0710.4162/ms.tex
1: \documentclass[prd,twocolumn]{revtex4}
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: \newcommand{\wh}[1]{\textcolor{blue}{(\bf #1)}}
13: \hyphenation{biases arc-min}
14: 
15: \setlength{\tabcolsep}{10pt}
16: 
17: 
18: \begin{document}
19: 
20: \newcommand{\clee}{C_{\ell}^{EE}}
21: \newcommand{\clbb}{C_{\ell}^{BB}}
22: \newcommand{\cltens}{C_{\ell}^{BB,{\rm T}}}
23: \newcommand{\cllens}{C_{\ell}^{BB,{\rm L}}}
24: \newcommand{\xef}{x_e^{\rm fid}}
25: \newcommand{\xet}{x_e^{\rm true}}
26: \newcommand{\dz}{\Delta z}
27: \newcommand{\zmax}{z_{\rm max}}
28: \newcommand{\zmin}{z_{\rm min}}
29: \newcommand{\zmid}{z_{\rm mid}}
30: \newcommand{\lmax}{\ell_{\rm max}}
31: \newcommand{\lcdm}{$\Lambda$CDM}
32: \newcommand{\wmap}{\emph{WMAP}}
33: \newcommand{\planck}{\emph{Planck}}
34: \newcommand{\cmbpol}{\emph{CMBPol}}
35: 
36: \newcommand{\mnras}{Mon. Not. R. Astron. Soc.}
37: \newcommand{\aap}{Astron. Astrophys.}
38: \newcommand{\apjs}{Astrophys. J. Suppl. Ser.}
39: 
40: 
41: \title{Impact of reionization on CMB polarization tests of slow-roll inflation}
42: %\shorttitle{}
43: 
44: \author{Michael J. Mortonson$^{1,2}$ and Wayne Hu$^{1,3}$}
45: \affiliation{$^{1}$Kavli Institute for Cosmological Physics, 
46: Enrico Fermi Institute, University of Chicago, Chicago, IL 60637\\
47: $^{2}$Department of Physics,  University of Chicago, Chicago, IL 60637\\
48: $^{3}$Department of Astronomy and Astrophysics, University of Chicago, Chicago, IL 60637
49: }
50: 
51: \date{\today}
52: 
53: \begin{abstract}
54: Estimates of inflationary parameters from the
55: CMB $B$-mode polarization spectrum on the largest scales depend on knowledge of
56:  the reionization history, especially at low
57: tensor-to-scalar ratio. 
58: Assuming an incorrect reionization history in the analysis of 
59: such polarization data can strongly bias the inflationary parameters. 
60: One consequence is that the single-field slow-roll consistency relation 
61: between the tensor-to-scalar 
62: ratio and tensor tilt might be excluded with high significance even if 
63: this relation holds in reality. 
64: We explain the origin of the bias and present case studies 
65: with various tensor amplitudes and noise characteristics. 
66: A more model-independent approach can account for uncertainties
67:  about reionization, and we show that parametrizing 
68: the reionization history by a set of its principal components with respect to 
69: $E$-mode polarization removes the bias in inflationary parameter measurement with little
70: degradation in precision.
71: \end{abstract}
72: 
73: 
74: %\keywords{cosmic microwave background --- cosmology: theory --- large-scale structure of universe}
75: 
76: 
77: \maketitle
78: 
79: 
80: % =====================================================
81: \section{Introduction}
82: \label{sec:intro}
83: 
84: 
85: Temperature and polarization power spectra of the cosmic microwave background
86: (CMB) are consistent with predictions
87: of the simplest inflationary models~\cite{Gut81,AlbSte82,Lin82,Sat81}:
88: a nearly flat geometry,
89: superhorizon correlations as probed by the spectrum of acoustic peaks, 
90: and primordial scalar perturbations that are adiabatic, Gaussian, and close to 
91: scale-invariant~\cite{HuWhi96c,SpeZal97,HuSpeWhi97,Peietal03,Speetal07}.
92: One of the key remaining signatures of inflation, tensor 
93: perturbations (i.e.\ gravitational waves)~\cite{KamKosSte97,SelZal97}, 
94: has yet to be detected. 
95: Depending on the amplitude of 
96: the tensor perturbations, which is not well constrained theoretically,
97: it may be possible to measure the angular power spectrum of the inflationary 
98: gravitational waves in the $B$-mode component of the CMB polarization 
99: on large scales.
100: Non-detection of the tensor spectrum does not necessarily rule out 
101: inflation, but upper limits on $r$ can be used to exclude particular 
102: models of inflation and limit its energy scale.
103: Many experiments have been proposed to search for this 
104: signal~\cite{Planck,CMBTaskForce,Oxletal04,LawGaiSei04,Ruhl:2004kv,Mafetal05,Yooetal06,Kogut:2006nd,Macetal07,Polenta07}.
105: 
106: Measurement of  tensor perturbations in the $B$-mode 
107: polarization power spectrum would 
108: test models of inflation by constraining inflationary parameters.
109: These parameters include the tensor-to-scalar ratio, $r$, and the 
110: tensor spectral index, $n_t$, which are related by a consistency relation 
111: under the simplest single-field slow-roll inflationary scenarios.
112: If the tensor spectrum can be detected, precise measurements over a wide 
113: range of scales could test the consistency relation.
114: 
115: 
116: CMB constraints on $r$ and $n_t$ depend on the ability to 
117: accurately determine the large-scale power in $B$-modes due to tensor perturbations, 
118: independent of the effects of other cosmological parameters. 
119: On the largest scales, the tensor $B$-mode spectrum
120:  depends not only on inflationary parameters 
121: but also on the reionization history of the universe~\cite{Zal97}. 
122: The main impact of reionization on the spectrum is through the total optical 
123: depth, $\tau$.  The 3-year \wmap\ measurements of $E$-mode polarization 
124: determine $\tau$ to an accuracy of about $30\%$~\cite{Pagetal07,Speetal07}, 
125: and future CMB experiments should constrain $\tau$ at the~$5-10\%$~level~\cite{Holetal03,KeaMil06,Planck,MorHu07b}.
126: 
127: However, the detailed evolution of the reionization history also significantly 
128: affects the large-scale polarization spectra.  Uncertainty in this history leads
129: to added uncertainty in inflationary 
130: parameters.  Moreover, incorrect inferences due to an
131: oversimplified treatment of reionization may bias estimates of 
132: inflationary parameters. 
133: For unbiased estimation of the optical depth from the $E$-mode reionization peak, 
134: the solution is to use a complete, principal-component-based 
135: description of reionization effects  when 
136: estimating parameters from CMB polarization data~\cite{HuHol03,MorHu07b}.
137: In this paper, we extend this approach to tensor $B$-mode polarization
138: and show that it is equally if not more effective in ensuring accurate measurements
139: with little loss in precision.
140: 
141: 
142: The outline of the paper is as follows.
143: We discuss the effects of reionization and inflationary parameters on 
144: the polarization power spectra and the large-scale degeneracy between 
145: these parameters in \S~\ref{sec:param}. A brief overview of the principal 
146: component parametrization of the reionization history follows in \S~\ref{sec:pcs}. 
147: In \S~\ref{sec:mcmc}, we describe our Markov Chain Monte Carlo analysis of 
148: simulated polarization data and give the resulting constraints on $\tau$, 
149: $r$, and $n_t$, which we discuss further in \S~\ref{sec:discuss}.
150: 
151: 
152: 
153: 
154: 
155: 
156: % ****************************************
157: \begin{figure}
158: \centerline{\psfig{file=f1.eps, width=3.0in}}
159: \caption{$B$-mode tensor spectra 
160:  illustrating the degeneracy 
161: between $\tau$ and $n_t$ for large-scale measurements, 
162: with angular power spectra plotted in the upper panel and fractional 
163: deviations from the base model in the lower panel. 
164: For the base model (\emph{solid}), $r=0.03$, $\tau=0.1$, and 
165: $n_t=-0.00375$. The other two models have 
166: $\{\tau,n_t\}=\{0.12,-0.00375\}$ (\emph{long dashed}) and 
167: $\{0.12,0.13\}$ (\emph{short dashed}), with a pivot scale of 
168: $k_{\rm pivot}=0.01~{\rm Mpc}^{-1}$.
169: The reionization history here is assumed to be instantaneous. 
170: Cosmic variance of $\cltens$ for the base model, which excludes the variance
171: from lensing, is shown by the shaded band in 
172: the lower panel.
173: }
174: \label{fig:degen}
175: \end{figure}
176: % ****************************************
177: 
178: 
179: 
180: % =====================================================
181: \section{Polarization dependence on reionization and inflationary parameters}
182: \label{sec:param}
183: 
184: Like the scalar $E$-mode polarization power spectrum, the tensor $B$-mode spectrum
185: $\cltens$ consists of 
186: two main components: one from recombination that peaks at $\ell \sim 100$, 
187: and the other from the epoch of reionization, 
188: which peaks near $\ell \sim 5$ and dominates at $\ell \lesssim 20$ (Fig.~\ref{fig:degen}).   
189: In the tensor power spectrum, these components arise from wavenumbers
190: $k \sim 0.01$ Mpc$^{-1}$  and $\sim 0.0007$ Mpc$^{-1}$ respectively.   They provide a lever arm of 
191: over a decade in physical scale for measuring the tensor tilt. This
192: lever arm is slightly shortened compared with the ratio of angular scales due
193: to the closer distance to reionization.   Nonetheless, these features provide an opportunity
194: to test the slow-roll inflationary consistency relation~\cite[e.g.,][]{Lidetal97},
195: \begin{equation}
196: n_t = -r/8,
197: \label{eq:consrel}
198: \end{equation}
199:  between the tensor-to-scalar ratio $r$
200: and tensor tilt $n_t$. Deviations from the consistency relation and
201: running of the tilt come in at second order in the slow-roll parameters.  Specifically, $r$ is four times the ratio of the 
202: tensor power spectrum amplitude (for one component of gravitational waves) 
203: to the scalar curvature power spectrum amplitude, in accordance with the 
204: definition used by CAMB and \wmap.
205: 
206: Note that we quote $r$ at a normalization scale of $k_{\rm pivot} = 0.01$ Mpc$^{-1}$
207: whereas 3-year \wmap\ results quote it at $k_{\rm pivot}=0.002$ Mpc$^{-1}$ and CAMB defaults 
208: to $k_{\rm pivot}=0.05$ Mpc$^{-1}$.  Our choice, corresponding to the recombination peak at 
209: $\ell \approx 100$, better reflects the scale at which the tensor spectrum can be best measured.
210: We will use subscripts to denote choices of scale other than $0.01$~Mpc$^{-1}$, 
211: e.g. $r_{0.05}$ for normalization at $k_{\rm pivot}=0.05$~Mpc$^{-1}$, so that
212: \begin{equation}
213: r_{k_{\rm pivot}/{\rm Mpc}^{-1}} = r \left(\frac{k_{\rm pivot}}{0.01~{\rm Mpc}^{-1}}\right)^{n_t+1-n_s},
214: \label{eq:rscale}
215: \end{equation}
216: assuming no running of the scalar or tensor spectral indices.
217: Since we only consider small deviations from scale-invariance, the tensor-to-scalar 
218: ratio varies little with the normalization scale so any corrections to 
219: Eq.~(\ref{eq:consrel}) due to changing $k_{\rm pivot}$ are correspondingly small.
220: 
221: 
222: Cosmic variance and reionization history uncertainties in the interpretation of the reionization peak
223: limit the ability to measure the spectrum from this technique.    
224: In Fig.~\ref{fig:degen}, we show that
225: variation  in the optical depth to reionization, $\tau$, can mimic changes to 
226: $n_t$ within the precision of cosmic variance of the individual $\ell$ modes,
227: \begin{equation}
228: {\Delta \cltens \over \cltens} \approx \sqrt{2 \over 2\ell +1}\,.
229: \end{equation}
230: 
231: 
232: The example of Fig.~\ref{fig:degen} shows that there are at least two ways of breaking this 
233: degeneracy. One is by measuring the tensor spectrum at $\ell>100$, where 
234: the tilt of the spectrum matters but the optical depth has no effect 
235: other than an overall rescaling of the spectrum by $e^{-2\tau}$, which 
236: we absorb by changing the scalar amplitude $A_s$ to keep
237: $A_s e^{-2\tau}$ fixed.
238: Just as in the scalar case, if the tensor spectrum can be precisely
239: measured beyond the recombination peak, the global constraint on $n_t$ will not
240: be sensitive to reionization.  Unlike the scalar case, $\cltens$ falls sharply at higher
241: $\ell$ and becomes masked by $B$-modes 
242: generated by lensing of $E$-modes~\cite{ZalSel98} ($\cllens$; see Fig.~\ref{fig:cl}).  The power in the lensing 
243: $B$-modes is expected to be greatest at $\ell \sim 1000$, a smaller scale 
244: than the reionization and recombination peaks of the tensor spectrum. 
245: The relative amplitude of the tensor and lensing contributions to $\clbb$ depends on the 
246: tensor-to-scalar ratio, which the 3-year \wmap\ 
247: data restrict to $r\lesssim 0.3$~\cite{Speetal07}.
248: 
249: If $r$ is near the current upper limits,  the lensing spectrum can be statistically subtracted to a large extent.  Nevertheless, instrumental and foreground limitations
250: may still prevent the extraction of information from scales beyond the recombination peak for
251: next generation experiments such as {\it Planck} \cite{Bowetal04,Tucetal05,VerPeiJim06, AmbCooKap07}.
252: At much lower $r$, the best CMB constraints on $n_t$ will come from the combination
253: of the recombination and reionization peaks unless the lensing signal can be
254: subtracted directly from the polarization maps~\cite{HuOka01,KnoSon02,KesCooKam02,SelHir03,Marian:2007sr}.
255: 
256: The $\tau-n_t$ degeneracy is also broken through the constraint 
257: on $\tau$ from the $E$-mode reionization peak, which depends on
258: $\tau$ but not $n_t$ as long as $r$ is small enough that $\clee$ is 
259: dominated by scalar perturbations.  
260: To break the degeneracy in this way, it is important to have accurate constraints
261: on $\tau$ that do not depend on overly simplistic assumptions about the reionization 
262: history.  
263: 
264: The shape of the 
265: reionization peak depends on the history of the spatially-averaged 
266: ionized fraction, $x_e(z)$.  In Fig.~\ref{fig:cl}, we illustrate two models with
267: similar optical depths but different reionization histories.  
268: A model-dependent analysis of the polarization 
269: power spectrum that uses an incorrect form of $x_e(z)$ can result 
270: in significant bias in the total optical depth to 
271: reionization~\cite{Kapetal03,Holetal03,Coletal05,MorHu07b}.
272: To the extent that the constraint on $n_t$ 
273: relies on measurements of the reionization peak of $\cltens$, it may be biased as well in such an analysis.  We shall see that
274: the use of model-independent principal components of the reionization history 
275: protects against such biases at little cost to the precision of
276: the inflationary test.
277: 
278: 
279: 
280: 
281: 
282: 
283: % ****************************************
284: \begin{figure}
285: \centerline{\psfig{file=f2.eps, width=3.0in}}
286: \caption{$E$- and $B$-mode polarization angular power spectra and 
287: reionization histories (\emph{inset}) for an extended, 
288: ``double'' reionization history with $\tau = 0.090$ (\emph{thick curves}) 
289: and a  nearly-instantaneous reionization model with optical depth 
290: $\tau = 0.105$ (\emph{thin}). 
291: For the $B$-modes, both the tensor spectra [with tensor-to-scalar 
292: ratio $r=0.3$ (\emph{solid}) and $r=0.03$ (\emph{dashed})] 
293: and lensing spectra are plotted. Dotted curves are the assumed 
294: \planck\ and \cmbpol\ noise power spectra, from top to bottom 
295: at low $\ell$ respectively.
296: %\vskip 0.25cm
297: }
298: \label{fig:cl}
299: \end{figure}
300: % ****************************************
301: 
302: 
303: 
304: 
305: 
306: 
307: % =====================================================
308: \section{Principal Component Parametrization of Reionization}
309: \label{sec:pcs}
310: 
311: Following~\cite{HuHol03,MorHu07b}, we parametrize the reionization history 
312: as a free function of redshift by decomposing $x_e(z)$ into its
313: principal components with respect to the $E$-mode polarization of the 
314: CMB:
315: \begin{equation}
316: x_e(z)=\xef(z)+\sum_{\mu}m_{\mu}S_{\mu}(z),
317: \label{eq:mmutoxe}
318: \end{equation}
319: where the principal components, $S_{\mu}(z)$, 
320: are the eigenfunctions of the Fisher matrix that describes 
321: the dependence of $\clee$ on $x_e(z)$, $m_{\mu}$ are the amplitudes of the 
322: principal components for a particular reionization history, and 
323: $\xef(z)$ is the fiducial model at which the Fisher matrix is computed. 
324: In practice, we construct the principal components assuming $E$-mode
325: information only by taking the tensor-to-scalar ratio $r=0$.  For allowed
326: values of $r$, the information on the ionization history from the tensors
327: is subdominant to the scalars. 
328: The inverses of the eigenvalues of the Fisher matrix give the 
329: estimated variances of the principal components, $\sigma_{\mu}^2$, which 
330: determine the ordering of the components by requiring that 
331: $\sigma_{\mu}^2 < \sigma_{\mu+1}^2$.
332: The main advantage of using principal components as a basis for $x_e(z)$ is 
333: that only a small number of the components are required to completely 
334: describe the effects of reionization on large-scale CMB polarization, 
335: so we obtain a very general parametrization of the reionization history at 
336: the expense of only a few additional parameters.
337: 
338: The principal components are defined over a limited range in redshift, 
339: $\zmin<z<\zmax$, with $x_e=0$ at $z>\zmax$ and $x_e=1$ at $z<\zmin$. 
340: We take $\zmin=6$, $\zmax=30$, and constant $\xef(z)=0.15$ here, 
341: and refer to~\cite{MorHu07b} for 
342: further discussion of the choices of these and other parameters 
343: related to the principal components.
344: 
345: 
346: 
347: % ****************************************
348: \begin{figure}
349: \centerline{\psfig{file=f3.eps, width=3.0in}}
350: \caption{The five lowest-variance principal components of $x_e(z)$ 
351: over redshifts $6<z<30$, with increasing variance from top to bottom.
352: %\vskip 0.25cm
353: }
354: \label{fig:pc}
355: \end{figure}
356: % ****************************************
357: 
358: 
359: For complete representation of the effects of 
360: reionization between $\zmin$ and $\zmax$ on the low-$\ell$ $E$-mode 
361: spectrum to better accuracy than 
362: cosmic variance, no more than the first five principal components are 
363: needed (assuming $\zmax\lesssim 30$)~\cite{HuHol03,MorHu07b}. 
364: Due to projection effects~\cite{HuWhi97c}, 
365: the accuracy to which the lowest-variance principal components reconstruct 
366: $\cltens$ at low $\ell$ is even better than for the scalar $\clee$.
367: In the MCMC analysis presented in the following section, we always use 
368: the five lowest-variance principal components of $x_e(z)$ with $\zmax=30$, which 
369: we show in Fig.~\ref{fig:pc} (see~\cite{MorHu07b} for 
370: the effects of using a different number of components to analyze $E$-mode data). 
371: The amplitudes of these components then serve to 
372: parametrize general reionization histories in the analysis of 
373: CMB polarization data. 
374: 
375: 
376: 
377: % =====================================================
378: \section{Markov Chain Monte Carlo Constraints}
379: \label{sec:mcmc}
380: 
381: We use Markov Chain Monte Carlo (MCMC) analysis to explore the joint 
382: effects of the reionization history and inflationary 
383: parameters on CMB polarization power spectra~\cite[see e.g.][]{Chretal01,KosMilJim02,Dunetal04}. 
384: Chains of Monte Carlo samples are generated using the publicly available 
385: code CosmoMC~\cite{LewBri02} \footnote{{\tt http://cosmologist.info/cosmomc/}},
386: which includes the code CAMB~\citep{Lewetal00} for computing 
387: theoretical angular power spectra at each point in the  
388: parameter space. We have modified both codes to allow 
389: specification of an arbitrary reionization history calculated from a set 
390: of principal component amplitudes using Eq.~(\ref{eq:mmutoxe}), as described in~\cite{MorHu07b}. 
391: 
392: For chains in which $x_e(z)$ is parametrized by its principal components, 
393: the parameters that we vary include the amplitudes of the first five components, 
394: the tensor-to-scalar ratio at $k_{\rm pivot}=0.05$~Mpc$^{-1}$, and the tilt of the tensor spectrum:
395: $\{m_1,m_2,m_3,m_4,m_5,r_{0.05},n_t\}$. When using the simple instantaneous 
396: form of $x_e(z)$, the chain parameters are instead $\{\tau,r_{0.05},n_t\}$ with 
397: optical depth taking the place of principal components. 
398: We compute $r$ at $k_{\rm pivot}=0.01$~Mpc$^{-1}$ as a derived parameter using 
399: Eq.~(\ref{eq:rscale}) after generating each chain.
400: 
401: When analyzing the parameter chains, we impose priors on the 
402: principal component amplitudes corresponding to physical values of the 
403: ionized fraction, $0 \leq x_e \leq 1$~\cite{MorHu07b}. 
404: These priors are conservative in the sense that all excluded models are unphysical, 
405: but the models we retain are not necessarily physical.
406: 
407: We use only polarization for parameter constraints, 
408: and assume that the values of the 
409: standard \lcdm\ parameters (besides $\tau$) are fixed by measurements of the 
410: CMB temperature anisotropies. 
411: To a good approximation the effect of $x_e(z)$ on the large-scale 
412: polarization is independent of the other parameters~\citep{MorHu07b}. 
413: 
414: Specifically, we take  
415: $\Omega_bh^2=0.0222$, $\Omega_ch^2=0.106$, $100\theta=1.04$ (corresponding 
416: to $h=0.73$), $A_s e^{-2\tau}=1.7 \times 10^{-9}$, and $n_s=0.96$, 
417: consistent with the \wmap\ 3-year temperature power spectrum.
418: When computing the optical depth to reionization we 
419: take the  helium fraction to be $Y_p=0.24$ and assume that helium is neutral. 
420: With these parameter values, the reionization optical depth out to $\zmin = 6$ 
421: is fixed at $\tau(\zmin)\approx 0.04$. 
422: The remaining contribution to the total optical depth from $\zmin < z < \zmax$ 
423: is determined by the 
424: values of $\{m_{\mu}\}$ for each sample in the chains.
425: The default bin width for the fiducial models and principal components 
426: is $\dz=0.25$, which is small enough that 
427: numerical effects related to binning should be negligible. 
428: 
429: 
430: For each scenario that we study, we run four separate chains 
431: until the Gelman and Rubin convergence statistic $R$, 
432: corresponding to the ratio of the variance of parameters between chains to 
433: the variance within each chain, satisfies 
434: $R-1<0.01$~\citep{GelRub92,BroGel98}. The convergence diagnostic  
435: of~\cite{RafLew92} is used to determine how much each chain must be thinned 
436: to obtain independent samples.  
437: 
438: 
439: % ****************************************
440: \begin{table*}
441: \caption{Constraints on $\tau$, $r$, and $n_t$ for simulated data based 
442: on the double reionization history with optical depth $\tau^{\rm fid}=0.09$.
443: }
444: \begin{center}
445: \begin{tabular}{lllllrlr}
446: \hline
447: \hline
448:  & & & & Use PCs &  &  &  \\
449: \multicolumn{1}{c}{$r_{0.05}^{\rm fid}$} & \multicolumn{1}{c}{$r^{\rm fid}$} & \multicolumn{1}{c}{$n_t^{\rm fid}$} & Noise & of $x_e(z)$? & \multicolumn{1}{c}{$\tau$} & \multicolumn{1}{c}{$r$} & \multicolumn{1}{c}{$n_t$} \\
450: \hline
451: 0.3 & 0.299 & $-0.0375$ & CV & Yes & $0.091\pm0.005$ & $0.299\pm0.004$ & $-0.037\pm0.018$ \\
452: 0.3 & 0.299 & $-0.0375$ & CV & No & $0.118\pm0.003$ & $0.298\pm0.005$ & $-0.021\pm0.018$ \\
453: 0.3 & 0.299 & $-0.0375$ & Planck & Yes & $0.094\pm0.009$ & $0.321\pm0.095$ & $0.041\pm0.196$ \\
454: 0.3 & 0.299 & $-0.0375$ & Planck & No & $0.113\pm0.007$ & $0.390\pm0.096$ & $0.398\pm0.164$ \\
455: 0.03 & 0.0283 & $-0.00375$ & CV & Yes & $0.091\pm0.005$ & $0.0283\pm0.0006$ & $0.001\pm0.051$ \\
456: 0.03 & 0.0283 & $-0.00375$ & CV & No & $0.121\pm0.003$ & $0.0288\pm0.0006$ & $0.162\pm0.049$ \\
457: 0.03 & 0.0283 & $-0.00375$ & CMBPol & Yes & $0.092\pm0.007$ & $0.032\pm0.009$ & $0.069\pm0.184$ \\
458: 0.03 & 0.0283 & $-0.00375$ & CMBPol & No & $0.122\pm0.005$ & $0.038\pm0.010$ & $0.491\pm0.165$ \\
459: \hline
460: \hline
461: \end{tabular}
462: \end{center}
463: \label{tab:rnttau}
464: \end{table*}
465: % ****************************************
466: 
467: 
468: 
469: % =====================================================
470: \subsection{Simulated data}
471: \label{sec:data}
472: 
473: We use CAMB to generate model $E$- and $B$-mode polarization 
474: power spectra. We take the data to be the exact values of a given model
475: $\hat\clee =\clee$ and $\hat\clbb =\clbb$, 
476: neglecting cosmic variance and noise, so we expect constraints to be 
477: centered on the fiducial parameter values rather than displaced by 
478: $\sim 1~\sigma$. These constraints can be thought of as the average over 
479: many possible realizations of the data. (See~\cite{MorHu07b} for a discussion of the  
480: effects of cosmic variance when using realizations drawn from $C_{\ell}$ instead of 
481: taking $C_{\ell}$ as the data.)
482: 
483: For the $j$th sample in a chain, the likelihood is
484: \begin{eqnarray}
485: -\ln L_{(j)}&=&\sum_{\ell=2}^{\lmax}\left(\ell+\frac{1}{2}\right)
486: f_{\rm sky}^2 \\
487: && \times
488: \left(\frac{\hat{{\bf C}}_{\ell}^{EE}}{{\bf C}_{\ell(j)}^{EE}}
489: +\frac{\hat{{\bf C}}_{\ell}^{BB}}{{\bf C}_{\ell(j)}^{BB}}
490: +\ln\frac{{\bf C}_{\ell(j)}^{EE}{\bf C}_{\ell(j)}^{BB}}
491: {\hat{{\bf C}}_{\ell}^{EE}\hat{{\bf C}}_{\ell}^{BB}}-2\right), \nonumber
492: \label{eq:like}
493: \end{eqnarray}
494: where 
495: $\hat{{\bf C}}_{\ell}=\hat{C}_{\ell}+N_{\ell}$ is the sum of the 
496: simulated data and noise spectra, and 
497: ${\bf C}_{\ell(j)}=C_{\ell(j)}+N_{\ell}$ is the 
498: theoretical spectrum calculated with the 
499: parameter values at step $j$ in the chain plus the noise spectrum. 
500: For consistency with
501: CosmoMC, the likelihood contains an extra factor of $f_{\rm sky}$, 
502: the fraction of sky observed \footnote{The extra factor
503: of $f_{\rm sky}$ is meant to approximately model the loss of information due to mode coupling 
504: with incomplete sky coverage, but the correct approach depends on the nature of the information extracted.}.
505: 
506: We set $N_{\ell}=0$ to simulate measurements limited only by cosmic variance.
507: For more realistic scenarios based on \planck\ and \cmbpol, we model noise as
508: \begin{equation}
509: N_{\ell} = \left(\frac{w_p^{-1/2}}{\mu{\rm K\mbox{-}rad}}\right)^2 
510: \exp\left[\frac{\ell(\ell+1)(\theta_{\rm FWHM}/{\rm rad})^2}{8\ln 2}\right],
511: \end{equation}
512: where $w_p^{-1/2}$ is the polarization noise level and 
513: $\theta_{\rm FWHM}$ is the beam width.
514: 
515: 
516: We typically compute the likelihood up to $\lmax=1000$. The $E$-mode spectrum 
517: and $B$-mode tensor spectrum are fixed at $\ell \gtrsim 100$ by setting 
518: $A_s e^{-2\tau}$ constant in the Monte Carlo chains. The impact on 
519: $r$ and $n_t$ constraints of the $B$-mode spectrum at multipoles 
520: greater than a few hundred is negligible in the presence of the 
521: lensing spectrum and experimental noise for all allowed values of $r$.
522: 
523: The fiducial reionization history is the extended, double reionization model with 
524: polarization spectra plotted in Fig.~\ref{fig:cl}, for which $\tau=0.090$. 
525: This function is chosen because it is not well described by the instantaneous 
526: reionization model, so biases due to assuming instantaneous $x_e(z)$ 
527: should be readily apparent in the parameter constraints. 
528: 
529: 
530: We consider two fiducial values of the tensor-to-scalar 
531: ratio, $r_{0.05}=0.3$ and $r_{0.05}=0.03$. Using the consistency relation [Eq.~(\ref{eq:consrel})] 
532: to set $n_t$ and taking the fiducial scalar tilt $n_s=0.96$, these tensor-to-scalar 
533: ratios correspond to $r=0.299$ and $r=0.0283$ at $k_{\rm pivot}=0.01$~Mpc$^{-1}$, respectively.
534: 
535: 
536: Assuming a power-law primordial spectrum, the larger fiducial tensor-to-scalar 
537: ratio, $r_{0.05}=0.3$, is approximately the 68\% 
538: upper limit on $r$ from 3-year \wmap\ data, although 
539: when large-scale structure data are included it is ruled out at 
540: about 95\% CL~\cite{Speetal07}.
541: A tensor spectrum with this value of $r$ should be 
542: detectable by the \planck\ satellite~\cite{Planck}. 
543: The smaller value of $r$ may be out of 
544: the reach of \planck\ but accessible to a next-generation CMB satellite, 
545: such as the proposed \cmbpol\ \cite{CMBTaskForce}. 
546: In addition to CV-limited 
547: measurements, we consider a noise spectrum based on 
548: \planck\ with sensitivity $w_p^{-1/2}=81~\mu{\rm K}'$ and beam size 
549: $\theta_{\rm FWHM}=7.1'$ for the $r_{0.05}=0.3$ simulated data~\cite{Planck,Albetal06}, and 
550: low-resolution \cmbpol-like noise with $w_p^{-1/2}=20~\mu{\rm K}'$ and
551: $\theta_{\rm FWHM}=60'$ for $r_{0.05}=0.03$~\cite{CMBTaskForce} 
552: (dotted curves in Fig.~\ref{fig:cl}).
553: We assume $f_{\rm sky}=0.8$ for both \planck\ and \cmbpol\ and take $f_{\rm sky}=1$ 
554: for the more idealized, CV-limited data.
555: In all cases we neglect foregrounds, assuming that they can be
556: adequately subtracted using polarization data from frequency channels 
557: not used for cosmological parameter estimation.
558: 
559: The $B$-mode lensing spectrum has a significant impact on constraints on $r$ and 
560: $n_t$, but including lensing significantly slows down the computation of 
561: the angular power spectra in CAMB. However, $\cllens$ is nearly independent of 
562: the parameters that we vary in the Monte Carlo chains. Rather than computing 
563: the effects of lensing directly, then, we treat the lensing 
564: spectrum as a fixed contribution to the noise power spectrum. 
565: Tests comparing this approximation to an analysis including the
566: full effects of lensing show that the constraints obtained for $r$ and $n_t$ are 
567: the same. Using $\tau$-dependent lensing spectra computed for each 
568: Monte Carlo sample appears to improve the constraint on $\tau$, but this is 
569: an artifact due to fixing $A_s e^{-2\tau}$ and other parameters that also
570: affect $\cllens$, mainly  $\Omega_m h^2$~\cite{SmiHuKap06}.
571: 
572: % ****************************************
573: \begin{figure}
574: \centerline{\psfig{file=f4.eps, width=3.5in}}
575: \caption{2D marginalized 68 and 95\% contours for the optical depth ($\tau$),
576: tensor-to-scalar ratio ($r$), and tensor spectral index ($n_t$). 
577: For the simulated polarization spectra, we take $x_e(z)$ to be a double  
578: reionization history with $\tau=0.090$ (see Fig.~\ref{fig:cl}).
579: The fiducial tensor-to-scalar ratio is set to $r_{0.05}=0.3$ ($r=0.299$ at $k=0.01$~Mpc$^{-1}$), 
580: and the fiducial tensor tilt is assumed 
581: to obey the consistency relation, $n_t=-r/8$ (\emph{right panels, line}). 
582: Crosses indicate these fiducial parameter values. 
583: For the thick contours, the chains include the five 
584: lowest-variance principal components of $x_e(z)$.  Thin contours 
585: show the constraints when $x_e(z)$ is instead assumed to be 
586: instantaneous and parametrized only by $\tau$.
587: The lensing spectrum is treated as a contribution to the noise.  
588: \emph{Top panels}: CV-limited data; \emph{bottom}: including \planck-like noise. 
589: }
590: \label{fig:rnttau03}
591: \end{figure}
592: % ****************************************
593: 
594: 
595: % ****************************************
596: \begin{figure}
597: \centerline{\psfig{file=f5.eps, width=3.5in}}
598: \caption{Same as Fig.~\ref{fig:rnttau03}, but with fiducial 
599: tensor-to-scalar ratio $r_{0.05}=0.03$ ($r=0.0283$ at $k=0.01$~Mpc$^{-1}$). 
600: \emph{Top panels}: CV-limited data; 
601: \emph{bottom}: including \cmbpol-like noise.
602: }
603: \label{fig:rnttau003}
604: \end{figure}
605: % ****************************************
606: 
607: 
608: 
609: % =====================================================
610: \subsection{Results}
611: \label{sec:results}
612: 
613: 
614: Table~\ref{tab:rnttau} lists the 1D marginalized constraints on 
615: $\tau$, $r$, and $n_t$ from the MCMC 
616: analysis for each case study. 
617: The constraints on $\tau$, $r$, and $n_t$ are plotted in 
618: Figs.~\ref{fig:rnttau03} and~\ref{fig:rnttau003} as 2D contours 
619: after marginalizing over all other parameters. The two sets of contours in each panel
620: use the same simulated data, but different parametrizations of $x_e(z)$. 
621: For the thick contours, the five lowest-variance principal components of $x_e(z)$ 
622: are included in the Monte Carlo chains, while the thin contours come from 
623: chains that treat $x_e(z)$ as instantaneous reionization with only one 
624: parameter, $\tau$. For the principal component chains, $\tau$ is 
625: derived from the principal component amplitudes, $\{m_{\mu}\}$.
626: 
627: Since the constraint on optical depth comes primarily from the 
628: reionization peak
629: of the $E$-mode spectrum, estimates of $\tau$ in all cases in 
630: Tab.~\ref{tab:rnttau} are affected similarly by 
631: the parametrization of the reionization history. As noted in~\cite{Kapetal03,LewWelBat06}, using the 
632: one-parameter instantaneous $x_e(z)$ when that model is not sufficient to describe 
633: the true reionization history (in this case, the double reionization history 
634: represented by the thick curves in Fig.~\ref{fig:cl}) can lead to a significant 
635: bias in the constraint on $\tau$.  Using a set of principal components to 
636: parametrize $x_e(z)$ removes the bias in the optical depth~\cite{MorHu07b}.
637: 
638: The impact of the reionization history assumptions for inflationary parameters
639: depends more strongly on the values of $r$ and the noise. 
640: We consider first the most optimistic scenario for measurement of the 
641: $B$-mode tensor spectrum: a CV-limited measurement of 
642: $\cltens$ with approximately the largest amplitude currently allowed, 
643: $r_{0.05}=0.3$. With this tensor-to-scalar ratio, the tensor spectrum 
644: dominates over the $B$-mode lensing spectrum for $\ell \lesssim 150$ 
645: (see Fig.~\ref{fig:cl}), so the recombination peak of $\cltens$ is 
646: free from contamination over roughly a decade in $\ell$.  Moreover, the lensing
647: spectrum may be statistically subtracted well beyond the multipole where
648: it contributes equal power.
649: 
650: Since the amplitude of the tensor spectrum depends on $\tau$, $r$, and $n_t$, 
651: we expect some degeneracy between these parameters, as described in \S~\ref{sec:param}. 
652: Because of this degeneracy, a bias in $\tau$ can generate biases in the inflationary 
653: parameters as well, mainly $n_t$ (since $r$ at $k=0.01$~Mpc$^{-1}$ is tightly constrained 
654: by measuring $\cltens$ at $\ell\approx 100$).  However, $\tau$ only 
655: affects the low-$\ell$ reionization peak (with our assumption that 
656: $A_s e^{-2\tau}$ is held fixed), so these biases will only 
657: be significant if the reionization peak contributes significantly to 
658: constraints on the tensor spectrum. In this optimistic scenario with $r_{0.05}=0.3$ and 
659: cosmic variance-limited measurements, enough of the tensor recombination peak 
660: is observable so that additionally measuring
661: the reionization peak has a negligible effect on 
662: constraints on $r$ and $n_t$; the values of these parameters are 
663: determined almost entirely by $\clbb$ at $20 \lesssim \ell \lesssim 500$. 
664: As a result, the bias in the inflationary parameters 
665: due to incorrectly assuming instantaneous reionization is small; as 
666: the contours in the top right panel of Fig.~\ref{fig:rnttau03} show, the true parameter values 
667: remain within the 68\% CL even with this bias.
668: 
669: 
670: Now consider the same tensor-to-scalar ratio but more realistic assumptions 
671: about the experimental noise. Using our assumed noise spectrum for \planck\ 
672: as described in \S~\ref{sec:data}, we find constraints on $\tau$, $r$, and $n_t$ 
673: as shown in the bottom panels of Fig.~\ref{fig:rnttau03}.
674: The most obvious difference from the CV-limited case is that 
675: the parameter uncertainties are larger, especially for the inflationary 
676: parameters. There is also a difference in the instantaneous reionization biases: 
677: while the optical depth bias is similar to the CV-limited case, 
678: $n_t$ is biased much more with \planck-like noise than with a CV-limited measurement. 
679: The additional bias is due to the greater dependence on the reionization peak of $\cltens$ 
680: for $n_t$ constraints since the noise associated with \planck\ 
681: makes measurement of the recombination peak much more difficult.  
682: 
683: Note that while $r$ at $k_{\rm pivot}=0.01$~Mpc$^{-1}$ is still determined 
684: fairly accurately in this scenario, 
685: if the tensor-to-scalar ratio were quoted elsewhere, e.g. as $r_{0.002}$ or $r_{0.05}$, 
686: a significant bias would appear in that parameter as well.
687: For example, Fig.~\ref{fig:rbias} shows that $r_{0.0007}$, normalized at the 
688: approximate scale of the reionization peak, lies significantly 
689: below the fiducial value in most of our test cases.
690: The choice of scale also affects the uncertainty in the tensor-to-scalar ratio; 
691: while the error we quote for $n_t$ agrees with similar studies, e.g.~\cite{VerPeiJim06}, 
692: the error on $r$ differs in general since degeneracy with $n_t$ due to a 
693: different choice of $k_{\rm pivot}$ can greatly increase $\sigma_r$.
694: 
695: 
696: % ****************************************
697: \begin{figure}
698: \centerline{\psfig{file=f6.eps, width=3.0in}}
699: \caption{1D marginalized distributions of $r_{0.0007}$, the tensor-to-scalar ratio on scales
700: near the reionization peak, for each 
701: of the four MCMC scenarios in Figs.~\ref{fig:rnttau03} and~\ref{fig:rnttau003}.
702: The fiducial values of $r_{0.0007}$ are indicated by vertical dashed lines, and the normalization
703: is arbitrary.
704: }
705: \label{fig:rbias}
706: \end{figure}
707: % ****************************************
708: 
709: 
710: 
711: With the larger bias in $n_t$ in the presence of 
712: \planck-like noise, the true parameter values are excluded at the 
713: 95\% CL. Moreover, the consistency relation for single-field slow-roll 
714: inflation (plotted as a line in the $r-n_t$ plots) 
715: is excluded at about the same confidence level. An incorrect assumption about 
716: the reionization history could therefore lead to wrongly rejecting the simplest
717: class of inflationary models. 
718: 
719: As the thick contours in Fig.~\ref{fig:rnttau03} show, 
720: when general reionization histories are considered by parametrizing 
721: $x_e(z)$ by its principal components, the biases in $\tau$, $r$, and $n_t$ are 
722: all removed. The constraints on $r$ and $n_t$ still may not be very strong for 
723: an experiment like \planck, but at least when using the principal components 
724: of $x_e(z)$ one would not be led to exclude the true underlying model 
725: of inflation.  Furthermore, the precision of the constraints is not substantially
726: degraded by the principal component parametrization.
727: 
728: Note that the exact magnitude and direction of the parameter biases depend 
729: on what the true reionization history actually is. The goal here is not to 
730: make specific predictions about these biases, 
731: but rather to give an idea of how large they
732: could be and to show that they can be avoided by allowing for 
733: general models of reionization.
734: 
735: 
736: 
737: How does uncertainty about reionization influence constraints on inflationary 
738: parameters if the amplitude of the tensor spectrum is smaller? In general, 
739: we expect the importance of the reionization peak of $\cltens$ relative 
740: to the recombination peak to increase as the tensor spectrum drops further 
741: below the dominant source of ``noise'' on large scales, whether that be the 
742: noise spectrum of an experiment or the $B$-mode lensing signal. 
743: To check this expectation, we use simulated data with a fiducial tensor-to-scalar 
744: ratio of $r_{0.05}=0.03$. The resulting constraints on $\tau$, $r$, and $n_t$, 
745: plotted as for the $r_{0.05}=0.3$ results, are shown in Fig.~\ref{fig:rnttau003} 
746: for both cosmic variance-limited measurements 
747: (top panels) and the \cmbpol-like noise spectrum described in 
748: \S~\ref{sec:data} (bottom panels).
749: 
750: With a CV-limited experiment and $r_{0.05}=0.03$, 
751: due to the greater influence of lensing relative to the tensor spectrum, 
752: the reionization peak ($\ell<20$) and recombination peak at $\ell\lesssim 300$ 
753: contribute roughly equally to the constraint on $n_t$. 
754: Because of this, assuming the wrong reionization history biases $n_t$ more 
755: than for $r_{0.05}=0.3$, and in this case 
756: the consistency relation would be excluded at more than the $95\%$ CL.
757: 
758: 
759: If we add the \cmbpol\ noise spectrum ($w_p^{-1/2}=20~\mu{\rm K}'$, 
760: $\theta_{\rm FWHM}=60'$), both biases and errors on 
761: parameters are larger. The constraints in this case are quite similar 
762: to the constraints in the case of {\it Planck} with $r_{0.05}=0.3$ (except with all 
763: $r$ values reduced by a factor of 10), which makes sense 
764: since the low-$\ell$ amplitude of \cmbpol\ noise relative to $\cltens$ with $r_{0.05}=0.03$ is
765: similar to that of \planck\ noise to $\cltens$ with $r_{0.05}=0.3$ (see Fig.~\ref{fig:cl}).
766: Note that our assumptions about the \cmbpol\ noise spectrum are 
767: on the conservative end of the range usually considered for such an 
768: experiment; better sensitivity ($\sim 1~\mu{\rm K}'$) and/or resolution 
769: ($\sim 1-10$~arcmin) would enable measurement of the tensor $B$-modes to 
770: smaller scales so that constraints on inflationary parameters would 
771: be closer to the CV-limited scenario.
772: 
773: 
774: For even lower values of $r$ than those considered here, the impact 
775: of reionization on constraints on inflationary parameters is likely to 
776: be the same or greater. Even for an idealized, cosmic variance-limited 
777: experiment, the $B$-mode signal due to lensing becomes a significant 
778: contaminant of $\cltens$ at $r<0.03$.  Due to the shape of the spectra, if 
779: any part of $\cltens$ is detectable above the lensing $B$-modes and noise 
780: for low tensor-to-scalar ratios it will be the reionization peak. 
781: Any bias in $\tau$ due to incorrect modeling of the reionization history 
782: will then cause the inferred value of $n_t$ to be biased. 
783: Of course, at very low $r$ the uncertainties are so large that
784: the parameters are not usefully constrained at all.
785: In this case, it may be possible to measure $n_t$ using 
786: direct observations by future gravitational-wave experiments to better 
787: accuracy than what is possible with CMB polarization~\cite{SmiPeiCoo06,ChoEfs06}.
788: 
789: 
790: 
791: Finally, note that although the choice of normalization scale $k_{\rm pivot}=0.01$~Mpc$^{-1}$ 
792: is intended to decorrelate $r$ from the other parameters, there is some remaining 
793: degeneracy with $n_t$. The direction of this degeneracy depends on whether or not 
794: $\cltens$ can be measured on scales smaller than $k_{\rm pivot}$. For CV-limited 
795: measurement of a $B$-mode spectrum with $r=0.3$, these smaller scales are 
796: observable. A larger tensor tilt that increases the power at $\ell \gtrsim 100$ 
797: can be compensated for by lowering the value of $r$, so $r$ and $n_t$ are anticorrelated 
798: in this case (top right panel of Fig.~\ref{fig:rnttau03}). 
799: If the tensor spectrum on such small scales is hidden by 
800: lensing $B$-modes or noise, then the low-$\ell$ side of the recombination peak becomes more 
801: important. Increasing tensor tilt lowers the power at $20 \lesssim \ell \lesssim 100$, 
802: so $r$ is correlated with $n_t$ so as to match the spectrum of the data on these scales 
803: (lower right panel of Fig.~\ref{fig:rnttau03} and right panels of Fig.~\ref{fig:rnttau003}).
804: 
805: 
806: 
807: 
808: % =====================================================
809: \section{Discussion}
810: \label{sec:discuss}
811: 
812: 
813: The value of the optical depth to reionization estimated from the 
814: CMB $E$-mode polarization spectrum on large scales can be biased by 
815: adopting a model that has insufficient freedom to 
816: describe the true reionization history.  Likewise, the use of 
817: simple reionization models can bias inflationary parameters such 
818: as the tensor-to-scalar ratio and tensor tilt that depend on the 
819: large-scale amplitude of the $B$-mode spectrum of primordial 
820: gravitational waves. In each case, the problem can be solved by 
821: using a more general parametrization of the reionization history. 
822: We have shown that using a small but complete set of the 
823: principal components of the reionization history effectively 
824: yields unbiased constraints on both reionization and inflationary parameters.
825: 
826: 
827: Measurements of $r$ and $n_t$ are only affected by 
828: the assumed form of the reionization history 
829: if the reionization peak of the tensor $B$-mode spectrum 
830: at the very largest scales is needed to precisely constrain the
831: parameters.  If, instead, good constraints can 
832: be obtained using only the $B$-mode recombination peak at intermediate scales, 
833: then assumptions about reionization do not affect tests of the 
834: consistency relation between $r$ and $n_t$.  They would
835: instead appear as false evidence for running of the tensor tilt in violation of slow-roll
836: expectations.
837: Measurement of the recombination peak however is inhibited by experimental noise and contamination from $E$-mode power converted to $B$-mode power 
838: by gravitational lensing, 
839: both of which become more important at smaller scales.
840: 
841: 
842: To study the potential impact of reionization on 
843: parameter constraints from $B$-mode polarization,
844: we have employed a Markov Chain Monte Carlo analysis of simulated 
845: CMB polarization power spectra and compared results for two 
846: descriptions of reionization: a simple, one-parameter, instantaneous 
847: reionization model, and a parametrization using 
848: principal components of the reionization history 
849: with respect to the $E$-mode polarization power spectrum. 
850: 
851: By varying the properties of the simulated polarization power spectra, including 
852: the fiducial tensor-to-scalar ratio and the noise spectrum, we 
853: have determined over what range of scales CMB polarization data is most important 
854: for constraining inflationary parameters in various scenarios. 
855: In particular, the question of whether the large-scale reionization peak 
856: of $\cltens$ or the smaller-scale 
857: recombination peak is more important determines the severity of bias in 
858: inflationary parameters when reionization is modeled incorrectly. 
859: 
860: If the tensor-to-scalar ratio is near the current upper limit of $r\sim 0.3$ 
861: and measurements of $B$-mode polarization are limited only by cosmic 
862: variance, then the spectrum on scales $20 \lesssim \ell \lesssim 500$ 
863: dominates constraints on $r$ and $n_t$ and incorrect assumptions about 
864: reionization do not strongly bias the results. If the true tensor-to-scalar 
865: ratio is more than a factor of a few smaller than this upper bound, however, 
866: then lensing $B$-modes limit the information that can be extracted from 
867: the recombination peak of the tensor spectrum alone. Furthermore, all-sky
868: experiments in the foreseeable future are likely to have noise that 
869: exceeds the lensing signal, making tests of inflation even more 
870: reliant on the reionization peak of the tensor $B$-modes on large scales.
871: In all of these cases, a general parametrization of reionization 
872: such as that provided by principal components allows the use of the 
873: $B$-mode reionization peak for inflationary parameter constraints without 
874: significantly worsening the errors on those parameters.
875: 
876: 
877: \vfill
878: \acknowledgments 
879: We thank S. DeDeo, C. Dvorkin, H. Peiris, A. Upadhye, and S. Wang
880: for useful conversations.
881: This work was supported by the KICP through the grant NSF PHY-0114422 and
882: the David and
883: Lucile Packard Foundation. 
884: MJM was additionally supported 
885: by a 
886: National Science Foundation Graduate Research Fellowship. 
887: WH was additionally supported by  the DOE through 
888: contract DE-FG02-90ER-40560.
889: 
890: 
891: 
892: \bibliographystyle{apsrev}
893: 
894: \begin{thebibliography}{54}
895: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
896: \expandafter\ifx\csname bibnamefont\endcsname\relax
897:   \def\bibnamefont#1{#1}\fi
898: \expandafter\ifx\csname bibfnamefont\endcsname\relax
899:   \def\bibfnamefont#1{#1}\fi
900: \expandafter\ifx\csname citenamefont\endcsname\relax
901:   \def\citenamefont#1{#1}\fi
902: \expandafter\ifx\csname url\endcsname\relax
903:   \def\url#1{\texttt{#1}}\fi
904: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
905: \providecommand{\bibinfo}[2]{#2}
906: \providecommand{\eprint}[2][]{\url{#2}}
907: 
908: \bibitem[{\citenamefont{{Guth}}(1981)}]{Gut81}
909: \bibinfo{author}{\bibfnamefont{A.~H.} \bibnamefont{{Guth}}},
910:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{23}}, \bibinfo{pages}{347}
911:   (\bibinfo{year}{1981}).
912: 
913: \bibitem[{\citenamefont{{Albrecht} and {Steinhardt}}(1982)}]{AlbSte82}
914: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Albrecht}}} \bibnamefont{and}
915:   \bibinfo{author}{\bibfnamefont{P.~J.} \bibnamefont{{Steinhardt}}},
916:   \bibinfo{journal}{Physical Review Letters} \textbf{\bibinfo{volume}{48}},
917:   \bibinfo{pages}{1220} (\bibinfo{year}{1982}).
918: 
919: \bibitem[{\citenamefont{{Linde}}(1982)}]{Lin82}
920: \bibinfo{author}{\bibfnamefont{A.~D.} \bibnamefont{{Linde}}},
921:   \bibinfo{journal}{Physics Letters B} \textbf{\bibinfo{volume}{108}},
922:   \bibinfo{pages}{389} (\bibinfo{year}{1982}).
923: 
924: \bibitem[{\citenamefont{{Sato}}(1981)}]{Sat81}
925: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{{Sato}}},
926:   \bibinfo{journal}{\mnras} \textbf{\bibinfo{volume}{195}},
927:   \bibinfo{pages}{467} (\bibinfo{year}{1981}).
928: 
929: \bibitem[{\citenamefont{{Hu} and {White}}(1996)}]{HuWhi96c}
930: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}} \bibnamefont{and}
931:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{White}}},
932:   \bibinfo{journal}{\prl} \textbf{\bibinfo{volume}{77}}, \bibinfo{pages}{1687}
933:   (\bibinfo{year}{1996}), \eprint{arXiv:astro-ph/9602020}.
934: 
935: \bibitem[{\citenamefont{{Spergel} and {Zaldarriaga}}(1997)}]{SpeZal97}
936: \bibinfo{author}{\bibfnamefont{D.~N.} \bibnamefont{{Spergel}}}
937:   \bibnamefont{and}
938:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Zaldarriaga}}},
939:   \bibinfo{journal}{Physical Review Letters} \textbf{\bibinfo{volume}{79}},
940:   \bibinfo{pages}{2180} (\bibinfo{year}{1997}),
941:   \eprint{arXiv:astro-ph/9705182}.
942: 
943: \bibitem[{\citenamefont{{Hu} et~al.}(1997)\citenamefont{{Hu}, {Spergel}, and
944:   {White}}}]{HuSpeWhi97}
945: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}},
946:   \bibinfo{author}{\bibfnamefont{D.~N.} \bibnamefont{{Spergel}}},
947:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{White}}},
948:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{51}}, \bibinfo{pages}{3288}
949:   (\bibinfo{year}{1997}), \eprint{arXiv:astro-ph/9605193}.
950: 
951: \bibitem[{\citenamefont{{Peiris} et~al.}(2003)\citenamefont{{Peiris},
952:   {Komatsu}, {Verde}, {Spergel}, {Bennett}, {Halpern}, {Hinshaw}, {Jarosik},
953:   {Kogut}, {Limon} et~al.}}]{Peietal03}
954: \bibinfo{author}{\bibfnamefont{H.~V.} \bibnamefont{{Peiris}}},
955:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Komatsu}}},
956:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Verde}}},
957:   \bibinfo{author}{\bibfnamefont{D.~N.} \bibnamefont{{Spergel}}},
958:   \bibinfo{author}{\bibfnamefont{C.~L.} \bibnamefont{{Bennett}}},
959:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Halpern}}},
960:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Hinshaw}}},
961:   \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{{Jarosik}}},
962:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Kogut}}},
963:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Limon}}},
964:   \bibnamefont{et~al.}, \bibinfo{journal}{\apjs}
965:   \textbf{\bibinfo{volume}{148}}, \bibinfo{pages}{213} (\bibinfo{year}{2003}),
966:   \eprint{arXiv:astro-ph/0302225}.
967: 
968: \bibitem[{\citenamefont{{Spergel} et~al.}(2007)\citenamefont{{Spergel}, {Bean},
969:   {Dor{\'e}}, {Nolta}, {Bennett}, {Dunkley}, {Hinshaw}, {Jarosik}, {Komatsu},
970:   {Page} et~al.}}]{Speetal07}
971: \bibinfo{author}{\bibfnamefont{D.~N.} \bibnamefont{{Spergel}}},
972:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Bean}}},
973:   \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{{Dor{\'e}}}},
974:   \bibinfo{author}{\bibfnamefont{M.~R.} \bibnamefont{{Nolta}}},
975:   \bibinfo{author}{\bibfnamefont{C.~L.} \bibnamefont{{Bennett}}},
976:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Dunkley}}},
977:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Hinshaw}}},
978:   \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{{Jarosik}}},
979:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Komatsu}}},
980:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Page}}},
981:   \bibnamefont{et~al.}, \bibinfo{journal}{\apjs}
982:   \textbf{\bibinfo{volume}{170}}, \bibinfo{pages}{377} (\bibinfo{year}{2007}),
983:   \eprint{arXiv:astro-ph/0603449}.
984: 
985: \bibitem[{\citenamefont{{Kamionkowski}
986:   et~al.}(1997)\citenamefont{{Kamionkowski}, {Kosowsky}, and
987:   {Stebbins}}}]{KamKosSte97}
988: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kamionkowski}}},
989:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Kosowsky}}},
990:   \bibnamefont{and}
991:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Stebbins}}},
992:   \bibinfo{journal}{Physical Review Letters} \textbf{\bibinfo{volume}{78}},
993:   \bibinfo{pages}{2058} (\bibinfo{year}{1997}),
994:   \eprint{arXiv:astro-ph/9609132}.
995: 
996: \bibitem[{\citenamefont{{Seljak} and {Zaldarriaga}}(1997)}]{SelZal97}
997: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{{Seljak}}} \bibnamefont{and}
998:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Zaldarriaga}}},
999:   \bibinfo{journal}{Physical Review Letters} \textbf{\bibinfo{volume}{78}},
1000:   \bibinfo{pages}{2054} (\bibinfo{year}{1997}),
1001:   \eprint{arXiv:astro-ph/9609169}.
1002: 
1003: \bibitem[{\citenamefont{{The Planck Collaboration}}(2006)}]{Planck}
1004: \bibinfo{author}{\bibnamefont{{The Planck Collaboration}}}
1005:   (\bibinfo{year}{2006}), \eprint{arXiv:astro-ph/0604069}.
1006: 
1007: \bibitem[{\citenamefont{{Oxley} et~al.}(2004)\citenamefont{{Oxley}, {Ade},
1008:   {Baccigalupi}, {deBernardis}, {Cho}, {Devlin}, {Hanany}, {Johnson}, {Jones},
1009:   {Lee} et~al.}}]{Oxletal04}
1010: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{{Oxley}}},
1011:   \bibinfo{author}{\bibfnamefont{P.~A.} \bibnamefont{{Ade}}},
1012:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Baccigalupi}}},
1013:   \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{{deBernardis}}},
1014:   \bibinfo{author}{\bibfnamefont{H.-M.} \bibnamefont{{Cho}}},
1015:   \bibinfo{author}{\bibfnamefont{M.~J.} \bibnamefont{{Devlin}}},
1016:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{{Hanany}}},
1017:   \bibinfo{author}{\bibfnamefont{B.~R.} \bibnamefont{{Johnson}}},
1018:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{{Jones}}},
1019:   \bibinfo{author}{\bibfnamefont{A.~T.} \bibnamefont{{Lee}}},
1020:   \bibnamefont{et~al.}, in \emph{\bibinfo{booktitle}{Infrared Spaceborne Remote
1021:   Sensing XII, Proceedings of the SPIE}}, edited by
1022:   \bibinfo{editor}{\bibfnamefont{M.}~\bibnamefont{{Strojnik}}}
1023:   (\bibinfo{year}{2004}), vol. \bibinfo{volume}{5543}, pp.
1024:   \bibinfo{pages}{320--331}, \eprint{arXiv:astro-ph/0501111}.
1025: 
1026: \bibitem[{\citenamefont{{Yoon} et~al.}(2006)\citenamefont{{Yoon}, {Ade},
1027:   {Barkats}, {Battle}, {Bierman}, {Bock}, {Brevik}, {Chiang}, {Crites},
1028:   {Dowell} et~al.}}]{Yooetal06}
1029: \bibinfo{author}{\bibfnamefont{K.~W.} \bibnamefont{{Yoon}}},
1030:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1031:   \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{{Barkats}}},
1032:   \bibinfo{author}{\bibfnamefont{J.~O.} \bibnamefont{{Battle}}},
1033:   \bibinfo{author}{\bibfnamefont{E.~M.} \bibnamefont{{Bierman}}},
1034:   \bibinfo{author}{\bibfnamefont{J.~J.} \bibnamefont{{Bock}}},
1035:   \bibinfo{author}{\bibfnamefont{J.~A.} \bibnamefont{{Brevik}}},
1036:   \bibinfo{author}{\bibfnamefont{H.~C.} \bibnamefont{{Chiang}}},
1037:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Crites}}},
1038:   \bibinfo{author}{\bibfnamefont{C.~D.} \bibnamefont{{Dowell}}},
1039:   \bibnamefont{et~al.}, in \emph{\bibinfo{booktitle}{Millimeter and
1040:   Submillimeter Detectors and Instrumentation for Astronomy III, Proceedings of
1041:   the SPIE}}, edited by
1042:   \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{{Zmuidzinas}}},
1043:   \bibinfo{editor}{\bibfnamefont{W.~S.} \bibnamefont{{Holland}}},
1044:   \bibinfo{editor}{\bibfnamefont{S.}~\bibnamefont{{Withington}}},
1045:   \bibnamefont{and} \bibinfo{editor}{\bibfnamefont{W.~D.}
1046:   \bibnamefont{{Duncan}}} (\bibinfo{year}{2006}), vol. \bibinfo{volume}{6275},
1047:   \eprint{arXiv:astro-ph/0606278}.
1048: 
1049: \bibitem[{\citenamefont{Bock et~al.}(2006)}]{CMBTaskForce}
1050: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Bock}} \bibnamefont{et~al.}
1051:   (\bibinfo{year}{2006}), \eprint{arXiv:astro-ph/0604101}.
1052: 
1053: \bibitem[{\citenamefont{{Maffei} et~al.}(2005)\citenamefont{{Maffei}, {Ade},
1054:   {Calderon}, {Challinor}, {de Bernardis}, {Dunlop}, {Gear},
1055:   {Giraud-H{\'e}raud}, {Goldie}, {Grainge} et~al.}}]{Mafetal05}
1056: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{{Maffei}}},
1057:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1058:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Calderon}}},
1059:   \bibinfo{author}{\bibfnamefont{A.~D.} \bibnamefont{{Challinor}}},
1060:   \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{{de Bernardis}}},
1061:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Dunlop}}},
1062:   \bibinfo{author}{\bibfnamefont{W.~K.} \bibnamefont{{Gear}}},
1063:   \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{{Giraud-H{\'e}raud}}},
1064:   \bibinfo{author}{\bibfnamefont{D.~J.} \bibnamefont{{Goldie}}},
1065:   \bibinfo{author}{\bibfnamefont{K.~J.~B.} \bibnamefont{{Grainge}}},
1066:   \bibnamefont{et~al.}, in \emph{\bibinfo{booktitle}{EAS Publications Series}}
1067:   (\bibinfo{year}{2005}), pp. \bibinfo{pages}{251--256}.
1068: 
1069: \bibitem[{\citenamefont{{Lawrence} et~al.}(2004)\citenamefont{{Lawrence},
1070:   {Gaier}, and {Seiffert}}}]{LawGaiSei04}
1071: \bibinfo{author}{\bibfnamefont{C.~R.} \bibnamefont{{Lawrence}}},
1072:   \bibinfo{author}{\bibfnamefont{T.~C.} \bibnamefont{{Gaier}}},
1073:   \bibnamefont{and}
1074:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Seiffert}}}, in
1075:   \emph{\bibinfo{booktitle}{Millimeter and Submillimeter Detectors for
1076:   Astronomy II, Proceedings of the SPIE}}, edited by
1077:   \bibinfo{editor}{\bibfnamefont{C.~M.} \bibnamefont{{Bradford}}},
1078:   \bibinfo{editor}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1079:   \bibinfo{editor}{\bibfnamefont{J.~E.} \bibnamefont{{Aguirre}}},
1080:   \bibinfo{editor}{\bibfnamefont{J.~J.} \bibnamefont{{Bock}}},
1081:   \bibinfo{editor}{\bibfnamefont{M.}~\bibnamefont{{Dragovan}}},
1082:   \bibinfo{editor}{\bibfnamefont{L.}~\bibnamefont{{Duband}}},
1083:   \bibinfo{editor}{\bibfnamefont{L.}~\bibnamefont{{Earle}}},
1084:   \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{{Glenn}}},
1085:   \bibinfo{editor}{\bibfnamefont{H.}~\bibnamefont{{Matsuhara}}},
1086:   \bibinfo{editor}{\bibfnamefont{B.~J.} \bibnamefont{{Naylor}}},
1087:   \bibnamefont{et~al.} (\bibinfo{year}{2004}), vol. \bibinfo{volume}{5498}, pp.
1088:   \bibinfo{pages}{220--231}.
1089: 
1090: \bibitem[{\citenamefont{{MacTavish} et~al.}(2007)\citenamefont{{MacTavish},
1091:   {Ade}, {Battistelli}, {Benton}, {Bihary}, {Bock}, {Bond}, {Brevik}, {Bryan},
1092:   {Contaldi} et~al.}}]{Macetal07}
1093: \bibinfo{author}{\bibfnamefont{C.~J.} \bibnamefont{{MacTavish}}},
1094:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1095:   \bibinfo{author}{\bibfnamefont{E.~S.} \bibnamefont{{Battistelli}}},
1096:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{{Benton}}},
1097:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Bihary}}},
1098:   \bibinfo{author}{\bibfnamefont{J.~J.} \bibnamefont{{Bock}}},
1099:   \bibinfo{author}{\bibfnamefont{J.~R.} \bibnamefont{{Bond}}},
1100:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Brevik}}},
1101:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{{Bryan}}},
1102:   \bibinfo{author}{\bibfnamefont{C.~R.} \bibnamefont{{Contaldi}}},
1103:   \bibnamefont{et~al.} (\bibinfo{year}{2007}), \eprint{arXiv:0710.0375}.
1104: 
1105: \bibitem[{\citenamefont{{Ruhl} et~al.}(2004)\citenamefont{{Ruhl}, {Ade},
1106:   {Carlstrom}, {Cho}, {Crawford}, {Dobbs}, {Greer}, {Halverson}, {Holzapfel},
1107:   {Lanting} et~al.}}]{Ruhl:2004kv}
1108: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Ruhl}}},
1109:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1110:   \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{{Carlstrom}}},
1111:   \bibinfo{author}{\bibfnamefont{H.-M.} \bibnamefont{{Cho}}},
1112:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{{Crawford}}},
1113:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Dobbs}}},
1114:   \bibinfo{author}{\bibfnamefont{C.~H.} \bibnamefont{{Greer}}},
1115:   \bibinfo{author}{\bibfnamefont{N.~w.} \bibnamefont{{Halverson}}},
1116:   \bibinfo{author}{\bibfnamefont{W.~L.} \bibnamefont{{Holzapfel}}},
1117:   \bibinfo{author}{\bibfnamefont{T.~M.} \bibnamefont{{Lanting}}},
1118:   \bibnamefont{et~al.}, in \emph{\bibinfo{booktitle}{Millimeter and
1119:   Submillimeter Detectors for Astronomy II, Proceedings of the SPIE}}, edited
1120:   by \bibinfo{editor}{\bibfnamefont{C.~M.} \bibnamefont{{Bradford}}},
1121:   \bibinfo{editor}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1122:   \bibinfo{editor}{\bibfnamefont{J.~E.} \bibnamefont{{Aguirre}}},
1123:   \bibinfo{editor}{\bibfnamefont{J.~J.} \bibnamefont{{Bock}}},
1124:   \bibinfo{editor}{\bibfnamefont{M.}~\bibnamefont{{Dragovan}}},
1125:   \bibinfo{editor}{\bibfnamefont{L.}~\bibnamefont{{Duband}}},
1126:   \bibinfo{editor}{\bibfnamefont{L.}~\bibnamefont{{Earle}}},
1127:   \bibinfo{editor}{\bibfnamefont{J.}~\bibnamefont{{Glenn}}},
1128:   \bibinfo{editor}{\bibfnamefont{H.}~\bibnamefont{{Matsuhara}}},
1129:   \bibinfo{editor}{\bibfnamefont{B.~J.} \bibnamefont{{Naylor}}},
1130:   \bibnamefont{et~al.} (\bibinfo{year}{2004}), vol. \bibinfo{volume}{5498}, pp.
1131:   \bibinfo{pages}{11--29}, \eprint{arXiv:astro-ph/0411122}.
1132: 
1133: \bibitem[{\citenamefont{Kogut et~al.}(2006)}]{Kogut:2006nd}
1134: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Kogut}} \bibnamefont{et~al.},
1135:   \bibinfo{journal}{New Astron. Rev.} \textbf{\bibinfo{volume}{50}},
1136:   \bibinfo{pages}{1009} (\bibinfo{year}{2006}),
1137:   \eprint{arXiv:astro-ph/0609546}.
1138: 
1139: \bibitem[{\citenamefont{{Polenta} et~al.}(2007)\citenamefont{{Polenta}, {Ade},
1140:   {Bartlett}, {Br{\'e}elle}, {Conversi}, {de Bernardis}, {Dufour}, {Gervasi},
1141:   {Giard}, {Giordano} et~al.}}]{Polenta07}
1142: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Polenta}}},
1143:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1144:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Bartlett}}},
1145:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Br{\'e}elle}}},
1146:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Conversi}}},
1147:   \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{{de Bernardis}}},
1148:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Dufour}}},
1149:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Gervasi}}},
1150:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Giard}}},
1151:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Giordano}}},
1152:   \bibnamefont{et~al.}, \bibinfo{journal}{New Astronomy Review}
1153:   \textbf{\bibinfo{volume}{51}}, \bibinfo{pages}{256} (\bibinfo{year}{2007}).
1154: 
1155: \bibitem[{\citenamefont{{Zaldarriaga}}(1997)}]{Zal97}
1156: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Zaldarriaga}}},
1157:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{55}}, \bibinfo{pages}{1822}
1158:   (\bibinfo{year}{1997}), \eprint{arXiv:astro-ph/9608050}.
1159: 
1160: \bibitem[{\citenamefont{{Page} et~al.}(2007)\citenamefont{{Page}, {Hinshaw},
1161:   {Komatsu}, {Nolta}, {Spergel}, {Bennett}, {Barnes}, {Bean}, {Dor{\'e}},
1162:   {Dunkley} et~al.}}]{Pagetal07}
1163: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Page}}},
1164:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Hinshaw}}},
1165:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Komatsu}}},
1166:   \bibinfo{author}{\bibfnamefont{M.~R.} \bibnamefont{{Nolta}}},
1167:   \bibinfo{author}{\bibfnamefont{D.~N.} \bibnamefont{{Spergel}}},
1168:   \bibinfo{author}{\bibfnamefont{C.~L.} \bibnamefont{{Bennett}}},
1169:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Barnes}}},
1170:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Bean}}},
1171:   \bibinfo{author}{\bibfnamefont{O.}~\bibnamefont{{Dor{\'e}}}},
1172:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Dunkley}}},
1173:   \bibnamefont{et~al.}, \bibinfo{journal}{\apjs}
1174:   \textbf{\bibinfo{volume}{170}}, \bibinfo{pages}{335} (\bibinfo{year}{2007}),
1175:   \eprint{arXiv:astro-ph/0603450}.
1176: 
1177: \bibitem[{\citenamefont{{Holder} et~al.}(2003)\citenamefont{{Holder}, {Haiman},
1178:   {Kaplinghat}, and {Knox}}}]{Holetal03}
1179: \bibinfo{author}{\bibfnamefont{G.~P.} \bibnamefont{{Holder}}},
1180:   \bibinfo{author}{\bibfnamefont{Z.}~\bibnamefont{{Haiman}}},
1181:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kaplinghat}}},
1182:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Knox}}},
1183:   \bibinfo{journal}{\apj} \textbf{\bibinfo{volume}{595}}, \bibinfo{pages}{13}
1184:   (\bibinfo{year}{2003}), \eprint{arXiv:astro-ph/0302404}.
1185: 
1186: \bibitem[{\citenamefont{{Mortonson} and {Hu}}(2007)}]{MorHu07b}
1187: \bibinfo{author}{\bibfnamefont{M.~J.} \bibnamefont{{Mortonson}}}
1188:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}},
1189:   \bibinfo{journal}{\apj (in press)}  (\bibinfo{year}{2007}),
1190:   \eprint{arXiv:0705.1132}.
1191: 
1192: \bibitem[{\citenamefont{{Keating} and {Miller}}(2006)}]{KeaMil06}
1193: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{{Keating}}} \bibnamefont{and}
1194:   \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{{Miller}}},
1195:   \bibinfo{journal}{New Astronomy Review} \textbf{\bibinfo{volume}{50}},
1196:   \bibinfo{pages}{184} (\bibinfo{year}{2006}), \eprint{arXiv:astro-ph/0508269}.
1197: 
1198: \bibitem[{\citenamefont{{Hu} and {Holder}}(2003)}]{HuHol03}
1199: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}} \bibnamefont{and}
1200:   \bibinfo{author}{\bibfnamefont{G.~P.} \bibnamefont{{Holder}}},
1201:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{68}},
1202:   \bibinfo{pages}{023001} (\bibinfo{year}{2003}),
1203:   \eprint{arXiv:astro-ph/0303400}.
1204: 
1205: \bibitem[{\citenamefont{{Lidsey} et~al.}(1997)\citenamefont{{Lidsey}, {Liddle},
1206:   {Kolb}, {Copeland}, {Barreiro}, and {Abney}}}]{Lidetal97}
1207: \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{{Lidsey}}},
1208:   \bibinfo{author}{\bibfnamefont{A.~R.} \bibnamefont{{Liddle}}},
1209:   \bibinfo{author}{\bibfnamefont{E.~W.} \bibnamefont{{Kolb}}},
1210:   \bibinfo{author}{\bibfnamefont{E.~J.} \bibnamefont{{Copeland}}},
1211:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{{Barreiro}}},
1212:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Abney}}},
1213:   \bibinfo{journal}{Reviews of Modern Physics} \textbf{\bibinfo{volume}{69}},
1214:   \bibinfo{pages}{373} (\bibinfo{year}{1997}), \eprint{arXiv:astro-ph/9508078}.
1215: 
1216: \bibitem[{\citenamefont{{Zaldarriaga} and {Seljak}}(1998)}]{ZalSel98}
1217: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Zaldarriaga}}}
1218:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{{Seljak}}},
1219:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{58}},
1220:   \bibinfo{pages}{023003} (\bibinfo{year}{1998}),
1221:   \eprint{arXiv:astro-ph/9803150}.
1222: 
1223: \bibitem[{\citenamefont{{Bowden} et~al.}(2004)\citenamefont{{Bowden}, {Taylor},
1224:   {Ganga}, {Ade}, {Bock}, {Cahill}, {Carlstrom}, {Church}, {Gear}, {Hinderks}
1225:   et~al.}}]{Bowetal04}
1226: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Bowden}}},
1227:   \bibinfo{author}{\bibfnamefont{A.~N.} \bibnamefont{{Taylor}}},
1228:   \bibinfo{author}{\bibfnamefont{K.~M.} \bibnamefont{{Ganga}}},
1229:   \bibinfo{author}{\bibfnamefont{P.~A.~R.} \bibnamefont{{Ade}}},
1230:   \bibinfo{author}{\bibfnamefont{J.~J.} \bibnamefont{{Bock}}},
1231:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Cahill}}},
1232:   \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{{Carlstrom}}},
1233:   \bibinfo{author}{\bibfnamefont{S.~E.} \bibnamefont{{Church}}},
1234:   \bibinfo{author}{\bibfnamefont{W.~K.} \bibnamefont{{Gear}}},
1235:   \bibinfo{author}{\bibfnamefont{J.~R.} \bibnamefont{{Hinderks}}},
1236:   \bibnamefont{et~al.}, \bibinfo{journal}{\mnras}
1237:   \textbf{\bibinfo{volume}{349}}, \bibinfo{pages}{321} (\bibinfo{year}{2004}),
1238:   \eprint{arXiv:astro-ph/0309610}.
1239: 
1240: \bibitem[{\citenamefont{{Verde} et~al.}(2006)\citenamefont{{Verde}, {Peiris},
1241:   and {Jimenez}}}]{VerPeiJim06}
1242: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Verde}}},
1243:   \bibinfo{author}{\bibfnamefont{H.~V.} \bibnamefont{{Peiris}}},
1244:   \bibnamefont{and}
1245:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Jimenez}}},
1246:   \bibinfo{journal}{Journal of Cosmology and Astro-Particle Physics}
1247:   \textbf{\bibinfo{volume}{1}}, \bibinfo{pages}{19} (\bibinfo{year}{2006}),
1248:   \eprint{arXiv:astro-ph/0506036}.
1249: 
1250: \bibitem[{\citenamefont{{Tucci} et~al.}(2005)\citenamefont{{Tucci},
1251:   {Mart{\'{\i}}nez-Gonz{\'a}lez}, {Vielva}, and {Delabrouille}}}]{Tucetal05}
1252: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Tucci}}},
1253:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Mart{\'{\i}}nez-Gonz{\'a}le%
1254: z}}}, \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{{Vielva}}},
1255:   \bibnamefont{and}
1256:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Delabrouille}}},
1257:   \bibinfo{journal}{\mnras} \textbf{\bibinfo{volume}{360}},
1258:   \bibinfo{pages}{935} (\bibinfo{year}{2005}), \eprint{arXiv:astro-ph/0411567}.
1259: 
1260: \bibitem[{\citenamefont{{Amblard} et~al.}(2007)\citenamefont{{Amblard},
1261:   {Cooray}, and {Kaplinghat}}}]{AmbCooKap07}
1262: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Amblard}}},
1263:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Cooray}}}, \bibnamefont{and}
1264:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kaplinghat}}},
1265:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{75}},
1266:   \bibinfo{pages}{083508} (\bibinfo{year}{2007}),
1267:   \eprint{arXiv:astro-ph/0610829}.
1268: 
1269: \bibitem[{\citenamefont{{Hu} and {Okamoto}}(2002)}]{HuOka01}
1270: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}} \bibnamefont{and}
1271:   \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{{Okamoto}}},
1272:   \bibinfo{journal}{\apj} \textbf{\bibinfo{volume}{574}}, \bibinfo{pages}{566}
1273:   (\bibinfo{year}{2002}), \eprint{arXiv:astro-ph/0111606}.
1274: 
1275: \bibitem[{\citenamefont{Knox and Song}(2002)}]{KnoSon02}
1276: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Knox}} \bibnamefont{and}
1277:   \bibinfo{author}{\bibfnamefont{Y.-S.} \bibnamefont{Song}},
1278:   \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{89}},
1279:   \bibinfo{pages}{011303} (\bibinfo{year}{2002}),
1280:   \eprint{arXiv:astro-ph/0202286}.
1281: 
1282: \bibitem[{\citenamefont{{Kesden} et~al.}(2002)\citenamefont{{Kesden}, {Cooray},
1283:   and {Kamionkowski}}}]{KesCooKam02}
1284: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kesden}}},
1285:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Cooray}}}, \bibnamefont{and}
1286:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kamionkowski}}},
1287:   \bibinfo{journal}{\prl} \textbf{\bibinfo{volume}{89}}, \bibinfo{pages}{1304}
1288:   (\bibinfo{year}{2002}).
1289: 
1290: \bibitem[{\citenamefont{Seljak and Hirata}(2004)}]{SelHir03}
1291: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Seljak}} \bibnamefont{and}
1292:   \bibinfo{author}{\bibfnamefont{C.~M.} \bibnamefont{Hirata}},
1293:   \bibinfo{journal}{Phys. Rev.} \textbf{\bibinfo{volume}{D69}},
1294:   \bibinfo{pages}{043005} (\bibinfo{year}{2004}),
1295:   \eprint{arXiv:astro-ph/0310163}.
1296: 
1297: \bibitem[{\citenamefont{Marian and Bernstein}(2007)}]{Marian:2007sr}
1298: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Marian}} \bibnamefont{and}
1299:   \bibinfo{author}{\bibfnamefont{G.~M.} \bibnamefont{Bernstein}}
1300:   (\bibinfo{year}{2007}), \eprint{arXiv:0710.2538}.
1301: 
1302: \bibitem[{\citenamefont{{Kaplinghat} et~al.}(2003)\citenamefont{{Kaplinghat},
1303:   {Chu}, {Haiman}, {Holder}, {Knox}, and {Skordis}}}]{Kapetal03}
1304: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kaplinghat}}},
1305:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Chu}}},
1306:   \bibinfo{author}{\bibfnamefont{Z.}~\bibnamefont{{Haiman}}},
1307:   \bibinfo{author}{\bibfnamefont{G.~P.} \bibnamefont{{Holder}}},
1308:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Knox}}}, \bibnamefont{and}
1309:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{{Skordis}}},
1310:   \bibinfo{journal}{\apj} \textbf{\bibinfo{volume}{583}}, \bibinfo{pages}{24}
1311:   (\bibinfo{year}{2003}), \eprint{arXiv:astro-ph/0207591}.
1312: 
1313: \bibitem[{\citenamefont{{Colombo} et~al.}(2005)\citenamefont{{Colombo},
1314:   {Bernardi}, {Casarini}, {Mainini}, {Bonometto}, {Carretti}, and
1315:   {Fabbri}}}]{Coletal05}
1316: \bibinfo{author}{\bibfnamefont{L.~P.~L.} \bibnamefont{{Colombo}}},
1317:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Bernardi}}},
1318:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Casarini}}},
1319:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Mainini}}},
1320:   \bibinfo{author}{\bibfnamefont{S.~A.} \bibnamefont{{Bonometto}}},
1321:   \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{{Carretti}}},
1322:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Fabbri}}},
1323:   \bibinfo{journal}{\aap} \textbf{\bibinfo{volume}{435}}, \bibinfo{pages}{413}
1324:   (\bibinfo{year}{2005}), \eprint{arXiv:astro-ph/0408022}.
1325: 
1326: \bibitem[{\citenamefont{{Hu} and {White}}(1997)}]{HuWhi97c}
1327: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}} \bibnamefont{and}
1328:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{White}}},
1329:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{56}}, \bibinfo{pages}{596}
1330:   (\bibinfo{year}{1997}), \eprint{arXiv:astro-ph/9702170}.
1331: 
1332: \bibitem[{\citenamefont{Christensen et~al.}(2001)\citenamefont{Christensen,
1333:   Meyer, Knox, and Luey}}]{Chretal01}
1334: \bibinfo{author}{\bibfnamefont{N.}~\bibnamefont{Christensen}},
1335:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Meyer}},
1336:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Knox}}, \bibnamefont{and}
1337:   \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Luey}},
1338:   \bibinfo{journal}{Class. Quant. Grav.} \textbf{\bibinfo{volume}{18}},
1339:   \bibinfo{pages}{2677} (\bibinfo{year}{2001}),
1340:   \eprint{arXiv:astro-ph/0103134}.
1341: 
1342: \bibitem[{\citenamefont{Kosowsky et~al.}(2002)\citenamefont{Kosowsky,
1343:   Milosavljevic, and Jimenez}}]{KosMilJim02}
1344: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Kosowsky}},
1345:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Milosavljevic}},
1346:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Jimenez}},
1347:   \bibinfo{journal}{Phys. Rev.} \textbf{\bibinfo{volume}{D66}},
1348:   \bibinfo{pages}{063007} (\bibinfo{year}{2002}),
1349:   \eprint{arXiv:astro-ph/0206014}.
1350: 
1351: \bibitem[{\citenamefont{Dunkley et~al.}(2005)\citenamefont{Dunkley, Bucher,
1352:   Ferreira, Moodley, and Skordis}}]{Dunetal04}
1353: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Dunkley}},
1354:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Bucher}},
1355:   \bibinfo{author}{\bibfnamefont{P.~G.} \bibnamefont{Ferreira}},
1356:   \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Moodley}}, \bibnamefont{and}
1357:   \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Skordis}},
1358:   \bibinfo{journal}{Mon. Not. Roy. Astron. Soc.}
1359:   \textbf{\bibinfo{volume}{356}}, \bibinfo{pages}{925} (\bibinfo{year}{2005}),
1360:   \eprint{arXiv:astro-ph/0405462}.
1361: 
1362: \bibitem[{\citenamefont{{Lewis} and {Bridle}}(2002)}]{LewBri02}
1363: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Lewis}}} \bibnamefont{and}
1364:   \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{{Bridle}}},
1365:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{66}},
1366:   \bibinfo{pages}{103511} (\bibinfo{year}{2002}),
1367:   \eprint{arXiv:astro-ph/0205436}.
1368: 
1369: \bibitem[{\citenamefont{{Lewis} et~al.}(2000)\citenamefont{{Lewis},
1370:   {Challinor}, and {Lasenby}}}]{Lewetal00}
1371: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Lewis}}},
1372:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Challinor}}},
1373:   \bibnamefont{and}
1374:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Lasenby}}},
1375:   \bibinfo{journal}{\apj} \textbf{\bibinfo{volume}{538}}, \bibinfo{pages}{473}
1376:   (\bibinfo{year}{2000}), \eprint{arXiv:astro-ph/9911177}.
1377: 
1378: \bibitem[{\citenamefont{{Gelman} and {Rubin}}(1992)}]{GelRub92}
1379: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Gelman}}} \bibnamefont{and}
1380:   \bibinfo{author}{\bibfnamefont{D.~B.} \bibnamefont{{Rubin}}},
1381:   \bibinfo{journal}{Statistical Science} \textbf{\bibinfo{volume}{7}},
1382:   \bibinfo{pages}{457} (\bibinfo{year}{1992}), ISSN \bibinfo{issn}{0883-4237}.
1383: 
1384: \bibitem[{\citenamefont{{Brooks} and {Gelman}}(1998)}]{BroGel98}
1385: \bibinfo{author}{\bibfnamefont{S.~P.} \bibnamefont{{Brooks}}} \bibnamefont{and}
1386:   \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Gelman}}},
1387:   \bibinfo{journal}{Journal of Computational and Graphical Statistics}
1388:   \textbf{\bibinfo{volume}{7}}, \bibinfo{pages}{434} (\bibinfo{year}{1998}),
1389:   ISSN \bibinfo{issn}{1061-8600}.
1390: 
1391: \bibitem[{\citenamefont{Raftery and Lewis}(1992)}]{RafLew92}
1392: \bibinfo{author}{\bibfnamefont{A.~E.} \bibnamefont{Raftery}} \bibnamefont{and}
1393:   \bibinfo{author}{\bibfnamefont{S.~M.} \bibnamefont{Lewis}}, in
1394:   \emph{\bibinfo{booktitle}{Bayesian Statistics}}, edited by
1395:   \bibinfo{editor}{\bibfnamefont{J.~M.} \bibnamefont{Bernado}}
1396:   (\bibinfo{publisher}{OUP}, \bibinfo{year}{1992}), p. \bibinfo{pages}{765}.
1397: 
1398: \bibitem[{\citenamefont{{Albrecht} et~al.}(2006)\citenamefont{{Albrecht},
1399:   {Bernstein}, {Cahn}, {Freedman}, {Hewitt}, {Hu}, {Huth}, {Kamionkowski},
1400:   {Kolb}, {Knox} et~al.}}]{Albetal06}
1401: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Albrecht}}},
1402:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Bernstein}}},
1403:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Cahn}}},
1404:   \bibinfo{author}{\bibfnamefont{W.~L.} \bibnamefont{{Freedman}}},
1405:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Hewitt}}},
1406:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}},
1407:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Huth}}},
1408:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kamionkowski}}},
1409:   \bibinfo{author}{\bibfnamefont{E.~W.} \bibnamefont{{Kolb}}},
1410:   \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{{Knox}}},
1411:   \bibnamefont{et~al.} (\bibinfo{year}{2006}), \eprint{arXiv:astro-ph/0609591}.
1412: 
1413: \bibitem[{\citenamefont{{Smith}
1414:   et~al.}(2006{\natexlab{a}})\citenamefont{{Smith}, {Hu}, and
1415:   {Kaplinghat}}}]{SmiHuKap06}
1416: \bibinfo{author}{\bibfnamefont{K.~M.} \bibnamefont{{Smith}}},
1417:   \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{{Hu}}}, \bibnamefont{and}
1418:   \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{{Kaplinghat}}},
1419:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{74}},
1420:   \bibinfo{pages}{123002} (\bibinfo{year}{2006}{\natexlab{a}}),
1421:   \eprint{arXiv:astro-ph/0607315}.
1422: 
1423: \bibitem[{\citenamefont{{Lewis} et~al.}(2006)\citenamefont{{Lewis}, {Weller},
1424:   and {Battye}}}]{LewWelBat06}
1425: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Lewis}}},
1426:   \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{{Weller}}}, \bibnamefont{and}
1427:   \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{{Battye}}},
1428:   \bibinfo{journal}{\mnras} \textbf{\bibinfo{volume}{373}},
1429:   \bibinfo{pages}{561} (\bibinfo{year}{2006}), \eprint{arXiv:astro-ph/0606552}.
1430: 
1431: \bibitem[{\citenamefont{{Chongchitnan} and {Efstathiou}}(2006)}]{ChoEfs06}
1432: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{{Chongchitnan}}}
1433:   \bibnamefont{and}
1434:   \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{{Efstathiou}}},
1435:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{73}},
1436:   \bibinfo{pages}{083511} (\bibinfo{year}{2006}),
1437:   \eprint{arXiv:astro-ph/0602594}.
1438: 
1439: \bibitem[{\citenamefont{{Smith}
1440:   et~al.}(2006{\natexlab{b}})\citenamefont{{Smith}, {Peiris}, and
1441:   {Cooray}}}]{SmiPeiCoo06}
1442: \bibinfo{author}{\bibfnamefont{T.~L.} \bibnamefont{{Smith}}},
1443:   \bibinfo{author}{\bibfnamefont{H.~V.} \bibnamefont{{Peiris}}},
1444:   \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{{Cooray}}},
1445:   \bibinfo{journal}{\prd} \textbf{\bibinfo{volume}{73}},
1446:   \bibinfo{pages}{123503} (\bibinfo{year}{2006}{\natexlab{b}}),
1447:   \eprint{arXiv:astro-ph/0602137}.
1448: 
1449: \end{thebibliography}
1450: 
1451: \end{document}
1452: