, in the case of 
392: non-detections of absorption), provided the covering factor $f$ is 
393: known. Unfortunately, the radio continuum from quasars tends to arise from a 
394: far more extended region than the optical or UV emission and it may thus transpire 
395: that some fraction of the radio continuum may ``leak out'' from around the cloud 
396: and result in an incorrectly estimated spin temperature. Similarly, the line of 
397: sight along which the 21cm optical depth has been measured need not be the same 
398: as that for which one has an estimate for the HI column density from the Lyman-$\alpha$ 
399: line. The fraction of radio emission originating in compact components spatially
400: coincident with the UV point source can usually, however, be measured from VLBI 
401: observations (when such observations are available at sufficiently low frequencies).
402: One can then estimate $f$ and hence, the spin temperature; this is done next, for the 
403: ten sources of this paper, in order of increasing right ascension. The results are 
404: summarised in Table~2 in which Cols.~4 to 8 contain, respectively, the velocity 
405: width $\Delta V$ used to compute $T_{\rm s}$, the spectral RMS over this (or a slightly 
406: smaller) velocity width, the peak optical depth $\tau_{\rm max}$, 
407: $(1/f)\int \tau {\mathrm d}V$, and the spin temperature $T_{\rm s}$ (or limits on the 
408: last three quantities).
409: 
410: 1. VLBA maps of PKS~0215+015 at 2.3 and 8.4 GHz (from the Radio Reference Frame Image 
411: Database (RRFID)) show that the source is exceedingly compact, with most of the 
412: flux contained within the central $30$~milli-arcseconds. The quasar also has a flat 
413: spectrum 
414: between 8.4 GHz and 610 MHz, again implying that it is quite likely to be compact 
415: even at the lower frequencies. It is hence 
416: likely that the covering factor $f$ is close to unity. The HI column density of 
417: the $z = 1.345$ DLA is $\NHI = 8 \times 10^{19}$~\cm~(\cite{lanzetta95}); the 
418: spin temperature of the absorber is then $T_{\rm s} (3 \sigma) > 1020$~K (using $\Delta V 
419: = 7.8$~\kms).
420: 
421: 2. The $z = 3.178$ DLA towards PKS~0335$-$122 is the highest redshift system of 
422: our sample, with $\NHI = 6 \pm 1 \times 10^{20}$~\cm~(\cite{ellison2001}). Unfortunately, 
423: no VLBI information exists in the literature on the background quasar; however, this 
424: source has a fairly flat spectrum between 5 GHz and 1.4 GHz, with fluxes of 380 mJy 
425: and 420 mJy at these two frequencies. This spectral flatness implies that the flux 
426: is likely to arise from a fairly compact source component, which is hence likely 
427: to be covered by the DLA. This gives a lower limit to the covering factor, 
428: $f \ge 0.7$, which yields $T_{\rm s} (3 \sigma) > 2740$~K; if $f = 1$, one has $T_{\rm s} 
429:  (3 \sigma) > 3920$~K ($\Delta V = 15$~\kms).
430: 
431: 3. In the case of PKS~0336$-$014, 5 GHz VLBI observations (\cite{gurvits94}) have 
432: shown that the source is highly compact at this frequency, with the entire 
433: flux contained within the central $\sim 7$~milli-arcseconds. However, the source 
434: also has a negative spectral index, indicating the presence of some extended emission. 
435: In the absence of 
436: VLBI information at lower frequencies, we will assume a covering factor of unity.
437: The HI column density of the $z = 3.0619$ DLA is $\NHI = 1.6 \pm 0.13 \times 10^{21}$~\cm~
438: (\cite{prochaska01}) giving a spin temperature $T_{\rm s} (3 \sigma) > 9240$~K ($\Delta V = 
439: 20$~\kms). Carilli et al.  (1996) estimated $T_{\rm s} > 2500$~K for this DLA.
440: 
441: 4. PKS~0537$-$286 has an exceedingly flat spectrum between 8.4~GHz and 327~MHz; for 
442: example, the QSO flux density was measured to be 0.9~Jy at 8~GHz (Parkes survey, 
443: \cite{wright90}), 1.2~Jy at 4.85~GHz (\cite{griffith94}), 838~mJy at 
444: 1.4~GHz (NRAO-VLA Sky Survey, \cite{condon98}) and 1.05~Jy in our GMRT 
445: 327~MHz image. This implies that the source is likely to be highly compact. 
446: VLBA observations at 5~GHz (\cite{fomalont00}) have shown that the flux within 
447: the central 10~milli-arcseconds is $\sim 500$~mJy, implying a covering factor 
448: greater than 0.5. In fact, these observations recovered the entire 5~GHz flux of 
449: 1.2~Jy on the shortest VLBA baseline ($2 \times 10^6$ wavelengths, 
450: \cite{hirabayashi00}), implying that the entire source is contained within 
451: $\sim  0.1$~arcseconds. It is thus quite likely that the covering factor is close 
452: to unity; we will use $f = 1$ in this analysis. The HI column density of the 
453: $z = 2.9742$ DLA has been found to be $2.0 \times 10^{20}$~\cm~(\cite{ellison2001});
454: we then obtain a $3\sigma$ lower limit of $1890 $~K on the spin temperature of the 
455: absorbing gas ($\Delta V = 10$~\kms).
456: 
457: 5. VLBI mapping of QSO~0957+561 has been carried out on a number of occasions, at
458: various frequencies at and above 1.6~GHz (see, for example, \cite{garrett94,campbell95}).
459: The flux of component A has been measured to be only about 32~mJy at 1.6~GHz 
460: (\cite{garrett94}); clearly, most of the flux seen in the 610~MHz GMRT map comes 
461: from extended emission which is unlikely to be covered by the DLA. The covering 
462: factor is thus quite uncertain for this absorber, although it is likely to be 
463: low. Given this, limits on the spin temperature derived from the present data are 
464: unlikely to be reliable. In fact, it is known that the DLA does not 
465: cover the second image (B) of this gravitational lens system (\cite{rao2000}).
466: We will hence not include this system in the later analysis and also do not 
467: list limits on the spin temperature and the optical depth in Table~2.
468: 
469: 6. PKS~1354+258 is also an extended source at 610~MHz, with no VLBI information 
470: available in the literature; the DLA covering factor is hence again quite uncertain. 
471: However, the extremely high column density of the absorber ($N_{\rm HI} = 3.2 \pm 0.2
472: \times 10^{21}$~ cm$^{-2}$; \cite{rao2000}) results in high lower limits on the 
473: spin temperature even in the case of low covering factors. For example, 
474: $f = 0.1$ gives $T_{\rm s} (3 \sigma) > 1100$~K, while $f= 1$ would imply $T_{\rm s} (3 \sigma) 
475: > 7000 $~K (assuming velocity widths of $8$~\kms~and $20$~\kms~respectively. 
476: We will however again not quote a spin temperature in Table~2 due 
477: to the uncertainties in the covering factor.
478: 
479: 7. No information exists in the literature regarding the VLBI structure of 
480: PKS~1354$-$107. However, the source has a strongly inverted spectrum and 
481: is thus likely to be very compact. We will hence assume a covering factor 
482: of unity. Combining this with the HI column density of $6 \times 10^{20}$~\cm~
483: (\cite{ellison2001}) gives $T_{\rm s} (3 \sigma) > 955 $~K ($\Delta V = 6.5$~\kms).
484: 
485: 8. The HI column density of the $z = 0.5318$ DLA towards PKS~1629+12 is $\NHI = 
486: 2.8 \times 10^{20}$~\cm~(\cite{nestor2001}). PKS~1629+12 is a well-known compact 
487: steep spectrum source with an extremely one-sided structure (\cite{saikia90}). 
488: The 2.3 GHz VLBI map of Dallacasa et al. (1998) shows two components with 
489: fluxes of 110 mJy and 150 mJy. Only the 110~mJy western component is seen in their 
490: 8.4 GHz map, with almost the same flux; they hence identify this as the flat 
491: spectrum core of the source. The separation between the two components is 
492: 1.14~arcsec (i.e. $\sim 6.4$~kpc at $z = 0.5318$, using $H_0 = 65$~\kms~Mpc$^{-1}$ 
493: and an $\Omega = 1$, $\Lambda = 0$ cosmology). On the other hand, the core flux density 
494: was measured to be $179$~mJy at 408~MHz from a $\sim 0.6 \times 0.5$~arcsec 
495: resolution Merlin image (\cite{saikia90}). Using the 2.3~GHz and 408~MHz flux 
496: densities to estimate the core spectral index, we obtain a flux density of 
497: $\sim 142$~mJy at the redshifted 21cm line frequency of $\sim 927$~MHz. If the 
498: DLA covers only the core, the spin temperature of the absorber is $T_{\rm s} \sim 20$~K, 
499: a remarkably small value. However, the 21cm absorption is spread over 
500: $\sim 40$~km~s$^{-1}$ and is hence unlikely to arise from a single line of sight. 
501: Further, the 408~MHz Merlin map (\cite{saikia90}) shows that the entire flux at 
502: this frequency is contained within $\sim 3$~arcsecs ($\sim 16.8$~kpc at $z = 0.5318$). 
503: It is thus possible that the DLA covers some of the extended flux of the background 
504: source, besides the core. VLBI observations in the redshifted 21cm line would be 
505: necessary to detect the source components towards which the absorption occurs, 
506: and hence, the covering factor. For now, we note that a covering factor of 
507: unity results in $T_{\rm s} = 310~K$; this is a definite {\it upper limit} on the spin 
508: temperature, which must then lie between 20~K and 310~K.
509: 
510: 9. PKS~2128$-$123 has a very flat spectrum between 8.4 GHz and 750~MHz, 
511: with most measurements between these frequencies yielding a flux density of
512: around 2~Jy (see the NED\footnote{NASA/IPAC Extragalactic Database} for details).
513: VLBA maps at 2.3 and 8.4 GHz show that most of this flux is contained within 
514: $\sim 20$~milli-arcseconds, implying that the covering factor is likely to be close to 
515: unity. The HI column density of the $z \sim 0.4298$ absorber is $\NHI = 2.5 
516: \pm 0.06 \times 10^{19}$~\cm~(\cite{ledoux2002}); the $3\sigma$~lower limit on the spin 
517: temperature is then $T_{\rm s} (3 \sigma) > 980$~K ($\Delta V = 7$~\kms).
518: 
519: 10. Finally, TXS~2342+342 has a fairly flat spectrum between 1.4~GHz and 8~GHz 
520: (\cite{wilkinson98}; \cite{white92}), indicating that the quasar is very
521: compact at these frequencies. However, as Carilli et al. (1996) point 
522: out, the spectrum rises slowly down to the redshifted 21cm frequency of 
523: 356~MHz, indicating the presence of a more extended source component. 
524: Carilli et al. (1996) argue that the DLA is likely to at least cover 
525: the compact component and the covering factor is hence larger than 0.5; 
526: it is, however, unclear whether the extended flux is also covered. We will 
527: use $f = 1$ in the analysis (following \cite{carilli96}); the HI column 
528: density of $\NHI = 2 \pm 0.5 \times 10^{21}$~\cm~ then yields $T_{\rm s} (3 \sigma) > 3585$~K 
529: ($\Delta V = 15$~\kms). Carilli et al. (1996) estimated $T_{\rm s} > 1800$~K for 
530: this DLA.
531: 
532: 
533: \section{Discussion}
534: \label{sec:dis}
535: 
536: \begin{table*}[t!]
537: \label{tab:basicdata}
538: \begin{center}
539: \caption{Damped Ly-$\alpha$ Systems with 21cm Observations}
540: \begin{tabular}{@{}|c|c|c|c|c|c|c|c|c|c|c|}
541: \hline
542: Name & $z_{\rm abs}$ & $\NHI$ & $f$ & $ (1/f) \int \tau {\mathrm d}V$ & ${\Delta V_{21}}^{  a}$ 
543: & $T_{\rm s}^{  b}$ & Absorber & $b^d$& Refs.\\
544:   &&$10^{20}$~\cm&  & \kms & \kms &K & ID & kpc & \\
545: &  &&&&&&&& \\
546:   \hline
547: &  &&&&&&&&\\
548:   0738+313 (A)  &0.09123 &$15\pm 2$& 1   &$1.0\pm0.01$&40  &$825\pm120$& Dwarf  &$<4$ & 1,2,3 \\
549: PKS~0439$-$433$^c$&0.10097& ?    & 1     &$0.07\pm0.02$     &10  & ? &Spiral ?& 8.0 & 4,5   \\
550:   0738+313 (B)  &0.22125 &$7.9\pm1.4$& 1 &$0.488\pm0.004$&30  &$890\pm165$& Dwarf  & 20  & 1,3,6   \\
551:   PKS~0952+179 &0.23779 &$21\pm2.5$ &0.25 &$0.56\pm0.02$ &30&$2055\pm330$ & LSB    &$<7$ & 7,8,9  \\
552: PKS~1413+135$^{c\star}$&0.24671&200$^\star$& 0.12&59.43 &39&185& Spiral & ?   & 10,11,12\\
553:   PKS~1127$-$145&0.3127  &$51\pm9$ & 1 &$3.074\pm0.002$&120 &$910\pm160$& LSB    &$<10$& 8,9,13 \\
554:   PKS~1229$-$021&0.39498 &$5.6\pm0.5$& 0.5 &$1.81\pm0.01$ &110 &$170\pm15$ & Spiral &  11 & 14,15,16\\
555:   QSO~0248+430  &0.3941  &  ?        & ?   & 2.99     &40  & ?       &  -     & -   & 17   \\
556:   PKS~2128$-$123&0.4298  &$0.25\pm0.06$& 1 &$<0.014$  &-   &$>$ 980 &  -     & -   & 18,19\\
557:   3C196         &0.43670&$6.3\pm1.6$& ?   & ?        &250 & -       & Spiral & 9.2 & 14,20,21\\
558: PKS~1243$-$072  &0.4367 & ?         & 1   &0.38      &23  & ?       & Spiral & 11.3& 17,22 \\
559:   3C446$^c$     &0.4842 &6.3        & 0.1 &$<0.22$   &-   &$>$ 1570 & -      & -   & 13,23 \\ 
560:   A0~0235+164   &0.524  &50        & 1  &$13\pm0.6$  &125 &210  & Spiral &$<11.2$&  24,25,26 \\
561:   B2~0827+243   &0.5247 &$2.0\pm0.2$& 0.67&$0.33\pm0.03$&50  &$330\pm70$  & Spiral & 34  & 7,8,9 \\
562:   PKS~1629+12 &0.5318  &2.8        & 1   &$0.494\pm0.001$     &40  & 310  & Spiral & 17  & 9,19 \\
563: PKS~0118$-$272$^c$&0.5579  & ?         & 0.5 &$<0.12$   &-   & ?       &  -     & -   & 7,27 \\
564:   3C286      &0.69215 &$20\pm1.6$  &1 ?&$0.91\pm0.09$ &18 & $1205\pm240$& LSB    & 19  & 15,28,29 \\
565:   PKS~0454+039    &0.8596  &$5\pm0.16$ & 1   &$<0.32$   &-   &$>$855 & Dwarf  & 6.4 & 15,30,31 \\
566:   PKS~0215+015$^c$&1.3449  &0.8        & 1   &$<0.043$ &-   &$>$ 1020 & -      & -   &  19,32 \\
567:   QSO~0957+561A   &1.3911  &$2.1\pm0.5$& ?   &-        &-   & -       & -      & -   & 8,19 \\
568:   PKS~1354+258    &1.4205  &$32\pm2$   & -   &-        &-   & -       & -      & -   & 8,19 \\
569:   MC3~1331+170    &1.77636 &$15\pm1.5$ & 0.8 &$0.5\pm0.1$&25  &$1645\pm615$ & - & -   & 31,33,34 \\
570:   PKS~1157+014    &1.944   &$63\pm10$  & 0.55&$4 \pm 0.2$ &60 &$865 \pm190$ & - & -   & 31,35,36 \\
571:   PKS~0458$-$02   &2.03945 &$45\pm8$  & 1    &$6.4\pm0.4$ &30 &$385\pm100$& -      & -   & 34,37,47\\
572:   PKS~0528$-$2505 &2.8110  &$22\pm1$   & 1   &$<1.70$  &-   &$>$ 710  & -      & -   & 38,39,40 \\
573:   TXS~2342+342    &2.9084  &$20\pm0.5$ & 1   &$<0.306$ &-   &$>$ 3585 & -      & -   & 19,41 \\
574:   PKS~1354$-$107  &2.966   &6          & 1   &$<0.345$ &-   &$>$ 955 & -      & -   & 19,42 \\
575:   PKS~0537$-$286  &2.9742  &2          & 1   &$<0.058$ &-   &$>$ 1890 & -      & -   & 19,42 \\
576:   PKS~0336$-$014  &3.0619  &$16\pm1.3$ & 1   &$<0.095$ &-   &$>$ 9240 & -      & -   & 19,41,46\\
577:   PKS~0335$-$122  &3.178   &$6\pm1$    & 1   &$<0.084$ &-   &$>$ 3920 & -      & -   & 19,42\\
578:   PKS~0201+113    &3.3875  &$18\pm3$   & 1   &$<0.30$  &-   &$>$ 3290 & -      & -   & 41,44,45 \\
579: &  &&&&&&& &\\
580:   \hline
581:   \end{tabular}
582: \end{center}
583: \vskip 0.05in
584:   ${}^\star$~ The $z = 0.24671$ absorber towards PKS~1413+135 may be an ``associated''
585: absorption system, rather than an intervening one (\cite{perlman02}). 
586: Further, its HI column density is estimated from an X-ray spectrum (\cite{stocke92}).\\
587:   ${}^a$~ The velocity width $\Delta V$ quoted here is the entire velocity range
588:   over which absorption is seen.\\
589:   ${}^b$~The spin temperature values (or $3\sigma$ limits) quoted here have been
590:   consistently (re-)computed from the published data using Eq.~(\ref{eqn:tspin}), and the latest available values for all parameters in the
591:   equation. The values quoted here and the errors have been rounded off to the 
592:   nearest 5~K. The errors on $\NHI$ and hence, on $T_{\rm s}$, have been quoted when 
593:   available in the original references; in the case of asymmetric errors, the 
594:   larger value has been quoted. In certain instances, the values may differ 
595:   slightly from that quoted by the authors in the original reference. The 
596:   only exception is PKS~0201+113, for which multiple and quite different 
597:   values are quoted in the literature. We use the value from Kanekar \& 
598:   Chengalur (1997), modified by the new HI column density of $1.8 \pm 0.3 
599:   \times 10^{21}$~\cm, from Ellison et al. (2001b). \\
600:   ${}^c$~Candidate \dla systems, based on IUE spectra, metal lines or X-ray spectra.\\
601:   ${}^d$~The impact parameter $b$ has been calculated using an $\Omega = 1$, 
602: $\Lambda = 0$ Universe, with $H_0 = 65$~\kms~Mpc$^{-1}$.\\
603: \vskip 0.05in
604: 
605: ${}^{1}$~\cite{rao98}, ${}^{2}$~\cite{chengalur99}, ${}^{3}$~\cite{cohen2001},
606: ${}^{4}$~\cite{petitjean96}, ${}^{5}$~\cite{kanekar2001a}, ${}^{6}$~\cite{kanekar2001b},
607: ${}^{7}$~\cite{kanekar2001c}, ${}^{8}$~\cite{rao2000}, ${}^{9}$~\cite{nestor2001},
608: ${}^{10}$~\cite{stocke92}, ${}^{11}$~\cite{carilli92}, ${}^{12}$~\cite{perlman02}.
609: ${}^{13}$~\cite{chengalur2000}, ${}^{14}$~\cite{kanekar2002b}, ${}^{15}$~\cite{boisse98}, 
610: ${}^{16}$~\cite{lebrun97}, ${}^{17}$~\cite{lane2002}, ${}^{18}$~\cite{ledoux2002}, 
611: ${}^{19}$~This paper, ${}^{20}$~\cite{briggs2001}, ${}^{21}$~\cite{cohen96}, 
612: ${}^{22}$~\cite{kanekar2002a}, ${}^{23}$~\cite{lanzetta95}, ${}^{24}$~\cite{roberts76}, 
613: ${}^{25}$~\cite{burbidge96}, ${}^{26}$~\cite{cohen99}, ${}^{27}$~\cite{vladilo97}, 
614: ${}^{28}$~\cite{steidel94}, ${}^{29}$~\cite{davis78}, ${}^{30}$~\cite{steidel95}, 
615: ${}^{31}$~\cite{briggs83}, ${}^{32}$~\cite{turnshek2002}, ${}^{33}$~\cite{wolfe79}, 
616: ${}^{34}$~\cite{pettini94}, ${}^{35}$~\cite{briggs84}, ${}^{36}$~\cite{wolfe81}, 
617: ${}^{37}$~\cite{wolfe85}, ${}^{38}$~\cite{carilli96}, ${}^{39}$~\cite{morton80}, 
618: ${}^{40}$~\cite{foltz88}, ${}^{41}$~\cite{white93}, ${}^{42}$~\cite{ellison2001}, 
619: ${}^{43}$~\cite{ellison2001b}, ${}^{44}$~\cite{kanekar97}, ${}^{45}$~\cite{briggs97},  
620: ${}^{46}$~\cite{prochaska01}, ${}^{47}$~\cite{pw99}.
621: 
622: \end{table*}
623: 
624: 	Neutral gas in the galaxy is believed to exist in two stable phases,
625: a cold dense phase (the cold neutral medium, CNM), with typical temperatures
626: of $\sim 100$~K, and a warm rarefied phase (the warm neutral medium, WNM),
627: with typical temperatures of $\sim 8000$~K. It should be emphasised that 
628: such a multi-phase medium exists only over a finite range of interstellar pressures 
629: (\cite{wolfire95}); at higher pressures only the CNM phase exists, while at lower 
630: pressures only the WNM phase is found. Further, although the multi-phase structure 
631: results from a balance 
632: between the heating and cooling rates, which themselves depend on various 
633: quantities such as the radiation field, the metallicity,  the dust content, 
634: etc., it has been shown that a two-phase pressure equilibrium is possible 
635: over fairly wide ranges of the above quantities (and, of course, the 
636: interstellar pressure; \cite{wolfire95}). On the observational front, the 
637: existence of neutral gas in a multi-phase medium has been established
638: not just for gas in the solar neighborhood, but also in high velocity
639: clouds (e.g. \cite{braun00}), dwarf galaxies (e.g. \cite{young97}) and
640: DLAs (\cite{lane2000,kanekar2001b}). In the latter case of the two DLAs at
641: $z=0.0912$ and $z=0.2212$ toward B0738+313, it has been explicitly
642: shown that the observed high spin temperatures are a consequence of 
643: $70 - 80$~\% of the gas being in the warm phase. In this regard, 
644: the multi-phase structure of at least these two systems is more like that 
645: of dwarf galaxies (see, for example, \cite{young97}) than large spiral disks.
646: 
647: 	Table~3 contains a list of the 31 known candidate and 
648: confirmed DLAs for which 21cm observations are available in the published literature
649: (this is an update of the list in \cite{chengalur2000}). Note that the 
650: Lyman-$\alpha$ line has so far not been observed for the DLAs at $z = 0.24671$ 
651: towards PKS~1413+135, $z = 0.3941$ towards QSO~0248+430 and $z=0.43669$ towards 
652: PKS~1243$-$072; however, all three systems show strong 21cm absorption, implying 
653: (for reasonable values of $T_{\rm s}$) HI column densities higher than $10^{20}$~\cm~
654: (\cite{carilli92}; \cite{kanekar2002a}; \cite{lane2002}) and have hence been 
655: included in the present sample. Further, 
656: the $z = 0.24671$ absorber towards PKS~1413+135 may be an associated system 
657: (\cite{perlman02}); the HI column density listed in Table~3 is estimated from the 
658: deficit of soft X-rays (\cite{stocke92}). No similar direct estimate exists for the HI 
659: column densities of the systems towards PKS~1243$-$072 and QSO~0248+430. We also note 
660: that two candidate DLAs (on the basis of metal lines), at $z = 0.5579$ towards 
661: PKS~0118$-$272 (\cite{kanekar2001c}) and $z = 0.101$ towards PKS~0439$-$433 
662: (\cite{kanekar2001a}) have, for completeness, been included in this table; however, 
663: they will not be used in the later analysis, as their HI column densities (and hence, 
664: spin temperatures) are unknown. Cols.~3, 4, 5, 6 and 7 of the table contain the HI 
665: column density (in units of $10^{20}$~\cm), the adopted covering factor $f$, the ``true'' 
666: equivalent width ($(1/f) \int \tau {\mathrm d}V$), the velocity spread of 21cm 
667: absorption (between nulls, in \kms) and the spin temperature. Besides these, Cols.~8 
668: and 9 list the nature of the objects identified as the DLA hosts and their impact 
669: parameters ($b$, in kpc) to the QSO line of sight. Of course, it is possible that 
670: the latter identifications are in error in some cases if, for example, the absorption 
671: arises in an object closer to the QSO but too faint to be identified in optical/IR 
672: images. With this caveat in mind, we note that seven of the DLA hosts have been 
673: identified with large, luminous galaxies, with luminosities close to $L_\star$.
674: 
675: \begin{figure}
676: \centering
677: \epsfig{file=tvsz.eps,height=3.3truein,width=3.3truein}
678: \caption{The spin temperature (Log[$T_{\rm s}$]) of the 24 DLAs of the complete sample 
679: with $T_{\rm s}$ estimates as a function of redshift $z$. Note that seven DLAs listed 
680: in Table~3 have not been included here for reasons listed in the text. Objects 
681: identified with spiral galaxies are plotted using squares; the filled square 
682: shows the $z = 0.24671$ spiral towards PKS~1413+135, which may be an associated 
683: system. Other detections are shown as filled triangles, with non-detections shown 
684: as open triangles, with arrows. }
685: \label{fig:tvsz}
686: \end{figure}
687: 
688: \begin{figure}
689: \centering
690: \epsfig{file=eq_wid.eps,height=3.3truein,width=3.3truein}
691: \caption{The {\it true} equivalent width ($\int \tau \mathrm{d} V$/$f$) as a 
692: function of the HI column density for the 24 DLAs of the complete sample for 
693: which we can calculate $T_{\rm s}$ values. Objects identified with spiral galaxies are
694: plotted using squares; the filled square shows the $z = 0.24671$ spiral towards 
695: PKS~1413+135, which may be an associated system. Other detections are shown as 
696: filled triangles, with non-detections shown as open triangles, with arrows. The crosses 
697: show measurements in the Galaxy by Colgan et al. (1988) and Payne et al. (1982), using 21cm 
698: emission/absorption studies.}
699: \label{fig:eq_wid}
700: \end{figure}
701: 
702: \begin{figure}
703: \centering
704: \epsfig{file=imp.eps,height=3.3truein,width=3.3truein}
705: \caption{The impact parameter $b$ (in kpc) as a function of spin temperature for 
706: 10 DLAs at $z < 1$. The open squares represent objects identified with 
707: spiral galaxies; the remaining detections are shown as filled triangles, with 
708: non-detections shown as open triangles, with arrows. }
709: \label{fig:ts-b}
710: \end{figure}
711: 
712: Fig.~\ref{fig:tvsz} shows a plot of the measured spin temperatures as 
713: a function of redshift for the 24 DLAs of the complete sample which have reliable 
714: estimates of the spin temperature (note that this plot includes the $z = 0.24671$ absorber 
715: towards PKS~1413+135 which may be an associated system). The seven systems 
716: in Table~3 which are not included in Fig.~\ref{fig:tvsz} are the two 21cm absorbers 
717: towards PKS~1243$-$072 and QSO~0248+430 for which the 
718: HI column density is unknown (see above), the DLAs towards PKS~1354+258 and 
719: QSO~0957+561 for which the covering factor is highly uncertain (see 
720: Sect.~\ref{sec:Ts}), the two candidate DLAs towards PKS~0118$-$272 and PKS~0439$-$433 
721: and, finally, the $z = 0.437$~absorber towards 3C196, where the lines of sight to the 
722: optical QSO and the radio continuum trace very different paths through the absorbing 
723: galaxy. The dotted line in Fig.~\ref{fig:tvsz} is at 350~K -- more than 80~\% 
724: of spin temperature measurements in M~31 and the Milky Way have values $< 350$~K 
725: (\cite{braun92}). It can be seen that the spin temperatures of the majority of 
726: DLAs are considerably higher than those typical of lines of sight through the 
727: disks of the Milky Way or nearby spiral galaxies. The only exceptions to the above are 
728: the five systems at low $z$ which have been identified with spiral disks and 
729: the DLA at $z \sim 2.04$ towards PKS~0458$-$020, for which $T_{\rm s} \sim 385\pm 100$~K; we 
730: discuss the latter separately below. Note that it is not presently possible to estimate 
731: the spin temperatures of the remaining two $L \sim L_\star$ DLAs at $z \sim 0.437$ 
732: towards PKS~1243$-$072 and 3C196. Finally, the 21cm absorption profile for PKS~1629+12 
733: reported in this paper continues the  trend (noted earlier in Chengalur \& Kanekar (2000) 
734: and Kanekar \& Chengalur (2001)) of low spin temperatures being found in all DLAs identified 
735: as $L \sim L_\star$ galaxies. 
736: 
737: Fig.~\ref{fig:eq_wid} presents the data in an alternative way, as a 
738: scatter plot of the ``true'' HI 21cm equivalent width (i.e. $(1/f) \int \tau 
739: {\mathrm d} V$) against the HI column density.  Both of these are directly
740: observable quantities. The crosses show the same quantities for Galactic HI
741: clouds (\cite{colgan88}; \cite{payne82}), measured using 21cm 
742: absorption/emission studies. HI column density estimates using observations 
743: of the Lyman-$\alpha$ line have been found to be in good agreement with 
744: estimates from 21cm emission studies (\cite{dickey90}). Fig.~\ref{fig:eq_wid}
745: hence allows a direct comparison of the same observables for
746: DLAs and galactic clouds. It can be clearly seen that, at a given HI column 
747: density, the majority of DLAs tend to have smaller 21cm equivalent widths than those
748: found for lines of sight through the Galaxy. The only exceptions are DLAs 
749: that are known to be associated with spiral galaxies (shown as squares) 
750: and the $z \sim 2.04$ DLA towards PKS~0458$-$020,  (which we discuss in 
751: more detail below), which do lie within the range of Galactic values. The 
752: conclusion would seem to be that the majority of DLAs contain larger fractions 
753: of the warm (and thus weakly absorbing) phase of neutral hydrogen than typical 
754: in nearby spirals, as earlier pointed out by Carilli et al. (1996) and Kanekar 
755: \& Chengalur (2001).
756: 
757: It is possible that the high $T_{\rm s}$ values seen in DLAs might arise due to 
758: absorption originating in the outskirts of large disks, which would also 
759: provide a higher absorption cross-section than the inner regions of these 
760: galaxies. Fig.~\ref{fig:ts-b} shows a plot of $T_{\rm s}$ versus impact parameter 
761: $b$ for the 10 DLAs of the sample (all with $z < 1$) with optical identifications,
762: as well as estimates of both $T_{\rm s}$ and impact parameter; absorbers identified with 
763: spiral galaxies are shown as open squares. No correlation can be seen 
764: to exist between $T_{\rm s}$ and $b$. On the contrary, the largest impact parameter 
765: of the sample ($b = 34$~kpc) is seen in a low $T_{\rm s}$ system, the $z = 0.5247$ DLA 
766: towards B2~0827+243, while four of the high $T_{\rm s}$ DLAs have $b < 10$~kpc. The low 
767: redshift sample clearly does not support the hypothesis that the high $T_{\rm s}$ seen in
768: low redshift DLAs is because the absorption arises from lines of sight which pass
769: through the outer regions of disk galaxies. Instead, it appears that 
770: the spin temperature correlates with the nature of the absorbing galaxy. We discuss 
771: next the $z \sim 2.04$ DLA towards PKS~0458$-$020, the only high $z$ DLA with a 
772: spin temperature $T_{\rm s} < 400$~K.
773: 
774: The $z = 2.03945$ DLA towards PKS~0458$-$020 is as yet the highest redshift DLA 
775: with a confirmed detection of 21cm absorption (\cite{wolfe85}). Briggs et al. (1989) 
776: found the VLBI 21cm absorption profile to be very similar to the single dish profile 
777: and used this to argue that the absorber shows little structure on the scale of
778: the extended radio continuum emission of the background source; these observations 
779: hence imply that the transverse size of the absorber is larger than $8\; h^{-1}$~kpc.
780: Further, the low ionization metal lines of the DLA are clearly asymmetric (\cite{pw97}), 
781: which can be explained if the absorber is a rapidly rotating disk (although such 
782: asymmetries are also explanable in other models; see, for example, \cite{haehnelt98}, 
783: \cite{mcdonald99}). The spin temperature of the absorber was originally estimated to 
784: be $T_{\rm s} = 594 \pm 297 $~K (\cite{wolfe85}), the large error arising due to uncertainties 
785: in fitting the Lyman-$\alpha$ profile ($\NHI = 8 \pm 4 \times 10^{21}$~\cm; \cite{wolfe85}). 
786: Recently, however, the HI column density has been measured far more accurately by 
787: Pettini et al. (1994), who quote $\NHI = 4.5 \pm 0.8 \times 10^{21}$~\cm. Combining this 
788: with the HI 21cm absorption profile (and taking $f = 1$, as confirmed by the 
789: VLBI observations) yields $T_{\rm s} = 385 \pm 100$~K, 
790: a value not too far from those obtained in local spirals; this is the first case of a 
791: DLA at $z \ga 0.6$ with a spin temperature less than $400$~K. Interestingly enough, 
792: the 21cm equivalent width of the absorber too lies close to the range of Galactic 
793: values shown in Fig.~\ref{fig:eq_wid}. We note, however, that the metallicity of the 
794: absorber is far lower than solar values ([Zn/H] = $-1.16\pm 0.09$; \cite{pw99}), 
795: implying that the DLA has not undergone much chemical evolution. Detailed analysis 
796: of the formation of a multi-phase medium in proto-galaxies (\cite{spaans97a}; 
797: \cite{spaans97b}) indicate that the epoch of formation of a multi-phase ISM is a strong 
798: function of the galactic mass; galaxies with baryonic mass $\sim 5 \times 10^{11}$~M$_\odot$ 
799: form the cold phase by $ z \sim 3$ while dwarf galaxies, with baryonic mass  
800: $\sim 3\times 10^9$~M$_\odot$, typically form the cold phase only by $z\sim 1$. The
801: relatively low measured spin temperature in the $z \sim 2.04$ DLA thus makes it likely that 
802: the absorber is a large galaxy.  The VLBI observations and the low spin temperature
803: are hence both consistent with the DLA being a massive galaxy, with the high mass 
804: providing a sufficiently high central pressure to compensate for the low
805: metallicity and produce a sizeable fraction of the cold phase of HI.
806: 
807: \begin{table}
808: \label{table:cnm}
809: \begin{center}
810: \caption{The CNM fraction in high $z$ DLAs.}
811: \vskip 0.1in
812: \begin{tabular}{@{}|l|c|c|c|c|}
813: \hline
814: Name & $z_{\rm abs}$ & $\NHI $ & $\NHI $ & $f_{\rm CNM}$ \\
815: && (200~K)&(Total) & \\
816: &&$ \times 10^{20}$& $\times  10^{20}$& \\
817: &&&& \\
818: \hline
819: &&&& \\
820: PKS~0528$-$250  & 2.8110 & $<$6.2  & $22 \pm 1$  & $<$0.28 \\
821: TXS~2342+342    & 2.9084 & $<$0.7  & $20\pm0.5$  & $<$0.03 \\
822: PKS~1354$-$107  & 2.966  & $<$1.3  & 6.0         & $<$0.21 \\
823: PKS~0537$-$286  & 2.9742 & $<$0.2  & 2.0         & $<$0.09 \\
824: PKS~0336$-$014  & 3.0619 & $<$0.2  & $16 \pm1.3$ & $<$0.01 \\
825: PKS~0335$-$122  & 3.178  & $<$0.2  & $6\pm1 $    & $<$0.04 \\
826: PKS~0201+113    & 3.3875 & $<$1.1  & $18 \pm 3$  & $<$0.06 \\
827: &&&& \\
828: \hline
829: \end{tabular}
830: \end{center}
831: \end{table}
832: 
833: 	The evolution of spin temperature with redshift has been the subject 
834: of interest for some time. The original observations of high spin temperatures 
835: at high redshift led to the suggestion that the elevation of the spin 
836: temperature was perhaps due to evolutionary effects. More recently, 
837: the detection of low redshift DLAs ($z \la 0.2$) with high $T_{\rm s}$ values 
838: (\cite{lane98,chengalur99}) indicated that a simple trend did not exist in 
839: the evolution of $T_{\rm s}$ with redshift. However, Chengalur \& Kanekar (2000) noted the 
840: following trend: DLAs with both low and high spin temperatures are found at
841: low redshifts ($z \la 1$), while only high $T_{\rm s}$ DLAs are found at high redshifts 
842: ($z \ga 3$). The present, larger sample of Fig.~\ref{fig:tvsz} shows this trend 
843: clearly, with the exceedingly high values of spin temperature at $z \ga 3$ indicating 
844: that these systems contain very small fractions of the CNM phase of HI. 
845: 
846: The non-detection of HI absorption in the present fairly deep search can be used 
847: to place stringent constraints on the fractional content of CNM in the high $z$ 
848: absorbers. To do so, we note that the allowed values of CNM temperatures in the 
849: Milky Way lie between 40~K and 200~K (\cite{wolfire95}). We hence use 200~K (in 
850: Eq.~(\ref{eqn:tspin})) as an upper limit to the temperature of the CNM phase 
851: in the seven $z \ga 3$ DLAs to estimate upper limits on the CNM fraction in these 
852: systems. The original high resolution spectra (see Table~1) are used for this 
853: purpose, not the smoothed spectra used to estimate the spin temperature. 
854: The results are shown in Table~4 where the third column gives the $3\sigma$ upper 
855: limit to the HI column density in 200~K gas (note that this limit 
856: scales linearly with the assumed CNM temperature and would hence be even lower for 
857: lower temperature CNM). The last column in this table gives the upper limit to the 
858: fraction of the HI in CNM, $f_{\rm CNM}$; it can be seen that five of the seven systems 
859: at $z \ga 3$ have CNM fractions less than 10~\%, with three of these having 
860: $f_{\rm CNM} < 5$~\%. Clearly, most of the gas in the high $z$ DLAs is warm, 
861: the typical fraction of CNM gas is even smaller than that in the two high 
862: $T_{\rm s}$ DLAs at low redshift with measured CNM fractions (\cite{lane2000,kanekar2001b}) 
863: or in local dwarf galaxies (\cite{young97}).
864: 
865: An independent line of evidence for high temperatures at high redshift comes from 
866: observations of the $n$(CII)/$n$(CI) ratio (\cite{liszt2002}), which is quite different 
867: in the CNM and the WNM. Liszt (2002) used observations of this ratio in 11 DLAs to 
868: suggest that high redshift ($z>2.3$) DLAs are dominated by warm gas, with at most 
869: a few per cent of the neutral HI in the CNM. This is in agreement with our results. 
870: Besides this, it has been observed that high $z$ DLAs typically have very low 
871: molecular gas fractions; this can be understood in models in which the absorbing 
872: gas is at high temperature (\cite{petitjean2000}; but see also \cite{liszt2002}). 
873: We note, in passing, that it is exceedingly interesting that the $z = 3.0619$ DLA 
874: towards PKS~0336$-$014 has $T_{\rm s} > 9240$~K, implying that its gas is significantly hotter 
875: than even the standard WNM in the Milky Way (which has temperatures in the range 
876: $5000 - 8000$~K; \cite{wolfire95}). 
877: 
878: 	Of the 24 DLAs with $T_{\rm s}$ measurements, 16 are at $z < 2$ and 8 at $z > 2$; we
879: will call these the low and high $z$ samples respectively. The sample is of a sufficiently
880: large size to carry out a statistical comparison to test the 
881: likelihood that the systems in the two samples stem from the same parent distribution. 
882: The Gehan test (\cite{gehan65}; \cite{miller81}) is appropriate for this comparison 
883: as it can be directly used in the case of limits on the measured quantity (in our 
884: case, the spin temperature). When applied to our data, this test rules out the possibility 
885: of the two samples being drawn from the same 
886: distribution at the $\sim 99$\% confidence level. (If one chooses to treat the $3\sigma$
887: upper limits as detections, a Kolmogorov-Smirnov rank 1 test rules out the above 
888: hypothesis at the 95~\% confidence level.) It is thus very unlikely that the high and 
889: low $z$ DLAs are systems of similar ``character''. We note that $z = 2$ was used 
890: as the redshift of separation of the two samples as this is the approximate redshift 
891: where star formation is believed to peak; it is hence plausible that systems at $z > 2$ 
892: will show differences from those at lower redshifts. Since we have only 4 systems with 
893: $T_{\rm s}$ estimates in the redshift range $1 < z < 2.8$ (the $z \sim 1.4$ DLAs 
894: towards PKS~1354+258 and QSO~0957+561A do not have good estimates of $T_{\rm s}$), it is 
895: not possible to independently determine the threshold redshift with the present sample.
896: 
897: In conclusion, the sample of damped Lyman-$\alpha$ systems which have been searched 
898: for 21cm absorption now consists of 31 systems; estimates of the spin temperature 
899: are available in 24 cases, of which 16 DLAs are at $z < 2$ and 8 at $z > 2$. All 
900: of the low $T_{\rm s}$, low $z$ DLAs have been identified with large, luminous galaxies, while all 
901: DLAs at $z < 1$ with high spin temperature ($T_{\rm s} \ga 1000$~K) have been identified 
902: either with LSBs or dwarfs. It thus appears likely that the spin temperature depends
903: mainly on the nature of the absorbing galaxy, as suggested earlier by Chengalur \& 
904: Kanekar (2000). We use a recent measurement of the HI column density in the $z = 2.03945$ DLA 
905: towards PKS~0458$-$020 to derive a relatively low $T_{\rm s}$ value in the absorber; this is 
906: the first high $z$ DLA to show $T_{\rm s} < 400$~K. The low metallicity allied with a 
907: low spin temperature suggest that the DLA is a massive galaxy, consistent
908: with VLBI limits of its transverse size. The new deep 
909: GMRT 21cm spectra of this paper have also been used to estimate the fraction of HI 
910: in the cold phase, $f_{\rm CNM}$, in the absorbers at $z \ga 3$; it was found that 
911: $f_{\rm CNM} < 0.3$ in all seven absorbers, with $f_{\rm CNM} < 0.1$ in five of the seven 
912: cases. These results are in good agreement with estimates of $f_{\rm CNM} \sim few$~\%, 
913: from observations of the $n$(CII)/$n$(CI) ratio (\cite{liszt2002}). Finally, we have 
914: used the above spin temperature estimates to test the likelihood that the $z > 2$ 
915: and $z < 2$ DLAs are drawn from the same parent population and rule out this 
916: hypothesis at the $\sim 99$\% confidence level. 
917: 
918: 
919: \begin{acknowledgements}
920: We thank the referee, Wendy Lane, for numerous suggestions and comments on an 
921: earlier version which have significantly improved this paper. These observations 
922: would not have been possible without the many years of dedicated effort put in 
923: by the GMRT staff in order to build the telescope. The GMRT is operated by the 
924: National Centre for Radio Astrophysics of the Tata Institute of Fundamental Research.
925: This research has made use of the United States Naval Observatory (USNO) Radio 
926: Reference Frame Image Database (RRFID) and the NASA/IPAC Extragalactic Database 
927: (NED) which is operated by the Jet Propulsion  Laboratory, California Institute 
928: of Technology, under Contract with the National Aeronautics and Space Administration.
929: 
930: \end{acknowledgements}
931: 
932: \begin{thebibliography}{}
933: \bibitem[Aldcroft et al. 1994]{aldcroft94} Aldcroft T. L., Bechtold J. \& Elvis M.,
934: 	1994, ApJS 93, 1
935: \bibitem[Barthel et al. 1990]{barthel90} Barthel P. D., Tytler D. R. \& Thomson B.,
936: 	1990, A\&AS 82, 339
937: \bibitem[Boisse et al. 1998]{boisse98} Boiss\'{e} P., le Brun V., Bergeron J. \& 
938: 	Deharveng J-M., 1998, A\&A 333, 841
939: \bibitem[Braun \& Walterbos 1992]{braun92}Braun R. \& Walterbos R. A. M.,
940: 	1992, ApJ 386, 120.
941: \bibitem[Braun \& Burton 2000]{braun00} Braun R. \& Burton W. B., 2000,
942: 	A\&A 354, 853.
943: \bibitem[Briggs et al. 1983]{briggs83} Briggs F. H. \& Wolfe A. M., 1983, ApJ 268, 76
944: \bibitem[Briggs et al. 1984]{briggs84} Briggs F. H., Turnshek D. A. \& Wolfe A. M., 
945: 	1984, ApJ 287, 549
946: \bibitem[Briggs et al. 1989]{briggs89} Briggs F. H., Wolfe A. M., Liszt H. S., 
947: 	Davis M. M. \& Turner K. L., 1989, ApJ 341, 650 
948: \bibitem[Briggs et al. 1997]{briggs97} Briggs F. H., Brinks E. \& Wolfe A. M., 
949: 	1997, AJ 113, 467
950: \bibitem[Briggs et al. 2001]{briggs2001} Briggs F. H., de Bruyn A. G. \& Vermeulen 
951: 	R. C., 2001, A\&A 373, 113
952: \bibitem[le Brun et al. 1997]{lebrun97} le Brun V., Bergeron J., Boiss\'{e} P. \& 
953: 	Deharveng J-M., 1997, A\&A 321, 733
954: \bibitem[Burbidge et al. 1996]{burbidge96} Burbidge E. M., Beaver E. A., 
955: 	Cohen R. D., Junkkarinen V. T. \& Lyons R. W., 1996, AJ 112, 2533
956: \bibitem[Campbell et al. 1995]{campbell95} Campbell R. M., Lehar J., Corey B. E.,
957: 	Shapiro I. I. \& Falco E. E., 1995, AJ 110, 2566
958: \bibitem[Carilli et al. 1992]{carilli92} Carilli C. L., Perlman E. S. \& 
959: 	Stocke J. T., 1992, ApJ 400, L13
960: \bibitem[Carilli et al. 1996]{carilli96} Carilli C. L., Lane W., de Bruyn A. G., 
961: 	Braun R. \& Miley G. K., 1996, AJ 111, 1830
962: \bibitem[Cohen et al. 1996]{cohen96} Cohen R. D., Beaver E. A., Diplas A. et al., 
963: 	1996, ApJ 456, 132
964: \bibitem[Cohen et al. 1999]{cohen99} Cohen R. D., Burbidge E. M., 
965: 	Junkkarinen V. T., Lyons R. W. \& Madejski G., 1999, BAAS 194.7101
966: \bibitem[Cohen 2001]{cohen2001} Cohen J. G., 2001, AJ 121, 2895
967: \bibitem[Colgan et al. 1988]{colgan88} Colgan S. W. J., Salpeter E. E. \& Terzian Y., 
968: 	1988, ApJ 328, 275
969: \bibitem[Condon et al. 1998]{condon98} Condon J. J., Cotton W. D., Greisen E. W. et al., 
970: 	1998, AJ 115, 1693
971: \bibitem[Chengalur \& Kanekar 1999]{chengalur99} Chengalur J. N. \& Kanekar N., 
972: 	1999, MNRAS 302, L29
973: \bibitem[Chengalur \& Kanekar 2000]{chengalur2000} Chengalur J. N. \& Kanekar N., 
974: 	2000, MNRAS 318, 303
975: \bibitem[Dallacasa et al. 1998]{dallacasa98} Dallacasa D., Bondi M., Alef W. \&
976:  	Mantovani F., 1998, A\&AS 129, 219
977: \bibitem[Davis \& May 1978]{davis78} Davis M. M. \& May L. S., 1978, ApJ 219, 1
978: \bibitem[Dickey \& Lockman 1990]{dickey90} Dickey J. \& Lockman F. J., 1990,
979: 	 ARA\&A 28, 215
980: \bibitem[Ellison et al. 2001a]{ellison2001} Ellison S. L., Yan L., Hook I. M. et al., 
981: 	2001a, A\&A 379, 393
982: \bibitem[Ellison et al. 2001b]{ellison2001b} Ellison S. L., Pettini M., Steidel 
983: 	C. C. \& Shapley A. E., 2001b, ApJ 549, 770
984: \bibitem[Foltz et al. 1988]{foltz88} Foltz C. B., Chaffee Jr. F. H. \&
985: 	Black J. H., 1988, ApJ 324, 267
986: \bibitem[Fomalont et al. 2000]{fomalont00} Fomalont E. B., 2000, ApJS 131, 95
987: \bibitem[Garrett et al. 1994]{garrett94} Garrett M. A., Calder R. J., Porcas R. W. et 
988: 	al., 1994, MNRAS 270, 457
989: \bibitem[Gehan 1965]{gehan65} Gehan E. A., 1965, Biometrika 52, 203
990: \bibitem[Griffith et al. 1994]{griffith94} Griffith M. R., Wright A. E., Burke B. F. \&
991: 	Ekers R. D., 1994, ApJS 90, 179
992: \bibitem[Gurvits et al. 1994]{gurvits94} Gurvits L. I., Schilizzi R. T., Barthel P. D. 
993: 	et al, 1994, A\&A 291, 737
994: \bibitem[Haehnelt et al. 1998]{haehnelt98} Haehnelt M. G., Steinmetz M. \& Rauch M., 
995: 	1998, ApJ 495, 64
996: \bibitem[Hirabayashi et al. 2000]{hirabayashi00} Hirabayashi H., Fomalont E. B., 
997: 	Horiuchi S. et al., 2000, PASJ 52, 997
998: \bibitem[Kanekar \& Chengalur 1997]{kanekar97} Kanekar N. \& Chengalur J. N., 
999: 	1997, MNRAS 292, 831
1000: \bibitem[Kanekar et al. 2001a]{kanekar2001a} Kanekar N., Chengalur J. N., 
1001: 	Subrahmanyan R. \& Petitjean P., 2001a, A\&A 367, 46
1002: \bibitem[Kanekar et al. 2001b]{kanekar2001b} Kanekar N., Ghosh T. \& Chengalur J. N., 
1003: 	2001b, A\&A 373, 394
1004: \bibitem[Kanekar \& Chengalur 2001]{kanekar2001c} Kanekar N. \& Chengalur J. N., 2001, 
1005: 	A\&A 369, 42
1006: \bibitem[Kanekar et al. 2002]{kanekar2002a} Kanekar N., Athreya R. \& Chengalur J. N., 
1007: 	2002, A\&A 382, 838
1008: \bibitem[Kanekar \& Chengalur 2002]{kanekar2002b} Kanekar N. \& Chengalur J. N., 2002, in 
1009: 	preparation
1010: \bibitem[Lane et al. 1998]{lane98} Lane  W., Smette A., Briggs F. H. et al., 1998, 
1011: 	AJ 116, 26.
1012: \bibitem[Lane et al. 2000]{lane2000} Lane  W., Briggs F. H. \& Smette A., 2000, ApJ 532, 146
1013: \bibitem[Lane \& Briggs 2002]{lane2002} Lane  W. \& Briggs F. H., 2002, ApJ 561, L27
1014: \bibitem[Lanzetta et al. 1995]{lanzetta95} Lanzetta K. M., Wolfe A. M. \& 
1015: 	Turnshek D. A., 1995, ApJ 440, 435
1016: \bibitem[Ledoux et al. 2002]{ledoux2002} Ledoux C., Bergeron J. \& Petitjean P., 
1017: 	2002, A\&A 385, 802
1018: \bibitem[Liszt 2002]{liszt2002} Liszt H., 2002, A\&A 389, 393
1019: \bibitem[McDonald \& Miralda-Escude 1999]{mcdonald99} McDonald P. \& Miralda-Escude J., 
1020: 	1999, ApJ 519, 486
1021: \bibitem[Miller 1981]{miller81} Miller Jr., R. G., 1981, Survival Analysis (John 
1022: 	Wiley \& Sons)
1023: \bibitem[Morton et al. 1980]{morton80} Morton D. C., Jian-sheng C., Wright A. E., 
1024: 	Peterson B. A. \& Jauncey D. L., 1980, MNRAS 193, 399
1025: \bibitem[Nestor et al. 2001]{nestor2001} Nestor D. B., Rao S. M., Turnshek D. A. 
1026: 	et al., 2001, in ``Extragalactic Gas at Low Redshift'', J. S. Mulchaey 
1027: 	\& J. T. Stocke (eds.), ASP Conference Series 254, 34
1028: \bibitem[Payne et al. 1982]{payne82} Payne H. E., Salpeter E. E. \& Terzian Y., 
1029: 	1982, ApJS 48, 199
1030: \bibitem[Perlman et al. 2002]{perlman02} Perlman E. S., Stocke J. T., Carilli C. L.
1031: 	et al., 2002, AJ, 124, 2401
1032: \bibitem[Petitjean et al. 1996]{petitjean96} Petitjean P., Th\'{e}odore
1033: 	B., Smette A. \& Lespine Y., 1996, A\&A 313, L25
1034: \bibitem[Petitjean et al. 2000]{petitjean2000} Petitjean P., Srianand R. \& Ledoux C., 
1035: 	2000, A\&A 364, L26
1036: \bibitem[Pettini et al. 1994]{pettini94} Pettini M., Smith L. J., Hunstead R. W. \& 
1037: 	King D. L., 1994, ApJ 426, 79
1038: \bibitem[Pettini et al. 1997]{pettini97} Pettini M., King D. L., Smith L. J. \&
1039: 	Hunstead R. W., 1997, ApJ 478, 536
1040: \bibitem[Prochaska \& Wolfe 1997]{pw97} Prochaska J. X. \& Wolfe A. M., 1997, ApJ 487, 73
1041: \bibitem[Prochaska \& Wolfe 1999]{pw99} Prochaska J. X. \& Wolfe A. M., 1999, ApJS 121, 369
1042: \bibitem[Prochaska et al. 2001]{prochaska01} Prochaska J. X., Wolfe A. M., Tytler D. 
1043: 	et al., 2001, ApJS 137, 21
1044: \bibitem[Rao \& Turnshek 1998]{rao98} Rao S. M. \& Turnshek D. A., 1998, ApJ
1045: 	500, L115
1046: \bibitem[Rao \& Turnshek 2000]{rao2000} Rao S. M. \& Turnshek D. A., 2000, ApJS 130, 1
1047: \bibitem[Roberts et al. 1976]{roberts76} Roberts M. S., Brown R. L., Brundage W. D. 
1048: 	et al., 1976, AJ 81, 293.
1049: \bibitem[Rohlfs 1986]{rohlfs86} Rohlfs K., 1986, Tools of Radio
1050:         Astronomy (Springer-Verlag, Berlin Heidelberg)
1051: \bibitem[Saikia et al. 1990]{saikia90} Saikia D. J., Junor W., Cornwell T. J., 
1052: 	Muxlow T. W. B. \& Shastri P., 1990, MNRAS 245, 408
1053: \bibitem[Schaye 2001]{schaye01} Schaye J., ApJ 559, L1
1054: \bibitem[Spaans \& Carollo 1997]{spaans97a} Spaans M. \& Carollo C. M., 1997, ApJ 482, L3
1055: \bibitem[Spaans \& Norman 1997]{spaans97b} Spaans M. \& Norman C. A., 1997, ApJ 483, 87
1056: \bibitem[Steidel et al. 1994]{steidel94} Steidel C. C., Pettini M., Dickinson M. \&
1057: 	Persson S. E., 1994, ApJ 437, 75
1058: \bibitem[Steidel et al. 1995]{steidel95} Steidel C. C., Bowen D. V., Blades J. C. \& 
1059: 	Dickson M., 1995, ApJ 440, L45
1060: \bibitem[Stocke et al. 1992]{stocke92} Stocke J. T., Wurtz R., Wang Q., Elston R. \&
1061:   	Jannuzi B. T., 1992, ApJ 400, L17
1062: \bibitem[Turnshek \& Rao 2002]{turnshek2002} Turnshek D. A. \& Rao S. M., 2002, 
1063: 	ApJ 572, L7
1064: \bibitem[Vladilo et al. 1997]{vladilo97} Vladilo G., Centuri\'{o}n M.,
1065:         Falomo R. \& Molaro P., 1997, A\&A 327, 47
1066: \bibitem[White \& Becker 1992]{white92} White R. L. \& Becker  R. H., 1992, ApJS
1067: 	79, 331
1068: \bibitem[White et al. 1993]{white93} White R. L., Kinney A. L. \& Becker R. H., 
1069: 	1993, ApJ 407, 456
1070: \bibitem[Wilkinson et al. 1998]{wilkinson98} Wilkinson P. N., Browne I. W. A., 
1071: 	Patnaik A. R., Wrobel J. M. \& Sorathia B., 1998, MNRAS 300, 790
1072: \bibitem[Wolfe \& Davis 1979]{wolfe79} Wolfe A. M. \&  Davis M. M., 1979, AJ 84, 699
1073: \bibitem[Wolfe et al. 1981]{wolfe81} Wolfe A. M., Briggs F. H. \& Jauncey D. L., 
1074: 	1981, ApJ 248, 460
1075: \bibitem[Wolfe et al. 1985]{wolfe85} Wolfe A. M., Briggs F. H., Turnshek D. A. et al., 
1076: 	1985, ApJ 294, L67
1077: \bibitem[Wolfe et al. 1986]{wolfe86} Wolfe A. M., Turnshek D. A., Smith H. E. \& 
1078: 	Cohen R. D., 1986, ApJS 61, 249.
1079: \bibitem[Wolfire et al. 1995]{wolfire95} Wolfire M. G., Hollenbach D., McKee C. F., 
1080: 	Tielens A. G. G. M. \& Bakes E. L. O., 1995, ApJ 443, 152.
1081: \bibitem[Wright \& Otrupcek 1990]{wright90} Wright A. \& Otrupcek R. (eds.), 1990, 
1082: 	Parkes catalogue, Australia Telescope National Facility
1083: \bibitem[Young \& Lo 1997]{young97} Young L. M. \& Lo K. Y., 1997, ApJ 490, 710
1084: \end{thebibliography}
1085: 
1086: 
1087: \end{document}
1088: