1: \documentstyle[emulateapj,graphicx]{article}

2: %\documentstyle[12pt,aasms4,graphicx]{article}

3: %\documentstyle[12pt,aaspp4,graphicx]{article}

4: \def\etal  {{et~al.}\ }

5: \def\ni{\noindent}        %No Indent%

6: \def\ub{\underbar}

7: \def\hi{\noindent \hangindent=2.5em}

8: \def\et{{\it et\thinspace al.}}    %et al.%

9: \def\pc{{\rm\,pc}}

10: \def\kpc{{\rm\,kpc}}

11: \def\Mpc{{\rm\,Mpc}}

12: \def\mpc{{\rm\,Mpc}}

13: \def\hnot{{\rm\,km/s/Mpc}}

14: \def\kmsec{{\rm\,km/s}}

15: \def\kms{{\rm\,km/s}}

16: \def\msun{{\rm\,M_\odot}}

17: \def\lsun{{\rm\,L_\odot}}

18: \def\mdot{{\rm\,M_\odot}}

19: \def\vol#1  {{{#1}{\rm,}\ }}

20: \def\lya{{\rm Ly}$\alpha$\ }

21: \def\mytau{\tau_{{\rm Ly}\alpha}}

22: \def\hi{{H\thinspace I}\ }

23: \def\hii{{H\thinspace II}\ }

24: \def\hei{{He\thinspace I}\ }

25: \def\heii{{He\thinspace II}\ }

26: \def\heiii{{He\thinspace III}\ }

27: \def\nhi{N_{HI}}

28: \def\nheii{N_{\heii}}

29: \def\jhi{J_{\hi}}

30: \def\jhiu{J_{-21}}

31: \def\jheii{J_{\heii}}

32: \def\shi{\sigma_{\hi}(\nu)}

33: \def\shia{\bar\sigma_{\hi}}

34: \def\etal{et al.\ }

35: \def\Sec{\S }

36: \def\eqs{eqs.\ }

37: \def\cf{{cf.}\ }

38: \def\gsim{\;\rlap{\lower 2.5pt

39:  \hbox{$\sim$}}\raise 1.5pt\hbox{$>$}\;}

40: \def\lsim{\;\rlap{\lower 2.5pt

41:    \hbox{$\sim$}}\raise 1.5pt\hbox{$<$}\;}

42: \def\refset{\parindent=0pt\hangafter=1\hangindent=1em}

43: \newcount\refno

44: \refno = 1

45: \newcount\rfno

46: \rfno = 1

47: \def\eq{$^{\the\refno\ }$\advance\refno by 1}

48: \def\ad{\advance\rfno by 1}

49: 

50: \def\clock{\count0=\time \divide\count0 by 60

51:      \count1=\count0 \multiply\count1 by -60 \advance\count1 by \time

52: \number\count0:\ifnum\count1<10{0\number\count1}\else\number\count1\fi}

53: \def\draft{

54:    \rightline{DRAFT: \today \quad\quad \clock}}

55: \def\myputfigure#1#2#3#4#5%

56: {\hskip0.03\textwidth\vskip#5pt%\hfill

57: \makebox[0pt]{\hskip#2in

58: \includegraphics[width=#3\textwidth]{#1}}\vskip#4pt\hfill}

59: \newcommand{\beq}{\begin{equation}}

60: \newcommand{\eeq}{\end{equation}}

61: \makeatletter

62: \newenvironment{tablehere}

63:   {\def\@captype{table}}

64:   {}

65: \makeatother

66: 

67: 

68: \begin{document}

69: \title{Constraining Reionization with the Evolution of the \\ Luminosity

70: Function of Lyman $\alpha$ Emitting Galaxies}

71: 

72: \author{Zolt\'{a}n Haiman}

73: \affil{Department of Astronomy, Columbia University, 550 West 120th Street,

74: New York, NY 10027} \and

75: %\vspace{-0.5\baselineskip}

76: \author{Renyue Cen} \affil{Department of Astrophysical Sciences,

77: Princeton University, Peyton Hall, Ivy Lane, Princeton, NJ 08544}

78: 

79: 

80: \begin{abstract}

81: At redshifts beyond $z\gsim 6$, as the mean fraction of neutral

82: hydrogen $\langle x_{\rm HI} \rangle$ in the intergalactic medium

83: (IGM) increases, the line flux of \lya emitters can be significantly

84: suppressed, which can result in a decrease in the observed number of

85: emitters at a given \lya flux.  However, cosmological HII regions

86: surrounding the \lya emitting galaxies alleviate these effects.  We

87: use simple models of the \lya line suppression that incorporate the

88: presence of HII regions to predict the overall effect of the \lya

89: absorption on the \lya luminosity function.  We find, in agreement

90: with other recent studies, that unless ionizing sources are unusually

91: strongly clustered, a fully neutral IGM may be inconsistent with the

92: large abundance of confirmed $z=6.5$ \lya emitters.  However, the

93: presence of local HII regions prohibits placing a tight constraint on

94: the mean neutral fraction. We find $\langle x_{\rm HI} \rangle \lsim

95: 0.25$; the presence of strong winds and/or the clustering of ionizing

96: sources would further weaken this constraint. We conclude that the

97: evolution of the \lya LF is consistent with reionization occurring

98: near this redshift. Finally, we suggest that a measurement of observed

99: \lya line width as a function of the \lya luminosity, in a future,

100: larger sample of \lya emitters, may serve as a robust diagnostic of

101: the neutral fraction in the IGM.

102: \end{abstract}

103: 

104: \keywords{cosmology: theory -- cosmology: early universe -- galaxies:

105: formation -- galaxies: high-redshift -- galaxies: quasars: general --

106: galaxies: quasars: absorption lines}

107: 

108: \section{Introduction}

109: 

110: The recent detection of Gunn--Peterson (GP) troughs in the spectra of

111: the SDSS quasars (Becker et al. 2001; White et al. 2003) places a

112: relatively direct lower limit on the mean (volume averaged) neutral

113: fraction of the IGM at $z\sim 6$. While this limit is relatively weak,

114: $\langle x_{\rm HI} \rangle\gsim 10^{-3}$, the spectral imprint

115: (Mesinger \& Haiman 2004) and size (Wyithe \& Loeb 2004a) of

116: cosmological HII regions around the SDSS quasars indicate that the

117: neutral fraction around $z\sim 6$ is significantly higher, $\langle

118: x_{\rm HI} \rangle\gsim 0.1$.  The rapid evolution of $\langle x_{\rm

119: HI} \rangle$ with redshift, inferred from the spectra of several

120: quasars between $5.5\lsim z\lsim 6.5$ (Fan \etal\ 2002; Cen \&

121: McDonald 2002; Lidz et al. 2002; but see Songaila 2004), and also the

122: thermal state of the IGM at $3\lsim z \lsim 4$ (Hui \& Haiman 2003),

123: indicate and that a percolation of discrete HII regions is occurring

124: around this epoch.

125: 

126: Ly$\alpha$ emission lines from high--redshift sources can serve as

127: another probe of the ionization state of the IGM. The damping wing of

128: the GP absorption from the IGM can cause a characteristic absorption

129: feature (Miralda-Escud\'e 1998).  In a significantly neutral IGM, the

130: absorption can produce conspicuous effects, i.e. attenuating the

131: emission line, making it asymmetric, and shifting its apparent peak to

132: longer wavelengths (Haiman 2002; Santos 2004).  These effects

133: generally increase with $\langle x_{\rm HI} \rangle$, and it has been

134: suggested that a strong drop in the abundance of \lya emitters beyond

135: the reionization redshift can serve as a diagnostic of a nearly

136: neutral IGM (Haiman \& Spaans 1999).  Recently, Malhotra \& Rhoads

137: (2004; hereafter MR04) and Stern et al. (2004) carried out the first

138: application of this technique, by comparing the luminosity functions

139: (LFs) of \lya emitters at $z=5.7$ and $z=6.5$. The LF shows no

140: evolution in this range, and this has been interpreted as evidence

141: against percolation taking place near $z\sim 6$.

142: 

143: A potential caveat for this interpretation is that $z\sim 6$ galaxies

144: will be surrounded by their own local cosmological HII regions, which

145: can significantly reduce the attenuation of the \lya line flux (Cen \&

146: Haiman 2000; Madau \& Rees 2000).  It has been shown that this can

147: render \lya lines detectable even if the galaxies are embedded in a

148: fully neutral IGM (Haiman 2002; Santos 2004; Cen et al. 2004),

149: especially when the increased transmission due to the clustering of

150: ionizing sources is taken into account (Furlanetto et al. 2004; Gnedin

151: \& Prada 2004; Wyithe \& Loeb 2004b), and/or if the \lya emission has

152: a significant recession velocity with respect to the absorbing gas.

153: 

154: In this {\it Letter}, we study the impact of IGM absorption on the

155: \lya LF.  Our analysis is similar to that of the recent analysis by

156: MR04; the improvement is that we model the attenuation of Ly$\alpha$

157: lines including cosmological HII regions.  While very little

158: attenuation (by a factor of $\lsim 2$) can be tolerated by the lack of

159: evolution of the LF over the range $5.7\lsim z\lsim 6.5$, we find that

160: the presence of local HII regions allows a neutral fraction as high as

161: $\langle x_{\rm HI} \rangle=0.25$. On the other hand, $\langle x_{\rm

162: HI} \rangle\sim 1$ is allowed only if the ionizing sources are

163: unusually strongly clustered. In either case, we argue that the

164: present \lya LFs are consistent with reionization occurring near

165: $z\sim 6$.\footnote{Throughout this paper we assume a cold--dark

166: matter cosmology ($\Lambda$CDM), with ($\Omega_\Lambda$, $\Omega_M$,

167: $\Omega_b$, $H_0$) = (0.73, 0.27, 0.044, 71 km s$^{-1}$ Mpc$^{-1}$),

168: consistent with the recent results from {\it WMAP} (Spergel et

169: al. 2003).  Unless stated otherwise, all lengths are quoted in

170: comoving units.}

171: 

172: \section{The Observed Ly$\alpha$ Luminosity Function}

173: 

174: Following Malhotra \& Rhoads (2004; hereafter MR04) we take as basic

175: inputs to our analysis the \lya LFs at $z=5.7$ and $z=6.5$.  We adopt

176: the forms of the LF at these two redshifts from MR04.  Note that at

177: both redshifts, the LFs are inferred by culling several independent

178: datasets (Hu et al. 2002; Rhoads et al. 2003; Kodaira et al. 2004).

179: At $z=5.7$, the LF is comparatively well determined and well fit by a

180: Schechter function with relatively small errors in normalization

181: ($\Phi_\star$) and characteristic luminosity ($L_\star$), while the

182: faint--end slope ($\alpha$) is less well--determined.  Here we adopt

183: the best--fit values of $(\Phi_\star,L_\star,\alpha) = (10^{-4}~{\rm

184: Mpc^{-3}},10^{43}~{\rm erg~s^{-1}}, -1.5)$ (displayed by the

185: short--dashed curve in Fig.~\ref{fig:LFs} below). Within the

186: uncertainties quoted in MR04, the values of these parameters do not

187: significantly change our conclusions below.

188: 

189: \myputfigure{fig1.eps}{3.2}{0.45}{-10}{-10} \figcaption{The

190: suppression of the total Ly$\alpha$ line flux, as a function of the

191: observed line luminosity, for an IGM with $\langle x_{\rm HI} \rangle

192: = 0.1$ (or $\Gamma_{12}=0.0015$; long--dashed curve), and for and IGM

193: with $\langle x_{\rm HI} \rangle = 1.0$ (or $\Gamma_{12}=0$; solid

194: curve).

195: \label{fig:suppression}}

196: \vspace{\baselineskip}

197: 

198: At $z=6.5$, the LF is less well determined, as shown by four

199: data--points with errors in Figure~\ref{fig:LFs} below.  Of these four

200: data--points, the most constraining is from the deep Subaru field,

201: which revealed the presence of a handful of emitters at $z\sim 6.5$

202: (Kodaira et al. 2003), and has the smallest error bar.  The single,

203: faint lensed galaxy at this redshift found by Hu et al. (2002;

204: represented by the lowest--luminosity point in Fig.~\ref{fig:LFs}

205: below) implies a surprisingly high abundance of the faintest emitters,

206: but with a single source, it has a large uncertainty.  Finally, we

207: follow MR04 and conservatively assume that the intrinsic LF

208: (i.e. prior to processing of the line through the IGM) is the same at

209: $z=6.5$ as at $z=5.7$.  More realistic models, based on the

210: hierarchical growth of structures, combined with \lya line processing

211: by gas and dust interior to the galaxies, predict that the number of

212: emitters were smaller at earlier epochs (Haiman \& Spaans 1999; Le

213: Delliou et al. 2004), which would strengthen the limits on $\langle

214: x_{\rm HI}\rangle$ below.

215: 

216: \section{The Attenuation of \lya Emission Lines}

217: \label{sec:transmission}

218: 

219: We follow Cen \& Haiman (2000), and use a simple model to find the

220: attenuation of the Ly$\alpha$ line as a function of wavelength.  The

221: model is straightforward, and we only briefly recap the main

222: features. We start with a Gaussian emission line originating at

223: $z_s=6.5$, with a line--width of $\Delta v=300~{\rm km~s^{-1}}$,

224: typical of normal galaxies (narrower lines would be attenuated more

225: severely; see discussion below). We assume that the source is

226: surrounded by a spherical Str\"omgren sphere that propagated into the

227: IGM with a neutral fraction $\langle x_{\rm HI}\rangle$.  The IGM is

228: assumed to have a clumping factor

229: $C\equiv\langle\rho^2\rangle/\langle\rho\rangle^2=10$ with a

230: log-normal distribution.  We assume the presence of a uniform ionizing

231: background, characterized by an ionizing rate $\Gamma_{12}$

232: (ionizations per $10^{12}$s per atom, yielding the given value of

233: $\langle x_{\rm HI}\rangle$ in ionization equilibrium for the IGM.

234: 

235: We assume a Salpeter IMF, and relation (Kennicutt 1983) between

236: star--formation rate and \lya luminosity (SFR=$1 \msun$/yr)

237: $\rightarrow$ ($L({\rm Ly}\alpha)=10^{42}$ erg/s). We conservatively

238: assume $f_{\rm esc}$ = 1 (escape fraction of ionizing radiation) in

239: computing $R_s$, and a lifetime $t_s=10^8$ yrs for each source.  With

240: all other parameters in the model fixed, we vary $\Gamma_{12}$ (or

241: equivalently $\langle x_{\rm HI} \rangle$), and compute the factor by

242: which the total line flux is suppressed relative to the unabsorbed

243: Gaussian line. This factor is shown in Figure~\ref{fig:suppression} as

244: a function of the intrinsic luminosity of the source. Faint sources

245: are more severely attenuated: by a factor of up to $\sim 15$ for $\log

246: L($\lya$/{\rm erg~s^{-1}})=40$ [or up to $\sim 6$ for $\log

247: L($\lya$/{\rm erg~s^{-1}})=42$, the present lower limit for the

248: detection of a \lya emitter at $z=6.5$] in a neutral universe.  Bright

249: ($\log L($\lya$/{\rm erg~s^{-1}})=43$) sources, however, are only

250: attenuated by a factor of $\sim 1.5$ even if $\langle x_{\rm HI}

251: \rangle$ is as high as 0.1.

252: 

253: The attenuation factors in Figure~\ref{fig:suppression} are computed

254: by assuming that the HII regions are being created by a single

255: ionizing source (the \lya emitter itself). These HII regions (with

256: typical sizes of a few Mpc in the observed range of source

257: luminosities) may contain many other undetected galaxies, which make

258: the HII regions larger (Furlanetto et al. 2004; Gnedin \& Prada 2004;

259: Wyithe \& Loeb 2004b) and also more highly ionized.  Therefore, the

260: results shown in Figure~\ref{fig:suppression} should be viewed as

261: conservative upper limits.  These results show that the HII regions of

262: individual galaxies can significantly reduce the attenuation,

263: especially for relatively bright sources.

264: 

265: \section{The Impact on the Ly$\alpha$ Luminosity Function}

266: 

267: We next use the attenuation factors computed above, and obtain the

268: reduction in the cumulative \lya LF.  The resulting LF is shown in

269: Figure~\ref{fig:LFs}, for three values of $\langle x_{\rm HI}

270: \rangle$: 0, 0.1, and 1 (top to bottom).  The plot also shows the data

271: for the $z=6.5$ LF with error bars (from MR04).  The short--dashed

272: curve is a reproduction of the $z=5.7$ LF (a Schechter function),

273: which, by assumption, would be identical to the $z=6.5$ LF if the IGM

274: contained no neutral HI.

275: 

276: Note that the suppression of the line flux depends on the luminosity

277: of the source, but the reduction of the LF in Figure~\ref{fig:LFs} is

278: uniform, by a nearly constant factor. The steepening of the LFs at the

279: bright end compensates for the reduced effect of line attenuation; the

280: two conspire to give a nearly constant suppression of the cumulative

281: LF.

282: 

283: \myputfigure{fig2.eps}{3.2}{0.45}{-10}{-10}

284: \figcaption{The Ly$\alpha$ luminosity functions for a

285: fully ionized IGM (no suppression, short--dashed curve), for an IGM with

286: $\langle x_{\rm HI} \rangle = 0.1$ (or $\Gamma_{12}=0.0015$;

287: long--dashed curve), and for and IGM with $\langle x_{\rm HI} \rangle = 1.0$

288: (or $\Gamma_{12}=0$; solid curve).

289: \label{fig:LFs}}

290: \vspace{\baselineskip}

291: 

292: \section{Results and Discussion}

293: 

294: We can now assign a likelihood to each value of $\langle x_{\rm

295: HI}\rangle$, by comparing the model LFs and the data shown in

296: Figure~\ref{fig:LFs} . For each value of $\langle x_{\rm HI} \rangle$,

297: we obtain the mean expected number of \lya emitters above the flux

298: thresholds shown by the data points (here we use the effective volume

299: probed by each survey, listed in Table 2 in MR04).  We assume Poisson

300: fluctuations to compute the likelihood in each bin.  We find that the

301: lowest--luminosity data--point has an a--priori Poisson likelihood

302: that is exceedingly low, even in a fully ionized universe. As

303: Figure~\ref{fig:LFs} shows, $\lsim 0.03$ objects are expected in the

304: small survey volume in which the single lensed \lya emitter,

305: represented by this data--point, was discovered. However, given the

306: difficulties in obtaining an accurate effective volume from the poorly

307: constraint lensing configuration for this object, we follow MR04 and

308: omit this source from our analysis.  The result, i.e. the product of

309: the Poisson probabilities for the three higher luminosity bins, are

310: shown in Figure~\ref{fig:probs} as a function of $\langle x_{\rm HI}

311: \rangle$.

312: 

313: We find that $\langle x_{\rm HI}\rangle=1$ is ruled out at high

314: significance, in agreement with the conclusions reached by MR04 and

315: Stern et al. (2004) (but see discussion below).  On the other hand, at

316: the 99.7\% confidence level, we find only the relatively weak

317: constraint $\langle x_{\rm HI} \rangle < 0.25$.  The upper dashed line

318: corresponds to the limit of no HI absorption and shows that the data

319: is consistent (at the $\sim 10\%$ confidence level) with the intrinsic

320: $z=5.7$ and $z=6.5$ LFs being identical (but note that the IGM would

321: have to be ionized to an unrealistically high level to actually

322: approach this limit).

323: 

324: \myputfigure{fig3.eps}{3.2}{0.45}{-10}{-10} \figcaption{The

325: probability of each mean neutral fraction in the IGM, given the

326: observational constraints from the $z=6.5$ Ly$\alpha$ luminosity

327: function.  The model assumes that the intrinsic LF is the same as at

328: $z=5.7$ (see text for discussion).\label{fig:probs}}

329: \vspace{\baselineskip}

330: 

331: As mentioned above, HII regions can be enlarged due to the clustering

332: of ionizing sources, and also due to their being elongated along our

333: line of sight (e.g. Gnedin \& Prada 2004).  These effects reduce the

334: \lya attenuation, and weaken the upper limit on $\langle x_{\rm

335: HI}\rangle$.  It is interesting to ask whether {\em any} constraint

336: can be placed on the neutral fraction, in the presence of

337: clustering. To address this question, we re--computed the attenuation

338: factors shown in Figure~\ref{fig:suppression}, and the implied LFs

339: shown in Figure~\ref{fig:LFs}, fixing $\langle x_{\rm HI}\rangle=1$,

340: but boosting the ionizing emissivity of each source by a constant

341: factor of $B$.  We find that the boosting factor required to render

342: this model acceptable at the 99.7\% confidence level is approximately

343: $B=60$ (note that the increased luminosity is required to reduce both

344: the GP damping wing, and the residual HI inside the HII region).

345: 

346: In order to assess whether the ionizing emissivity can be boosted by

347: this factor, we computed the total mass in dark matter halos in the

348: mass range $M_{\rm min}$ to $M_{\rm s}$ and within a sphere of radius

349: $R$ around a halo of mass $M_{\rm s}$. We included the spatial

350: correlations between halos with a prescription for linear bias (see

351: eq. 15 in Haiman et al. 2001).  For the likely halo mass range of

352: $10^{11}{\rm M_\odot}\lsim M \lsim 10^{12}{\rm M_\odot}$ (Rhoads et

353: al. 2003) and HII sphere radii ($2\lsim R \lsim 6$ Mpc; e.g. Haiman

354: 2002) of \lya emitters at $z=6.5$, we find that the total collapsed

355: halo mass, down to $2\times 10^{8}{\rm M_\odot}$ (corresponding to a

356: virial temperature of $10^4$K at $z=6.5$) is increased by a factor of

357: $2-18$ (the satellite halos have linear bias parameters of $2\lsim

358: b\lsim 10$). Unless the star formation efficiency is preferentially

359: higher in the low--mass ``satellite'' galaxies, it is unlikely that

360: the ionizing emissivity is boosted by the required factor of $60$.  A

361: recent numerical simulation by Gnedin \& Prada (2004, see their

362: fig. 3) however, suggests that at least $10\%$ of galaxies may

363: transmit their \lya lines without any absorption at $z=6.5$.

364: 

365: 

366: \myputfigure{fig4.eps}{3.2}{0.45}{-10}{-10} \figcaption{The FWHM of

367: the transmitted \lya line as a function of the luminosity of the

368: source.  The upper (lower) panel assumes an intrinsic Gaussian line

369: shape with a width of 300 (30) km/s.  On both panels, the dashed curve

370: assumes an ionizing background of $\Gamma_{12}=0.1$ (or $\langle

371: x_{\rm HI} \rangle\sim 10^{-3}$), and the solid curve assumes

372: $\Gamma_{12}=0$ ($\langle x_{\rm HI} \rangle = 1$).  The feature at

373: $\sim 10^{41-42}~{\rm erg~s^{-1}}$ is unique to a significantly

374: neutral IGM.

375: \label{fig:fwhm}}

376: \vspace{\baselineskip}

377: 

378: Another effect that can further weaken the upper limits on the neutral

379: fraction is a systematic redshift of the \lya emission lines relative

380: to the absorbing gas.  Lyman break galaxies at lower redshift

381: reveal such systematic shift by several hundred km/s, attributed to

382: galactic winds (Shapley et al. 2003).  If velocity offsets as high as

383: this are common for \lya emitters at $z\sim 6$, this would render \lya

384: line attenuation from the residual HI within the HII region

385: effectively negligible, and will also significantly reduce the GP

386: damping wing absorption.  By repeating our analysis above to obtain a

387: likelihood for a neutral universe, we find that a redshift by $600$

388: km/s is required to for $\langle x_{\rm HI} \rangle = 1$ to yield an

389: acceptable fit for the LF.  Shapley et al. (2003) find a decreasing

390: velocity offset with increasing \lya equivalent width, with $\langle

391: \Delta v\rangle<500$ km/s for $W_\alpha>50$\AA\ (see their fig.~11),

392: suggesting that high--$z$ \lya emitters may not have the requisite

393: $600$ km/s offsets. 

394: 

395: Finally, we propose a different diagnostic of the neutral fraction

396: that could be available from a future, larger sample of $z\gsim 6$

397: \lya emitters.  For faint \lya emitters, the line attenuation is

398: dominated by the residual HI inside their cosmic HII regions, whereas

399: for bright emitters, the GP damping wing is dominant (Haiman 2002).

400: In Figure~\ref{fig:fwhm}, we show the predicted full width at half

401: maximum (FWHM) of the transmitted line, as a function of \lya

402: luminosity (for two different intrinsic line--widths). The figure

403: reveals a clear imprint of this transition at $\sim 10^{41-42}~{\rm

404: erg~s^{-1}}$.  This feature occurs only if the IGM contains a

405: significant amount of neutral hydrogen (solid curves), and we checked

406: that it is present for different assumed intrinsic line--shapes

407: (e.g. a top--hat or a Lorentzian, which have less/more extended wings

408: than a Gaussian). This suggests that a measurement of observed \lya

409: line width as a function of the \lya luminosity may therefore serve as

410: a more robust diagnostic of the neutral fraction.

411: 

412: 

413: \section{Conclusions}

414: 

415: In this {\it Letter}, we have considered the attenuation of the line

416: flux of \lya emitting galaxies at $z=6.5$.  We included the effect of

417: HII regions around these galaxies, which reduce the line

418: attenuation. We computed the overall impact of the IGM on the \lya

419: luminosity function, and found that the presence of the HII regions

420: makes a large mean neutral fraction in the IGM acceptable. The best

421: upper limit is at most as strong as $\langle x_{\rm HI} \rangle \lsim

422: 0.25$. However, a fully neutral IGM may be acceptable only in the

423: presence of an unusually strong clustering of ionizing sources, and/or

424: a strong systematic recession velocity of the \lya emitting gas.  In

425: either case, the present determination of the $z\sim 6.5$ \lya LF is

426: consistent with reionization occurring at this epoch. The LF of a

427: larger sample of \lya emitters, or a systematic study of their

428: linewidths and shapes, as a function of luminosity, should

429: nevertheless provide a robust diagnostic of the ionization state of

430: the IGM in the future.

431: 

432: \acknowledgments{This work was supported by NASA through grants

433: NAG5-26029, HST-GO-09793.18 (ZH) and NAG5-13381 (RC), and by NSF

434: through grants AST-0307200, AST-0307291 (ZH), and AST-0206299,

435: AST-0407176 (RC).}

436: 

437: \begin{thebibliography}{DUM}

438: \bibitem[Becker et al. 2001]{betal01} Becker, R.H., \etal\ 2001, AJ, 122, 2850 

439: \bibitem[Cen \& Haiman 2000]{ch00} Cen, R., \& Haiman, Z. 2000, ApJ, 542, L75 

440: \bibitem[Cen, Haiman \& Mesinger 2004]{chm04} Cen, R., Haiman, Z., \& Mesinger, A. 2004, ApJ, in press (astro-ph/0403419) 

441: \bibitem[Cen \& McDonald 2002]{cm02} Cen, R., \& McDonald, P. 2002, ApJ, 570, 457 

442: \bibitem[Fan et al.2002]{f02} Fan, X., \etal\ 2002, AJ, 123, 1247 

443: \bibitem[Furlanetto, Hernquist \& Zaldarriaga 2004]{fetal04} Furlanetto, S. R., Hernquist, L., \& Zaldarriaga, M. 2004, MNRAS, 354, 695

444: \bibitem[Gnedin \& Prada 2004]{gp04} Gnedin, N. Y., \& Prada, F. 2004, ApJ, 608, L77 

445: \bibitem[Gunn \& Peterson 1965]{gp65} Gunn, J. E., \& Peterson, B. A., 1965, ApJ, 142, 1633

446: \bibitem[Haiman 2002]{h02} Haiman, Z. 2002, ApJ, 576, L1

447: \bibitem[Haiman, Abel \& Madau 2001]{ham01} Haiman, Z., Abel, T., \& Madau, P. 2001, ApJ, 551, 599

448: \bibitem[Haiman \& Cen 2002]{hc02} Haiman, Z., \& Cen, R. 2002, ApJ, 578,702 

449: \bibitem[Haiman \& Spaans 1999]{hs99} Haiman, Z., \& Spaans, M. 1999, ApJ, 518, 138 

450: \bibitem[Hu et al. 2002]{hetal02} Hu, E. M., et al. 2002, ApJ, 568, L75 [Erratum 2002 ApJ, 576, L99] 

451: \bibitem[Hui \& Haiman 2003]{hh03} Hui, L., \& Haiman, Z. 2003, ApJ, 596, 9

452: \bibitem[Kennicutt 1983]{k83} Kennicutt, R. C., Jr. 1983, ApJ, 272, 54

453: \bibitem[Kodaira et al. 2003]{ketal03} Kodaira, K. et al. 2003, PASJ, 55, L17 

454: \bibitem[Le Delliou et al. 2004]{letal04} Le Delliou, M. et al. 2004, A\&A, submitted, astro-ph/0405304

455: \bibitem[Lidz et al. 2002]{letal02} Lidz, A., Hui, L., Zaldarriaga, M., \& Scoccimarro, R. 2002, \apj, 579, 491

456: \bibitem[Madau \& Rees 2000]{mr00} Madau, P., \& Rees, M. J. 2000, ApJ, 542, L69 

457: \bibitem[Malhotra \& Rhoads 2004]{mr04} Malhotra, S. \& Rhoads, J. 2004, ApJL, in press, astro-ph/0407408 

458: \bibitem[Miralda-Escud\'e 1998]{me98} Miralda-Escud\'{e}, J. 1998, ApJ, 501, 15 

459: \bibitem[Rhoads et al. 2003]{retal03} Rhoads, J., et al. 2003, AJ, 125, 1006

460: \bibitem[Santos 2004]{s04} Santos, M. R. 2004, MNRAS, 349, 1137

461: \bibitem[Shapiro \& Giroux 1987]{sg87} Shapiro, P., \& Giroux, M. L. 1987, ApJ, 321, L107 

462: \bibitem[Shapley et al. 2003]{setal03} Shapley, A. E., Steidel, C. C., Pettini, M., \& Adelberger, K. L. 2003, ApJ, 588, 65

463: \bibitem[Songaila 2004]{song04} Songaila, A. 2004, \aj, 127, 2598

464: \bibitem[Spergel et al.(2003)]{spergel03} Spergel, D.N. et al. 2003, AJS, 148, 175 

465: \bibitem[Stern et al. 2004]{setal04} Stern, D., et al. 2004, ApJ, submitted, astro-ph/0407409 

466: \bibitem[White et al. 2003]{white03} White, R. L., Becker, R. H., Fan, X., \& Strauss, M. A. 2003, \aj, 126, 1

467: \bibitem[Wyithe \& Loeb 2004a]{wl04a} Wyithe, J. S. B., \& Loeb, A. 2004a, \nat, 427, 815

468: \bibitem[Wyithe \& Loeb 2004b]{wl04b} Wyithe, J. S. B., \& Loeb, A. 2004b, ApJ, submitted, astro-ph/0407162

469: \end{thebibliography}

470: \end{document}

471: \bye

472:  

473: