1: \documentclass[twocolumn,showpacs,amsmath,amssymb,prl]{revtex4}
2: \usepackage{graphicx}% Include figure files
3: \usepackage{dcolumn}% Align table columns on decimal point
4: \usepackage{bm}% bold math
5:
6: \begin{document}
7: \input{epsf}
8:
9: \title{Possibility of Precise Measurement of the Cosmological Power\\
10: Spectrum With a Dedicated 21cm Survey After Reionization} \author{Abraham
11: Loeb$^{1}$ \& J. Stuart B. Wyithe$^2$} \affiliation{$^1$ Astronomy
12: Department, Harvard University, 60 Garden Street, Cambridge, MA 02138, USA}
13: \affiliation{$^2$School of Physics, University of Melbourne, Parkville,
14: Victoria, Australia}
15:
16: \begin{abstract}
17:
18: Measurements of the 21cm line emission by residual cosmic hydrogen after
19: reionization can be used to trace the power spectrum of density
20: perturbations through a significant fraction of the observable volume of
21: the Universe. We show that a dedicated 21cm observatory coule probe a
22: number of independent modes that is two orders of magnitude larger than
23: currently available, and enable a cosmic-variance limited detection of the
24: signature of a neutrino mass $\sim 0.05$ eV. The evolution of the linear
25: growth factor with redshift could also constrain exotic theories of gravity
26: or dark energy to an unprecedented precision.
27:
28: \end{abstract}
29:
30: \pacs{98.80-k, 95.30Dr, 95.55Jz}
31:
32: \date{\today}
33:
34: \maketitle
35:
36: Recently, there has been much interest in the feasibility of mapping the
37: three-dimensional (3D) distribution of cosmic hydrogen through its resonant
38: spin-flip transition at a rest-frame wavelength of
39: 21cm\cite{Furlanetto,BarLoeb}. Several experiments are currently being
40: constructed (such as
41: MWA\footnote{http://www.haystack.mit.edu/ast/arrays/mwa/},
42: LOFAR\footnote{http://www.lofar.org/}, PAPER
43: \footnote{http://astro.berkeley.edu/~dbacker/EoR/},
44: 21CMA\footnote{http://web.phys.cmu.edu/~past/}) and more ambitious
45: designs are being planned (SKA\footnote{http://www.skatelescope.org/})
46: to detect the theoretically-predicted emission signal.
47:
48: Measurements of the power-spectrum of 21cm brightness fluctuations could
49: constrain the initial conditions from inflation as well as the nature of
50: the dark matter and dark energy. The 21cm fluctuations are expected to
51: simply trace the primordial power-spectrum of matter density perturbations
52: either before the first galaxies had formed (at redshifts $z\gtrsim
53: 20$)\cite{LZ04,Lewis} or after reionization ($1\lesssim z\lesssim 6$) --
54: when only dense pockets of self-shielded hydrogen (such as damped
55: Ly$\alpha$ systems) survive \cite{WL07,Pen}. During the epoch of
56: reionization, the fluctuations are mainly shaped by the topology of ionized
57: regions \cite{McQuinn,Trac,Iliev}, and thus depend on astrophysical
58: details. However, even during this epoch, the line-of-sight anisotropy of
59: the 21cm power spectrum due to peculiar velocities, can in principle be
60: used to separate the implications for fundamental physics from the unknown
61: details of the astrophysics \cite{BL04,McQuinn}. In what follows, we will
62: focus our discussion on the post-reionization epoch \cite{WL07,Pen} which
63: offers two advantages. First, it is least contaminated by the Galactic
64: synchrotron foreground (whose brightness temperature scales with the
65: redshift under consideration as $(1+z)^{2.6}$ \cite{Furlanetto}). Second,
66: because the UV radiation field is nearly uniform after reionization, it
67: should not imprint any large-scale features on the 21cm power spectrum that
68: would mimic cosmological signatures. On large spatial scales the 21cm
69: sources are expected to have a linear bias analogous to that inferred from
70: galaxy redshift surveys. Since a 21cm survey maps the global hydrogen
71: distribution without resolving individual galaxies, the 21cm bias is
72: expected to be modest compared to surveys that select for the brightest
73: galaxies at the same redshifts.
74:
75: In general, cosmological surveys are able to measure the power-spectrum of
76: primordial density fluctuations, $P(k)$, to a precision that is ultimately
77: limited by cosmic variance, namely the number of independent Fourier modes
78: that fit within the survey volume. 21cm observations are advantageous
79: relative to existing data sets because they access a 3D volume instead of
80: the 2D surface probed by the cosmic microwave background (CMB), and they
81: extend to a sufficiently high redshift (well beyond the horizon of galaxy
82: redshift surveys \cite{SDSS}) where most of the comoving volume of the
83: observable Universe resides. At these high redshifts, small-scale modes
84: are still in the perturbative (linear growth) regime where their analysis
85: is straightforward. The expected 21cm power extends down to the
86: pressure-dominated (Jeans) scale of the cosmic gas which is orders of
87: magnitude smaller than the comoving scale at which the CMB anisotropies are
88: damped by photon diffusion \cite{LZ04}.
89:
90: Altogether, the above factors make 21cm surveys an ideal cosmological probe
91: of fundamental physics \cite{Kl}. To illustrate this point, we show in
92: this {\it Letter} that a dedicated 21cm observatory would enable a
93: determination of the matter power-spectrum at redshifts $z\lesssim 6$ to an
94: unprecedented precision. In our numerical examples, we adopt the standard
95: set of cosmological parameters \cite{Spergel}.
96:
97: \begin{figure} [ht]
98: \centerline{\epsfxsize=3.2in \epsfbox{fig1.eps}}
99: \caption{The fraction of the total comoving volume of the observable
100: Universe that is available up to a redshift $z$.}
101: \label{fig1}
102: \end{figure}
103:
104: \paragraph*{Number of Modes.}
105: The limitation of existing redshift surveys of galaxies
106: \cite{Tegmark,Lahav} is apparent in Fig.~\ref{fig1} which plots the
107: comoving volume of the Universe out to a redshift $z$ as a function of $z$.
108: State-of-the-art galaxy redshift surveys, such as the spectroscopic sample
109: of luminous red galaxy (LRGs) in the {\it Sloan Digital Sky Survey (SDSS)}
110: \cite{SDSS}, extend only out to $z\sim 0.5$ (over $\sim10\%$ of the sky)
111: and probe $\sim 0.01\%$ of the observable Universe.
112:
113: The CMB fluctuations probe a thin shell on the 2D surface of the sky. The
114: number of modes with a comoving wave number $k\equiv2\pi/\lambda$ between
115: $k$ and $k+dk$ that fit on this 2D surface is, $dN_{\rm CMB} = \pi k dk
116: \left[\mathcal{A}/(2\pi)^2\right]$, where $\mathcal{A}=D^2d\Omega$, $D$ is
117: the comoving distance to the last scattering surface at $z\sim 10^3$ and
118: $d\Omega$ is the solid angle of the survey field. Redshift surveys of
119: galaxies and 21cm surveys probe a 3D comoving volume $\mathcal{V}$ and
120: potentially access a larger number of modes, $dN_{\rm 3D} = 2\pi k^2 dk
121: \left[\mathcal{V}/(2\pi)^3\right]$.
122:
123: Figure~\ref{fig2} compares $N_{\rm CMB}=(k/10)\,dN_{\rm CMB}/dk$ with
124: $N_{\rm 3D}=(k/10)\,dN_{\rm 3D}/dk$, for future 21cm surveys after
125: reionization \cite{WL07}. The CMB data set is assumed to cover a fraction
126: $f_{\rm sky}=0.65$ of the sky (excluding the region around Milky Way
127: galaxy). For comparison, we also show the corresponding number of modes
128: within the same $k$ interval in the spectroscopic LRG sample of {\it SDSS}
129: \cite{SDSS}, which covers $\sim3700$ square degrees out to $z\sim0.5$ or a
130: volume of $\mathcal{V}=0.72h^{-3}$Gpc$^3$ (where $h\approx0.7$ is the
131: Hubble constant in units of 100${\rm km~s^{-1}~Mpc^{-1}}$).
132:
133: 21cm observatories that are currently under construction (such as MWA) will
134: survey only a few percent of the sky and process only $\sim15$\% of the
135: available frequency range (band-pass). In this {\it Letter}, we consider
136: future 21cm surveys that would potentially cover $f_{\rm sky}=0.65$ with a
137: processed frequency-bandwidth spanning a redshift range of a factor of 3 in
138: $(1+z)$ centered on $z=1.5$, $z=3.5$ and $z=6.5$ \footnote{The factor of 3
139: in $(1+z)$ corresponds to the largest frequency bandwidth over which a
140: low-frequency dipole antenna has suitable sensitivity.}. With an array
141: design based on MWA in which the effective area of each tile of 16
142: dipole antennae equals its physical area, the value of $f_{\rm sky}=0.65$
143: corresponds to $\sim33$ correlated primary beams or fields.
144:
145: \begin{figure} [ht]
146: \centerline{\epsfxsize=3.2in \epsfbox{fig2.eps}}
147: \caption{The number of modes $N$ within a wave number bin of width $\Delta
148: k=k/10$ centered on $k$, that are available in different cosmological
149: surveys. The thick dashed grey line corresponds to the spectroscopic LRG
150: sample of {\it SDSS} \cite{SDSS}, while the thick solid line (marked HRG)
151: corresponds to a future spectroscopic survey at $2.5<z<3.5$ covering 1000
152: square degrees with a a co-moving galaxy density equal to the LRG
153: sample. The thick dark line corresponds to a CMB data set with $f_{\rm
154: sky}=0.65$. The thin lines show the number of modes accessible in a 21cm
155: survey (including the limit on large scale modes due to foreground removal
156: \cite{McQuinn}) covering $f_{\rm sky}=0.65$ within a redshift range
157: spanning a factor of 3 in $(1+z)$, and centered on $z=1.5,3.5$ and
158: $6.5$. For $z\leq1.5$, we have truncated the curves at $k=0.2~{\rm
159: Mpc^{-1}}$ to illustrate the smaller range of $k$ accessible within the
160: linear regime at lower redshifts \cite{eisen}. }
161: \label{fig2}
162: \end{figure}
163:
164: \paragraph*{Results.} The fractional uncertainties in the 21cm
165: power-spectrum $P_{21}$ for a cosmic-variance limited survey, $(\Delta
166: P_{21}/P_{21})=1/\sqrt{N}$, are presented in the inset of Fig.~\ref{fig3}
167: (straight lines). Also shown in both the inset and main panel are the noise
168: curves for observations using a design based on the so-called MWA5000
169: experiment, which is assumed to have 10 times the collecting area of MWA
170: (as described in Refs. \cite{McQuinn,WLG}). MWA5000 would be cosmic
171: variance limited in an integration time of $\sim 10^3$ hours at wave
172: numbers near $k\sim0.1$Mpc$^{-1}$. Since the dipoles of each antenna tile
173: look at $\sim \pi$ steradians, simultaneous processing of multiple primary
174: beams would allow a survey of area $f_{\rm sky}=0.65$ at $10^3$ hours of
175: integration per pointing within a few years. In computing the thermal
176: noise of the observatory, we have adopted a model for the bias factor
177: $b_{21}$ of the 21cm sources ($b_{21}^2\equiv P_{21}/P$) from
178: Ref. \cite{WL07}. The noise curves include a limit on large scale modes
179: due to foreground removal. Current estimates for MWA show that foreground
180: removal should be effective for modes over a frequency range $\lesssim
181: 6$MHz which is ${1\over 4}$ of the total 24MHz processed bandwidth
182: \cite{McQuinn}. However, improvements on this range would provide access to
183: a larger $N$ as well as to lower-$k$ modes. In Fig.~\ref{fig3} and
184: subsequently, we assume a scenario in which the foreground can be removed
185: on scales of up to ${1\over 12}$ of the total processed bandwidth, namely
186: ${1\over 12}\times (\sqrt{3}-1/\sqrt{3})[1400{\rm MHz}/(1+z)]=
187: 30[(1+z)/4.5]^{-1}{\rm MHz}$.
188:
189: For comparison, we also show the noise curves for the {\it SDSS}-LRG survey
190: (thick grey line), including the effects of Poisson shot-noise\footnote{We
191: ignore Poisson fluctuations in the 21cm power spectra since the
192: contributing galaxies are expected to be of much lower mass than LRGs, and
193: there should be a large number of them per resolution element of the
194: survey.}, $(\Delta P_{\rm gal}/P_{\rm gal})=[1+(b^2P(k)n_{\rm
195: gal})^{-1}]/\sqrt{N}$, for a galaxy number density of $n_{\rm
196: gal}=46748/(0.72h^{-3})$Gpc$^{-3}$ and a bias factor of $b_{\rm gal}=2$
197: \cite{SDSS}. We find that the potential 21cm constraints on the matter
198: power-spectrum are 1--2 orders of magnitude better than a low-redshift
199: galaxy survey like {\it SDSS}-LRG.
200:
201: \begin{figure} [ht]
202: \centerline{\epsfxsize=3.2in \epsfbox{fig3.eps}}
203: \caption{ The fractional change in the amplitude of the power-spectrum
204: owing to the presence of a massive neutrino (horizontal grey lines,
205: asymptoting towards a constant at high $k$ values). The case shown,
206: $f_\nu=0.004$, corresponds to $m_\nu=0.05$eV. For comparison, the limits
207: imposed by cosmic variance on measurements of the power-spectrum from the
208: {\it SDSS}-LRG and a future 1000 square degree galaxy survey at $2.5<z<3.5$
209: are marked by the thick dashed and solid grey lines respectively. The {\it
210: U}-shaped error curves correspond to an all-sky 21cm survey [$f_{\rm
211: sky}=0.65$ over a redshift range spanning a factor of 3 in $(1+z)$] with
212: MWA5000 and a $10^3$ hour integration per field (line styles for
213: $z=1.5,3.5,6.5$ as in Figure~\ref{fig2}). The noise is plotted in bins of
214: size $\Delta k/k = 0.1$. The inset shows these results on logarithmic axes
215: that span a larger dynamic range of achievable precision. The straight
216: thin lines in the inset show the cosmic-variance uncertainty in the
217: power-spectrum measurement owing only to the number of available modes.
218: For $z\leq1.5$, we have truncated the curves at $k=0.2~{\rm
219: Mpc^{-1}}$ to illustrate the smaller range of $k$ accessible within the
220: linear regime at lower redshifts \cite{eisen}. }
221: \label{fig3}
222: \end{figure}
223:
224: As an example for the potential use of a 21cm survey, we show in
225: Fig.~\ref{fig3} the expected relative changes in the amplitude of the
226: power-spectrum owing to the presence of a massive neutrino \cite{EH}. At
227: wave numbers much larger than the neutrino free-streaming wave number
228: ($k_{\rm fs}= 0.1\Omega_m(0)h\sqrt{f_{\nu}}~{\rm Mpc^{-1}}$), the
229: suppression of the power-spectrum is given by \cite{neutrino,Hannestad},
230: \begin{equation}
231: \frac{P(k,f_\nu)}{P(k,f_{\nu}=0)}=(1-f_\nu)^3\left[1.9\times10^5
232: g(0)\Omega_m(0){f_\nu \over N_\nu}\right]^{-6f_\nu/5},
233: \end{equation}
234: where $g(z)\approx \Omega_m^{0.2}/[1+0.003(\Omega_\Lambda/\Omega_m)^{4/3}]$
235: is the growth function of the gravitational potential for matter and
236: vacuum density parameters of
237: $\Omega_m(z)=[1+\Omega_\Lambda/\Omega_m(0)(1+z)^3]^{-1}$ and
238: $\Omega_\Lambda=(1-\Omega_m)$ \cite{KGB}; $f_\nu\equiv
239: [\Omega_\nu(0)/\Omega_m(0)] =0.08 N_\nu (m_\nu/1~{\rm eV})$ is the
240: present-day mass fraction of the matter density carried by $N_\nu$ neutrino
241: species of particle mass $m_\nu$. In this {\it Letter} we conservatively
242: assume a non-degenerate hierarchy of neutrino masses with $N_\nu=1$ and
243: $m_\nu$ denoting the largest mass eigenstate. Figure~\ref{fig3} shows the
244: case corresponding to $f_\nu=0.004$
245: ($m_\nu=0.05$eV).
246:
247: While Fig.~\ref{fig3} indicates that the cosmic variance in a galaxy
248: redshift survey (such as the {\it SDSS}-LRG survey) is sufficiently small
249: to detect the suppression of power at $k\gg k_{\rm fs}$ due to a neutrino
250: mass of $m_\nu=0.05$eV, uncertainties in other cosmological parameters and
251: parameter degeneracies reduce this sensitivity by an order of
252: magnitude\cite{EH,Slosar}. Thus, in order to avoid possible systematic
253: offsets between the power-spectrum amplitude observed by different
254: techniques (such as galaxy surveys, CMB maps, and Ly$\alpha$ forest data)
255: at different $k$ values, it is desirable for the 21cm survey to be
256: self-contained and cover a sufficiently broad range of wave numbers that
257: probe the curvature of the neutrino effect in Fig. \ref{fig3}. For this to
258: be achieved, the removal of the Galactic synchrotron foreground would need
259: to be effective over large frequency intervals of up to $\sim
260: 30[(1+z)/4.5]^{-1}{\rm MHz}$. Foreground removal across such intervals
261: would provide access to the required range of scales over which a constant
262: offset (in the form of a linear bias factor) would not affect the $m_\nu$
263: measurement. The desired wave number to be reached by foreground removal
264: corresponds to the scale where cosmic variance is larger than the change
265: induced by a massive neutrino. Currently, the detailed properties of the
266: foreground are not well measured. The feasibility of foreground removal
267: over a broad frequency interval will remain uncertain until data from the
268: first generation of 21cm observatories is analysed.
269:
270: Figure~\ref{fig3} illustrates that foreground removal to $\sim
271: 30[(1+z)/4.5]^{-1}{\rm MHz}$ would be sufficient to detect the modification
272: of the power spectrum due to the minimum neutrino mass of $0.047\pm 0.01$eV
273: indicated by the latest atmospheric neutrino data \cite{neutrino,Maltoni}
274: at all spatial scales where the effect is larger than cosmic variance. The
275: advantage of the large survey volume is evident since it allows the
276: modification of shape to be measured in addition to the suppression
277: detected by the {\it SDSS}-LRG survey (the latter being cosmic variance
278: limited on scales where the shape is measured). In a follow-up paper
279: \cite{Visb}, we will address the level of degeneracy with other
280: cosmological parameters. Already, Fisher-matrix studies \cite{Mao,McQuinn}
281: have demonstrated the improved capabilies of 21cm observations during the
282: epoch of reionization, where contamination from astrophysical sources needs
283: to be removed through the angular dependence of the 21cm power-spectrum
284: \cite{BL04}.
285:
286: A 21cm survey measures the modulation in the cumulative 21cm emission from a
287: large number of galaxies, as its coarse angular and redshift resolution
288: is not capable of resolving the 21cm sources individually \cite{WL07,WLG}.
289: The damped Ly$\alpha$ systems which contain most of the hydrogen mass in
290: the Universe at $z\lesssim 6$, are expected to be hosted by abundant low
291: mass galaxies \cite{Wolfe} and thus have a weak bias relative to the
292: underlying matter distribution on large spatial scales. This weak bias is
293: not expected to introduce a feature to the power-spectrum that is
294: degenerate with the neutrino signature (as would be the case prior to
295: reionization). Although the comoving wave number at which non-linear
296: evolution becomes important increases from $k\sim0.1h^{-1}$Mpc$^{-1}$ at
297: $z=0.3$ to $\sim0.5h^{-1}$Mpc$^{-1}$ at $z=3$ \cite{eisen}, the constraints
298: on $m_\nu$ can be potentially improved by accounting for the related
299: non-linear effects \cite{Saito}.
300:
301: %In difference from low-$z$ galaxy surveys, a future 21cm survey could
302: %measure evolution of the power spectrum over a broad range of redshifts.
303: %For $k\gg k_{\rm fs}$, the growth of the power-spectrum is given by
304: %\cite{neutrino,Hannestad}, $P(k,z) =
305: %\left[{g(z)}/{(1+z)g(0)}\right]^{2-6f_\nu/5}P(k,0)$. This, in principle,
306: %yields a novel way for measuring $f_\nu$ from the relation,
307: %\begin{equation}
308: %f_\nu={5\over 3} - {5\over 6}{(d\ln{P_{21}}/d\ln{Y}) - 2
309: %(d\ln{b_{21}}/d\ln{Y}) \over (d\ln{g}/d\ln{Y}) +1},
310: %\label{growth}
311: %\end{equation}
312: %where $Y\equiv(1+z)$. However, the associated constraints are sensitive to
313: %the evolution in $b_{21}$. For $z\gg1$, $d\ln{g}/d\ln{Y}\approx 6/(1+z)^3$,
314: %$d\ln{P_{21}}/d\ln{Y}\approx 2$ and $d\ln{g}/d\ln{Y}$ is small. Thus, the
315: %accuracy achievable on $f_\nu$ would be limited by the bias evolution to a
316: %level $\Delta f_\nu\approx 5(d\ln{b_{21}}/d\ln{Y})/3$. Unfortunately, this
317: %implies that the bias would need to be constant (or at least known
318: %\cite{Peebles}) to better than the percent level per logarithm of redshift
319: %in order to constrain $f_\nu$ effectively from the growth factor evolution.
320:
321: By measuring the evolution of the growth factor with redshift to the
322: exquisite precision implied by Fig. \ref{fig3}, it would also be possible
323: to constrain alternative theories of gravity or dark energy well beyond the
324: capabilities of existing data sets \cite{Hui}.
325: %Based on Eq. (\ref{growth}),
326: The evolution of the growth factor would be limited by the knowledge of the
327: bias of the 21cm sources, $b_{21}$. The limit on the uncertainty in growth
328: factor would satisfy $[d\ln{g}/d\ln{(1+z)}]\approx
329: [d\ln{b_{21}}/d\ln{(1+z)}]$.
330:
331: Finally, we note that a precise $P(k)$ measurement at multiple redshifts
332: would also allow to determine the redshift evolution of the baryonic
333: acoustic oscillations (BAO) in the 21cm power spectrum \cite{BL05,WL07}.
334: The BAO scale constitutes a standard ruler \cite{eisen,BlGl} that can be
335: used to measure the equation of state of the dark energy \cite{WLG,Pen},
336: constrain $1-\Omega_{\rm tot}$, and further remove degeneracies between
337: $m_\nu$ and other cosmological parameters \cite{Wong}.
338:
339: \paragraph*{Hardware.}
340: %\paragraph*{Discussion.}
341: %Our results indicate that a dedicated 21cm observatory would be able to
342: %probe a number of independent modes that is two orders of magnitude larger
343: %than available from existing data sets.
344: The required 21cm observatory could be similar in antenna design to the
345: planned MWA but would require expanded collecting area and computer
346: resources to account for the increased cross-correlation requirements and
347: the analysis of multiple beams. A suitable observatory would contain ten
348: times the number of tiles in MWA. The computational load increases as the
349: square of the number of tiles in a telescope, and linearly with both the
350: amount of processed band-pass and the number of cross-correlated primary
351: beams. Thus, an all-sky survey over a frequency range covering a factor of
352: 3 in $(1+z)$ requires an overall improvement by $\sim 10^4$ in computer
353: power relative to MWA\footnote{The cost of the antenna hardware is expected
354: to be in the range of hundreds of million of dollars, an order of magnitude
355: higher than for MWA. Improved telescope designs \cite{Mao} and advances in
356: computer technology could help to realize our projected 21cm observatory.}.
357:
358: \bigskip
359: \paragraph*{Acknowledgments.}
360: We thank M. McQuinn and M. Zaldarriaga for helpful comments. We also thank
361: the Harvard-Australia foundation for funding a visit of A.L. to Australia,
362: during which this work was performed.
363:
364:
365: \begin{references}
366:
367: \bibitem{Furlanetto} S. R. Furlanetto, S.~P., Oh, \& F. H. Briggs,
368: Phys. Rep. {\bf 433}, 181 (2006).
369:
370: \bibitem{BarLoeb} R. Barkana, \& A. Loeb, Reports of Progress in
371: Physics {\bf 70}, 627 (2007).
372:
373: \bibitem{LZ04} A. Loeb, \& M. Zaldarriaga, Phys. Rev. Lett. {\bf 92},
374: 211301 (2004).
375:
376: \bibitem{Lewis} A. Lewis, \& A. Challinor, Phys. Rev. {\bf D 76},
377: 083005 (2007).
378:
379: \bibitem{WL07} S. Wyithe, \& A. Loeb, ArXiv e-prints, 708, arXiv:0708.3392
380: (2007).
381:
382: \bibitem{Pen} T.-C. Chang, U.-L. Pen, J.~B. Peterson, \& P. McDonald, rXiv
383: e-prints, 709, arXiv:0709.3672 (2007).
384:
385: \bibitem{McQuinn} M. McQuinn, O. Zahn, M. Zaldarriaga, L. Hernquist,
386: \& S. R. Furlanetto, Astrophys. J. {\bf 653}, 815 (2006).
387:
388: \bibitem{Trac} M. G. Santos, A. Amblard, J. Pritchard, H. Trac,
389: R. Cen, \& A. Cooray ArXiv e-prints, 708, arXiv:0708.2424 (2007).
390:
391: \bibitem{Iliev} I. T. Iliev, G. Mellema, U.-L. Pen, \& P. R. Shapiro,
392: ArXiv e-prints, 712, arXiv:0712.1356 (2007).
393:
394: \bibitem{BL04} R. Barkana, \& A. Loeb, A. Astrophys. J. Lett.
395: {\bf 624}, L65 (2005).
396:
397: \bibitem{SDSS} D. J. Eisenstein, et al., Astrophys. J. {\bf 633}, 560
398: (2005); N. Padmanabhan, et al., Mon. Not. R. Astr. Soc. {\bf 378}, 852
399: (2007).
400:
401: \bibitem{Kl} M. Kleban, K. Sigurdson, \& I. Swanson, JCAP {\bf 708}, 9
402: (2007).
403:
404: \bibitem{Spergel} D. N. Spergel, et al. Astrophys. J. Suppl. {\bf
405: 170}, 377 (2007).
406:
407: \bibitem{Tegmark} M. Tegmark, M., et al., Phys. Rev {\bf D 74}, 123507
408: (2006).
409:
410: \bibitem{Lahav} {\O}. Elgar{\o}y, \& O. Lahav, Physica Scripta {\bf 127},
411: 105 (2006).
412:
413: \bibitem{WLG} J.~S.~B. Wyithe, A. Loeb, \& P. M. Geil,
414: Mon. Not. R. Astr. Soc., in press (2007); ArXiv e-prints, 709,
415: arXiv:0709.2955.
416:
417: \bibitem{KGB} L.~A. Kofman, N.~Y. Gnedin, \& N.~A. Bahcall,
418: Astrophys. J. {\bf 413}, 1 (1993).
419:
420: \bibitem{EH} W. Hu, D. J. Eisenstein, \& M. Tegmark, Phys. Rev. Lett. {\bf
421: 80}, 5255 (1998).
422:
423: \bibitem{Slosar} U. Seljak, A. Slosar, \& P. McDonald, J. of Cosm. and
424: Astro-Particle Phys. {\bf 10}, 14 (2006); A. Slosar, P. McDonald, \&
425: U. Seljak, New Astron. Rev. {\bf 51}, 327 (2007).
426:
427: \bibitem{neutrino} J. Lesgourgues, \& S. Pastor, Phys. Rep. {\bf 429}, 307
428: (2006).
429:
430: \bibitem{Hannestad} S. Hannestad, Ann. Rev. of Nucl. and Part. Sci. {\bf
431: 56}, 137 (2006).
432:
433: \bibitem{Maltoni} M. Maltoni, T. Schwetz, M. A. Tortola, \& J. W. F. Valle,
434: New J. Phys. {\bf 6}, 122 (2004).
435:
436: \bibitem{Visb} E. Visbal, A. Loeb, \& J. S. B. Wyithe, in preparation
437: (2008).
438:
439: \bibitem{Mao} Y. Mao, M. Tegmark, M. McQuinn, M. Zaldarriaga, \& O. Zahn,
440: ArXiv e-prints, 802, arXiv:0802.1710 (2008).
441:
442: \bibitem{Wolfe} A.~M. Wolfe, E. Gawiser, \& J. X. Prochaska,
443: Ann. Rev. Astr. Astrophys. {\bf 43}, 861 (2005).
444:
445: \bibitem{eisen} H.-J. Seo, \& D.~J. Eisenstein, Astrophys. J. {\bf 633},
446: 575 (2005).
447:
448: \bibitem{BlGl} Blake, C., \& Glazebrook, K., Astrophys. J. {\bf 594}, 665 (2003)
449:
450: \bibitem{Saito} S. Saito, M. Takada, \& A. Taruya, ArXiv e-prints, 801,
451: arXiv:0801.0607 (2008).
452:
453: %\bibitem{Peebles} M. Tegmark, \& P.~J.~E. Peebles, Astrophys. J. {\bf 500},
454: %L79 (1998).
455:
456: \bibitem{Hui} S. Wang, L. Hui, M. May, \& Z. Haiman, Phys. Rev. {\bf
457: D 76}, 063503 (2007).
458:
459: \bibitem{BL05} R. Barkana, \& A. Loeb, Mon. Not. R. Astron. Soc. {\bf
460: 363}, L36 (2005).
461:
462: \bibitem{Wong} S. Hannestad, \& Y.~Y.~Y. Wong, J. of Cosm. and
463: Astro-Part. Phys. {\bf 7}, 4 (2007).
464:
465: \end{references}
466:
467: \end{document}
468:
469:
470:
471:
472:
473:
474:
475:
476:
477:
478:
479: