1: %%\documentclass[manuscript]{aastex}

2: %\documentclass[12pt,preprint]{aastex}  % This is needed for e-submission

3: 

4: \documentclass{article}

5: \usepackage{emulateapj}

6: \usepackage{apjfonts}

7: \usepackage{graphics}

8: 

9: \newenvironment{inlinefigure}{%

10: \def\@captype{figure}%

11: \noindent\begin{minipage}{0.999\linewidth}\begin{center}}

12: {\end{center}\end{minipage}\smallskip}

13: 

14: \newlength{\colwidth}

15: \setlength{\colwidth}{\textwidth}

16: \addtolength{\colwidth}{-\columnsep}

17: \setlength{\colwidth}{0.5\colwidth}

18: 

19: \newcommand{\cm}{\rm cm} 

20: \newcommand{\s}{{\rm s}}

21: \newcommand{\kms}{{\rm km}\,{\rm s}^{-1}}

22: \newcommand{\K}{{\rm K}}

23: \newcommand{\pc}{{\rm pc}}

24: \newcommand{\kpc}{{\rm kpc}}

25: \newcommand{\erg}{{\rm erg}}

26: \newcommand{\Msun}{{{\rm M}_\odot}}

27: 

28: \newcommand{\HI}{\ion{H}{1}} 

29: \newcommand{\HeII}{\ion{He}{2}} 

30: \newcommand{\Zn}{\rm Zn}

31: \renewcommand{\H}{\rm H}

32: 

33: \newcommand{\lya}{Ly$\alpha$} 

34: \newcommand{\lyb}{Ly$\beta$}

35: 

36: \lefthead{Schaye}

37: \righthead{A physical limit on the \HI\ column density}

38: 

39: %\shorttitle{A physical limit on the \HI\ column density}

40: %\shortauthors{Schaye}

41: 

42: \begin{document}

43: 

44: \submitted{Accepted for publication in the Astrophysical Journal Letters}

45: 

46: \title{A physical upper limit on the \HI\ column density of gas clouds} 

47: \author{Joop~Schaye}

48: \affil{School of Natural Sciences, Institute for Advanced

49: Study, Einstein Drive, Princeton NJ 08540, schaye@ias.edu}

50: 

51: \begin{abstract}

52: An intriguing fact about cosmic gas clouds is that they all appear to

53: have neutral (atomic) hydrogen column densities smaller than

54: $10^{22}~\cm^{-2}$. Observations of damped \lya\ (DLA) absorption

55: systems further indicate that the maximum $N_{HI}$ decreases with

56: increasing metallicity. It is generally assumed that this trend is due

57: to a dust-induced selection bias: DLA systems with high $N_{HI}$ and

58: high metallicity contain so much dust that the background QSO becomes

59: too dim to be included in optically selected surveys. Here, it is

60: argued that this explanation may not be viable. Instead, it is

61: proposed that conversion to molecular hydrogen determines the maximum

62: \HI\ column density. Molecular hydrogen forms on the surface of dust

63: grains and is destroyed by photodissociation. Therefore, the molecular

64: fraction correlates with both the dust content and, because of

65: self-shielding, the total hydrogen column density, and anticorrelates

66: with the intensity of the incident UV-radiation. It is shown that the

67: first relation can account for the observed anticorrelation between

68: the maximum $N_{HI}$ and metallicity.

69: \end{abstract}

70: 

71: \keywords{galaxies: formation --- galaxies: ISM --- intergalactic

72: medium ---  ISM: clouds --- ISM: molecules --- quasars: absorption

73: lines}

74: 

75: \section{Introduction}

76: Hydrogen is the most abundant element in the universe and consequently

77: the total hydrogen content is one of the most basic characteristics of

78: gas clouds. Hydrogen is most easily observed in atomic, neutral form

79: through the \lya\ (1216~\AA) or the 21cm transition, either in

80: absorption or emission. Because the line-of-sight extent of a gas

81: cloud is difficult to measure and because the transverse size is

82: generally unknown in case of absorption studies, it is not the mass,

83: but the column density that is the main observable quantity

84: characterizing the total gas content of a cloud. Perhaps the most

85: basic observation about the distribution of neutral hydrogen column

86: densities is that clouds with $N_{HI} > 10^{22}~\cm^{-2}$ ($ >

87: 80~\Msun\,\pc^{-2}$) appear to be extremely rare, if they exist at

88: all.\footnote{In some extreme environments, such as broad absorption

89: line QSOs, hydrogen columns in excess of $10^{22}~\cm^{-2}$ have been

90: inferred from soft X-ray absorption measurements (e.g., Mathur, Elvis,

91: \& Singh 1995). However, these observations cannot distinguish between

92: atomic and molecular hydrogen.}

93: 

94: A well-known, possible explanation for the observed cut-off in the

95: distribution of absorption line column densities is selection bias: if

96: the dust-to-gas ratio is roughly constant, then higher column density

97: clouds contain more dust, leading to a stronger extinction of

98: background sources. Thus, the presence of dust may cause sight lines

99: through high column density clouds to be missing from magnitude

100: limited surveys (e.g., Ostriker \& Heisler 1984; Wright 1990; Fall \&

101: Pei 1993). If dust-bias is important, then there should be an

102: anticorrelation between the maximum column density and the dust-to-gas

103: ratio (and hence metallicity $Z$) of gas clouds. Damped Ly$\alpha$ (DLA)

104: systems, i.e., absorbers with $N_{HI} > 2\times 10^{20}~\cm^{-2}$, do

105: indeed show such an anticorrelation: 

106: Boiss\'e et al.\ (1998; see also Prantzos \& Boissier 2000) found that

107: observed DLA systems satisfy $[\Zn/\H] + \log(N_{HI}) < 20.5$ and

108: interpreted this as a dust-induced selection effect.

109: 

110: The presence of dust in DLA systems has been established using two

111: independent types of observations. First, it has been shown that QSOs

112: with DLA systems in the foreground appear redder than those without

113: DLA systems in the foreground (Fall, Pei, \& McMahon 1989).

114: Second, the relative abundances of refractory

115: elements, such as chromium and iron, are significantly lower than

116: those of elements such as zinc, which are thought to be only lightly

117: depleted on dust grains (e.g., Pettini et al.\ 1997a).

118: 

119: Although dust seems to be present in at least a subset of DLA systems,

120: the idea that dust-bias can explain the absence of high $N_{HI}$, high

121: $Z$ systems faces at least three potential problems. First, 21cm

122: emission line studies of nearby galaxies do not reveal higher \HI\

123: column densities than are found in absorption line studies. Radial

124: profiles of the neutral hydrogen surface density of disk galaxies

125: generally show a maximum value of $\max(N_{HI}) \la

126: 10^{21}~\cm^{-2}$ ($\la 8~\Msun\,\pc^{-2}$) and always less than

127: $10^{22}~\cm^{-2}$ (e.g., Cayatte et al.\ 1994; Rhee \& van Albada

128: 1996). Second, preliminary results from the survey of

129: radio-selected QSOs by Ellison et 

130: al.\ (2001), which is free from the dust-bias that affects optically

131: selected samples, do not provide evidence for a previously unrecognized

132: population of $N_{HI} > 10^{21}~\cm^{-2}$ absorbers.  Third, the

133: metallicities and dust-to-gas ratios of DLA systems are typically more

134: than an order of magnitude below the Galactic values (e.g., Pettini et

135: al.\ 1997a, 1997b) and Prochaska \& Wolfe (2001) have

136: argued that the implied extinction corrections are far too small to

137: explain the obscuration threshold proposed by Boiss\'e et al.

138: 

139: In this letter, a simple physical explanation for the cut-off in the

140: distribution of \HI\ column densities is proposed.  Although the total

141: hydrogen column density of a gas cloud will always increase as its

142: density increases, the same is not true for the neutral hydrogen

143: column density. As the (column) density increases, the fraction of

144: hydrogen in molecular form increases and eventually the \HI\ column

145: density will stop increasing.  Furthermore, the cross section for DLA

146: absorption provided by self-gravitating clouds with high molecular

147: fractions is low because such clouds are compact and unstable to star

148: formation.  Because hydrogen molecules are formed on the surface of

149: dust grains, the molecular fraction correlates with the dust-to-gas

150: ratio (and thus metallicity). Hence, the conversion of \HI\ to $\H_2$

151: naturally explains the observed anticorrelation between the maximum

152: \HI\ column density and metallicity. In the remainder of this letter

153: it will be shown that this idea works quantitatively.

154: 

155: Earlier, related work on the conditions required for the conversion of

156: \HI\ to $\H_2$ includes Hollenbach, Werner, \& Salpeter (1971),

157: Federman, Glassgold, \& Kwan (1979) and Franco \& Cox (1986).

158: 

159: \section{Method}

160: \label{sec:method}

161: In this section I will describe and discuss the method used to derive

162: the maximum neutral hydrogen column density as a function of

163: metallicity for a given dust-to-metals ratio and incident radiation

164: field.

165: 

166: Consider a sight line through a gas cloud of arbitrary shape. Let

167: $n_H$ and $L$ be the characteristic density and size of the

168: absorber. For now, let us assume that the cloud is 

169: self-gravitating, i.e., the pressure of the medium external to the

170: cloud is low compared to its central pressure, and supported by

171: thermal pressure. Such clouds will generally be close to local

172: hydrostatic equilibrium, i.e., the characteristic size will be close

173: to the local Jeans length $L_J$.  If $L \gg L_J$, then the cloud will

174: expand or evaporate and equilibrium will be restored on a sound

175: crossing timescale. If $L \ll L_J$, then the cloud is Jeans unstable

176: and will fragment or shock to the virial temperature, equilibrium will

177: then be restored on a dynamical timescale. Schaye (2001a) used this

178: argument to derive the properties of \lya\ forest absorbers, to

179: explain the shape of their column density distribution function, and

180: to compute their contribution to the cosmic baryon density. The

181: results agreed very well with both observations and hydrodynamical

182: simulations. Because the densities of interest here are much higher

183: than those of \lya\ forest absorbers, the timescales for the

184: restoration of hydrostatic equilibrium are much shorter and hence one

185: would expect that the same argument works at least as well for

186: self-gravitating DLA systems as for \lya\ forest absorbers.

187: 

188: In local hydrostatic equilibrium, the total hydrogen column density is

189: given by the following expression (Schaye 2001a)\footnote{This

190: expression differs by a factor $\pi^{1/2}$ from equation A5 of

191: Schaye (2001a), who performed a purely dimensional analysis. Since

192: this factor is usually included in the definition of the Jeans length,

193: I include it here to obtain a more conservative upper limit on the

194: maximum column density.}:

195: \begin{equation}

196: N_{H,J} \equiv n_H L_J 

197: = \left ({\pi\gamma k \over \mu m_H^2 G}\right )^{1/2} 

198: (1-Y)^{1/2} f_g^{1/2} n_H^{1/2} T^{1/2},

199: \label{eq:NJ}

200: \end{equation}

201: where $\gamma = 5/3$ is the ratio of specific heats for a monatomic

202: gas, $Y=0.24$ is the baryonic mass fraction in helium, $f_g$ is the

203: fraction of mass in gas (excluding stars) and the other symbols have

204: their usual meanings. To be conservative, $f_g$ will be set equal to

205: unity. 

206: 

207: For a given metallicity, dust-to-metals ratio and incident radiation

208: field, the neutral hydrogen column density is computed as a function

209: of the density as follows. First, a grid of $(N_H,n_H)$ models is

210: computed using the publicly available photoionization package

211: CLOUDY\footnote{\texttt{http://www.pa.uky.edu/$\sim$gary/cloudy/}}

212: (version 94; see Ferland 2000 for details), modeling the absorbers as

213: slabs of constant density illuminated from two

214: sides\footnote{Illumination from two sides is approximated by doubling

215: all column densities of a plane parallel cloud illuminated from one

216: side.}. The temperature is \emph{not} a free parameter. For each model

217: the thermal equilibrium temperature, the mean molecular weight $\mu$,

218: and the neutral hydrogen column density are computed self

219: consistently. Second, the solutions

220: $(N_H,n_H,T(N_H,n_H),\mu(N_H,n_H))$ are selected for which

221: equation~(1) is satisfied (i.e., $N_H = N_{H,J}$) and which are stable

222: ($dP/dn_H > 0$). Third, from these solutions the maximum $N_{HI}$ is

223: determined. The whole procedure is then repeated for different

224: metallicities to derive the maximum neutral hydrogen column density as

225: a function of the metallicity.

226: 

227: Figure~1 illustrates the results for a model with metallicity 0.1

228: solar and a dust-to-metals ratio of 0.5 times the Galactic value,

229: values typical for DLA systems (e.g., Pettini et al.\ 1997ab; Vladilo

230: 1998). The dust was assumed to have ISM properties and consists of a

231: mixture of graphites and silicates (see the CLOUDY documentation for

232: details). The gas clouds were exposed to the model for the UV/X-ray

233: background radiation at $z=3$ of Haardt \& Madau (2001)\footnote{The

234: data and a description of the input parameters can be found at

235: \texttt{http://pitto.mib.infn.it/$\sim$haardt/refmodel.html}. The

236: cloud is also exposed to the cosmic microwave background and to a

237: cosmic ray density of $2\times 10^{-9}~\cm^{-3}$, but these have no

238: significant effect on the results.}, which includes contributions from

239: QSOs and galaxies and yields \HI\ and \HeII\ photoionization rates

240: of $1.15\times 10^{-12}~\s^{-1}$ and $1.96\times 10^{-14}~\s^{-1}$

241: respectively.

242: \begin{inlinefigure}

243: \centerline{\resizebox{0.96\colwidth}{!}{\includegraphics{f1.eps}}}

244: \figcaption[f1.eps]{Hydrogen column densities as a function of

245: density. The model assumes local hydrostatic and thermal equilibrium,

246: metallicity $Z=0.1 Z_\odot$, dust-to-metal ratio 0.5 times the

247: Galactic value, and the model for the $z=3$ UV/X-ray background

248: radiation of Haardt \& Madau (2001).  There is a gap around $n_H \sim

249: 10^{-1}~{\rm cm}^{-2}$ because there are no stable solutions in this

250: region.}

251: \end{inlinefigure}

252: 

253: \begin{figure*}[t]

254: \begin{center} 

255: \resizebox{0.96\textwidth}{!}{\includegraphics{f2.eps}}

256: %\resizebox{0.8\textwidth}{!}{\includegraphics{f2.eps}}

257: \figcaption[f2.eps]{Metallicity as a function of neutral hydrogen

258: column density. Model curves indicate the maximum $N_{HI}$ (solid

259: lines) and the H\,I column density for which the molecular

260: fraction is 10 percent (dashed lines). Panel (a): From top to bottom

261: the curves correspond to a dust-to-metals ratio of 0.1, 0.5 and 1.0

262: times the Galactic value respectively and the $z=3$ UV/X-ray

263: background of Haardt \& Madau (2001). Panel (b): Curves are for a

264: dust-to-metals ratio of half the Galactic value and the UV/X-ray

265: background used in panel (a) multiplied by, from top to bottom,

266: factors of 3, 1, and 1/3 respectively.  Data points are the observed

267: $z\ge 1$ DLA systems taken from Pettini et al.\ (1997b; 2000),

268: Prochaska \& Wolfe (1999), de la Varga et al.\ (2000), and Molaro et

269: al.\ (2000). The 1-$\sigma$ errors are typically about 0.1~dex for

270: both $N_{HI}$ and [Zn/H].} 

271: \end{center}

272: \end{figure*}

273: 

274: Note that the curves in Fig.~1 do not show the evolution of an

275: individual cloud. Each point along the curves corresponds to a cloud

276: in local hydrostatic and thermal equilibrium, and higher densities

277: correspond to lower cloud masses. If a cloud accretes matter, it will

278: become Jeans unstable and fragment into pieces of higher $N_H$ and

279: lower mass which will again be close to the equilibrium curve.  The

280: model clouds become self-shielded at $n_H \sim

281: 10^{-2}~\cm^{-3}$. There is a gap at $n_H\sim 10^{-1}~\cm^{-3}$

282: because clouds in this range are unstable ($dP/dn_H < 0$).  At lower

283: densities the clouds are warm ($T\sim 10^4~\K$) and extended ($L\ga

284: 10^3~\pc$), while clouds with higher densities are colder ($T \la

285: 10^3~\K$) and more compact ($L\la 10^2~\pc$). For $n_H\ga 10~\cm^{-3}$

286: the clouds are highly molecular and clouds with higher densities have

287: smaller \HI\ column densities. The physical properties of

288: self-shielded, self-gravitating gas clouds will be discussed in more

289: detail elsewhere.

290: 

291: The method described above assumes that the absorbers are

292: self-gravitating, an assumption that may not hold for DLA systems.

293: Indeed, it is well known that the interstellar medium of the Galaxy

294: has a multiphase structure and that \HI\ clouds are generally pressure

295: confined, although molecular clouds are not. Furthermore, it was

296: recently argued by Schaye (2001a) that at least a subset of DLA

297: systems appear to arise in galactic winds, in which case (ram)

298: pressure is likely to be more important than

299: self-gravity. Fortunately, the possibility that some (or even most)

300: DLA systems are confined by external pressure does not affect the

301: current analysis because a gas cloud with $L < L_J$ will always have a

302: smaller column density than a gas cloud in hydrostatic equilibrium

303: \emph{with the same density}. Pressure-confined solutions will

304: therefore lie below the solid curve in Fig.~1 and the assumption of

305: local hydrostatic equilibrium is thus conservative when estimating the

306: maximum $N_{HI}$. Recall that self-gravitating clouds containing stars

307: and/or dark matter also have lower column densities.

308: 

309: Higher neutral column densities are possible for clouds that are

310: self-gravitating and for which turbulent or magnetic pressure

311: dominates over thermal pressure, and for clouds that are rotationally

312: supported $\emph{and}$ oriented such that our line of sight is nearly

313: perpendicular to their spin axes. One would, however, expect these

314: caveats to be important mostly for clouds in regions of ongoing star

315: formation. In such regions the ISM is likely to be multiphase and the

316: \HI\ clouds pressure-confined, thus reducing their column densities

317: compared to clouds in hydrostatic equilibrium.  Finally, it should be

318: noted that although the integrated column density can be very large

319: for a sight line through an edge-on disk galaxy, this column will

320: generally be the sum of the column densities of a large number of

321: clouds.

322: 

323: \section{Results}

324: 

325: Figure~2a shows the observed [Zn/H] as a function of $N_{HI}$ for DLA

326: systems with redshift $z\ge 1$. The absence of points in the upper

327: right part of the diagram, indicating a metallicity-dependent $N_{HI}$

328: cut-off, was first noted by Boiss\'e et al.\

329: (1998) who interpreted it as a dust-induced selection bias.  The lack

330: of data points in the bottom left part of the figure is most likely a

331: selection effect: the corresponding zinc lines are too weak to be

332: detected. Indeed, while higher quality observations of [Si/H] and

333: [Fe/H] confirm the absence of high $N_{HI}$, high $Z$ systems, they do

334: not show a deficiency of low $N_{HI}$, low $Z$ systems (Prochaska \&

335: Wolfe 2001).  

336: 

337: The solid curves indicate the neutral hydrogen column

338: density in local hydrostatic and thermal equilibrium, computed using

339: the method described in section~\ref{sec:method}. From top to bottom

340: the curves correspond to dust-to-metals ratios of 0.1, 0.5 and 1.0

341: times the Galactic value respectively. Equilibrium solutions for

342: pressure-confined gas clouds and self-gravitating clouds that contain

343: a gravitationally significant amount of stars or dark matter are all

344: located to the left of these curves. 

345: From the figure it is clear that the models can naturally account for

346: the absence of high $N_{HI}$, high $Z$ systems. A higher metallicity

347: (dust-to-gas ratio) results in an increased $H_2$ formation rate and,

348: because of the increased cooling via metals and $H_2$, a lower

349: temperature. Both the higher $H_2$ formation rate and, via equation 1,

350: the lower temperature contribute to the decrease in the maximum

351: $N_{HI}$.

352: 

353: It is important to note that the maximum column densities are derived

354: for single clouds, i.e., regions along the sight line over

355: which the density is of the same order as the nearest (local) maximum,

356: whereas the data points correspond to the integrated column densities

357: of systems of clouds. Since essentially all DLA systems show

358: substructure in their low-ionization metal lines, which are thought to

359: trace the neutral hydrogen density, DLA systems must generally consist

360: of collections of clouds. Unfortunately, the strength of the DLA line

361: prohibits us from measuring the \HI\ columns of the individual

362: components. Hence, each observed $N_{HI}$ value must be considered as

363: an upper limit to the column density of the dominant component. If,

364: for example, the strongest component would account for 50 percent of

365: the neutral hydrogen column (and if all components had roughly the

366: same metallicity), then the data point should be shifted by $-0.3$ dex

367: along the $N_{HI}$-axis, which would place it comfortably to the left

368: of most models.

369: 

370: Although the models clearly show that there should be an

371: anticorrelation between metallicity and the maximum $N_{HI}$, the

372: exact $N_{HI}$ values are somewhat uncertain. First, although

373: self-gravitating clouds will generally not be far from local

374: hydrostatic equilibrium, there is no reason why they should be in

375: exact equilibrium. Second, the derived column densities are sensitive

376: to the assumed intensity of the incident ionizing radiation. Figure~2b

377: illustrates the effect of changing the amplitude of the UV/X-ray

378: background. From top to bottom the curves correspond to models that

379: use the same background radiation as was used in panel (A), but with

380: the amplitudes multiplied by factors of 3, 1, and 1/3 respectively.

381: The maximum $N_{HI}$ increases if the background is

382: stronger. Although these models are indicative of the uncertainty in

383: the mean background radiation, the ionizing radiation could be

384: stronger around DLA systems if they are located near regions

385: of star formation or QSOs. In such systems higher $N_{HI}$ values

386: would be possible, although it should be noted that absorbers in these

387: environments would likely be pressure confined.

388: 

389: Extrapolation of the solid curves in Fig.~2 suggests that large

390: $N_{HI}$ values are possible at zero metallicity. The reason is that

391: for the low-$Z$ models the $N_{HI}(n_H)$ curves flatten off

392: very slowly with increasing $n_H$. The maximum $N_{HI}$ is only

393: reached when the density is higher than is typical of molecular clouds

394: and when the characteristic mass of the cloud is lower than is typical

395: of stars. The models used here are probably inadequate for systems

396: which are nearly fully molecular. In any case, such systems are so

397: compact that the cross section for interception by a random sight line

398: is very small and they are highly unstable to star formation. Note

399: also that the extreme low metallicity models are not relevant for DLA

400: systems which all have $Z \gg 10^{-3}~Z_\odot$. Nevertheless, it is

401: instructive to look at a more robust quantity than the maximum

402: $N_{HI}$, such as the $N_{HI}$ corresponding to a fixed

403: molecular fraction. The dashed curves in Fig.~2 show the $N_{HI}$

404: at which the molecular fraction reaches 10

405: percent, much higher than is typical of DLA systems (e.g., Petitjean,

406: Srianand, \& Ledoux 2001). These curves do turn over at low $Z$. For example,

407: $\log N_{HI} = 21.8$ at zero metallicity for the ionizing background

408: used in Fig.~2a. Because the molecular fraction rises rather sharply

409: to $f(\H_2) \sim 10^{-1}$ (see Fig.~1) due to a thermal

410: instability, the curves corresponding to $f(\H_2)\sim

411: 10^{-5}$ (not plotted) are similar to the dashed curves in Fig.~2. 

412: 

413: Finally, the models cannot explain the apparent lack of $N_{HI} <

414: 10^{21}~\cm^{-2}$ systems with $Z > Z_\odot$. If

415: DLA systems with such high metallicities do occur in nature and if

416: their cross sections and lifetimes are non-negligible, then it may be

417: that a dust-induced selection bias still needs to be invoked to

418: account for the absence of such systems from current samples.

419: 

420: \section{Conclusions}

421: 

422: One of the most basic observational findings is that all gas clouds

423: seem to have neutral hydrogen column densities smaller than $10^{22}$

424: atoms per square centimeter. Studies of DLA systems have revealed that

425: the maximum $N_{HI}$ decreases with increasing

426: metallicity. It was argued that dust-induced selection bias may not

427: be a viable explanation for these observations.

428: Instead, it was proposed that clouds with $N_{HI} > 10^{22}~\cm^{-2}$ do

429: not occur because the clouds turn molecular before reaching such high

430: column densities. Furthermore, because clouds with high molecular

431: fractions are much more compact and short-lived than clouds with low

432: molecular fractions, the latter provide a much larger cross section

433: per unit mass for DLA absorption.  The maximum \HI\ column density is

434: a decreasing function of metallicity, mainly because the formation

435: rate of molecular hydrogen increases with the dust content of the

436: system.

437: 

438: It was shown that models of self-gravitating gas clouds in local

439: hydrostatic and thermal equilibrium, with a dust-to-metals ratio

440: somewhat lower than the Galactic value, exposed to a model of the

441: $z\sim 3$ UV/X-ray background radiation, can roughly account for the

442: observed anticorrelation between the maximum $N_{HI}$ value and

443: metallicity.  The assumption of pure gas clouds in hydrostatic

444: equilibrium is conservative because self-gravitating clouds containing

445: stars and/or dark matter, as well as clouds that are confined by

446: external pressure, have lower column densities. The model works

447: particularly well if one takes into account that observed DLA systems

448: consist of multiple components, and that it is the integrated \HI\

449: column density that is measured, whereas the model predicts the

450: maximum $N_{HI}$ for single gas clouds.

451: 

452: Finally, the models presented here predict that the molecular fraction

453: increases with increasing $N_{H}$ and dust-to-gas ratio (and thus

454: metallicity if the dust-to-metals ratio is roughly constant), but is

455: anticorrelated with the intensity of the incident UV radiation. These

456: correlations could be uncovered observationally if a sample is

457: constructed in which two out of the three parameters

458: $(N_{H},Z,I_{UV})$ are roughly constant. There could, however, still

459: be considerable scatter in these relations if only a fraction of

460: clouds are purely gaseous and self gravitating. The models also

461: predict a strong anticorrelation between molecular fraction and

462: temperature that is fairly insensitive to the values of the other

463: parameters.

464: 

465: \acknowledgments 

466: I would like to thank M.~Pettini for providing me with a compilation

467: of zinc abundances in DLA systems. It is a pleasure to thank

468: A.~Aguirre, E.~de~Blok, M.~Fall, S.~Ellison, M.~Pettini and G. Vladilo

469: for useful suggestions and discussions. This work was supported by a

470: grant from the W.~M.~Keck foundation.

471: 

472: 

473: \begin{thebibliography}{}

474: 

475: \bibitem[Boiss\'e et al.(1998)]{boisse98:dla} 

476: Boiss\'e, P., Le Brun, V., Bergeron, J., \& Deharveng, J.\ 1998, \aap,

477: 333, 841  

478: 

479: \bibitem[Cayatte et al.(1994)]{cayatte94:hi_sb} 

480: Cayatte, V., Kotanyi, C., Balkowski, C., \& van Gorkom, J.~H.\ 1994,

481: \aj, 107, 1003

482: 

483: \bibitem[de la Varga et al.(2000)]{delavarga00:dla} 

484: de la Varga, A., Reimers, D., Tytler, D., Barlow, T., \& Burles, S.\

485: 2000, \aap, 363, 69  

486: 

487: \bibitem[Ellison et al.(2001)]{ellison01:corals}

488: Ellison, S. L., Yan, L., Hook, I. M., Pettini, M., Wall, J. V., \&

489: Shaver, P. 2001, \aap, in press (astro-ph/0109205)

490: 

491: \bibitem[Fall \& Pei(1993)]{fall93:dladust} 

492: Fall, S.~M.~\& Pei, Y.~C.\ 1993, \apj, 402, 479 

493: 

494: \bibitem[Fall, Pei, \& McMahon(1989)]{fall89:dladust} 

495: Fall, S.~M., Pei, Y.~C., \& McMahon, R.~G.\ 1989, \apjl, 341, L5 

496: 

497: \bibitem[Federman, Glassgold, \& Kwan(1979)]{federman79:h2formation} 

498: Federman, S.~R., Glassgold, A.~E., \& Kwan, J.\ 1979, \apj, 227, 466 

499: 

500: \bibitem[Ferland(2000)]{ferland00:cloudy}

501: Ferland, G. J. 2000, Rev.\ Mexicana Astron.\ Astrofis.\ Ser.\ Conf., 9, 153

502: 

503: \bibitem[Franco \& Cox(1986)]{franco86:h2formation} 

504: Franco, J.~\& Cox, D.~P.\ 1986, \pasp, 98, 1076 

505: 

506: \bibitem[Haardt \& Madau(2001)]{haardt01:cuba}

507: Haardt, F., \& Madau, P. 2001, to be published in the proceedings of

508: XXXVI Rencontres de Moriond, astro-ph/0106018

509: 

510: \bibitem[Hollenbach, Werner, \& Salpeter(1971)]{hollenbach71:h2formation} 

511: Hollenbach, D.~J., Werner, M.~W., \& Salpeter, E.~E.\ 1971, \apj, 163, 165 

512: 

513: \bibitem[Mathur, Elvis, \& Singh(1995)]{mathur95:balqso} 

514: Mathur, S., Elvis, M., \& Singh, K.~P.\ 1995, \apjl, 455, L9 

515: 

516: \bibitem[Molaro et al.(2000)]{molaro00:q0000_zn} 

517: Molaro, P., Bonifacio, 

518: P., Centuri{\' o}n, M., D'Odorico, S., Vladilo, G., Santin, P., \& Di

519: Marcantonio, P.\ 2000, \apj, 541, 54

520: 

521: \bibitem[Ostriker \& Heisler(1984)]{ostriker84:dust} 

522: Ostriker, J.~P.~\& Heisler, J.\ 1984, \apj, 278, 1 

523: 

524: \bibitem[Petitjean, Srianand, \& Ledoux(2000)]{petitjean00:h2} 

525: Petitjean, P., Srianand, R., \& Ledoux, C. 2000, A\&A, 364, L26 

526: 

527: \bibitem[Pettini et al.(1997a)]{pettini97:dust} 

528: Pettini, M., King, D.~L., Smith, L.~J., \& Hunstead, R.~W.\ 1997a, \apj, 

529: 478, 536 

530: 

531: \bibitem[Pettini et al.(1997b)]{pettini97:dla} 

532: Pettini, M., Smith, L. J., King, D. L., \& Hunstead, R. W. 1997b, ApJ, 

533: 486, 665 

534: 

535: \bibitem[Pettini et al.(2000)]{pettini00:dla} 

536: Pettini, M., Ellison, S. L., Steidel, C. C., Shapley, A. E., \& Bowen,

537: D. V. 2000, ApJ, 532, 65 

538: 

539: \bibitem[Prantzos \& Boissier(2000)]{prantzos00:dla} 

540: Prantzos, N., \& Boissier, S.\ 2000, \mnras, 315, 82 

541: 

542: \bibitem[Prochaska \& Wolfe(1999)]{prochaska99:dla_metals} 

543: Prochaska, J.~X.~\& Wolfe, A.~M.\ 1999, \apjs, 121, 369 

544: 

545: \bibitem[Prochaska \& Wolfe(2001)]{prochaska01:dla_abundances} 

546: Prochaska, J. X., \& Wolfe, A. M. 2001, ApJ, in press

547: 

548: \bibitem[Rhee \& van Albada(1996)]{rhee96:hi_sb} 

549: Rhee, M.-H.~\& van Albada, T.~S.\ 1996, \aaps, 115, 407 

550: 

551: \bibitem[Schaye(2001a)]{schaye01:lya} 

552: Schaye, J. 2001a, ApJ, 559, 507

553: 

554: \bibitem[Schaye(2001b)]{schaye01:dla} 

555: Schaye, J. 2001b, ApJ, 559, L1

556: 

557: \bibitem[Vladilo(1998)]{vladilo98:dla} 

558: Vladilo, G.\ 1998, \apj, 493, 583 

559: 

560: \bibitem[Wright(1990)]{wright00:dust} 

561: Wright, E.~L.\ 1990, \apj, 353, 411 

562: 

563: \end{thebibliography}

564: 

565: \end{document}

566: