1: \documentclass[twocolumn,apjl]{emulateapj}

2: \usepackage{natbib}

3: \usepackage{epsfig}

4: \usepackage{bm}

5: \usepackage{color}

6: 

7: %\usepackage{hyperref}

8: %\bibliographystyle{apj_hyperref}

9: \providecommand{\eprint}[1]{\href{http://arxiv.org/abs/#1}{#1}}

10: \newcommand{\ISBN}[1]{\href{http://cosmologist.info/ISBN/#1}{ISBN: #1}}

11: \providecommand{\adsurl}[1]{\href{#1}{}}

12: 

13: \begin{document}

14: 

15: \newcommand{\clee}{C_{\ell}^{EE}}

16: \newcommand{\xef}{x_e^{\rm fid}}

17: \newcommand{\xet}{x_e^{\rm true}}

18: \newcommand{\dz}{\Delta z}

19: \newcommand{\zmax}{z_{\rm max}}

20: \newcommand{\zmin}{z_{\rm min}}

21: \newcommand{\zmid}{z_{\rm mid}}

22: \newcommand{\zre}{z_{\rm reion}}

23: \newcommand{\lcdm}{$\Lambda$CDM}

24: \newcommand{\wmap}{\emph{WMAP}}

25: \newcommand{\thetaa}{\theta_{A}}

26: \newcommand{\obhh}{\Omega_{b}h^2}

27: \newcommand{\ochh}{\Omega_{c}h^2}

28: \newcommand{\ascal}{A_{s}}

29: \newcommand{\nscal}{n_{s}}

30: \newcommand{\wh}[1]{\textcolor{blue}{(\bf #1)}}

31: 

32: 

33: 

34: \title{Reionization constraints from five-year WMAP data}

35: %\shorttitle{}

36: 

37: \author{Michael J. Mortonson$^{1,2}$ and Wayne Hu$^{1,3}$}

38: \affil{$^{1}$Kavli Institute for Cosmological Physics, 

39: Enrico Fermi Institute, University of Chicago, Chicago, IL 60637\\

40: $^{2}$Department of Physics,  University of Chicago, Chicago, IL 60637\\

41: $^{3}$Department of Astronomy and Astrophysics, University of Chicago, Chicago, IL 60637

42: }

43: 

44: 

45: \begin{abstract}

46: We study the constraints on reionization from five years of \wmap\ data, 

47: parametrizing the evolution of the average fraction of ionized hydrogen 

48: with principal components that provide a complete basis for describing the 

49: effects of reionization on large-scale $E$-mode polarization. 

50: Using Markov Chain Monte Carlo methods, we find that

51: the resulting model-independent estimate of the total optical depth 

52: is nearly twice as well determined as the estimate from 3-year \wmap\ data, 

53: in agreement with simpler analyses that assume instantaneous reionization. 

54: The mean value of the optical depth from principal components 

55: is slightly larger than the instantaneous value; we find 

56: $\tau=0.097\pm0.017$ using only large-scale polarization, and 

57: $\tau=0.101\pm0.019$ when temperature data is included.  Likewise,

58: scale invariant $n_s=1$ spectra are no longer strongly disfavored by WMAP alone.

59: Higher moments of the ionization history show less improvement in the 

60: 5-year data than the optical depth. By plotting the distribution of 

61: polarization power for models from the MCMC analysis, we show that 

62: extracting most of the remaining information about the shape of the 

63: reionization history from the CMB requires better measurements of $E$-mode 

64: polarization on scales of $\ell\sim 10-20$.  Conversely, the quadrupole and octopole polarization

65: power is already predicted to better than cosmic variance given {\it any}

66: allowed ionization history

67: at $z<30$ so that more precise measurements will test the $\Lambda$CDM paradigm.

68: \end{abstract}

69: 

70: 

71: \keywords{cosmic microwave background --- cosmology: theory --- large-scale structure of universe}

72: 

73: 

74: % =====================================================

75: \section{Introduction}

76: \label{sec:intro}

77: 

78: The amplitude of fluctuations in the $E$-mode component of 

79: cosmic microwave background (CMB)

80: polarization on large scales provides the current 

81: best constraint on the Thomson scattering optical depth to 

82: reionization, $\tau$. Assuming that the universe was reionized 

83: instantaneously, \cite{Dunetal08} estimate the total optical 

84: depth to be $\tau=0.087\pm0.017$ 

85: using five years of \wmap\ data.

86: Theoretical studies suggest that the process of reionization 

87: was too complex to be well described 

88: as a sudden transition~\citep[e.g.,][]{BarLoe01}.  

89: Previous studies have examined how the constraint on $\tau$ depends on 

90: the evolution of the globally-averaged ionized fraction during 

91: reionization, $x_e(z)$, for a variety of specific theoretical scenarios. 

92: If the assumed form of $x_e(z)$ is incorrect, 

93: the estimated value of $\tau$ can be biased; this bias can be lessened by 

94: considering a wider variety of reionization histories at the expense of 

95: increasing the uncertainty in $\tau$~\citep{Kapetal03,Holetal03,Coletal05}.

96: 

97: For the previous release of three years of \wmap\ data, using a more 

98: model-independent analysis of reionization based on principal components 

99: does not change the basic conclusions of studies that assume 

100: instantaneous reionization or other simple models; however, as the 

101: polarization power spectrum becomes better determined, it is 

102: increasingly important to adopt an approach with sufficient freedom to 

103: approximate a variety of possible reionization scenarios in order 

104: to minimize parameter biases \citep{MorHu08a,MorHu08b}. 

105: In this letter, we study how model-independent constraints on 

106: reionization have changed with the addition of two more years of 

107: data from \wmap\ and on what scales further measurements would have

108: the largest impact on ionization constraints. We review the principal component 

109: parametrization of the ionization history in \S~\ref{sec:pcs}. 

110: We then present the current constraints on the principal component 

111: amplitudes and the optical depth to reionization (\S~\ref{sec:tau})

112: and the range of polarization power spectra for general models that 

113: are currently allowed by the data (\S~\ref{sec:cl}). 

114: We discuss these results in \S~\ref{sec:disc}.

115: 

116: 

117: 

118: % ****************************************

119: \begin{figure*}[t]

120: \centerline{\includegraphics[width=3.5in]{f1a.eps}

121: \includegraphics[width=3.5in]{f1b.eps}}

122: \caption{

123: Marginalized 2D 68\%  and 95\%  CL contours for the optical 

124: depth to reionization ($\tau$) and the amplitudes 

125: of the 5 lowest-variance principal components of $x_e(z)$ 

126: ($m_{\mu},~\mu=1-5$). 

127: Panels along the diagonals show the 1D posterior probability 

128: distributions. Constraints are plotted for both 3-year 

129: (\emph{red dotted lines}) 

130: and 5-year (\emph{blue shading, solid lines}) \wmap\ data. 

131: In the left plot, only the 

132: low-$\ell$ reionization peak in the $E$-mode polarization power spectrum 

133: is used for parameter constraints, and all parameters besides the 5 

134: PC amplitudes and $\tau$ are held fixed. For the constraints in the right 

135: plot, we use both temperature and polarization data and allow five 

136: additional parameters to vary: $\obhh$, $\ochh$, $\thetaa$, 

137: $\ascal$, and $\nscal$.

138: The plot boundaries for the PC amplitudes correspond to physicality priors 

139: that exclude models that are unable to satisfy $0\leq x_e \leq 1$ for any 

140: combination of the higher-variance ($\mu \geq 6$) PCs.}

141: \vskip 0.25cm

142: \label{fig:pctau2d}

143: \end{figure*}

144: % ****************************************

145: 

146: 

147: 

148: % =====================================================

149: \section{Ionization Principal Components}

150: \label{sec:pcs}

151: 

152: We parametrize the reionization history 

153: as a free function of redshift by decomposing $x_e(z)$ into its

154: principal components (PCs) with respect to the $E$-mode polarization of the 

155: CMB~\citep{HuHol03,MorHu08a}:

156: \begin{equation}

157: x_e(z)=\xef(z)+\sum_{\mu}m_{\mu}S_{\mu}(z),

158: \label{eq:mmutoxe}

159: \end{equation}

160: where the principal components, $S_{\mu}(z)$, 

161: are the eigenfunctions of the Fisher matrix that describes 

162: the dependence of $\clee$ on $x_e(z)$, $m_{\mu}$ are the amplitudes of the 

163: principal components for a particular reionization history, and 

164: $\xef(z)$ is the fiducial model at which the Fisher matrix is computed.  The

165: components are rank ordered by their Fisher-estimated variances.

166: The lowest-variance eigenmode ($\mu=1$) is an 

167: average of the ionized fraction over the entire redshift range, weighted toward 

168: high $z$.  The $\mu=2$ mode measures the difference between 

169: the amount of ionization at high $z$ and at low $z$, and higher modes 

170: follow this pattern with weighted averages of $x_e(z)$ that oscillate with 

171: higher and higher frequency in redshift.

172: The main advantage of using principal components as a basis for $x_e(z)$ is 

173: that only a small number of the components are required to completely 

174: describe the effects of reionization on large-scale CMB polarization, 

175: so we obtain a very general parametrization of the reionization history at 

176: the expense of only a few additional parameters.

177: 

178: The principal components are defined over a limited range in redshift, 

179: $\zmin<z<\zmax$, with $x_e=0$ at $z>\zmax$ and $x_e=1$ at $z<\zmin$. 

180: We take $\zmin=6$, since the absence of Gunn-Peterson absorption in 

181: the spectra of quasars at $z\lesssim 6$ indicates that the universe is 

182: nearly fully ionized at lower redshifts \citep{FanCarKea06}. 

183: In the MCMC analysis presented here, we always use 

184: the five lowest-variance principal components of $x_e(z)$ with $\zmax=30$, constructed

185: around a constant fiducial model of

186: $\xef(z)=0.15$.  

187: The amplitudes of these components then serve to 

188: parametrize general reionization histories in the analysis of 

189: CMB polarization data. 

190: We refer the reader to~\cite{MorHu08a} for 

191: further discussion of these choices and the demonstration that five components 

192: suffice to describe the $E$-mode spectrum to better than cosmic variance precision.

193: 

194: We impose priors on the 

195: principal component amplitudes corresponding to physical values of the 

196: ionized fraction, $0 \leq x_e \leq 1$, according to the conservative

197: approach of \cite{MorHu08a}.  All excluded models are unphysical, 

198: but the models we retain are not necessarily strictly physical.

199: Finally, we neglect helium reionization, which is a small correction at the current 

200: level of precision but will be more important for 

201: future analyses \citep[e.g.,][]{ColPie08}.

202: 

203: 

204: % ****************************************

205: \begin{figure*}[t]

206: \centerline{\includegraphics[width=3.0in]{f2a.eps}

207: \includegraphics[width=3.0in]{f2b.eps}}

208: \caption{

209: Median (\emph{blue solid curve}) and 68\% and 95\% CL regions 

210: (\emph{blue shaded regions}) of polarization power spectra 

211: for any ionization history at $z<30$ allowed by the 5-year \wmap\ data,

212: assuming the standard $\Lambda$CDM paradigm. As in 

213: Fig.~\ref{fig:pctau2d}, only polarization data 

214: are used in the left panel and both 

215: temperature and polarization data are included in the right panel.

216: \emph{Red dashed curves}: cosmic variance (68\% and 95\% CL) around the median 

217: model.}

218: \vskip 0.25cm

219: \label{fig:cleedist}

220: \end{figure*}

221: % ****************************************

222: 

223: 

224: 

225: 

226: 

227: % =====================================================

228: \section{Optical Depth Constraints}

229: \label{sec:tau}

230: 

231: We examine the implications of the \wmap\ 5-year data for general 

232: models of reionization parametrized by principal components using a

233: Markov Chain Monte Carlo analysis that mirrors our previous study 

234: of the 3-year data in \cite{MorHu08a}. We consider constraints from 

235: either large-scale polarization alone, with parameters that do not directly 

236: affect reionization fixed to values 

237: that fit the temperature data (``EE''), or from the full set of 

238: temperature and polarization data, varying the parameters of the 

239: ``vanilla'' \lcdm\ model (baryon density $\obhh$, cold dark matter density 

240: $\ochh$, acoustic scale $\thetaa$, scalar amplitude $\ascal$, and 

241: scalar spectral tilt $\nscal$) in addition to the reionization PC amplitudes 

242: (``TT+TE+EE''). In both cases, the total optical depth to reionization, 

243: $\tau$, is a derived parameter.

244: 

245: The MCMC constraints on principal component amplitudes and the 

246: derived optical depth for both of these cases are plotted 

247: in Fig.~\ref{fig:pctau2d}, along with the previous constraints 

248: from 3-year \wmap\ data \citep{MorHu08a}. While there are some 

249: improvements in all 5 of the individual components when considering $EE$ alone, 

250: these changes are not as large as the improvement in the optical 

251: depth constraint when all of the data are considered.  

252:  Adding both temperature data and extra parameters in going from EE only 

253: to TT+TE+EE has the net effect of slightly strengthening constraints on the higher ranked

254: PC amplitudes, although there is very little effect on $\tau$.

255: The additional constraining power for both 3-year and 5-year data 

256: comes mainly from the measured temperature power spectrum at 

257: $\ell\sim 10-100$, which excludes models with additional Doppler effect 

258: contributions due to narrow features in the ionization 

259: history \citep{MorHu08a}.

260: 

261: Modeling reionization as an instantaneous transition 

262: at some redshift $\zre$, \cite{Dunetal08} estimate the optical depth 

263: from the 5-year \wmap\ data to be $\tau=0.087\pm 0.017$, almost a factor 

264: of two more precise than the estimate from three years 

265: of data~\citep{Speetal07}.

266: For ionization histories parametrized by PCs, we find that the 

267: constraint on optical depth is $\tau=0.097\pm0.017$ for the EE case and 

268: $\tau=0.101\pm0.019$ for TT+TE+EE. As with the 3-year data, the error 

269: on $\tau$ is roughly 10\% larger with the inclusion of temperature data 

270: and a larger set of parameters, and the error in both cases is 

271: the same or only slightly larger than for the instantaneous reionization 

272: analysis. 

273: 

274: The central value of $\tau$ for more general ionization 

275: histories is higher than the instantaneous reionization value by 

276: $\sim 0.5-1~\sigma$; a similar shift toward larger optical depths was 

277: seen in PC analysis of the 3-year data \citep{MorHu08a}.  

278: The maximum likelihood model, however, has $\tau=0.088$ 

279: for EE data and $\tau=0.090$ for TT+TE+EE, much closer to the 

280: instantaneous reionization maximum likelihood optical depth of 

281: $\tau=0.089$ \citep{Dunetal08}.

282: The larger mean optical 

283: depth is at least partly due to having a large parameter volume

284: of models with finite ionization fraction

285: at high redshift that are still allowed by the data.   

286: With flat priors on the principal components, this volume effect

287: can boost the mean optical depth of models even though the mean likelihood

288: of low optical depth models remains the same.  Our assumption of 

289: full ionization at redshifts below $\zmin=6$ for all models also 

290: limits how small the optical depth can be. 

291: Relative to the best-fit instantaneous model 

292: with $\zre=11.0\pm 1.4$~\citep{Dunetal08}, there are simply more ways 

293: to increase $\tau$ than there are to decrease $\tau$ by changing the 

294: ionization history, given these priors and the current data.

295: 

296: For the TT+TE+EE analysis, the larger mean optical depth 

297: is accompanied by shifts in correlated parameters, particularly 

298: the spectral tilt: $n_s=0.990\pm0.024$ with $x_e(z)$ parametrized by PCs, 

299: and $n_s=0.960\pm0.015$ for instantaneous reionization \citep{Kometal08}. 

300: As with the optical depth, however, some of this shift is a parameter 

301: volume effect. The maximum likelihood model for the principal component 

302: analysis has $n_s=0.976$. The best fit scale invariant model  

303: (fixing $n_s=1$) is a poorer fit to the data by 

304: $\Delta \chi_{\rm eff}^2 \equiv -2 \ln (\mathcal{L}/\mathcal{L}_{\rm max}) \sim 1$, where $\mathcal{L}_{\rm max}$ is the maximum likelihood.

305: (The instantaneous reionization maximum likelihood model is also at 

306: $\Delta \chi_{\rm eff}^2 \approx 1$ relative to the best fit with principal 

307: components.)

308: As measurements of CMB polarization improve with future data, particularly

309: with detections in the $10 < \ell < 20$ range (see \S \ref{sec:cl}), 

310: the constraints on parameters such as optical depth and tilt should 

311: become less sensitive to our assumptions about the priors.

312: 

313: 

314: Unlike the total optical depth, constraints on the optical depth 

315: over more limited redshift ranges have only improved slightly. 

316: With three years of \wmap\ data, the 95\% upper limit on 

317: the optical depth from $z>20$ (allowing for a significant ionized 

318: fraction up to $z\sim 40$) was $\tau(z>20)<0.08$ \citep{MorHu08a}. The 

319: limit from 5-year data is $\tau(z>20)<0.07$. 

320: If we instead choose the dividing redshift to be the best-fit value of 

321: the redshift of instantaneous reionization, $\zre=11$, we find 

322: a similar constraint for the contribution to the optical depth 

323: from high redshift: $\tau(z>11)<0.07$. Compared to the 3-year data, 

324: there is also a more significant (but still weak) preference 

325: for nonzero optical depth from $6<z<11$.

326: 

327: 

328: % =====================================================

329: \section{Power spectra of allowed models}

330: \label{sec:cl}

331: 

332: To better understand at which scales the reionization peak of 

333: $E$-mode polarization is best constrained by the current data, 

334: we plot the 68\% and 95\% CL limits on $\clee$ from the Monte Carlo chains in 

335: Fig.~\ref{fig:cleedist}.  These limits reflect the range of

336: ensemble-averaged power allowed by the 5-year data and

337: the 

338: PC-parametrized reionization histories.

339: Since this parametrization is complete in the power spectrum, 

340: the range in Fig.~\ref{fig:cleedist} reflects

341: the allowed model power spectra for {\it any} ionization history at

342: $z<z_{\rm max}=30$.

343: 

344: 

345: At $\ell\lesssim 5$, the variation in allowed models is smaller 

346: than the uncertainty due to cosmic variance.   In other words,

347: the data at $\ell \sim 5$ in combination with any ionization history and

348: the power law initial power spectrum

349: make a prediction for the ensemble-averaged power at lower $\ell$ that is

350: sharper than can be measured.    Conversely, measurements that

351: violate this prediction at a statistically significant level require modifications to

352:  the

353: $\Lambda$CDM paradigm itself, much like low measurements of the temperature

354: quadrupole.  It is interesting that the maximum likelihood $E$-mode polarization

355: quadrupole reported by \cite{Noletal08}, 

356: $6 C_2^{EE}/2\pi \approx 0.15~\mu{\rm K}^2$, is in excess of the 95\% cosmic

357: variance region shown in Fig.~\ref{fig:cleedist}.

358: 

359: 

360: The uncertainty in the model space is largest at intermediate scales 

361: of $\ell \sim 10-20$, where the large-scale polarization power is expected to 

362: be smallest. There is substantial room for improved measurements of 

363: the spectrum on these scales before reaching the cosmic variance limit.

364: Tighter constraints on the $E$-mode power at $10<\ell <20$ would 

365: better determine the amplitude of principal components beyond the first; 

366: such measurements are necessary to be able to discriminate among 

367: different reionization histories with the same total optical depth.

368: Physically, these measurements would better constrain the ionization 

369: history at high redshifts ($z\gtrsim 15$).

370: 

371: On small scales ($\ell > 30-40$) the limits on power spectra of 

372: reionization models from the chains again become tighter than 

373: cosmic variance since the theoretical amplitude of the 

374: recombination peak is well determined due to constraints on 

375: parameters from the temperature data. 

376: 

377: 

378: Comparison of the two panels in Fig.~\ref{fig:cleedist} shows that 

379: the main effect of including temperature data in constraints on 

380: principal components is to eliminate models with large power at 

381: $10<\ell<30$.  As mentioned in the previous section, these models are 

382: excluded by the data due to their increased temperature fluctuations at 

383: $\ell \sim 10-100$.

384: Even with TT+TE+EE data, the range of power in 

385: models allowed by current data is a few times larger than 

386: cosmic variance.

387: 

388: 

389: 

390: 

391: % =====================================================

392: \section{Discussion}

393: \label{sec:disc}

394: 

395: 

396: The 5-year \wmap\ polarization data significantly improve the 

397: estimate of the total optical depth, reducing the error from 

398: $\sigma_{\tau}\approx 0.03$ to $\sigma_{\tau}\approx 0.017$. 

399: This improvement is seen in both a model-independent analysis 

400: using principal components of the ionization history and in 

401: an analysis that assumes instantaneous reionization, although there 

402: is a small shift in the central value with the model-independent 

403: method preferring a slightly higher mean around $\tau=0.1$.

404: 

405: As with the 3-year data, the $E$-mode reionization peak is currently 

406: best measured on the largest scales, $\ell\sim 5$. Determining 

407: details of the ionization history beyond the optical depth requires 

408: information about the full shape of the reionization peak, which 

409: can be obtained by supplementing the current observations with 

410: better measurements of the $E$-mode power on scales of $5<\ell<30$.

411: In particular, improved knowledge of the power on scales between 

412: the main reionization peak and the recombination peak at $\ell\sim 10-20$ 

413: would be the most useful for distinguishing models of 

414: reionization with different ionization histories but the same optical depth.

415: 

416: Conversely, the data along with {\it any} allowed ionization history at $z<30$ 

417: in the standard $\Lambda$CDM context already predict the 

418: ensemble-averaged polarization quadrupole

419: and octopole powers to better than cosmic variance.   More precise measurements

420: in this regime can test the standard model itself.

421: 

422: 

423: \acknowledgements {\it Acknowledgments}:  

424: %restore if we come up with list: We thank  ... for useful discussions.  

425: This work was supported by the KICP through the grant NSF PHY-0114422 and

426: the David and

427: Lucile Packard Foundation.

428: WH was additionally supported by  the DOE through

429: contract DE-FG02-90ER-40560.

430: 

431: 

432: 

433: %\bibliographystyle{apj_hyperref}

434: \bibliographystyle{apj}

435: %\bibliography{wmap5pc}

436: \begin{thebibliography}{13}

437: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

438: 

439: \bibitem[{{Barkana} \& {Loeb}(2001)}]{BarLoe01}

440: {Barkana}, R. \& {Loeb}, A. 2001, \physrep, 349, 125

441: 

442: \bibitem[{{Colombo} {et~al.}(2005){Colombo}, {Bernardi}, {Casarini}, {Mainini},

443:   {Bonometto}, {Carretti}, \& {Fabbri}}]{Coletal05}

444: {Colombo}, L.~P.~L., {Bernardi}, G., {Casarini}, L., {Mainini}, R.,

445:   {Bonometto}, S.~A., {Carretti}, E., \& {Fabbri}, R. 2005, \aap, 435, 413

446: 

447: \bibitem[{{Colombo} \& {Pierpaoli}(2008)}]{ColPie08}

448: {Colombo}, L.~P.~L. \& {Pierpaoli}, E. 2008, arXiv:0804.0278

449: 

450: \bibitem[{{Dunkley} {et~al.}(2008)}]{Dunetal08}

451: {Dunkley}, J. {et~al.} 2008, arXiv:0803.0586

452: 

453: \bibitem[{Fan {et~al.}(2006)Fan, Carilli, \& Keating}]{FanCarKea06}

454: Fan, X.-H., Carilli, C.~L., \& Keating, B. 2006, Ann. Rev. Astron. Astrophys.,

455:   44, 415

456: 

457: \bibitem[{{Holder} {et~al.}(2003){Holder}, {Haiman}, {Kaplinghat}, \&

458:   {Knox}}]{Holetal03}

459: {Holder}, G.~P., {Haiman}, Z., {Kaplinghat}, M., \& {Knox}, L. 2003, \apj, 595,

460:   13

461: 

462: \bibitem[{{Hu} \& {Holder}(2003)}]{HuHol03}

463: {Hu}, W. \& {Holder}, G.~P. 2003, \prd, 68, 023001

464: 

465: \bibitem[{{Kaplinghat} {et~al.}(2003){Kaplinghat}, {Chu}, {Haiman}, {Holder},

466:   {Knox}, \& {Skordis}}]{Kapetal03}

467: {Kaplinghat}, M., {Chu}, M., {Haiman}, Z., {Holder}, G.~P., {Knox}, L., \&

468:   {Skordis}, C. 2003, \apj, 583, 24

469: 

470: \bibitem[{{Komatsu} {et~al.}(2008)}]{Kometal08}

471: {Komatsu}, E. {et~al.} 2008, arXiv:0803.0547

472: 

473: \bibitem[{{Mortonson} \& {Hu}(2008{\natexlab{a}})}]{MorHu08a}

474: {Mortonson}, M.~J. \& {Hu}, W. 2008{\natexlab{a}}, \apj, 672, 737

475: 

476: \bibitem[{{Mortonson} \& {Hu}(2008{\natexlab{b}})}]{MorHu08b}

477: ---. 2008{\natexlab{b}}, \prd, 77, 043506

478: 

479: \bibitem[{Nolta {et~al.}(2008)}]{Noletal08}

480: Nolta, M.~R. {et~al.} 2008, arXiv:0803.0593

481: 

482: \bibitem[{{Spergel} {et~al.}(2007)}]{Speetal07}

483: {Spergel}, D.~N. {et~al.} 2007, \apjs, 170, 377

484: 

485: \end{thebibliography}

486: 

487: 

488: \end{document}

489: