0709.0654/ms.tex
1: 
2: %\documentclass[12pt,preprint]{aastex}
3: 
4: %% manuscript produces a one-column, double-spaced document:
5: 
6: % \documentclass[manuscript]{aastex}
7: 
8: %% preprint2 produces a double-column, single-spaced document:
9: 
10: \documentclass[apj]{emulateapj}
11: %\documentclass[preprint2]{aastex}
12: \usepackage{amssymb,amsmath}
13: %\usepackage{psfig}
14: 
15: \shorttitle{Mapping reionization with Ly$\alpha$ emission}
16: \shortauthors{Cantalupo, Porciani \& Lilly}
17: \slugcomment{To be published in the Astrophysical Journal}
18: 
19: \newcommand{\ergscm}{$\mathrm{erg}\mathrm{s}^{-1}\mathrm{cm}^{-2}$}
20: \newcommand{\ergscmarcsec}{$\mathrm{erg}\mathrm{s}^{-1}\mathrm{cm}^{-2}\mathrm{arcsec}^{-2}$}
21: \newcommand{\ergscmAA}{$\mathrm{erg}\mathrm{s}^{-1}\mathrm{cm}^{-2}\mathrm{\AA}^{-1}$}
22: \newcommand{\lya}{Ly$\alpha$\ }
23: \newcommand{\lyam}{\mathrm{Ly}\alpha }
24: \newcommand{\StART}{\texttt{StART}\ }
25: \newcommand{\mr}{\mathrm}
26: 
27: %\def\be{\begin{equation}}
28: %\def\ee{\end{equation}}
29:     
30: \begin{document}
31: 
32: \def\simlt{\mathrel{\rlap{\lower 3pt\hbox{$\sim$}} \raise
33:         2.0pt\hbox{$<$}}} \def\simgt{\mathrel{\rlap{\lower
34:         3pt\hbox{$\sim$}} \raise 2.0pt\hbox{$>$}}}
35: \def\simgt{\mathrel{\rlap{\lower 3pt\hbox{$\sim$}} \raise
36:         2.0pt\hbox{$>$}}} \def\simgt{\mathrel{\rlap{\lower
37:         3pt\hbox{$\sim$}} \raise 2.0pt\hbox{$>$}}}
38: 
39: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
40: 
41: \title{Mapping Neutral Hydrogen during Reionization \\
42:  with the \lya Emission from Quasar Ionization Fronts} 
43: 
44: \author{Sebastiano Cantalupo, Cristiano Porciani and Simon J. Lilly}
45: \affil{Institute for Astronomy, ETH Z\"urich, CH-8093 Z\"urich, Switzerland}
46:    
47: \email{cantalupo@phys.ethz.ch}
48: 
49: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
50: 
51: \begin{abstract}
52: 
53: We present a new method
54: to directly map the neutral-hydrogen distribution 
55: during the reionization epoch
56: and to constrain the emission properties 
57: of the highest-redshift quasars (QSOs).
58: %
59: As a tracer of HI, we propose to use
60: the \lya radiation produced by quasar ionization fronts (I-fronts)
61: that expand in the partially ionized intergalactic medium (IGM)
62: before reionization is complete.
63: %
64: These \lya photons are mainly generated by collisional excitations
65: of hydrogen atoms in the boundary of the rapidly expanding HII region.
66: %
67: The observable signal is produced 
68: by the part of the I-front 
69: that lies behind the QSO with respect to the observer.
70: %
71: The expected \lya flux depends 
72: on the properties of both the QSO (e.g. its total ionizing rate and
73: spectral hardness) and the surrounding medium (e.g. the local density
74: and the mean hydrogen neutral fraction).
75: %
76: Combining two radiative transfer models 
77: (one for the QSO ionizing radiation and one for the 
78: \lya photons), we estimate the expected
79: \lya spectral shape and surface brightness (${\rm SB}_{\lyam}$) 
80: for a large number of configurations where we varied both the properties
81: of the ionizing QSO and of the surrounding medium.
82: %
83: We find that the expected signal is observable as a 
84: single (broad) emission line with a characteristic width of 
85: $100-200$ km s$^{-1}$. 
86: %
87: The expected ${\rm SB}_{\lyam}$ produced at redshift $z\simeq 6.5$ 
88: within a fully neutral region (at mean density)
89: by a typical QSO I-front
90: lies in the range $10^{-21}-10^{-20}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$
91: and decreases proportionally to $(1+z)^2$ for a given QSO age.
92: %
93: QSOs with harder spectra may produce a significantly brighter emission
94: at early phases.
95: %
96: The signal may cover up to a few hundred square arcmin
97: on the sky
98: and should be already detectable with current
99: facilities by means of moderate/high resolution spectroscopy.
100: The detection of this \lya emission can shed new light on the 
101: reionization history, the age and the
102: emission properties of the highest-redshift QSOs.
103: 
104: \end{abstract}
105: 
106: \keywords{cosmology: theory - diffuse radiation - intergalactic medium - 
107: radiative transfer - line: profiles - quasars: general}
108: 
109: %--------------------------------------------------
110: \section{Introduction}
111: 
112: The epoch of reionization (EoR) marks the time at which 
113: the first luminous sources ionized the 
114: mostly neutral
115: intergalactic medium (IGM).
116: Uncovering when and how the EoR took place 
117: represents one of the most fundamental questions of modern cosmology
118: and considerable efforts 
119: are being made to understand this era.
120:  
121: The recent polarization measurements of the cosmic microwave background (CMB)
122: by the WMAP satellite suggest that the universe was mostly
123: neutral at redshift $z\simgt14$ (Page et al. 2006).
124: On the other hand, at $z<5$,
125: hydrogen absorption features in the spectra of luminous quasars
126: are resolved into individual \lya forest
127: lines (e.g. Rauch 1998) as expected in a highly ionized IGM.
128: %
129: At $z\gtrsim6$, quasar spectra start showing complete Gunn-Peterson 
130: absorption throughs (GPT) which might 
131: suggest a rapid increase in the neutral fraction of cosmic hydrogen
132: (Becker et al. 2001; Fan et al. 2003; White et al. 2003).  
133: %
134: However, the \lya transition has a large cross-section and 
135: the presence of GPTs only requires very low values for 
136: mean neutral fraction of cosmic hydrogen: 
137: $\langle x_{HI}\rangle \gtrsim10^{-3}$
138: (Fan et al. 2002; Songaila 2004; Oh \& Furlanetto 2005). 
139: 
140: Stronger constraints on $\langle x_{HI}\rangle$ 
141: can be obtained from the detailed spectral shape of the GPT, i.e.
142: from the size of the HII region produced by the QSO itself (White et al. 2003; 
143: Mesinger \& Haiman 2004; Wyithe \& Loeb 2004; Yu \& Lu 2005; Fan et al. 2006; 
144: Maselli et al. 2006; Bolton \& Haehnelt 2007).
145: % 
146: The physical scales of these regions are typically inferred to be $\sim5$ Mpc
147: at $z>6$ (Fan et al. 2006), an order of magnitude larger than the HII bubbles 
148: expected from clustered star-forming galaxies 
149: (Furlanetto, Zaldarriga \& Hernquist 2004). 
150: %
151: The size of the HII region around a QSO depends
152: on the neutral density of the surrounding medium but also
153: on the source luminosity and age (more specifically, on its light curve).
154: With some knowledge of the QSO parameters 
155: it is then possible to 
156: constrain the value of $\langle x_{HI}\rangle$. 
157: Whythe et al. (2005) applied this method to
158: seven QSOs at $6\lesssim z\lesssim 6.42$ and derived
159: $\langle x_{HI}\rangle\gtrsim0.1$. 
160: %
161: Similarly, from the size of the \lya and Ly$\beta$ troughs of a QSO at 
162: $z=6.28$, Mesinger \& Haiman (2004) found $\langle x_{HI}\rangle\gtrsim0.2$.
163: %
164: On the contrary, Fan et al. (2006) did not find strong evidence
165: for a mean hydrogen neutral fraction as high as 0.1 at $z=6.4$.
166: % 
167: Apart from systematics, the main uncertainty of this method
168: lies in the estimate of the QSO age.
169: As a matter of fact, the current observed sizes seem to be equally consistent
170: with both a significantly neutral and a highly ionized surrounding IGM 
171: (e.g. Bolton \& Haehnelt 2007). 
172: 
173: Additional (indirect) information
174: about the ionized fraction of the IGM 
175: can be obtained from the luminosity-function evolution
176: of \lya emitting galaxies (LAE) at $z>6$ 
177: (Miralda-Escud\'e 1998; Madau \& Rees 2000; Haiman 2002; 
178: Santos 2004; Taniguchi et al. 2005; Furlanetto et al. 2006; Malhotra \& Rhoads 2006; Kashikawa et al. 2006), 
179: and from their clustering properties (McQuinn et al. 2007).
180: %
181: The key idea is that the observed \lya flux from high-$z$ galaxies 
182: should strongly depend on the neutral fraction of their surrounding
183: medium. 
184: %
185: However, these observations are challenging and difficult to interpret.
186: %
187: Any change in the luminosity function due to the EoR must 
188: be distinguished from an intrinsic physical evolution of the sources, 
189: as well as the effects of dust or galactic winds. 
190: %
191: Similarly, clustering signatures of the EoR can only be detected
192: by comparing the spatial distribution of a given galaxy population
193: (e.g. Lyman-break galaxies) with that of the sub-sample of LAEs
194: (see Fan, Carilli \& Keating 2006 for a recent review).
195: 
196: 
197: A very promising method for studying the EoR
198: is the detection of the redshifted 21-cm emission from neutral hydrogen 
199: (Madau, Meiksin \& Rees 1997; see Furlanetto, Oh \& Briggs 2006 
200: for a recent and exhausitive review).
201: This technique has received a lot of attention recently as a concrete
202: possibility of studying the detailed history of the EoR.
203: From the technical point of view, however, it requires
204: an extremely challenging measure:  
205: the intrinsic 21-cm signal, detected as a 
206: brightness-temperature fluctuation against the CMB, is at least four orders of 
207: magnitude smaller than the emission from our Galaxy. 
208: Other important foregrounds
209: include radio recombination lines, terrestrial interference and ionospheric 
210: distortion. 
211: The hope is to separate the fluctuating spectral features due to
212: 21-cm emission from the smooth spectrum of the foregrounds.
213: Anyway, 
214: given the difficulties of high signal-to-noise imaging, attention has been 
215: focussed on statistical measurements, like the power-spectrum analysis. 
216: %
217: First-generation facilities (as LOFAR and MWA, now under construction)
218: should only detect a statistical signature of reionization
219: For a direct detection of the three-dimensional structure of HI and HII
220: regions (the so called  ``21-cm tomography'') 
221: we will probably have to wait for a subsequent generation of radio arrays with 
222: much larger collecting area (e.g. SKA).
223: 
224: 
225: Are there other possibilities to directly detect the high-z IGM in emission?
226: \lya is the strongest hydrogen emission line. 
227: This transition (from the 2P to the 1S level of
228: atomic hydrogen) has a spontaneous emission coefficient 
229: $A_{21}=6.25\times10^{8}$ s$^{-1}$. 
230: An efficient mechanism to populate the $n=2$ level, with consequent
231: \lya emission, is the HII recombination process. 
232: Unfortunately, the recombination rate is
233: typically very low and recombinaton radiation from optically thin hydrogen in the 
234: high-redshift IGM cannot be detected with current instruments
235: (see e.g. Hogan \& Weymann 1987; 
236: but also Baltz, Gnedin \& Silk 1998).
237: 
238: Optically thick and self-shielded HI clouds exposed to strong UV radiation are 
239: expected to re-emit a significant
240: part of the impinging flux in the form of ``fluorescent'' \lya emission 
241: (Gould \& Weinberg 1996; 
242: Cantalupo et al. 2005). A recent attempt to detect fluorescent emission, 
243: and thus HI protogalactic-clouds,
244: at intermediate redshift ($z\sim3$) has provided a number of promising
245: candidates (Cantalupo, Lilly \& Porciani 2007).
246: %
247: %
248: However, self-shielded clouds correspond to overdense 
249: regions with relatively small sizes that trace just a small
250: fraction of the neutral hydrogen distribution.
251: 
252: Under favourable conditions,
253: the excitation of atomic hydrogen due to collisions with energetic electrons 
254: populates the $n=2$ level in a much more efficient way than recombinations.
255: In this paper, we show that the corresponding \lya emission can be 
256: efficiently used to map the neutral hydrogen distribution during the EoR.
257: This emission should be already detectable with current observational
258: facilities and could be used to shed light on  
259: the bulk of neutral hydrogen during the EoR. 
260: 
261: The layout of the paper is as follows. 
262: In \S 2 we summarize the basic physics of \lya emission from 
263: recombination and collisional excitation processes. 
264: In \S 3 we show that collisional excitations should efficiently produce 
265: \lya emission within the ionization fronts of high-redshift QSOs.
266: We also explain how 
267: it is possible to map the
268: distribution of neutral hydrogen behind the QSO using this emission.
269: %
270: In \S 4 we present 
271: more sophisticated models of \lya emission based on numerical simulations
272: including a detailed treatment of radiative transfer.
273: In \S 5 we discuss the dependence of our results on model parameters and 
274: possible detection strategies. 
275: Finally, in \S 6, we summarize our results.
276: %
277: The reader who is not interested into the technical details of the model
278: can safely read \S 5.2 right after \S 3 (focussing also on Figures 7 and 8). 
279: 
280: Throughout this paper, we adopt a standard, flat $\Lambda$CDM cosmological 
281: model with mass-density parameter $\Omega_{\rm m}=0.24$, a baryonic contribution
282: of $\Omega_{\rm b}=0.044$, and a present-day Hubble constant 
283: of $H_0=100\,h$ km s$^{-1}$ Mpc$^{-1}$  with $h=0.73$. 
284: We also adopt the notation ``pMpc''  
285: to indicate proper distances measured in Mpc. 
286: 
287: %--------------------------------------------------
288: 
289: 
290: \section{The physics of Ly$\alpha$ emission}
291: 
292:  There are two main channels that drive the production of Ly$\alpha$ photons in the IGM:
293: i) recombination processes and ii) collisional excitations by free electrons. 
294: Both mechanisms directly arise from the ionization of the hydrogen atom.
295: 
296: \subsection{\lya from recombinations}
297: 
298: When an electron is captured by
299: a free proton, it can either directly populate the ground 1 $^2S$ level 
300: (with the emission of a Lyman continuum photon) 
301: or an excited state from which it cascades by downward transitions to the 
302: 1 $^2S$ level 
303: (with the emission of a continuum photon plus several transition lines). 
304: Cascades that populate the 2 $^2P$ state decay to the 1 $^2S$ level 
305: by emitting a Ly$\alpha$ photon, while
306: cascades to the 2 $^2S$ level subsequently decay via two-photon emission.
307: 
308: Electrons that populate $n\geq3$ levels may also decay to the ground state 
309: with the emission of
310: another Lyman-series line (with no two-photon or Ly$\alpha$ emission).  
311: However, even in the low-density IGM, 
312: the typical Lyman-line emitting regions have quite large optical
313: depths ($\tau$) in the Lyman resonance lines (Osterbrock 1989).  
314: Therefore, a good approximation 
315: (unless the medium is fully ionized) is to assume that every Lyman-line
316: photon from $n\geq3$ levels is converted 
317: into lower-series photons plus either Ly$\alpha$ or two-photon
318: radiation (Case B approximation, Baker \& Menzel 1938). 
319: In this case, it is convenient to write the effective 
320: Ly$\alpha$ emission coefficient from recombinations as:
321: \begin{equation}
322: \alpha^{\mr{eff}}_{\lyam}(T)=\epsilon^{\mr{B}}_{\lyam}(T) \alpha_{\mr{B}}(T)\ ,
323: \end{equation}
324: where $\alpha_{\mr{B}}(T)$ is the hydrogen total recombination coefficient excluding recombinations to the
325: ground level, 
326: and $\epsilon^{\mr{B}}_{\lyam}(T)$ is the fraction of those recombinations 
327: producing \lya photons.
328: Combining the tabulated values by Pengelly (1964; for $T>10^3\,\mathrm{K}$) 
329: and Martin (1988; for $T<10^3\,\mathrm{K}$), 
330: we have derived the following fitting formula for $\epsilon^{\mr{B}}_{\lyam}(T)$ 
331: (accurate to the 0.1\% in the temperature range 
332: $100\,\mathrm{K}<T<10^5\,\mathrm{K}$):
333: %
334: \begin{equation}
335: \epsilon^{\mr{B}}_{\lyam}(T)=0.686-0.106\log(T_4)-0.009\cdot(T_4)^{-0.44}\ ,
336: \end{equation}
337: %
338: where $T_4=T/10^4\,\mathrm{K}$. 
339: Note that the value of $\epsilon^{\mr{B}}_{\lyam}(T)$ varies very little
340: with temperature, ranging beetwen 0.68 and 0.61 for $T_4=1-5$. 
341: The temperature dependence of $\alpha^{\rm eff}_{\lyam}(T)$
342: is shown in Figure \ref{f1} as a dashed line.  
343: Note that the value of $\epsilon^{\mr{B}}_{\lyam}(T)$ varies very little
344: with temperature, ranging beetwen 0.68 and 0.61 for $T_4=1-5$. 
345: 
346:  The volume \lya emissivity due to radiative recombinations is given by:
347: %
348: \begin{equation}
349: \frac{4\pi j_{\lyam}}{h\nu_{\lyam}}=n_{\mr{e}} n_{\mr{p}} \alpha^{\mr{eff}}_{\lyam}
350: \end{equation}
351: %
352: where $j_{\lyam}$ is the emissivity (energy radiated per unit time, volume and
353: solid angle),
354: $h\nu_{\lyam}=10.2$ eV is the energy of a \lya photon and $n_e$, $n_p$ are, respectively, the
355: electron and proton number densities. Note that the coefficient
356:  $\alpha^{\mr{eff}}_{\lyam}$
357: is independent from the assumption regarding 
358: the fate of Lyman-continuum photons. 
359: Wherever these photons are absorbed, and also in the case they escape 
360: the medium, 
361: the local values of $n_e$ and $n_p$ will change accordingly, modifying the \lya emissivity.
362: 
363: \begin{figure}
364: \plotone{f1.eps}
365: \caption{Effective Ly$\alpha$ production rate from recombination processes ($\alpha^{eff}_{Ly\alpha}$, 
366: dashed line) and collisional excitations ($q^{eff}_{Ly\alpha}$, solid line) 
367: as a function of the gas (electronic) temperature.}
368: \label{f1}
369: \end{figure}
370: 
371: 
372: \subsection{\lya from collisional excitations}
373: 
374: Free electrons can interact one or
375: more times with neutral hydrogen atoms before eventually recombining. 
376: Given the relatively low densities of the IGM, 
377: we can imagine the free electrons colliding with
378: hydrogen in the ground state and transferring to it a fraction of their energy. %
379: If this energy (i.e. the electron temperature) is high enough, the collision
380: will excite the atomic levels with subsequent decay via line emission. 
381: The efficiency of the process is strongly temperature dependent. In particular, the collisional
382: excitation coefficient for the transition from the ground level (1) to the level $nl$ is given by:
383: %
384: \begin{equation}
385: q_{1,nl}=\frac{8.629\times10^{-6}}{\sqrt{T}}\frac{\Omega(1,nl)}{\omega_1} e^{-E_{1,n}/k_BT} \mathrm{cm^3 s^{-1}}\ ,
386: \end{equation}
387: %
388: where $\Omega(1,nl)$ is the temperature dependent effective collision strength, $\omega_1$ is the statistical
389: weight of the ground state and $E_{1,n}$ is the energy difference beetwen the ground and the $nl$ level.
390: We define the effective collisional excitation coefficient for 
391: \lya emission as:
392: %
393: \begin{equation}
394: q^{\mr{eff}}_{\lyam}=q_{1,2p}+q_{1,3s}+q_{1,3d}
395: \end{equation}
396: %
397: where we consider excitation processes up to the level $n=3$ that will 
398: eventually produce \lya radiation (in Case B 
399: approximation). 
400: The contribution from higher levels is completely negligible for the
401: electron temperatures we are interested in.
402: Similarly, we neglect angular-momentum-exchanging processes 
403: from proton and electron 
404: collisions (e.g. the 2s-2p and 2p-2s transition) that, in principle, might 
405: boost or suppress \lya emission. The typical densities of the IGM are
406: too low for these transitions to be important (Osterbrock 1989).
407: %
408:  We use the polynomial fits of Giovanardi, Natta \& Palla (1987)
409:  to compute $q^{\mr{eff}}_{\lyam}$.
410:  The result is shown as a solid line in Figure \ref{f1}.
411: 
412:  The total \lya emissivity coming from
413: radiative recombinations and collisional excitations is thus given by:
414: %
415: \begin{equation}\label{lya_em}
416: \frac{4\pi j_{\lyam}}{h\nu_{\lyam}}=n_{\mr{e}} n_{\mr{p}}\ \alpha^{\mr{eff}}_{\lyam} + 
417:  n_{\mr{e}} n_{\mr{HI}}\ q^{\mr{eff}}_{\lyam}\ ,
418: \end{equation}
419: %
420: where $n_{\mr{HI}}$ is the neutral hydrogen density.
421: %
422: Higher ionization fractions favour \lya emission from 
423: radiative recombinations, while
424: higher temperatures increase the contribution from collisional excitations 
425: provided that a significant fraction of hydrogen atoms remain neutral.
426: %
427: Therefore,
428: collisional excitation significantly contributes to the total \lya 
429: emissivity in relatively hot regions where $n_{\mr{e}}\sim N_{\mr{HI}}$,
430: i.e. where the neutral fraction ($x_{\mr{HI}}$) is about 0.5. 
431: %
432: In particular, if the temperature of these regions
433: is in the range $2-5\times10^4\,\mathrm{K}$, collisional excitation may 
434: increase the total \lya signal of several orders
435: of magnitudes with respect to radiative recombinations.
436: 
437: \section{\lya emission from quasar ionization fronts during reionization}\label{sec3}
438: 
439:  The two requirements for efficient \lya production via collisional excitation
440: ($T\sim$ a few $\times 10^4\,\mathrm{K}$ and $x_{\mr{HI}}\sim0.5$) 
441: are met within the ionization fronts (I-fronts) 
442: produced by powerful UV sources with a hard spectrum (like QSOs)
443: located in a mostly neutral IGM. 
444: %
445: An I-front constitutes the transition region between the inner HII zone 
446: and the outer (predominantly neutral) IGM.
447: In this transition layer, 
448: the hydrogen neutral fraction varies by several orders of magnitudes
449: within a length scale determined by the mean free path of the ionizing photons. 
450: The thickness of the I-front is generally negligible with respect to the characteristic 
451: radius of the HII region.
452: %
453: Sources with harder spectra produce thicker I-fronts 
454: (for a given IGM density)
455: because of the frequency dependence of the photo-ionization cross-section.
456: %
457: If the I-front is not in photo-ionization equilibrium, 
458: ionization fractions evolve with time
459: and the transition zone moves outward (the HII region expands).
460: The propagation of the I-front produced by powerful UV sources like QSO 
461: can be highly relativistic, racing ahead of the hydrodynamic response of the 
462: ionized gas for the entire lifetime of the source (Shapiro \& Giroux 1987).
463: 
464: \subsection{A simple estimate of the \lya signal}
465: 
466: As shown in equation (\ref{lya_em}),
467: the volume emissivity of \lya radiation within the I-front depends 
468: on the electron, proton and HI densities, and on the local temperature.
469: %
470: Detailed modelling of these quantities within an expanding I-front
471: requires an accurate 
472: numerical solution of the radiative transfer problem and is 
473: postponed to the next Section.
474: %
475: However, we can give an order-of-magnitude estimate of the expected 
476: amplitude of the signal following a simple analytical reasoning. 
477: 
478: Let us assume that a bright QSO turns on at redshift $z_{\mr in}>6$, 
479: before the reionization process is complete, 
480: and produces a rapidly expanding I-front that completely ionizes 
481: the surrounding IGM. 
482: %
483: Let us further assume, for simplicity, that the IGM is at mean cosmic
484: density, the hydrogen neutral fraction 
485: varies linearly (from 0 to 1) with distance within the 
486: thickness of the I-front, and that the I-front temperature is uniform. 
487: %
488:  The actual value of the I-front temperature depends on the detailed balance
489: between photoheating and cooling processes 
490: (see, e.g. Miralda-Escude \& Rees 1994). In the following we
491: estimate the expected Ly$\alpha$ signal for a given I-front temperature
492: and medium density
493: \begin{footnote}{
494: The appropriate scaling relations based on other parameters, like
495: the QSO age, ionizing rate and local overdensity will be derived in \S 5.}
496: \end{footnote}.
497: %
498: Typical values for this temperature can be inferred from 
499: the far-UV spectral indices of QSOs (see e.g., Telfer et al. 2002)
500: and lie in the range
501: $2-4\times10^4$ K if the IGM is at mean cosmic density 
502: (see e.g., Abel \& Haehnelt 1999).
503: %
504: At these temperatures
505: we can neglect \lya emission due to radiative recombinations and 
506: write the integrated \lya flux at the inner edge of the I-front as:
507: %
508: %\begin{equation}
509: \begin{eqnarray}\label{I1}
510: \lefteqn{I_{\lyam} \simeq\frac{1}{\pi}\int_{\mr{I-front}} \!\!\!\!\!n_{\mr{HI}}(r)n_{\mr{e}}(r)q^{\mr{eff}}_{\lyam}(T)\mathrm{d}r}\nonumber \\
511:           & & \simeq\frac{1}{\pi} n^{2}_{\mr{H}}\, C \,\chi_{\mr{e}}\, 
512: \left(\frac{S}{6}\right)\, q^{\mr{eff}}_{\lyam}(T)\ ,
513: \end{eqnarray}
514: %\end{equation}
515: %
516: where  $C\equiv \langle n_{\mr{H}}^2\rangle/\langle n_{\mr{H}}\rangle$ is the hydrogen 
517: clumping factor, $\chi_{\mr{e}}$ the 
518: factor that accounts for the contribution of ionized helium to the electron 
519: density, and $S$ denotes
520: the thickness of the I-front. 
521: The factor $1/6$ derives from the integration of $x_{\mr{HI}}(1-x_{\mr{HI}})$
522: over the front, while the factor $1/\pi$ accounts for the angular distribution
523: of the emitted photons (Gould \& Weinberg 1996).
524: %
525: The actual value of $S$ depends on several
526: factors, including the spectrum of ionizing radiation $F_\nu$ 
527: and the local density.
528: As a first-order approximation, we can assume that the
529: size of the I-front is given by the mean free path 
530: of the ionizing photons with 
531: mean frequency $\langle \nu \rangle=\int_{\nu_0}^\infty F_\nu\, d\nu
532: /\int_{\nu_0}^\infty (F_\nu/h_{\mr P}\nu)\, d\nu$
533: (with $\nu_0$
534: the hydrogen ionization threshold and $h_{\mr P}$ the Planck constant), 
535: i.e. $S\simeq(n_{\mr{H}} \sigma_{\langle \nu \rangle })^{-1}$, where
536: $\sigma_\nu$ is the (frequency dependent) photo-ionization cross-section.
537: Substituting $S$ in equation (\ref{I1}), 
538: replacing the hydrogen density with $n_{\mr{H}}=n_{\mr{H},0}(1+z)^3$
539:  (where $n_{H,0}$ is the comoving mean number density), and including the redshift dimming,
540: we obtain the observed integrated surface brightness (SB) in photons:
541: %
542: \begin{equation}
543: \Phi_{\lyam}\simeq\left(\frac{f_{\mr{esc}}\,C\,\chi_{\mr{e}}}{6\pi}\right)n_{\mr{H},0}\frac{q^{\mr{eff}}_{\lyam}(T)}
544: {\sigma_{\langle \nu \rangle }}
545: \end{equation} 
546: %
547: where $f_{\mr{esc}}$ is the fraction of \lya photons that manage to escape 
548: from the I-front along the line of sight.
549: Remarkably, the observed SB (in photon number) does not depend 
550: on the QSO redshift for a given temperature of I-front. 
551: 
552: How strong is this signal?
553: %
554: The exact value of $f_{\mr{esc}}$ is difficult to estimate as it depends on 
555: several factors, including
556: the size of the HII region and the value of the residual HI in proximity of 
557: the I-front. 
558: Detailed calculations (see \S\ref{secMod}) suggest that
559: $f_{\mr{esc}}\sim0.5$.
560: %
561: Typical values of the clumping factor at $z>6$, estimated
562: from simulations (e.g., Gnedin \& Ostriker 1997), are of order of 
563: $C\sim30$.
564: %
565: Therefore, assuming $\langle \nu \rangle\simeq3\,\nu_0$, $\chi_{\mr e}=1.2$ 
566: and fixing the temperature to $T\simeq3\times10^4$ K, we obtain:
567: %
568: \begin{equation}
569: \Phi_{\lyam}\simeq 200\ \mathrm{photons}\ 
570: \mathrm{s}^{-1}\ \mathrm{cm}^{-2}\ \mathrm{sr^{-1}}\ ,
571: \end{equation}
572: %
573: at the top of Earth's atmosphere, or, equivalently:
574: %
575: \begin{equation}
576: {\mr SB}_{\lyam}\simeq 10^{-20}\ 
577: \mathrm{erg}\ \mathrm{s}^{-1}\ \mathrm{cm}^{-2}\ \mathrm{arcsec^{-2}}\ ,
578: \end{equation}
579: %
580: for a QSO at $z\sim 6.5$.
581: 
582:  Although much fainter than the limit fluxes of present-day surveys for \lya emitters
583: at $z\sim6.5$ ($\sim 10^{-18}\mathrm{erg}\ \mathrm{s}^{-1}\ \mathrm{cm}^{-2}\ $
584: for apertures of order of few arcsec$^2$; Tran et al. 2004; Kashikawa 
585: et al. 2006), the emitting region can cover several hundred 
586: comoving Mpc$^2$. This corresponds to several hundred arcmin$^2$
587: on the sky at $z\sim6.5$ and makes the signal detectable with current
588: facilities by means of moderately-high resolution spectroscopy.
589: \footnote{
590: The detectability of this signal with present and future instruments will be 
591: discussed in \S 5.2.} 
592: 
593: Observations of this emission would provide direct evidence for 
594: the presence of an I-front, 
595: shedding light on the properties of both the ionizing source and the
596: surrounding medium. 
597: In particular, as we will show in \S 5,
598: the angular shape of the apparent I-front can constrain the
599: QSO age and its angular emission properties.
600: 
601: \begin{figure}
602: \epsscale{1.1}
603: \plotone{f2.eps}
604: \caption{This schematic cartoon illustrates how to map the HI distribution using the \lya emission
605: from the I-front of an high-z QSO.
606: The QSO and the observer are on the left side. The ionizing photons, in the plane
607: of the figure, propagate from left to right and create an I-front 
608: wherever they encounter
609: a partially neutral region. Within the I-front, the interactions between  
610: neutral hydrogen and the energetic electron released by photo-ionization
611: produce \lya photons via collisional excitation (yellow stripes). The \lya emission escapes
612: the I-front trough the HII region in the direction of the observer. 
613: The regions already ionized by local sources
614: (bounded by the dashed lines) do not produce \lya via collisional 
615: excitations and
616: will appear as holes in the 2-D \lya map, similarly to a 21-cm tomography.
617: This strictly applies to a young QSO with a relativistically expanding
618: I-front. For older quasars, the entire boundary of the HI region can emit
619: \lya photons.}
620: \label{f2}
621: \end{figure}
622: 
623: 
624: \subsection{\lya map of the HI distribution during reionization}
625: 
626: Detecting the \lya emission from the I-front makes it possible to
627: directly map the distribution of neutral hydrogen 
628: behind the QSO at the location of the I-front.
629: %
630: In fact,
631: the distribution of the \lya signal on the sky
632: traces the local density of HI as it was before the arrival of the I-front. 
633: %
634: We will illustrate this with a simple example.
635: Let us consider two
636: patches of the IGM (at mean density) that are simultanously reached
637: by the QSO I-front. Right before the arrival of the front, one of them
638: is still neutral ($x^{\mr a}_{\mr{HI}}=1$) and the other is
639: significantly ionized ($x^{\mr b}_{\mr{HI}}<0.1$) 
640: by a local faint source.
641: Let us further assume that the local source in the second region 
642: has a softer spectrum than a QSO and thus that
643: the (initial) temperature of the free electrons is low, $T\lesssim10^4$ K.
644: While the I-front crosses the two regions,
645: energetic electrons are released 
646: by hydrogen atoms
647: and collisional excite the remaining neutral hydrogen. 
648: The \lya emissivity reaches a maximum at the center of the I-front where the 
649: actual neutral fraction is about $0.5 x_{\mr{HI}}$
650: and the temperature peaks. 
651: However, the mean energy  
652: of the free electrons and the number density of neutral-hydrogen atoms 
653: in the second region 
654: are both reduced by a factor 
655: $x^{\mr a}_{\mr{HI}}/x^{\mr b}_{\mr{HI}}>10$ with respect to the first region:
656: the \lya emissivity is thus reduced by more than a factor of 100.
657: 
658: A schematic description of how this effect can be used to map the HI distribution  
659: is shown in Figure \ref{f2} for a uniform IGM. The locally ionized
660: regions will appear as ``holes'' in the 2-dimensional \lya map.
661: \footnote{This strictly applies to young QSOs with I-fronts that are expanding 
662: at ultrarelativistic speeds. For older QSOs, the front will move faster in
663: the regions pre-ionized by other local sources and will produce bright \lya
664: emission from the entire (irregular) boundary of the HI region.}
665: Given the large angular extension of the HII region, mapping the \lya emission from 
666: the I-front behind a high-redshift QSO can 
667: effectively constrain the topology and the history of the reionization
668: process.
669:  
670: \section{Modelling the \lya signal}\label{secMod}
671: 
672: The rate equations that regulate the temperature and  
673: the ionization-fraction profiles inside the expanding I-front cannot be solved 
674: analytically. 
675: These quantities depend
676: on the local ionizing spectrum that is determined by non-local
677: radiative-transfer (RT) effects. 
678: For instance, higher energy photons will be absorbed at higher
679: column densities, causing an effective hardening of the ionizing spectrum 
680: at larger distances from the source.
681: Therefore, detailed modelling of the \lya emissivity 
682: requires a full RT transfer calculation for the ionizing radiation.
683: 
684: %\ \\ 
685: \subsection{Continuum Radiative Transfer}
686: 
687: To follow the radiative transfer of the continuum ionizing radiation
688: from the QSO,
689: we use a three-dimensional, photon-conserving, and time-dependent 
690: code based on a ray-tracing algorithm
691: (Cantalupo \& Porciani, in preparation). 
692: This code has been developed to study
693: the propagation of I-fronts in cosmological simulations and it
694: includes an adaptive refinement scheme of the computational grids 
695: (in space and time) on the front.
696: %
697: Among other features, it accounts for the presence of
698: multiple ionizing sources and for the propagation of ionizing radiation
699: produced by recombinations. 
700: %
701: The time-dependent rate equations include the 
702: evolution of HI, HeI and HeII. 
703: The gas temperature is computed including the
704: energy input due to
705: photoionizations and collisional ionizations of 
706: the three species and all the relevant
707: cooling processes (recombinations, collisional excitations/ionizations, dielectronic recombination of HeII,
708: bremsstrahlung, Compton and Hubble cooling). 
709: %
710: 
711: For simplicity, in the present study, we consider a single (steady) ionizing QSO 
712: surrounded by a uniform (cosmologically expanding) medium.
713: %
714: We follow the evolution of the I-front over a few $10^8$ yr,
715: i.e up to distances of
716: 70-140 comoving Mpc from the source (depending on
717: the QSO luminosity). 
718: The adaptive mesh refinement (AMR) scheme 
719: allow us to resolve the I-fronts with a large number of 
720: (optically-thin) cells of a few comoving kpc at all times.
721: %
722: A more general configuration based on the density and velocity fields
723: extracted from a hydro-dynamical simulation will be analyzed in future work.
724: %
725: 
726: 
727: 
728: \subsection{Finite light-travel time effects}
729: 
730: High-redshift QSOs are expected to produce 
731: relativistically expanding I-fronts for an important fraction of their lifetime 
732: (Shapiro et al. 2006).
733: This is difficult to follow with most RT codes
734: (see Iliev et al. 2006 for a recent compilation),
735: including our own,
736: in which the ray-tracing algorithm assumes an infinite light speed and
737: ionization fronts tend to propagate faster than light at early times.
738: %
739: A quick fix is obtained by limiting the ray-tracing up to a maximum distance
740: but this would lead to the loss of photon conservation.
741: We thus decided to follow a novel approach.
742: %
743: %
744: For a steady ionizing source in a uniform medium,
745: the correct speed of the I-front at a given radius, $v_{\mr I}(r_{\mr I})$, 
746: can be expressed in terms of its counterpart obtained assuming
747: an infinite speed of light, $v_{\mr I,NR}(r_{\mr I})$, by the relation
748: %
749: \begin{equation}
750: v_{\mr I}(r_{\mr I})=\frac{v_{\mr I,NR}(r_{\mr I})}{1+v_{\mr I,NR}(r_{\mr I})/c}\ ,
751: \end{equation}
752: %
753: regardless of the difference in physical times at which the I-front reaches 
754: the radius $r_{\mr{I}}$ (White et al. 2003; Yu 2005; Shapiro et al. 2006).
755: %
756: Therefore, we can get the correct velocity of the I-front 
757: by redefining the time variable as follows.
758: We associate to each time step used in the radiative transfer code
759: $\Delta t_{\mr{NR}}$ an actual time step $\Delta t_{\mr{R}}$
760: simply given by:
761: %
762: \begin{equation}
763: \Delta t_{\mr{R}}=\Delta t_{\mr{NR}} (1+v_{\mr{I},\mr{NR}}/c)\ .
764: \end{equation}
765: %
766: Note that this is an approximate solution, since the optical
767: depth of the photons is still computed at the time of their emission
768: and does not take 
769: into account the changes taking place in the medium during their
770: propagation time.
771: 
772: 
773: 
774: \begin{figure}[t!]
775: \epsscale{0.9}
776: \plotone{f3.eps}
777: \caption{Schematic view of the apparent I-front shape (solid line) around an isotropically emitting QSO.
778: The dotted circles represent the I-fronts 
779: in the QSO rest-frame at different epochs. The departure from the
780: spherical shape is due to the relativistic expansion speed of the I-front c
781: ombined with finite light-travel effects.
782: The apparent shape is rotationally symmetric along the quasar line of sight.  
783: The positions of the apparent and rest-frame I-front coincide for 
784: $\theta=\pi/2$, while the apparent
785: expansion of the I-front at $\theta=0$ is superluminal.}
786: \label{fig3}
787: \end{figure}
788: 
789: \begin{figure}[t!]
790: \epsscale{1.2}
791: \plotone{f4.eps}
792: \caption{Time-evolution of the apparent I-front 
793: distance from the QSO (in physical Mpc)
794: as a function of the angle $\theta$ (defined as in 
795: Figure \ref{fig3}) and the mean neutral fraction of the environment 
796: $\langle x_{\mr{HI}}\rangle$ (measured before the arrival of the I-front)
797: obtained with our radiative transfer code 
798: including finite light-travel effects.
799: Dotted lines are for reference and indicate propagation at 
800: the speed of light (black) and $\propto t^{1/3}$ (red).
801: The bottom right panel shows the evolution of the ``Contact'' radius, 
802: i.e. the position where the I-front 
803: closest to the observer ($\theta=0$) is reached by 
804: a photon emitted by the I-front behind the quasar ($\theta=\pi$). 
805: Note that the apparent $\theta=\pi/2$ and the
806: rest-frame I-front expansions are identical.
807:  All the panels assume a QSO that turns on at $z_{in}=6.5$
808: with total ionizing rate $\dot{N_{\gamma}}=10^{57}$ ph s$^{-1}$, far-UV spectral index $\alpha=-1.7$, and a clumping factor $C=35$. 
809: These parameters are compatible with observed QSOs showing
810: a Gunn-Peterson trough at $z>6$ (Yu \& Lu 2005).}
811: \label{f4}
812: \end{figure}
813: 
814: 
815: 
816: \subsection{The apparent evolution of the I-front}
817: 
818: The observed shape of the I-front 
819: (i.e. the relation between the QSO age, $t$, and the I-front distance 
820: from the QSO, $r_{\mr I}(t)$, as a function of the viewing angle $\theta$ defined
821: in Figure 3)
822: will differ from the proper shape because of finite 
823: light-travel effects (see equations (1)-(3) in Yu 2005).
824: In fact,
825: the photons we receive from the part of the I-front that is
826: moving towards us travel a smaller distance 
827: than the photons emitted behind the quasar. 
828: 
829: In Figure \ref{f4}, we show
830: the time-evolution of the apparent I-front position
831: for three different values of 
832: $\theta$ and $\langle x_{\mr{HI}}\rangle$ (the mean neutral fraction
833: of the medium surrounding the ionization source before the latter turns on).
834: This is the output of a simulation where a
835: luminous QSO (with a total ionizing rate $\dot{N_{\gamma}}=10^{57}$ ph s$^{-1}$,
836: and far-UV energy spectral index $\alpha=-1.7$)
837: turns on at $z_{\mr in}=6.5$ in an uniform medium at mean density 
838: with a clumping factor $C=35$.
839: This parameter set is consistent with observations of QSOs 
840: showing a GPT at $z>6$ (Yu \& Lu 2005).
841: %
842: From now on, the evolution of the I-front 
843: in a fully neutral 
844: medium ($\langle x_{\mr{HI}}\rangle=1$) 
845: surrounding this source will constitute our reference model.
846: 
847: The time-evolution of the I-front position obtained from our code 
848: is in good agreement with the approximate analytical solution by Yu (2005) 
849: and Shapiro et al. (2006).
850: For a given QSO ionizing rate, the apparent position of the I-front closest
851: to the observer (red short-dashed line in Figure 4) is very sensitive
852: to both the QSO age and $\langle x_{\mr{HI}}\rangle$. This is not the case for
853: the apparent position of the \lya emitting I-front \emph{behind} the QSO
854: (blue long-dashed line in Figure 4),
855: that determines fairly well the QSO age (at least for $\langle x_{\mr{HI}}\rangle\gtrsim 0.1$).
856: Combining the information from the \lya emitting I-front and the
857: position of the GPT in the QSO spectrum, we can significantly constrain both the
858: QSO age and $\langle x_{\mr{HI}}\rangle$.
859: 
860: At any given time, it is useful to define the
861: ``contact'' radius ($R_{\mr{t}}$) as
862:  the position where the I-front closest to the observer ($\theta=0$) 
863: is reached by 
864: the photons emitted by the I-front behind the quasar ($\theta=\pi$).
865: This quantity determines the strength of the damping-wing absorption
866: for the observed \lya signal from the I-front at $\theta=\pi$.
867: The evolution of the contact radius for our reference model is shown in 
868: the bottom right panel of Figure \ref{f4}.
869: Note that $R_{\mr{t}}$ keeps nearly constant at early times and grows afterwards.
870: Therefore, even when the rest-frame radius of the HII region is extremely 
871: small, the
872: \lya photons produced behind the QSO (and emitted towards us) 
873: will encounter the edge of the
874: ionized bubble when it has grown up to a much larger scale, 
875: substantially reducing the damping
876: wing absorption. 
877: 
878: 
879: \subsection{\lya emissivity from the I-front}
880: 
881: 
882: \begin{figure*}
883: \epsscale{0.9}
884: \plottwo{f5a.eps}{f5b.eps}
885: \caption{Radial (rest-frame) temperature and ionization
886: profiles for our reference model at two different 
887: expansion epochs.
888: Solid, dotted and dash-dotted lines respectively show
889: the HI, HeII and HeIII fractions.
890: The long-dashed line indicates 
891: the temperature expressed in units of $T_5=(T/10^5\,\mathrm{K})$. 
892: The short-dashed line represents the
893: ratio $n_{\mr{e}}/n_{\mr{HI}}$ which is larger than 1 inside the HII region 
894: because of the contribution of partially or totally ionized Helium. 
895: Note that the scales and the x-ranges of the
896: two plots are very different.}
897: \label{f5} 
898: \end{figure*}
899: 
900: 
901: \begin{figure}
902: \plotone{f6.eps}
903: \caption{\lya emissivity from collisional excitations (black solid line) 
904: and recombinations (blue short-dashed line) for our reference model at
905: different expansion epochs of the QSO I-front. 
906: For reference, we also show the evolution of the
907: temperature profiles (red long-dashed line).}
908: \label{f6} 
909: \end{figure}
910: 
911: 
912: Using the time evolution of the I-front derived in the previous section,
913: we want now to compute the \lya signal produced  
914: at given distance (or, equivalently, QSO age) from the central source. 
915: 
916: In Figure \ref{f5} 
917: we show the temperature and ionization profiles for our reference model
918: at two different epochs in the QSO rest-frame.
919: %
920: At early times (left panel), the HII/HI transition
921: is very sharp and the temperature (red, long-dashed line) 
922: slowly decreases going outwards until it suddenly 
923: drops down in the neutral region. 
924: In particular,
925: there is a small peak in the temperature profile at the position of the 
926: I-front ($x_{\mr{HI}}\sim0.5$, where $T\sim3\times10^4$ K) 
927: due to the radiative transfer effects that increase
928: the hardness of the ionizing spectrum. 
929: %
930: The HeII region extends outward of the HII zone, 
931: contributing to the free electron density (blue dotted line)
932: and to the temperature. 
933: %
934: At later stages (right-panel), the HII/HI transition 
935: is less sharp and 
936: the inner HII region is colder, $T\sim10^4$ K. 
937: However, the temperature peak at the I-front is 
938: more pronounced, reaching a value of $T\sim2\times10^4$ K. 
939: The HeIII region is now well inside the 
940: HII zone, with an important effect on the temperature profile (note the 
941: temperature drop at $\sim3$ pMpc).
942: 
943: In Figure \ref{f6} we show 
944: the evolution of the \lya emissivity from collisional 
945: excitations (black solid line), 
946: and radiative recombinations (blue, short-dashed line). 
947: As expected, 
948: \lya photons 
949: generated by collisional excitations are only produced on the I-front,
950: while those produced by radiative recombinations are emitted within
951: the entire HII region.
952: Photons from collisional excitations dominate the total \lya emissivity,
953: but their contribution diminishes with time
954: as a consequence of the decreasing I-front temperature
955: (red long-dashed line).
956: 
957: 
958: \subsection{Ly$\alpha$ Radiative Transfer}
959: 
960: \begin{figure*}
961: \epsscale{1.05}
962: \plottwo{f7a.eps}{f7b.eps}
963: \caption{Spectral evolution of the \lya emission from the expanding 
964: I-front (behind the QSO) for our reference model.
965: The different lines corresponds to 
966: the five different epochs for which the emissivities are shown in 
967: Figure \ref{f6}.
968: The spectra have been obtained with our \lya RT code
969: for a non-relativistic front (left) and including relativistic expansion (right).
970: Absorption by (residual) HI along the line of sight is included.}
971: \label{f7}
972: \end{figure*}
973: 
974: The density of neutral hydrogen within the I-front is high enough to make the
975: medium extremely optically-thick 
976: ($\tau_{\lyam}\gtrsim10^4$ at the line center) to the \lya photons generated 
977: via collisional excitations.
978: Moreover, the residual neutral hydrogen along the line of sight 
979: may substantially change the observed \lya line shape and flux. 
980: %
981: A numerical treatment of the transfer of resonant line radiation is
982: required to fully account for these effects.
983: %
984: We thus compute
985: the observed properties of the \lya emission from both 
986: collisional excitations and radiative recombinations 
987: using an updated version of   
988: the three-dimensional \lya Monte Carlo code 
989: presented in Cantalupo et al. (2005). 
990: %
991: In particular, the new version includes two improvements.
992: First, it uses a more precise (and computationally less expensive) 
993: analytical fit of the Voigt-Hijertig function proposed by Tepper-Garcia (2006).
994: Second, it performs the line transfer in a medium
995: with a spatially varying temperature.
996: % 
997: The presence of temperature gradients
998: not only changes
999: the \lya emissivity but also the shape of the absorption cross-section and 
1000: thus the scattering process.
1001: %
1002: In particular, a \lya photon generated in a region with high temperature can 
1003: more easily penetrate a colder zone before getting eventually absorbed. 
1004: Thus, the photons produced in the (warm) I-front will diffuse more
1005: in the neutral (cold) region than expected for a single-temperature medium. 
1006: 
1007: Given the small thickness of the I-front compared with the radius of the HII 
1008: region, we can work in the plane-parallel approximation.
1009: In the absence of Hubble expansion and assuming no absorption from
1010: residual HI in the ionized region, the \lya spectrum emerging from the I-front
1011: would be characterized by the symmetric, double-peaked emission 
1012: typical of plane-parallel, uniform-density media (e.g., Neufeld 1990).
1013: % 
1014: In this case, \lya photons would be emitted with a cosine law
1015: over a net solid angle of $\pi$ (see e.g. Gould \& Weinberg 1996).
1016: % 
1017: However, 
1018: both the Hubble expansion and the presence of residual HI
1019: increase the number of scatterings for the photons in 
1020: the blue peak which are then 
1021: absorbed and re-emitted over a wider solid angle 
1022: and frequency interval.
1023: %
1024: The net effect is a suppression of the blue peak in the spectrum.
1025: \begin{footnote}{This neglects peculiar velocities that, in general, 
1026: modify the spectral shape of either the red or the blue peak, as 
1027: shown in Cantalupo et al. 2005.}
1028: \end{footnote}
1029: %
1030: The strength of this suppression depends on the residual HI fraction 
1031: and, therefore, on the I-front distance from the QSO
1032: (or, equivalently, on the QSO age). 
1033: In most cases, the HI residual fraction is
1034: high enough to erase the blue peak. 
1035: %
1036: On the other hand,
1037: the damping wing absorption from HI lying outside the ionized region 
1038: only reduces the \lya line flux by a few per cent at all epochs (without 
1039: altering the line-shape). 
1040: 
1041: In the left panel of Figure \ref{f7}, 
1042: we show the evolution of the observed \lya spectra 
1043: obtained from the 
1044: I-front profiles and emissivities presented in Figure \ref{f6}. 
1045: At early stages, the blue peak 
1046: is still detectable but, in general,
1047: it is strongly suppressed. 
1048: Therefore, the emission consists of a single redshifted 
1049: peak with an asymmetric tail on the red side
1050: (i.e. towards the direction of the cold neutral gas). 
1051: %
1052: The broadening of the observed peak is
1053: significant compared with the sharpness of the emissivity in Figure \ref{f6}.
1054: %
1055: Both the suppression of 
1056: the blue-peak and the damping-wing absorption from 
1057: external HI reduce the integrated surface brightness with respect to
1058: the static, plane-parallel case. 
1059: Our simulation suggests that the observable integrated SB roughly 
1060: corresponds to
1061: 58 per cent of the ideal case (solid line in Figure \ref{f8}). 
1062: 
1063: 
1064: \subsection{\lya Radiative Transfer and Relativistic I-fronts}
1065: 
1066: In the previous section we treated the I-front as static, i.e. we
1067: assumed that the
1068: \lya photons leave the I-front on a time scale which is short
1069: with respect to the characteristic ionization time of the IGM.  
1070: This assumption, however, is not valid 
1071: for young QSOs
1072: when the I-front expansion is ultra-relativistic.
1073: %
1074: In this case, as the \lya photons scatter, the medium becomes optically thinner and the 
1075: optically-thick edge of the I-front moves outward. 
1076: If the I-front speed is close to the speed of light,
1077: the \lya photons find themselves inside the HII region after few scatterings. 
1078: Once there,
1079: they will be scattered over a wider solid angle ($4\pi$ at maximum) 
1080: with respect to the semi-infinite slab case.
1081: Therefore, ultra-relativistic I-fronts should emit 
1082: a \lya flux which is reduced by up to
1083: a factor of $4$ with respect to the static plane-parallel emission
1084: and up to a factor of $2$ 
1085: with respect to the non-relativistic case discussed in the previous section.
1086: 
1087: A full treatment of this effect would require to perform the
1088: \lya RT on a time-evolving grid, 
1089: with a significant increase of the algorithm complexity and of the requested 
1090: computational time. 
1091: %
1092: However, the fact that the I-front profile shifts in a self-similar way
1093: on the relevant time scales gives us the possibility to properly treat \lya 
1094: scattering in these extreme conditions.
1095: %
1096: We thus perform the \lya RT in the rest-frame of the I-front
1097: (where the HI density is nearly time-independent). 
1098: %
1099: In this reference frame, \lya photons are preferentially emitted 
1100: in the direction opposite to the propagation of the I-front.
1101: % 
1102: The resulting time evolution of the line shape and of the integrated
1103: SB are shown in the right panel
1104: of Figure \ref{f7} and in Figure \ref{f8}, respectively.
1105: %
1106: Both the integrated SB and the line broadening 
1107: are decreased by a factor $\sim2$ during the early stages of
1108: the I-front expansion, while they 
1109: coincide with the non-relativistic results at later epochs.
1110: %
1111: Note that the asymmetry of the line profile is reversed (skewed towards
1112: the blue side) at very early times.
1113: 
1114: \begin{figure}
1115: \plotone{f8.eps}
1116: \caption{Integrated \lya SB of the expanding I-front (behind the QSO) of our reference model at different expansion epochs,
1117: as obtained by our \lya RT simulations (solid circles). The dashed line represent the expected SB from 
1118: a plane-parallel (and optically thick) medium given by the integrated \lya emissivity inside the I-front
1119:  ($SB^{pp}_{\lyam}$). The solid line represent $0.58\times SB^{pp}_{\lyam}$. Finally, the solid triangles shows the result
1120: of our \lya RT that includes the effect of the I-front relativistic expansion (see text for details).}
1121: \label{f8}
1122: \end{figure}
1123: 
1124: 
1125: 
1126: \section{Discussion}
1127: 
1128: \begin{figure}
1129: \epsscale{1.1}
1130: \plotone{f9.eps}
1131: \caption{
1132: Expected \lya Surface Brightness 
1133: produced by the QSO I-front as it crosses an initial neutral region of the IGM at mean density. 
1134: In each panel, the reference model (i.e, total ionizing rate $\dot{N_{\gamma}}=10^{57}$ ph s$^{-1}$, 
1135: far-UV spectral index $\alpha=-1.7$, initial redshift $z_{in}=6.5$, clumping factor $C=35$) 
1136: is represented by the black solid line. The effect of the variation of one parameter with respect to the reference model 
1137: is shown in each panel with the red dotted and the blue dashed lines representing, respectively: 
1138: $\dot{N}_{\gamma}=10^{56}$ and $\dot{N}_{\gamma}=10^{58}$ (upper-left panel);
1139: $\alpha=-2.5$ and $\alpha=-1.0$ (upper-right panel); 
1140: $C=20$ and $C=50$ (bottom-left panel); $z_{in}=8$ and $z_{in}=10$ (bottom-right panel).
1141:  The corresponding SB for a region with initial neutral fraction $x_{\mr{HI}}$ and overdensity
1142: $(1+\delta)$ will scale approximatively like $x^2_{HI}(1+\delta)^{1/2}$ (see text for details).
1143:  Note that the observed SB at very early stages (e.g., $t_{\mr{Q}}\lesssim5$ Myr for the reference model) may be substantially
1144: lower than the values shown in the figure because of the high-relativistic expansion of the I-front (see text for details).
1145:  }\label{fD}
1146: \end{figure}
1147: 
1148: 
1149: 
1150: The parameters of the ionizing source in
1151: our reference model have been chosen based on observations of
1152: QSOs showing a GP trough at $z\sim 6$.
1153: %
1154: In this case, the \lya SB produced by the I-front which propagates 
1155: within a neutral patch of the IGM at mean density
1156: always lies between $10^{-21}-10^{-20}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$ 
1157: (see Figure \ref{f8}).
1158: %
1159: How does the \lya SB change with varying the QSOs and the surrounding medium properties? 
1160: Moreover, what is the redshift dependence of our results?
1161: In order to answer these questions, we performed a series of simulations in which we kept fixed all
1162: the parameters of the reference model but one.
1163:  The resulting SBs (for a neutral patch of the IGM at mean density) are shown in Figure \ref{fD}.
1164: In each panel, we indicate on the top-left the varying parameter (see caption for details) and 
1165: the reference model is always represented by the black solid line.
1166: The SB is obtained from the integrated emissivity and corrected for absorption and
1167: RT effects, as described in \S 4. The correction for the 
1168: relativistic I-front expansion in the \lya RT is not included. However, as shown in \S 4.6, the
1169: decrease in the integrated SB is at maximum a factor of 2 when the expansion is ultra-relativistic and
1170: it is negligible at later stages. 
1171: 
1172:  From the top-left panel in Figure \ref{fD}, we notice that, for a given QSO age, $t_{\mr{Q}}$, the expected
1173: \lya SB is roughly proportional to $\dot{N}_{\gamma}^{1/3}$, unless the I-front expansion 
1174: is still ultra-relativistic (i.e. $t_{\mr{Q}}\lesssim 5$ Myr for our reference model). 
1175: Note that this corresponds to a linear scaling with the ionizing flux at the I-front position.
1176: After the ultra-relativistic period,
1177: and before recombinations start to play a significant role, 
1178: the time evolution of the \lya SB is well represented by a power law (SB $\propto t_{\mr{Q}}^{-1}$)
1179: independently of the QSO luminosity.
1180:  The slope of the power law is mainly determined by the spectral index of the QSO spectrum 
1181: (see top-right panel in Figure \ref{fD}), as expected by the strong temperature dependence
1182: of the collisional-excitation rate. In particular, we found that the SB evolves proportionally
1183: to $t_{\mr{Q}}^{-(\alpha+2.7)}$. 
1184: Varying the clumping factor ($20<C<50$) 
1185: has little effect on the SB evolution (bottom-left panel in Figure \ref{fD}). 
1186: This is because the increase in the collisional excitation rate due to a more clumpy medium
1187: is partially balanced by the fact that the medium becomes colder (mainly because of the
1188: cooling from collisional excitations themselves). 
1189: This self-regulation is more efficient as the I-front slows down. 
1190: 
1191: Finally, in the bottom-right panel of Figure \ref{fD}, we show that
1192: the expected SB$_{\lyam}$, for a given $t_{\mr{Q}}$, decreases proportionally to $(1+z)^{-2}$ (instead of
1193: $(1+z)^{-1}$ as expected for a medium with fixed temperature, see \S 3.1). 
1194: Note that, in order to evidence this relation, the y-axis of the figure has been properly 
1195: normalized. 
1196: 
1197: In the models discussed so far we only considered the \lya SB generated within a neutral patch
1198: of the IGM at mean density. As we already shown in \S 3, 
1199: the expected emission is proportional to the square of the initial neutral fraction ($x_{\mr{HI}}$) 
1200: (i.e., the medium neutral fraction before the arrival of the I-front).
1201: %
1202: In order to estimate how the expected SB changes with the local overdensity 
1203: ($\delta\equiv(\rho-\rho_0)/\rho_0$), 
1204: we performed a series of simulations placing slabs with different values of $(1+\delta)$ at a given distance
1205: from the QSO. We found that, in overdense regions ($\delta\lesssim50$),
1206: the expected \lya SB is nearly proportional to $(1+\delta)^{1/2}$. 
1207: In summary, assuming $20\lesssim C\lesssim 50$, 
1208: the \lya SB produced by a QSO I-front is of the order of:
1209: %
1210: \begin{equation}
1211: \begin{split}
1212: SB_{\lyam }&\sim \ 10^{-20} \cdot x^{2}_{\mathrm{HI}} (1+\delta)^{1/2} 
1213:                  \cdot \left[\frac{t_{\mr{Q}}}{10 \mathrm{Myr}}\right]^{-1} \\ 
1214: 	   &\times \left[\frac{\dot{N}_{\gamma}}{10^{57} \mathrm{s}^{-1}}\right]^{1/3} 
1215:                   \left[\frac{1+z}{7.5}\right]^{-2}  \mathrm{erg}\ \mathrm{s}^{-1} 
1216:                     \mathrm{cm}^{-2} \mathrm{arcsec}^{-2}\ , 
1217: \end{split}
1218: \label{SBdep}
1219: \end{equation}
1220: %
1221: for $\alpha\simeq-1.7$, $5\lesssim t_{\mr{Q}}\lesssim 100$ Myr, and
1222: $\delta\lesssim50$.
1223: Note, however, that QSOs with harder ionizing spectra ($\alpha\simeq
1224: -1$) produce significantly brighter I-fronts (blue dashed line in Figure 9).
1225: 
1226: We do not consider here the \lya SB produced within more overdense regions. In this case, 
1227: a proper treatment should also consider hydrodynamical effects.
1228: Fluorescent \lya emission from
1229: overdense ($\delta>100$) 
1230: regions could be as bright as the I-front emission at mean density. 
1231: %
1232: However, these dense regions will appear significantly more compact. 
1233: %
1234: An extensive study of the \lya emission produced by overdense regions 
1235: and the effect 
1236: of IGM inhomogeneities will be presented in a future paper (Cantalupo et al., in preparation). 
1237: 
1238: 
1239: 
1240: 
1241: \subsection{Uncertainties and limitations}
1242: 
1243: For very young QSOs, when the expansion-speed of the I-front ($v_{\mr I}$) is very close to the speed of light,
1244: our model might overestimate the actual \lya emission.
1245: %
1246: In this case,
1247: the photo-ionization time scale ($t_{\mr{ion}}$) at  the I-front 
1248: (roughly given by $t_{\mr{ion}}\sim S/v_{\mr I}$, where $S$ is the I-front thickness) can be
1249: shorter than the collisional-excitation time scale $t_{\mr{CE}}\sim(C n_{\mr{HI}}q(T))^{-1}$.
1250: %
1251: In our reference model, this effect is important for $t_{\mr{Q}}\lesssim 5$ Myr (and earlier at higher redshifts). 
1252: %
1253: In overdense regions the I-front is slower and collisional excitations are 
1254: faster with respect to the IGM at mean density. Therefore, the \lya emission produced in these regions
1255: is less sensitive to this effect. 
1256: %
1257: On the other hand, I-fronts expanding around young quasars 
1258: lie very close to the source. For instance, the 
1259: I-front behind a bright QSO at $z\sim6.5$ with $t_{\mr{Q}}\sim5$ Myr appears at 
1260: a proper distance of 1 Mpc (see Figure 4). If the opening angle of the QSO emission cone 
1261: is $45^o$ (with respect of the line of sight), the 
1262: total projected area covered by the I-front will extend over $\sim12$ arcmin$^2$ 
1263: (or $\sim 60$ arcmin$^2$ if the quasar emits isotropically).
1264: Therefore,  older QSOs provide a larger projected area for detecting the 
1265: I-front emission and a better constraint
1266: on the properties of both the quasar emission and the surrounding IGM.
1267: Moreover, older and luminous QSOs produce HII bubbles that are
1268: larger than possible pre-existing ionized regions generated 
1269: from clustered star-forming galaxies.
1270: 
1271: Our RT simulations do not consider X-ray radiation produced by the QSO.
1272: However, X-ray photons travel unimpeded well 
1273: beyond the position of the emitting I-front and do not modify the properties of the \lya signal. 
1274: %
1275: In the presence of pre-existing X-ray background, the initial gas temperature will be higher than  
1276: assumed in our models ($T\sim100$ K) and this will slightly increase 
1277: the \lya emission from collisional excitations (provided
1278: that the medium is still significantly neutral before the arrival of the I-front).
1279: 
1280: For simplicity, we have assumed a steady ionizing emission rate for the quasars. We also did not account
1281: for the possible contribution of other local sources that
1282: can change the apparent shape of the QSO I-front and the properties of the 
1283: surrounding medium (see e.g., Yu 2005; Yu \& Lu 2005; Wyithe \& Loeb 2007; Lidz et al. 2007). 
1284: Addressing these effects 
1285: requires performing high-resolution RT within a fully cosmological simulation and is beyond the scope of the present study.
1286: 
1287: 
1288: \subsection{Detectability}
1289: 
1290: To date, eight QSOs have been detected at redshift $z>6.1$ (Fan et al. 2003, 2004; Willot et al. 2007) and all of
1291: them have significant ($\Delta z>0.1$) dark absorption 
1292: troughs in their spectra in proximity of the QSO redshift.
1293:  Although this is not necessarily evidence for a surrounding  
1294: neutral medium, these objects represent the best targets for detecting the I-front \lya emission.
1295: 
1296: Let us take, as a practical example, the case of QSO J1148+5251. This quasar has the 
1297: most accurate redshift measurement($z=6.419\pm0.001$, Walter et al. 2003) and shows complete 
1298: Ly$\alpha$-Ly$\beta$ troughs corresponding to
1299: an apparent proximity region of $R_{\mr s}\sim5$ physical Mpc (Wyithe et al. 2004, Yu \& Lu 2005).
1300:  Assuming that the medium surrounding the QSO was still significantly neutral ($\langle x_{\mr{HI}}\rangle > 0.1$) 
1301: before the QSO turned on, from the estimated ionizing rate  
1302: ($\dot{N}_{\gamma}\sim 2\times 10^{57}$ ph s$^{-1}$, see e.g. Yu \& Lu 2005) we 
1303: expect $8\lesssim t_{\mr{Q}}\lesssim 30$ Myr (see Figure \ref{f4}).
1304:  This corresponds, for a QSO spectral index $\alpha=-1.7$, to
1305: $0.5\lesssim SB_{\lyam}\lesssim 1.5\times 10^{-20}$ \ergscmarcsec (see equation (\ref{SBdep}) 
1306: and Figure \ref{fD}) for   
1307: a neutral patch of the IGM at mean density. 
1308: 
1309:  Is this signal detectable? Although faint (about 3 orders of magnitude below the 
1310: sky background), the expected emission may extend over a projected area on the sky up to a few hundred square arcmin.
1311: By means of moderately-high resolution spectroscopy (R=1000-3000) and integrating the signal over a 
1312: fraction of the slit length, single neutral patches of the IGM (even at mean density) can be already
1313: detected from the ground with current facilities and long integration times ($T\sim 40$ hr)
1314: if they extends over few arcminutes scales (or smaller for slightly overdense regions).
1315: For instance, the expected signal-to-noise (S/N) ratio 
1316: corresponding to the above example and for a ground-based
1317: observation (in the atmospheric window at $\sim0.9\mu m$) 
1318: will be of the order of:
1319: %
1320: \begin{equation}
1321: \begin{split}
1322: S/N   &\sim 7\times  C_{\rm f} \left[\frac{D}{8\ \mathrm{m}}\right] \left[\frac{\zeta}{0.8}\right]  
1323:        \left[\frac{f}{0.25}\right]^{1/2} \\
1324:     & \left[\frac{\Delta_s}{3\mathrm{\AA}}\right]^{-1/2}
1325:        \left[\frac{T}{40\ \mathrm{hr}}\right]^{1/2} \left[\frac{\Delta\Omega}{180\ \mathrm{arcsec}^2}\right]^{1/2}\ , 
1326: \end{split}
1327: \end{equation}
1328: % 
1329: %
1330:  where $C_{\rm f}$ is the slit covering factor (i.e. the fraction of the slit where the \lya emission is present),
1331:  $D$ is the telescope diameter, $\zeta$ the atmospheric transmission, $f$ the system efficiency,
1332:  $\Delta_s$ the spectral bin, $T$ the integration time and $\Delta\Omega$ the area of the sky covered
1333: by the slit. For a slit width of $1''$, the above $S/N$ implies that we can detect
1334: a single neutral patch of the IGM (i.e., $C_{\rm f}=1$) at mean density if it has at least 
1335: a linear size of $l\sim3'$ (corresponding to $\sim1$ pMpc at $z\sim6.4$).
1336: 
1337:  Even if single neutral patches cannot be detected,
1338: a proper integration over the total slit length (that can be substantially increased, for instance, using
1339: a multi-slit plus filter technique, see e.g., Cantalupo et al. 2007) 
1340: may reveal at least the position of the \lya emitting I-front.
1341:  Once the I-front has been located, we can use its apparent distance from the QSO 
1342: to determine the QSO age and to constrain the value of $\langle x_{\mr{HI}}\rangle$
1343: (see Figure 4).
1344: Possible limitations are given by the unknown opening angle of the QSO emission (that can reduce
1345: the projected area over which the I-front may extend) and the difficulties of
1346: subtracting the sky lines.  
1347: 
1348: A significant improvement is expected from the next generation of space telescopes like JWST,
1349: although the background (i.e., the Zodiacal light) will still be 2 orders of magnitude 
1350: brighter than the expected signal.
1351: With JWST, we will be able to detect single neutral patches of the IGM on smaller scales 
1352: than allowed from the ground.
1353: Moreover observations will not be limited to the narrow redshift ranges permitted
1354: by the atmospheric windows. 
1355: %
1356: For a given QSO age, the \lya SB decreases proportionally to $(1+z)^2$ (instead of the usual $(1+z)^4$).
1357: The next generation of space telescopes will thus be able
1358: to map HI distribution during the EoR in a broad redshift range.
1359: 
1360: 
1361: \section{Summary and Conclusions}
1362: 
1363: We have presented a new method to directly map the neutral hydrogen 
1364: distribution during the reionization epoch, 
1365: and to measure the age of the highest-redshift QSOs.
1366: %
1367: We have shown that collisional excitations are an
1368: efficient mechanism to produce \lya photons, even in the low density IGM, 
1369: provided that nearly 50 per cent of the hydrogen is neutral and the
1370: local temperature is as high as $T\gtrsim2\times10^4$ K.
1371: %
1372: These conditions are achieved within the I-fronts produced by luminous UV
1373: sources, like QSOs, as they expand into the surrounding IGM.
1374: %
1375: %
1376: The observable \lya photons are
1377: those emitted by the I-front \emph{behind} the QSO 
1378: (with respect to the observer).
1379: These photons can cross the large HII region lying in front of
1380: their emission point without being significantly scattered.
1381: %
1382: The emerging signal
1383: traces the HI distribution at the location of the QSO I-front
1384: (similarly to future 21-cm tomography) since
1385: the expected SB
1386: roughly scales as the square of the (initial) local  
1387: HI fraction.
1388: %
1389: The angular distribution of the \lya emission and its
1390: distance from the QSO
1391: constrains both the properties of the source 
1392: (i.e. the QSO opening angle and age) 
1393: and of the surrounding medium (i.e., the 
1394: average neutral fraction).
1395:  
1396: Using detailed radiative transfer simulations that include finite light-speed
1397: effects, we have shown that the expected  emission appears as a 
1398: single (broad) line with a width of
1399: $100-200$ km s$^{-1}$.
1400: %
1401: The \lya SB of a typical QSO I-front
1402: that propagates at $z_{\rm Q}=6.5$ within
1403: a fully neutral patch of the IGM at mean density 
1404: lies in the range 
1405: ${\mr SB}_{\lyam}\sim10^{-21}-10^{-20}$ erg s$^{-1}$ cm$^{-2}$ arcsec$^{-2}$.
1406: %
1407: QSOs with very hard spectra ($\alpha\sim-1$) produce a significantly brighter 
1408: signal at early phases ($t_{\mr{Q}}\lesssim10$ Myr). 
1409: %
1410: Interestingly, for a given QSO age,
1411: the \lya SB of the I-front scales as $(1+z_{\rm Q})^{-2}$.
1412: %
1413: 
1414: The signal from neutral patches of the IGM extending over a few arcmin scales
1415: is already detectable by current ground facilities
1416: with the use of moderately/high resolution spectroscopy.
1417: %
1418: The next generation of space and ground based telescopes will allow us
1419: to draw high signal-to-noise maps of the HI distribution with higher 
1420: angular resolution.
1421: %
1422: Combining this information with
1423: a measure of the I-front position, 
1424: we will be able to simultaneously constrain the reionization history 
1425: and the emission properties of the highest-z QSOs.
1426: 
1427: \acknowledgements
1428: SC is supported by the Swiss National Science Fundation.
1429: 
1430: 
1431: 
1432: \begin{thebibliography}{}
1433: 
1434: \bibitem[Abel \& Haehnelt(1999)]{1999ApJ...520L..13A} Abel, T., \& 
1435: Haehnelt, M.~G.\ 1999, \apjl, 520, L13 
1436: 
1437: \bibitem[Baker \& Menzel(1938)]{1938ApJ....88...52B} Baker, J.~G., \& 
1438: Menzel, D.~H.\ 1938, \apj, 88, 52 
1439: 
1440: \bibitem[Becker et al.(2001)]{2001AJ....122.2850B} Becker, R.~H., et al.\ 
1441: 2001, \aj, 122, 2850 
1442: 
1443: \bibitem[Bolton \& Haehnelt(2007)]{2007MNRAS.374..493B} Bolton, J.~S., \& 
1444: Haehnelt, M.~G.\ 2007, \mnras, 374, 493 
1445: 
1446: \bibitem[Cantalupo et al.(2005)]{2005ApJ...628...61C} Cantalupo, S., 
1447: Porciani, C., Lilly, S.~J., \& Miniati, F.\ 2005, \apj, 628, 61 
1448: 
1449: \bibitem[Cantalupo et al.(2007)]{2007ApJ...657..135C} Cantalupo, S., Lilly, 
1450: S.~J., \& Porciani, C.\ 2007, \apj, 657, 135 
1451: 
1452: \bibitem[Fan et al.(2002)]{2002AJ....123.1247F} Fan, X., Narayanan, V.~K., 
1453: Strauss, M.~A., White, R.~L., Becker, R.~H., Pentericci, L., \& Rix, H.-W.\ 
1454: 2002, \aj, 123, 1247 
1455: 
1456: \bibitem[Fan et al.(2003)]{2003AJ....125.1649F} Fan, X., et al.\ 2003, \aj, 
1457: 125, 1649 
1458: 
1459: \bibitem[Fan et al.(2006)]{2006AJ....132..117F} Fan, X., et al.\ 2006, \aj, 
1460: 132, 117 
1461: 
1462: \bibitem[Fan et al.(2006)]{2006ARA&A..44..415F} Fan, X., Carilli, C.~L., \& 
1463: Keating, B.\ 2006, \araa, 44, 415 
1464: 
1465: \bibitem[Furlanetto et al.(2004)]{2004ApJ...613....1F} Furlanetto, S.~R., 
1466: Zaldarriaga, M., \& Hernquist, L.\ 2004, \apj, 613, 1 
1467: 
1468: \bibitem[Furlanetto et al.(2006)]{2006PhR...433..181F} Furlanetto, S.~R., 
1469: Oh, S.~P., \& Briggs, F.~H.\ 2006, \physrep, 433, 181 
1470: 
1471: \bibitem[Giovanardi et al.(1987)]{1987A&AS...70..269G} Giovanardi, C., 
1472: Natta, A., \& Palla, F.\ 1987, \aaps, 70, 269 
1473: 
1474: \bibitem[Gnedin \& Ostriker(1997)]{1997ApJ...486..581G} Gnedin, N.~Y., \& 
1475: Ostriker, J.~P.\ 1997, \apj, 486, 581 
1476: 
1477: \bibitem[Gould \& Weinberg(1996)]{1996ApJ...468..462G} Gould, A., \& 
1478: Weinberg, D.~H.\ 1996, \apj, 468, 462 
1479: 
1480: \bibitem[Gunn \& Peterson(1965)]{1965ApJ...142.1633G} Gunn, J.~E., \& 
1481: Peterson, B.~A.\ 1965, \apj, 142, 1633 
1482: 
1483: \bibitem[Haiman(2002)]{2002ApJ...576L...1H} Haiman, Z.\ 2002, \apjl, 576, 
1484: L1 
1485: 
1486: \bibitem[Hogan \& Weymann(1987)]{1987MNRAS.225P...1H} Hogan, C.~J., \& 
1487: Weymann, R.~J.\ 1987, \mnras, 225, 1P 
1488: 
1489: \bibitem[Iliev et al.(2006)]{2006MNRAS.371.1057I} Iliev, I.~T., et al.\ 
1490: 2006, \mnras, 371, 1057 
1491: 
1492: 
1493: \bibitem[Lidz et al.(2007)]{2007astro.ph..3667L} Lidz, A., McQuinn, M., 
1494: Zaldarriaga, M., Hernquist, L., \& Dutta, S.\ 2007, ArXiv Astrophysics 
1495: e-prints, arXiv:astro-ph/0703667 
1496: 
1497: \bibitem[Kashikawa et al.(2006)]{2006ApJ...648....7K} Kashikawa, N., et 
1498: al.\ 2006, \apj, 648, 7 
1499: 
1500: \bibitem[Madau \& Rees(2000)]{2000ApJ...542L..69M} Madau, P., \& Rees, 
1501: M.~J.\ 2000, \apjl, 542, L69 
1502: 
1503: \bibitem[Madau et al.(1997)]{1997ApJ...475..429M} Madau, P., Meiksin, A., 
1504: \& Rees, M.~J.\ 1997, \apj, 475, 429 
1505: 
1506: \bibitem[Malhotra \& Rhoads(2006)]{2006ApJ...647L..95M} Malhotra, S., \& 
1507: Rhoads, J.~E.\ 2006, \apjl, 647, L95 
1508: 
1509: \bibitem[Martin(1988)]{1988ApJS...66..125M} Martin, P.~G.\ 1988, \apjs, 66, 
1510: 125 
1511: 
1512: \bibitem[Maselli et al.(2007)]{2007MNRAS.376L..34M} Maselli, A., Gallerani, 
1513: S., Ferrara, A., \& Choudhury, T.~R.\ 2007, \mnras, 376, L34 
1514: 
1515: \bibitem[McQuinn et al.(2007)]{2007arXiv0704.2239M} McQuinn, M., Hernquist, 
1516: L., Zaldarriaga, M., \& Dutta, S.\ 2007, ArXiv e-prints, 704, 
1517: arXiv:0704.2239 
1518: 
1519: \bibitem[Mesinger \& Haiman(2004)]{2004ApJ...611L..69M} Mesinger, A., \& 
1520: Haiman, Z.\ 2004, \apjl, 611, L69 
1521: 
1522: \bibitem[Miralda-Escude \& Rees(1994)]{1994MNRAS.266..343M} Miralda-Escude, 
1523: J., \& Rees, M.~J.\ 1994, \mnras, 266, 343 
1524: 
1525: \bibitem[Miralda-Escude \& Rees(1998)]{1998ApJ...497...21M} Miralda-Escude, 
1526: J., \& Rees, M.~J.\ 1998, \apj, 497, 21 
1527: 
1528: \bibitem[Neufeld(1990)]{1990ApJ...350..216N} Neufeld, D.~A.\ 1990, \apj,
1529: 350, 216
1530: 
1531: \bibitem[Oh \& Furlanetto(2005)]{2005ApJ...620L...9O} Oh, S.~P., \& 
1532: Furlanetto, S.~R.\ 2005, \apjl, 620, L9 
1533: 
1534: \bibitem[Osterbrock(1989)]{1989agna.book.....O} Osterbrock, D.~E.\ 1989,  
1535: University Science Books, 1989
1536: 
1537: \bibitem[Page et al.(2007)]{2007ApJS..170..335P} Page, L., et al.\ 2007, 
1538: \apjs, 170, 335 
1539: 
1540: \bibitem[Pengelly(1964)]{1964MNRAS.127..145P} Pengelly, R.~M.\ 1964, 
1541: \mnras, 127, 145 
1542: 
1543: \bibitem[Santos(2004)]{2004MNRAS.349.1137S} Santos, M.~R.\ 2004, \mnras, 
1544: 349, 1137 
1545: 
1546: \bibitem[Shapiro \& Giroux(1987)]{1987ApJ...321L.107S} Shapiro, P.~R., \& 
1547: Giroux, M.~L.\ 1987, \apjl, 321, L107 
1548: 
1549: \bibitem[Shapiro et al.(2006)]{2006ApJ...648..922S} Shapiro, P.~R., Iliev, 
1550: I.~T., Alvarez, M.~A., \& Scannapieco, E.\ 2006, \apj, 648, 922 
1551: 
1552: \bibitem[Songaila(2004)]{2004AJ....127.2598S} Songaila, A.\ 2004, \aj, 127, 
1553: 2598 
1554: 
1555: \bibitem[Spergel et al.(2007)]{2007ApJS..170..377S} Spergel, D.~N., et al.\ 
1556: 2007, \apjs, 170, 377 
1557: 
1558: \bibitem[Taniguchi et al.(2005)]{2005PASJ...57..165T} Taniguchi, Y., et 
1559: al.\ 2005, \pasj, 57, 165 
1560: 
1561: \bibitem[Telfer et al.(2002)]{2002ApJ...565..773T} Telfer, R.~C., Zheng, 
1562: W., Kriss, G.~A., \& Davidsen, A.~F.\ 2002, \apj, 565, 773 
1563: 
1564: \bibitem[Tepper-Garc{\'{\i}}a \& 
1565: Fritze-v.~Alvensleben(2006)]{2006MNRAS.369.2025T} Tepper-Garc{\'{\i}}a, T., 
1566: \& Fritze-v.~Alvensleben, U.\ 2006, \mnras, 369, 2025 
1567: 
1568: \bibitem[Tran et al.(2004)]{2004ApJ...612L..89T} Tran, K.-V.~H., Lilly, 
1569: S.~J., Crampton, D., \& Brodwin, M.\ 2004, \apjl, 612, L89 
1570: 
1571: \bibitem[Walter et al.(2003)]{2003Natur.424..406W} Walter, F., et al.\ 
1572: 2003, \nat, 424, 406 
1573: 
1574: \bibitem[White et al.(2003)]{2003AJ....126....1W} White, R.~L., Becker, 
1575: R.~H., Fan, X., \& Strauss, M.~A.\ 2003, \aj, 126, 1 
1576: 
1577: \bibitem[White et al.(2005)]{2005AJ....129.2102W} White, R.~L., Becker, 
1578: R.~H., Fan, X., \& Strauss, M.~A.\ 2005, \aj, 129, 2102 
1579: 
1580: \bibitem[Willott et al.(2007)]{2007arXiv0706.0914W} Willott, C.~J., et al.\ 
1581: 2007, ArXiv e-prints, 706, arXiv:0706.0914 
1582: 
1583: \bibitem[Wyithe \& Loeb(2004)]{2004Natur.432..194W} Wyithe, J.~S.~B., \& 
1584: Loeb, A.\ 2004, \nat, 432, 194 
1585: 
1586: \bibitem[Wyithe et al.(2005)]{2005ApJ...628..575W} Wyithe, J.~S.~B., Loeb, 
1587: A., \& Carilli, C.\ 2005, \apj, 628, 575 
1588: 
1589: \bibitem[Wyithe \& Loeb(2007)]{2007MNRAS.374..960W} Wyithe, J.~S.~B., \& 
1590: Loeb, A.\ 2007, \mnras, 374, 960 
1591: 
1592: \bibitem[Yu(2005)]{2005ApJ...623..683Y} Yu, Q.\ 2005, \apj, 623, 683 
1593: 
1594: \bibitem[Yu \& Lu(2005)]{2005ApJ...620...31Y} Yu, Q., \& Lu, Y.\ 2005, 
1595: \apj, 620, 31 
1596: 
1597: 
1598: \end{thebibliography}
1599: 
1600: 
1601: 
1602: 
1603: 
1604: 
1605: \end{document}
1606: