1: \documentstyle[times,graphics,astrobib,amssymb]{mn2e}
2: %\documentstyle[times,graphics,astrobib,amssymb,referee]{mn2e}
3:
4:
5: \newcommand{\be}{\begin{equation}}
6: \newcommand{\e}{\end{equation}}
7: \newcommand{\bear}{\begin{eqnarray}}
8: \newcommand{\ear}{\end{eqnarray}}
9: \newcommand{\nline}{\nonumber \\}
10: \newcommand{\f}{\frac}
11: \newcommand{\de}{{\rm d}}
12: \newcommand{\del}{\partial}
13: %\newcommand{\la}{\langle}
14: %\newcommand{\ra}{\rangle}
15:
16: \begin{document}
17:
18: \title[Reionization sources]
19: {Searching for the reionization sources}
20: \author[Choudhury \& Ferrara]
21: {T. Roy Choudhury$^{1}$\thanks{E-mail: chou@ast.cam.ac.uk}~
22: and
23: A. Ferrara$^{2}$\thanks{E-mail: ferrara@sissa.it}\\
24: $^{1}$Institute of Astronomy, Madingley Road, Cambridge CB3 0HA, UK\\
25: $^{2}$SISSA/ISAS, via Beirut 2-4, 34014 Trieste, Italy}
26:
27:
28:
29: \maketitle
30:
31: \date{\today}
32:
33: \begin{abstract}
34: Using a reionization model simultaneously accounting for a number of experimental data sets, we investigate the nature and properties
35: of reionization sources. Such model predicts that hydrogen reionization starts at $z \approx 15$, is initially driven by metal-free
36: (PopIII) stars, and is 90\% complete by $z \approx 8$. We find that a fraction $f_\gamma >80$\% of the ionizing
37: power at $z \ge 7$ comes from haloes of mass $M< 10^9 M_{\odot}$ predominantly harboring PopIII stars; a
38: turnover to a PopII-dominated phase occurs shortly after, with this population, residing in $M>10^9 M_\odot$ haloes,
39: yielding $f_\gamma\approx 60$\% at $z=6$. Using Lyman-break broadband dropout techniques, $J$-band detection of sources
40: contributing to 50\% (90\%) of the ionizing power at $z \sim 7.5$ requires to reach a magnitude
41: $J_{110,{\rm AB}} = 31.2 (31.7)$, where $\sim 15 (30)$ (PopIII) sources/arcmin$^2$ are predicted. We conclude
42: that $z>7$ sources tentatively identified in broadband surveys are relatively massive ($M\approx 10^9 M_\odot$) and rare objects
43: which are only marginally ($\approx 1$\%) adding to the reionization photon budget.
44: \end{abstract}
45: \begin{keywords}
46: intergalactic medium cosmology: theory large-scale structure of Universe.
47: \end{keywords}
48: \section{Introduction}
49:
50:
51: The study of reionization received a big boost due to availability of a variety of observational data accumulated over the past few years
52: (for reviews see \citeNP{fck06} and \citeNP{cf06a}); additional progresses are soon expected from a number of different ground-based (LOFAR, MWA,
53: PAST, SKA, ALMA) and space-born (JWST, PLANCK, GLAST) experiments.
54:
55: The available results have allowed us to build self-consistent reionization scenarios that are able to account simultaneously
56: for a number of observables (redshift evolution of Lyman-limit absorption systems, GP and electron scattering optical depths,
57: mean temperature of the intergalactic medium (IGM), and cosmic star formation history;
58: \nocite{cf05,cf06b} Choudhury \& Ferrara 2005, 2006; hereafter CF05 and CF06 respectively). To summarize the emerging picture,
59: the most favorable model is one in which hydrogen reionization was an extended process starting around $z \approx 15$ and
60: being 90\% complete by $z \approx 8$. This (early) reionization model was also shown not to be in conflict with the Gunn-Peterson
61: optical depth evolution deduced from QSO absorption line experiments at $z \gtrsim 6$ both by using the statistics of dark gaps in
62: the Ly$\alpha$ transmitted flux \cite{gcf06} and through radiative transfer simulations of ionized regions around QSOs \cite{mgfc07}.
63: According to the model, reionization is initially driven by metal-free stars in low mass ($M < 10^8 M_{\odot}$) haloes; the
64: conditions for the formation of these objects are soon erased by the combined action of chemical and radiative feedback at $z < 10$.
65: As a consequence, the photoionizing power (and therefore integrated luminosity) of these sources is most significant around $z \approx 8-12$.
66:
67: In spite of this successful overall picture, relatively less attention has been devoted so far to the observable properties of the sources
68: responsible for cosmic reionization. Our main aim in this work is to fill such gap by providing quantitative guidelines for observers searching
69: for the first cosmic light sources. Specifically, we use the CF05 and CF06 model to estimate the IR fluxes and magnitude-limited counts
70: of the primary reionization sources.
71:
72: \begin{figure*}
73: \rotatebox{270}{\resizebox{0.4\textwidth}{!}{\includegraphics{halofraction.ps}}}
74: \caption{{\it Left}: Number of ionizing photons per H-atom contributed by haloes of different mass
75: in a fraction of the Hubble time $t_H(z)$ equal to the recombination time $t_{\rm rec}(z)$ as a function of
76: redshift $z$ for different halo masses and for different stellar populations. The solid and dot-dashed pair of
77: curves, from top to bottom, represent haloes of masses $10^7 M_{\odot} < M < 10^8 M_{\odot}$, $10^8 M_{\odot} < M < 10^9 M_{\odot}$,
78: and $M > 10^9 M_{\odot}$ respectively. The dashed line represent the evolution of the volume filling factor $Q_{\rm HII}$ of ionized regions.
79: {\it Right}: Cumulative fraction of the ionizing power $f_{\gamma}$ contributed by haloes of mass $>M$. The curves from
80: right to left correspond to $z = 6,7,8,9,10$ respectively.
81: }
82: \label{fig:halofrac}
83: \end{figure*}
84: %
85: %
86: \vspace{-0.5cm}
87: \section{Basic features of the model}
88: %
89: %
90: The main features of the semi-analytical model used in this work\footnote{
91: Throughout the paper, we use the best-fit cosmological parameters from the 3-year WMAP data \cite{sbd++06}, i.e.,
92: a flat universe with $\Omega_m = 0.24$, $\Omega_{\Lambda} = 0.76$, and $\Omega_b h^2 = 0.022$, and $h=0.73$. The
93: parameters defining the linear dark matter power spectrum are $\sigma_8=0.74$, $n_s=0.95$, $\de n_s/\de \ln k =0$.}
94: could be summarized along the following points (for a more detailed description
95: see CF05 and CF06). The model accounts for IGM inhomogeneities by adopting a lognormal distribution according to the method outlined in \citeN{mhr00};
96: reionization is said to be complete once all the low-density regions (say, with overdensities $\Delta < \Delta_{\rm crit} \sim 60$) are ionized.
97: The mean free path of photons is thus determined essentially by the distribution of high density regions. We follow the ionization and thermal histories
98: of neutral, HII and HeIII regions simultaneously and self-consistently, treating the IGM as a multi-phase medium.
99:
100: Three types of reionization sources have been assumed: (i) metal-free (i.e. PopIII) stars having a Salpeter IMF in the mass
101: range $1 - 100 M_{\odot}$: they dominate the photoionization rate at high redshifts; (ii) PopII stars with sub-solar metallicities
102: also having a Salpeter IMF in the mass range $1 - 100 M_{\odot}$; (iii) QSOs, which are
103: significant sources of hard photons at $z \lesssim 6$; they have negligible effects on the IGM at higher redshifts. Note that there
104: is no compelling reason to rule out the possibility that the IMF of PopIII stars was top-heavy; however, as this is not required by
105: current data, we limit ourselves to the most conservative model.
106:
107: Reionization by UV sources is accompanied by photo-heating of the gas, which can result in a suppression of star formation in low-mass
108: haloes. We compute such (radiative) feedback self-consistently from the evolution of the thermal properties of the IGM. Furthermore,
109: the chemical feedback inducing the PopIII $\rightarrow$ PopII transition is implemented according to the detailed study by Schneider
110: et al. (2006) in which a merger-tree ``genetic'' approach is used to determine the termination of PopIII star formation in a metal-enriched halo.
111:
112: The predictions of the model are compared with a wide range of observational data sets, namely, (i) redshift evolution of Lyman-limit absorption
113: systems \cite{smih94}, (ii) IGM Ly$\alpha$ and Ly$\beta$ optical depths \cite{songaila04}, (iii) electron scattering optical depth \cite{sbd++06},
114: (iv) temperature of the mean intergalactic gas \cite{stle99}, (v) cosmic star formation history \cite{nchos04}, and (vi) source number counts at $z
115: \approx 10$ from NICMOS HUDF \cite{bitf05}.
116:
117: The best-fit reionization model is characterized by a PopII (PopIII) star-forming efficiency $\epsilon_{*,{\rm II}} = 0.1 (\epsilon_{*,{\rm III}} = 0.03)$
118: and escape fraction $f_{\rm esc, II} = 0.01 (f_{\rm esc, III} = 0.68)$ (keeping in mind that $f_{\rm esc, II}$ and $f_{\rm esc, III}$
119: are not independent).\footnote{The values are slightly different from those quoted in CF06 because of an improved
120: likelihood analysis. The qualitative results and main conclusions are unaffected by this modification.} The resultant value of the electron scattering
121: optical depth is $\tau_{\rm el} = 0.1$
122:
123: The data constrain the reionization scenario quite tightly. We find that hydrogen reionization starts at $z \approx 15$ driven by metal-free
124: (PopIII) stars, and it is 90 per cent complete by $z \approx 8$. This can be seen from the dashed curve in Fig. \ref{fig:halofrac} which
125: represents the evolution of the volume filling factor $Q_{\rm HII}(z)$ of ionized regions. After a rapid initial phase, the growth of the
126: volume filled by ionized regions slows down at $z \lesssim 10$ due to the combined action of chemical and radiative feedback,
127: making reionization a considerably extended process completing only at $z \approx 6$.
128: %
129: %
130: \vspace{-0.5cm}
131: \section{Properties of the reionization sources}
132: %
133: %
134: \begin{figure*}
135: \rotatebox{270}{\resizebox{0.45\textwidth}{!}{\includegraphics{sourcefrac.ps}}}
136: \caption{Number density of reionization sources (PopII and PopIII) as a function of the limiting magnitude, $m_{\rm AB}$,
137: at $7 < z < 8$ observed in $J_{110}$ band (left panel) and at $9.5 < z < 10.5$ observed in $H_{160}$ band (right).
138: Halo masses corresponding to $m_{\rm AB}$ for the particular redshift under consideration [given by equations (\ref{eq:f_nu_0})
139: and (\ref{eq:m_ab})] are shown by the straight lines and the relevant values can be read off from the right vertical axis.
140: }
141: \label{fig:sourcefrac}
142: \end{figure*}
143: Having identified the most probable reionization scenario, we can now confidently
144: determine the properties of the reionization sources governing it. Let us start by
145: defining the quantity
146: \be
147: x_{\gamma}(z) \equiv \f{n_{\gamma}(z)}{n_H} \f{t_{\rm rec}(z)}{t_H(z)}
148: \e
149: as the number of ionizing photons per H-atom contributed by haloes in the mass range $[M_{\rm min}, M_{\rm max}]$
150: in a fraction of the Hubble time $t_H(z)$ equal to the recombination time $t_{\rm rec}(z)$.
151: By construction, the IGM is reionized when $x_{\gamma} \gtrsim 1$.
152: The quantity $n_H$ is the comoving number density of hydrogen
153: atoms while $n_{\gamma}$
154: is the time-integrated comoving photon density, calculated using the relation
155: \be
156: n_{\gamma}(z) = \int_0^{t(z)} \de t ~ \dot{n}_{\gamma}(M_{\rm min} : M_{\rm max}, t)
157: \e
158: where $\dot{n}_{\gamma}(M_{\rm min} : M_{\rm max}, t)$ is the ionizing photon comoving emissivity
159: from haloes within $[M_{\rm min}, M_{\rm max}]$.
160:
161: The plot of $x_{\gamma}(z)$ for different mass ranges is shown in Fig. \ref{fig:halofrac}. The lower mass
162: $10^7$-$10^8 M_{\odot}$ haloes dominate the photon production rate at early redshifts providing about 0.25 photon/H-atom on
163: the fractional recombination timescale. These objects produce the first ionized regions, are preferentially metal-free,
164: and therefore mostly harbor PopIII stars of high specific ionizing power.
165: At $z < 8$, the contribution from PopIII haloes decreases because their formation is hampered by the
166: heating associated with radiative feedback. As a result, the progress of ionization fronts relies on photons emitted by
167: more massive haloes with $ M> 10^9 M_{\odot}$. It can be seen from Fig.1 that such high mass haloes do not host PopIII stars
168: as they form from the merging of already polluted progenitors, a result of the ``genetic'' transmission of chemical feedback
169: (Schneider et al. 2006). This combination of radiative and chemical feedback makes the reionization process quite extended and its
170: completion has to wait until $z \approx 6$ when the PopII stars (and QSOs, not shown in the Figure; see CF06 for details) dominate
171: the photoionization rate.
172:
173: Additional insights on the source properties may be gained by considering the fractional instantaneous contribution of
174: haloes above a certain mass,
175: \be
176: f_{\gamma}(>M, z) \equiv \f{\dot{n}_{\gamma}(>M, z)}{\dot{n}_{\gamma}(z)},
177: \e
178: also shown for different redshifts in the right hand panel of Figure \ref{fig:halofrac}.
179: The results shown emphasize again that $>80$\% of the ionizing power at $z \ge 7$ is provided by haloes with masses $< 10^9 M_{\odot}$
180: which are predominantly harboring PopIII stars. A turnover to a PopII-dominated reionization phase occurs shortly after, with
181: this population, residing in $M>10^9 M_\odot$ haloes, producing $\approx 60$\% of the ionizing photons at $z=6$. In conclusion,
182: PopIII stars and small galaxies initiate reionization at high redshift and remain important until they are overcome by PopII stars
183: and QSOs below $z=7$.
184:
185: \vspace{-0.5cm}
186: \section{Source counts at high redshift}
187:
188: Having determined which haloes contribute most significantly to the ionizing power at a particular redshift,
189: we now address the detectability of these sources in broadband imaging surveys and the optimal strategies to do so. A halo of (dark matter) mass $M$ at a redshift
190: $z$ will emit a flux given by \cite{sf06}
191: \be
192: F_{\nu_0} = \f{\epsilon_{*} (\Omega_b/\Omega_m) M}{4 \pi d_L^2(z') \Delta \nu_0}
193: \int \de \nu' ~ l_{\nu'}(\Delta t)~ {\rm e}^{-\tau_{\rm eff}(\nu_0,z=0,z')}
194: \label{eq:f_nu_0}
195: \e
196: where $\epsilon_{*}$ is the star-forming efficiency of the population under consideration, $l_{\nu'}(\Delta t)$ is a template
197: specific luminosity for the stellar population of age $\Delta t=t_{z'}-t_{z''}$ (the time elapsed between the two redshifts),
198: $d_L(z')$ is the luminosity distance and $\Delta \nu_0$ is the instrumental bandwidth. The quantity $\tau_{\rm eff}(\nu_0,z=0,z')$ is
199: the effective optical depth due to intervening gas at $\nu_0$ between $z'$ and $z = 0$. The main contribution to $\tau_{\rm eff}(\nu_0,z=0,z')$ comes
200: from the combined blanketing of the Lyman series lines in the emitter rest frame wavelength range
201: $912{\rm \AA} <\lambda < 1216 {\rm \AA}$ and from the continuum absorption from neutral hydrogen in the range $\lambda <$ 912\AA.
202: Both the contributions to the opacity can be calculated self-consistently from the (modified) lognormal density distribution (CF05) assumed in our model.
203: For calculating the template luminosity $l_{\nu}$, we use stellar population models (having metallicity $Z = 0.004 = 0.2 Z_{\odot}$) of \citeN{bc03} for PopII stars and of \citeN{schaerer02}
204: for PopIII stars.
205: Note that the star-forming efficiency $\epsilon_{*}$ in the above equation
206: is fixed by constraining the reionization history and is
207: {\it not} a free parameter as far as the calculation of the source counts in this work is concerned.
208: The flux $F_{\nu_0}$ can be transformed into a magnitude in the AB system:
209: \be
210: m_{\rm AB} = -2.5 \log_{10} \left(\f{F_{\nu_0}}{{\rm erg~s}^{-1} {\rm cm}^{-2} {\rm Hz}^{-1}}\right) - 48.6
211: \label{eq:m_ab}
212: \e
213:
214: The number of sources within a redshift interval $[z_{\rm min}, z_{\rm max}]$ observed in a solid angle $\de \Omega$ having a flux larger than
215: $F_{\nu_0}$ is
216: \be
217: N(>{F_{\nu_0}})
218: = \int_{z_{\rm min}}^{z_{\rm max}}
219: \de z' \f{\de V}{\de z' \de \Omega}
220: \int_{F_{\nu_0}}^{\infty} \de F'_{\nu_0}
221: \f{\de n}{\de F'_{\nu_0}} (F'_{\nu_0}, z')
222: \e
223: where $\de V/\de z' \de \Omega$ denotes the comoving volume element per unit redshift per unit solid angle, and
224: \be
225: \f{\de n}{\de F'_{\nu_0}} (F'_{\nu_0}, z') = \int_{z'}^{\infty}
226: \de z'' \f{\de M}{\de F'_{\nu_0}}(F'_{\nu_0}, \Delta t)
227: \f{\de^2 n}{\de M \de z''}(M,z'')
228: \e
229: is the comoving number of objects at redshift $z'$ with observed flux within $[F'_{\nu_0}, F'_{\nu_0} + \de F'_{\nu_0}]$.
230: The quantity $\de^2 n/\de M \de z''$ gives the formation rate of haloes of mass $M$, which we obtain from our reionization model.
231:
232: For definiteness, let us consider two bands corresponding to the NICMOS observations of the HUDF \cite{bti++04,bitf05},
233: namely $J_{110}$ and $H_{160}$; these broadband filters are appropriate to detect Lyman-break dropout sources at $z \sim 7.5$ and $z \sim 10$, respectively.
234: The main results of the calculation are shown in Fig. \ref{fig:sourcefrac} where we plot the number of sources (i.e. galaxies
235: powered by either PopII or PopIII stars) observed in a particular redshift interval and band as a function of the limiting magnitude $m_{\rm AB}$.
236: The halo masses corresponding to $m_{\rm AB}$ for the particular redshift under consideration [given by equations (\ref{eq:f_nu_0}) and
237: (\ref{eq:m_ab})] are shown by the straight lines and the relevant values can be read off from the right vertical axis.
238:
239:
240: From Fig. 2, we deduce that, at the currently achieved sensitivity limit of $\sim 28$ AB magnitude, our best-fit model predicts
241: $\sim 10 (1)$ Lyman-break sources per arcmin$^2$ at $z \sim 7.5 (10)$ observable
242: in the $J_{110}$ ($H_{160}$) band. Note that all these sources are bright PopII star forming haloes having masses $\sim 10^{10} M_{\odot}$.
243: In the previous Section we found that such sources provide only a negligible ($\approx 1\%$) contribution to reionization at $z > 6$.
244:
245: Some of these sources have been tentatively identified in broadband observations. For example, observations of HUDF using the NICMOS filter
246: \cite{bti++04,bitf05} at AB magnitude limit $\sim 28$ reveal $\sim 1$ source per arcmin$^2$ at $z \sim 7.5$ and $< 1$ source
247: per arcmin$^2$ at $z \sim 10$ respectively. On the other hand, deep near-IR photometry of two lensing clusters (A1835 and AC114)
248: obtained with ISAAC/VLT \cite{rpslk06} seem to indicate $\sim 1$ source per arcmin$^2$ at $z \sim 7-10$ at a magnitude limit equivalent
249: to $H_{160, {\rm AB}} \sim 26$, which is considerably higher than the NICMOS results. Stringent constraints are difficult to obtain from theoretical models as there remains
250: considerable confusion regarding the actual number of reliable detections. However, although the actual source count is disputed,
251: it is most likely that these tentatively detected sources are bright massive haloes which are not significant for reionization.
252:
253: The sources responsible for the bulk of reionization are galaxies formed inside low mass $< 10^8 M_{\odot}$ haloes and powered by PopIII stars.
254: By inspecting the right hand panel of Fig. \ref{fig:halofrac} we conclude that in order to observe sources which contribute to 50\% (90\%)
255: of the ionizing power at $z \sim 7.5$, it is necessary to observe PopIII sources of mass $<10^{8.1} M_{\odot} (<10^{7.9} M_{\odot})$. This
256: in turn requires to reach a magnitude sensitivity of $J_{110,{\rm AB}} = 31.2 (31.7)$ [left panel of Fig. \ref{fig:sourcefrac}], currently
257: beyond the reach of any instrument. Interestingly, one can expect to observe $\sim 15 (30)$ PopIII sources per arcmin$^2$ once such sensitivities
258: are reached. Similar conclusions can be drawn from the right hand panel for sources at $z \sim 10$ observable in the $H_{160}$ band.
259: For detecting sources which contribute to 50\% (90\%) of the ionizing power, the magnitude sensitivity required would be
260: $H_{160,{\rm AB}} = 32.2 (32.5)$, which would again correspond to halo masses of $10^{7.8} M_{\odot} (10^{7.7} M_{\odot})$. Hence, the detection
261: of reionization sources would require sensitivities better than 32 AB magnitudes in this band.
262:
263: \vspace{-0.5cm}
264: \subsection{Variants of the best-fit model}
265: %
266: %
267: \begin{figure}
268: \rotatebox{270}{\resizebox{0.45\textwidth}{!}{\includegraphics{sourcefrac2.ps}}}
269: \caption{Number density of PopIII sources as a function of the limiting magnitude, $m_{\rm AB}$,
270: at $9.5 < z < 10.5$ observed in $H_{160}$ band. The plots are for three models
271: of PopIII star-formation as discussed in the text.
272: Halo masses corresponding to $m_{\rm AB}$ for the particular redshift under consideration [given by equations (\ref{eq:f_nu_0})
273: and (\ref{eq:m_ab})] are shown by the straight lines and the relevant values can be read off from the right vertical axis.
274: }
275: \label{fig:sourcefrac2}
276: \end{figure}
277:
278: In this subsection, we try to obtain some indication about how much can the source count estimates vary without violating
279: any other observational constraints. We first note that the star-forming efficiency of PopII sources are
280: quite stringently constrained by low-redshift observations, e.g., the cosmic SFR and the QSO absorption lines; hence their
281: numbers cannot vary significantly. The situation is markedly different for PopIII stars as there are very few observations
282: constraining their properties. In order to calculate the variation in PopIII source counts, we consider two extreme limits
283: of the parameter $\epsilon_{*,{\rm III}}$ which are essentially determined by the bounds on $\tau_{\rm el}$:
284:
285: (i) The high-$\epsilon_{*,{\rm III}}$ model: This model is characterized
286: by the parameters $\epsilon_{*,{\rm II}} = 0.1, \epsilon_{*,{\rm III}} = 0.15, f_{\rm esc, II} = 0.0, f_{\rm esc, III} = 0.4$.
287: The resultant $\tau_{\rm el}$ is 0.12 (the 1-$\sigma$ upper limit given by WMAP3; \citeNP{sbd++06}).
288: The initial stages of reionization in this case proceeds much faster
289: than in the best-fit model and 90 per cent of the IGM is
290: ionized by $z \approx 10.5$. However, because of the high efficiency
291: of the PopIII stars, the radiative feedback is more severe and hence the
292: reionization gets extended till $z \approx 6$
293: as in the best-fit model.
294:
295:
296: (ii) The low-$\epsilon_{*,{\rm III}}$ model: This model is characterized
297: by the parameters $\epsilon_{*,{\rm II}} = 0.1, \epsilon_{*,{\rm III}} = 0.01, f_{\rm esc, II} = 0.04, f_{\rm esc, III} = 0.06$.
298: The resultant $\tau_{\rm el}$ is 0.06 (the 1-$\sigma$ lower limit given by WMAP3; \citeNP{sbd++06}).
299: This model corresponds to the minimum ionizing power contribution
300: required from PopIII stars. This model has an important qualitative difference from the high-$\epsilon_{*,{\rm III}}$ model;
301: because of such low PopIII efficiency, almost
302: all of the reionization process is driven by the PopII stars. In fact, the
303: lack of ionizing power at high-$z$ makes reionization start much later. Also,
304: the effects of feedback are minor in this model; as a result, the reionization
305: occurs rapidly and shortly before $z=6$.
306:
307: We now proceed to discuss how these models differ in their predictions for PopIII source counts (Fig. \ref{fig:sourcefrac2}).
308: Since the effects of PopIII stars are most noticeable at $z \gtrsim 10$, we focus on the $H_{160}$ band.
309: The first important point is that the number of bright sources increases with $\epsilon_{*,{\rm III}}$, a
310: direct consequence of the higher specific SFR within dark matter haloes. This would imply that one can observe
311: most of the reionization sources at a magnitude limit of $H_{160, {\rm AB}} \sim 30.5$, a limit almost certainly
312: attained by near future experiments. However, the number of sources detected at fainter magnitudes is smaller for
313: the high-$\epsilon_{*,{\rm III}}$ model than for the other two (best-fit and high-$\epsilon_{*,{\rm III}}$) models. This is because
314: the efficient star formation at high-$z$ makes the radiative feedback more effective, hence effectively quenching star
315: formation in low-mass haloes and suppressing the number of faint sources.
316: Of course, one finds a rise in the number of faint sources for the
317: low-$\epsilon_{*,{\rm III}}$ model when the feedback has the least effect;
318: one should be able to observe, in principle, $\gtrsim 100$ sources at
319: $H_{160, {\rm AB}} \sim 33$. However, one should keep in mind that
320: for low values of
321: $\epsilon_{*,{\rm III}} \approx 0.01$,
322: PopIII stars are {\it not} the main driving force for reionization anyway, and the main focus should be the source counts for PopII sources. In fact,
323: we find that sources contributing to 90\% of the ionizing power at
324: $z \approx 7.5$ would be the
325: PopII stars with haloes of mass $< 10^8 M_{\odot}$; they
326: can be observed with a sensitivity limit of
327: $J_{110,{\rm AB}} \sim 33$, which is quite difficult with
328: near-future experiments.
329:
330:
331: It is thus clear from the above discussion that the luminosity
332: function of PopIII sources at $z \approx 10$ could be important
333: for constraining the star-forming efficiency of PopIII stars and also
334: for some indirect understanding of feedback effects.
335:
336: \vspace{-0.5cm}
337: \section{Discussion}
338:
339: We have used the self-consistent model of CF06, which is consistent
340: with a variety of observations, to estimate the number
341: of sources (galaxies) at $z \sim 7-10$. According to our analysis:
342: \begin{itemize}
343:
344: \item The best-fit model predicts $\gtrsim 1$ sources at $z \sim 7-10$
345: per arcmin$^2$ at a sensitivity limit of $\sim 28$ AB magnitude, roughly
346: consistent with present observations. However, these sources are
347: bright massive ($\gtrsim 10^{10} M_{\odot}$) haloes forming
348: PopII stars which have negligible contribution to reionization
349: of the IGM.
350:
351: \item Reionization at $z > 6$ is actually driven by PopIII
352: stars in low mass ($< 10^{8} M_{\odot}$) haloes. The required
353: sensitivity to detect these sources would be 31-32 AB magnitude.
354:
355: \item In case the star-forming efficiency of PopIII stars
356: is higher than what is assumed in our best-fit model, the reionization
357: sources can be detected with sensitivities of 30 AB magnitude; however,
358: the number of sources detected could be smaller than the
359: best-fit model because of negligible star formation in low-mass
360: haloes due to radiative feedback.
361:
362: \item For low star-forming efficiencies of PopIII stars ($\sim 0.01$), the
363: reionization is actually driven by PopII stars and occurs rapidly and
364: shortly before $z=6$. In this case, the sensitivity required
365: to observe the reionization sources
366: (PopII stars within haloes of masses $< 10^8 M_{\odot}$)
367: at $z \gtrsim 6$ would be 33 AB magnitude.
368:
369:
370: \end{itemize}
371:
372: The sensitivities required to observe the reionization sources
373: are expected to be achieved in future deep imaging surveys, particularly
374: the ones which take advantage of gravitational lensing magnification.
375: For example, the
376: NIR Wide Field Camera 3
377: (WFC3), scheduled to be
378: installed on HST in near future, promises to achieve a
379: sensitivity limit of $m_{\rm AB} \sim 31$ in the
380: $J_{110}$ and $H_{160}$ filters for a field of view of
381: $\sim 5$ arcmin$^2$ in about a few hundred hours
382: of observation time \cite{sle07}. According to our estimates, such
383: lensed surveys should be able
384: to detect quite a few PopIII reionization sources
385: (via Lyman-break dropout techniques) in the field of view.
386: A much better prospect of detecting these sources would
387: be through the Ultra-Deep Imaging
388: Survey using the JWST which too plans to achieve a
389: sensitivity limit of $m_{\rm AB} \gtrsim 31$ over 100-200 hours
390: of observation time per filter \cite{gmc++06}.
391: Direct detection of these sources would put stringent constraints
392: on reionization history, and in addition can be used for understanding
393: physical processes like feedback.
394:
395: Provided that the contamination problems due to bright atmospheric emission
396: lines, which could restrict visibility up to 50\% of the redshift range in the
397: $J$-band, could be maintained under control, a complementary approach for detecting
398: the reionization sources would be through narrow band surveys for Ly$\alpha$ emitters. A rough
399: estimate of the Ly$\alpha$ luminosity from a halo of mass $M$
400: forming PopIII stars is given by
401: \be
402: L_{\alpha} = \epsilon_{*,{\rm III}} (1 - f_{\rm esc,III})
403: \f{\Omega_b}{\Omega_m} M ~ c_{{\rm Ly}\alpha} ~ q(H)
404: \e
405: where $c_{{\rm Ly}\alpha} = 1.04 \times 10^{11}$ ergs and
406: $q(H)$ is the rate of hydrogen-ionizing photons per unit mass
407: of stars formed \cite{schaerer02}.
408: The above relation does not take into account the attenuation
409: arising from neutral hydrogen around the source, hence the luminosities
410: could possibly be overestimated. Under the above assumptions,
411: a $10^8 M_{\odot}$ halo would produce a luminosity of
412: $\sim 10^{41}$ erg s$^{-1}$ in our best-fit model. Such luminosities
413: seem to be well within the reach of lensed Ly$\alpha$ surveys.
414: For example, \citeN{ser++07} have
415: detected $\sim 2$ sources having luminosities $\sim 10^{41.5}$ erg s$^{-1}$
416: at $z \sim 10$ within a area of 0.3 arcmin$^2$ using a Keck survey of gravitationally-lensed sources. There is a possibility that these
417: sources could be PopIII stars forming in haloes
418: of masses $\gtrsim 10^8 M_{\odot}$ which do have $\lesssim 30$\% contribution
419: to the ionizing power at $z \approx 10$.
420: Future narrowband Ly$\alpha$ surveys
421: like DAzLE\footnote{http://www.ast.cam.ac.uk/$\sim$optics/dazle/} thus would
422: be extremely important for direct detection of primary reionization sources
423: at high redshift.
424:
425: Finally, it is worth mentioning that our models do not
426: include (i) star formation in minihaloes where molecular cooling
427: is efficient, or (ii) the possibility of a top-heavy IMF
428: for PopIII stars. The predictions regarding the source counts
429: could vary considerably depending on the details of the above two
430: processes, and hence future surveys present an excellent opportunity
431: to probe such effects.
432:
433: \vspace{-0.7cm}
434: \section*{Acknowledgments}
435: We are indebted to R. Salvaterra and R. Schneider for useful discussions and for sharing their results.
436:
437:
438: \vspace{-0.7cm}
439: \bibliography{mnrasmnemonic,astropap-mod,reionization}
440: \bibliographystyle{mnras}
441:
442: \end{document}
443:
444:
445: