1: \documentclass[floats,twocolumn,prl,aps,superscriptaddress,unsortedaddress]{revtex4}

2: \usepackage{amssymb,amsmath,epsfig}

3: 

4: 

5: %\def\msun{{\it h^{-1}} M_{\odot}}

6: \def\msun{M_{\odot}}

7: \def\lya{Ly$\alpha~$}

8: \def\lyb{Ly$\beta$}

9: 

10: 

11: \begin{document}

12: 

13: \title{Baryon Acoustic Oscillation Intensity Mapping as a Test of  Dark Energy}

14: 

15: % or \title{Baryon Acoustic Oscillation Intensity Mapping}

16: 

17: 

18: \author{Tzu-Ching Chang}\email{tchang@cita.utoronto.ca}

19: \affiliation{Inst. for

20:   Astronomy and Astrophysics, Academia Sinica, PO Box 23-141, Taipei

21:   10617, Taiwan}\affiliation{CITA, University of Toronto, 60

22:   St.George St., Toronto, ON, M5S 3H8, Canada }

23: 

24: \author{Ue-Li Pen}\email{pen@cita.utoronto.ca}

25: \affiliation{CITA, University of Toronto,

26:   60 St.George St., Toronto, ON, M5S 3H8, Canada

27: } 

28: \author{Jeffrey B. Peterson}\email{jbp@cmu.edu}

29: \affiliation{Department of Physics, Carnegie Mellon University,

30:   500 Forbes Ave, Pittsburgh, PA 15213, USA

31: }

32: \author{Patrick McDonald}\email{pmcdonal@cita.utoronto.ca}

33: \affiliation{CITA, University of Toronto,

34:   60 St.George St., Toronto, ON, M5S 3H8, Canada

35: } 

36: 

37: 

38: \begin{abstract}

39: The expansion of the universe appears to be accelerating, and the

40: mysterious anti-gravity agent of this acceleration has been called

41: ``dark energy''.  To measure the dynamics of dark energy, Baryon

42: Acoustic Oscillations (BAO) can be used. Previous discussions of the

43: BAO dark energy test have focused on direct measurements of redshifts

44: of as many as $10^9$ individual galaxies, by observing the 21cm line

45: or by detecting optical emission.  Here we show how the study of

46: acoustic oscillation in the 21 cm brightness can be accomplished by

47: economical three dimensional intensity mapping.  If our estimates

48: gain acceptance they may be the starting point for a new class of dark

49: energy experiments dedicated to large angular scale mapping of the

50: radio sky, shedding light on dark energy.

51: \end{abstract}

52: 

53: \pacs{}

54: \keywords{}

55: \maketitle

56: 

57: \def\citep#1{\cite{#1}}

58: 

59: {\it Introduction.---} To understand dark energy and sharply test

60: theories of its character, it is necessary to precisely measure the

61: last half of the expansion history. This period corresponds to the

62: redshift range $0 < z \lesssim 2$. Many techniques have been proposed

63: to study the late-stage expansion history, and some of the most

64: promising make use of Baryon Acoustic Oscillations

65: (BAO)\cite{Blake:2003}. In addition to producing CMB structure, the

66: acoustic oscillations also produced density structure in the atomic

67: gas and dark matter, which is still detectable today.  Many groups

68: have reported 2 to 3 $\sigma$ detections, in the low redshift

69: universe, of periodic structures in the density of galaxies at the

70: predicted wavelengths across the sky\cite{Percival:2007}.  Because the

71: acoustic waves are frozen in after recombination, the BAO peak

72: wavelengths can be used as a cosmological standard ruler: observation

73: of the angular size of the peak wavelengths across a range of

74: redshifts allows accurate measurement of the expansion history.

75: 

76: In this letter we present calculations of structure in the three

77: dimensional brightness due to the hyperfine transition of neutral

78: hydrogen (HI) at 21 cm wavelength.  We show that, via 21 cm emission,

79: baryon oscillations could be precisely measured, using a telescope

80: just 200 wavelengths in diameter, since each cosmic cell of the

81: appropriate scale contains more than $10^{12}\msun$ of emitting

82: neutral hydrogen gas.  We present forecasts on the dark energy

83: constraints from this new type of observation, which are competitive

84: with the best proposed dark energy experiments\cite{detf:2006}.  Where

85: needed in the analysis we adopt {\it WMAP3} values for the

86: cosmological parameters\citep{Spergel:2007}, $\Omega_m=0.24,

87: \Omega_b=0.04$, $\Omega_\Lambda=0.76$, and $h=0.73$, where $\Omega_m$,

88: $\Omega_b$, and $\Omega_\Lambda$ are the matter, baryon, and dark

89: energy fractional density, respectively, and $h$ is the dimensionless

90: Hubble parameter. We also use the {\it WMAP3} error limits for these

91: values.

92: 

93: A recent paper proposed a test of dark energy models by measurement of

94: baryon oscillations in the 21 cm brightness field at redshifts $z >3$

95: \cite{Wyithe:2007}.  Using only such high redshift data two

96: cosmological parameters---departures of spatial flatness $\Omega - 1$

97: and a slow change of dark energy equation of state---are nearly

98: indistinguishable. It is only when precise flatness $\Omega =1$ is

99: assumed that such high redshift data can be used to constrain dark

100: energy models. In this letter we use the more standard approach and

101: assume that $\Omega$ is not perfectly measured. Then data at lower

102: redshift become essential if dark energy models are to be tightly

103: constrained.

104: 

105: 

106: {\it Detectability.---} To measure the oscillations it is not

107: necessary to detect individual galaxies, but only to measure the

108: variation in HI mass on large scales.  A familiar analogy is the study

109: of galaxies via images of their surface brightness.  Except for the

110: nearest of galaxies, we do not detect the individual stars that

111: produce the optical emission, indeed the pixels of most galaxy images

112: represent the emission of millions of stars.  Never-the-less such

113: images allow study of galactic structure.

114: 

115: Above the third BAO peak nonlinear evolution attenuates BAO structure,

116: so the third peak is at the smallest spatial scale we need

117: consider. This peak has a wavelength 35$h^{-1}$ Mpc. A Nyquist sampled

118: map therefore needs pixels of size 18$h^{-1}$ Mpc. At $z=1.5$ the

119: corresponding angular wavelength is 20 arc minutes, which requires a

120: telescope of approximately 200 wavelength, or 100 meters, to resolve.

121: 

122: We base our estimate of neutral hydrogen mass contained in a

123: $18h^{-1}$ Mpc volume ( $M_{\rm HI} \sim 2 \times 10^{12} \msun$) on \lya

124: absorption studies of the cosmic density of neutral gas

125: \cite{Wolfe:2005}, which find $\Omega_{\rm HI} \sim 1 \times 10^{-3}$

126: at $z\sim 1$, with a weak dependence on redshift.

127: 

128: The average sky brightness temperature due to the 21cm line is about

129: $300 \mu$K.  The mean brightness temperature $T_b$ of the 21cm line

130: can be estimated using\cite{Barkana:2007}

131: %\begin{widetext}

132: \begin{equation}

133: T_b = 0.3\left({\Omega_{\rm HI} \over 10^{-3}}\right)%\left({h \over 0.73}\right) 

134: \left({\Omega_m + a^{3} \Omega_\Lambda\over 0.29}\right)^{-1/2}  

135: \left({1+z \over 2.5}\right)^{1/2} {\rm mK}.

136: \label{eq:21cm}

137: \end{equation}

138: %\end{widetext}

139: Most observation of the sky are measurements of variations in temperature $\Delta T_b \sim T_b \delta$,   

140: where $1+\delta = \rho_g / {\bar \rho_g}$ is the normalized neutral

141: gas density, and $a=(1+z)^{-1}$ is the scale factor.

142: 

143: On 18$h^{-1}$ Mpc scales spatial variations of the sky brightness due

144: to 21 cm emission in the cosmic web structure presents a sky noise at

145: the 150 $\mu$K level.  Figure~\ref{fig:hshs} shows the amplitude of

146: this cosmological large scale structure, while Figure~\ref{fig:pk_bao}

147: plots the small oscillations imprinted by BAO.  The variation of gas

148: density due to baryon oscillations is small compared to the

149: fluctuation of the large-scale structure.  To detect the BAO signals

150: beneath this noise Fourier analysis of a large regions of sky is

151: required.  The observations should cover as much sky as possible.

152: 

153: \begin{figure}[t]

154: %\psfig{file=pk_bao.eps,width=80mm}

155: %\plotone{hshs.ps}

156: \centerline{\epsfig{file=hshs.ps, width=\columnwidth}}

157: \caption{The power spectrum of neutral hydrogen hyperfine emission at

158:   $z=1.5$ is shown by the solid line.  The dashed line indicates the

159:   map noise from a 200-by-200-meter radio cylinder

160:   telescope operated in transit mode for 100 days, assuming a

161:   telescope system temperature of 50 K.  The horizontal dotted line

162:   indicates the non-linear scale at $z=1.5$. The vertical lines

163:   indicate the location of peaks in Figure~\ref{fig:pk_bao}

164: \label{fig:hshs}}

165: \end{figure}

166: 

167: 

168: \begin{figure}[t]

169: \centerline{\epsfig{file=wiggle.eps, width=\columnwidth}}

170: \caption{Baryon Acoustic Oscillations. To show

171:   BAO  we plot the ratio of matter power spectrum that

172:   contains contributions from baryons, calculated using a fitting

173:   formula\cite{Eisenstein:1998}, over a power spectrum with no

174:   baryons. To single out the oscillation, we further divide the

175:   ratio by a smooth fitting curve to the overall spectrum.  The error

176:   bars represent projections of the sensitivity possible with a

177:   200-by-200-meter cylinder telescope.}

178: \label{fig:pk_bao}

179: \end{figure}

180: 

181: For a filled aperture telescope the RMS variation in measured sky

182: temperature $\Delta T =T_{\rm sys} / \sqrt{\Delta f ~t_{\rm int}}$,

183: where $\Delta T$ is the temperature in Kelvin, $T_{\rm sys}$ is the

184: system temperature of the radio telescope in Kelvin, $\Delta f$ is the

185: observing bandwidth in Hertz and $t_{int}$ the integration time in

186: seconds.  Assuming current technology, which allows $T_{\rm sys} =50

187: $K, one needs an observing bandwidth-time product of $2 \times

188: 10^{11}$ to bring the thermal map noise down to $\sim 100 \mu$K, below

189: the cosmic structure noise.  In frequency space 18 $h^{-1}$ Mpc spans

190: roughly 3 MHz at $z=1.5$, so the required observation time is $\sim

191: 6\times 10^4$ seconds, about 18 hours per pixel.  Small regions of sky

192: might be mapped at this noise level using existing telescopes such as

193: the Green Bank Telescope (GBT), the Effelsberg 100-meter Radio

194: Telescope or the Parkes Telescope.

195: 

196: Brightness sensitivity is not improved by increase of telescope

197: aperture, instead multiple receivers, either on a single reflector or

198: on individual reflectors, can be used to increase the accumulated

199: integration time.

200: 

201: For sensitivity estimates, we consider an intensity mapping

202: telescope, dubbed IM, covering a square

203: aperture of size 200m $\times$ 200m.  We subdivide this area into 16

204: cylindrical sections, each 12.5m wide and 200m long, analogous to the

205: Hubble Sphere Hydrogen Survey (HSHS)\cite{Peterson:2006}.  The

206: cylinders are aligned North-South allowing the telescope to

207: instantaneously sample almost the entire meridian.  This geometry allows

208: all-sky coverage each 24 hour period in a telescope with no moving

209: parts. %It also allows simultaneous use of thousands of receivers, and

210: %the 200 m size allows a strongly tapered illumination pattern with a width

211: %scaled in proportion to wavelength.  This produces a nearly frequency

212: %independent set of beam patterns on the sky. This is useful for the

213: %foreground subtraction technique discussed below.  This layout allows

214: %accumulation of 10 minutes of integration time per cylinder per

215: % beam for each day for each pixel on the sky.  

216: The estimated sensitivity of the survey is shown in Figure~\ref{fig:hshs}.

217: 

218: {\it Foregrounds.---} The desired BAO signal is smaller than other

219: sources of sky brightness structure.  The largest interference is due

220: to synchrotron emission from the Galaxy. The spatial structure of this

221: emission exceeds the BAO signal by $10^3$.  Similar emission by

222: extragalactic radio sources distributed across the sky will contribute

223: sky brightness fluctuations on smaller scales.  However all these

224: foregrounds are spectrally smooth, with brightness temperature $T

225: \propto f^\alpha$, so one expects to be able to subtract them

226: spectrally. The image from each 3 MHz slice of the frequency spectrum

227: contains a unique BAO pattern, largely uncorrelated with the

228: neighboring slices.  In contrast, the foregrounds and backgrounds are

229: highly correlated. By subtracting the average of both neighboring

230: slices the BAO signal can be retained while the other emission is

231: removed.  Of course this subtractions will be imperfect, because of

232: imperfections in the telescope and spatial variations in the

233: astrophysical source spectrum. A similar challenge is faced by the

234: higher-redshift 21cm experiments probing the Cosmic Reionization era

235: and many foreground subtraction methods have been proposed

236: (\cite{Furlanetto:2006} and references therein).

237: 

238: The spectral index $\alpha$ for Galactic emission varies across the

239: sky.  The frequency-double-difference technique described above

240: suppresses the Galactic sky brightness but leaves a residual of order

241: $\Delta T = \Delta \alpha (\Delta f / f)^{2} T$ to first power in

242: $\Delta \alpha$, where $\Delta \alpha$ is the uncertainty in spectral

243: index and $\Delta f$ the frequency difference.  Using the {\it WMAP3}

244: measurement we estimate

245: 

246: 

247: \begin{equation} 

248: \Delta T \sim 1.3 ~\Delta \alpha \left({l \over

249:   200}\right)^{-0.5} \left({1+z \over 2.5}\right)^{3.2} \left({\Delta

250:   f \over f} \right)^2  ~~{\rm K},

251: \label{eq:fg}

252: \end{equation}

253: where $l$ is the spherical harmonic number.   The normalization factor

254: of 1.3 Kelvin temperature fluctuation is derived by extrapolating from

255: the {\it WMAP3} measurement of Galactic foreground variation at 20 GHz

256: on 3.7$^{\circ}$ degree scale, along with the spectral index of

257: synchrotron emission at mid-Galactic latitude\cite{Kogut:2007}.

258: Adopting $\Delta\alpha=1$, a conservative upper limit, we estimate a

259: residual Galactic interference signal which is plotted in Figure~\ref{fig:fg}.

260: 

261: \begin{table*}

262: \begin{ruledtabular}

263: \begin{tabular}{lcccccccccccccc}

264: $ $ & $\Omega_m h^2$ & $\Omega_b h^2$ & $\Omega_{\rm DE}$ & $a_p$ & $w_p$ &

265:  $w'$ & $\Omega_k$ & $\log_{10}(A)$ & $n_s$ & $h$ & $\Omega_m$ &

266: ${\rm FoM}$ \\

267: ${\rm reference\  value}$ & $0.128$ & $0.0223$ & $0.760$ & $$ & $-1.00$ & 

268: $0.00$ & $0.00$ & $-8.64$ & $0.950$ & $0.730$ & $0.240$ & $$ \\

269: P+IM & $0.0012$ & $0.00017$ & $0.0058$ & $0.65$ & $0.015$ & $0.16$ & $0.00072$ & $0.0077$ & $0.0062$ & $0.0075$ & $0.0054$ & $421.9$ \\ 

270: P+IMW & $0.0012$ & $0.00017$ & $0.010$ & $0.63$ & $0.023$ & $0.29$ & $0.00084$ & $0.0077$ & $0.0062$ & $0.015$ & $0.010$ & $152.3$ \\ 

271: P+II & $0.0012$ & $0.00017$ & $0.012$ & $0.79$ & $0.036$ & 

272: $0.52$ & $0.0031$ & $0.0076$ & $0.0061$ & $0.019$ & $0.012$ &$53.39$

273: \\

274: P+II+III & $0.0011$ & $0.00015$ & $0.0045$ & $0.76$ & $0.020$ 

275: & $0.18$ & $0.0021$ & $0.0063$ & $0.0046$ & $0.0070$ & $0.0043$ &

276: $279.5$ \\

277: P+II+III+IV & $0.00091$ & $0.00013$ & $0.0028$ & $0.75$ & $0.011$ &

278: $0.090$ & $0.0013$ & $0.0050$ & $0.0028$ & $0.0045$ &

279: $0.0026$ & $975.2$ \\

280: P+II+III+IV+IM & $0.00068$ & $0.00013$ & $0.0025$ & $0.69$ & $0.0082$ & $0.068$ & $0.00047$ & $0.0049$ & $0.0028$ & $0.0032$ & $0.0023$ & $1793.$ \\ 

281: P+II+III+IV+IMW & $0.00071$ & $0.00013$ & $0.0027$ & $0.71$ & $0.0095$ & $0.073$ & $0.00063$ & $0.0050$ & $0.0028$ & $0.0036$ & $0.0025$ & $1440.$ \\

282: \end{tabular}

283: \end{ruledtabular}

284: \caption{

285: Fiducial cosmological parameters and error bars for combining the Planck mission with the fiducial BAO

286: intensity mapping experiment such as IM (P+IM), with the Dark Energy Task Force\cite{detf:2006} (DETF) Stage II projects (P+II), plus the Stage III projects (P+II+III), plus the Stage IV projects (P+II+III+IV), and all above combined (P+II+III+IV+IM).  Here IM is the conservative best guess for intensity mapping experiment parameters, while IMW is the worse case scenario (see text).  DETF Stage II is well-defined which includes weak lensing (WL), cluster evolution (CL), and supernova (SN) experiments; for Stage III we use the optimistic photometric versions of WL, CL, and SN; for Stage IV we use the optimistic space-based WL, CL, and SN experiments.  The table

287: entries are shown using the units of the DETF Report: IM compares favorably

288: with other proposed Stage III projects.  The dark energy equation of

289: state is parametrized as $w=w_p+(a_p-a)w'$.  $\Omega_{\rm DE}$ and $\Omega_k$ are the dark energy and curvature density, respectively.  

290: $n_s$ is the spectral

291: index of the primordial power spectrum, and $A$ its normalization.

292: The Figure of Merit (FoM) refers to the inverse volume in the 

293: 95\% confidence ellipse.

294: \label{tab:par}}

295: \end{table*}

296: 

297: \begin{figure}%[h]

298: \centerline{\epsfig{file=foreground.eps, width=\columnwidth}}

299: \caption{The observable parameter space in redshift and in scale (k)

300:   for BAO. The shaded regions are observationally inaccessible (see

301:   text). The horizontal lines indicate the scale of the first three

302:   BAO wiggles, and the dashed lines show contours of constant spherical harmonic order $\ell$.} 

303: \label{fig:fg}

304: \end{figure}

305: 

306: Figure~\ref{fig:fg} 

307: shows various potential limitations to BAO

308: measurements using 21 cm sky brightness mapping.  The top exclusion 

309: comes from

310: the non-linear scale, which erases the baryon wiggles.  The left

311: exclusion comes from the finite volume, which sets a lower bound on

312: the lowest wavenumber that fits within a given volume.  The lower

313: right is a rough estimate of the impact of foregrounds using the

314: double frequency difference described above. The shaded region

315: reflects the region where the estimated residual foreground from

316: Eqn. (\ref{eq:fg}) exceeds the expected signal in

317: Eqn. (\ref{eq:21cm}).

318: 

319: %\begin{table}

320: %\begin{tabular}{cccc}

321: %z & radial error & transverse error & correlation \\

322: %0.25&0.00952&0.00595&-0.375\\

323: %0.75&0.00422&0.00269&-0.389\\

324: %1.25&0.00307&0.00206&-0.388\\

325: %1.75&0.00264&0.00186&-0.375\\

326: %2.25&0.00247&0.00183&-0.361\\

327: %\end{tabular}

328: %\caption{Distance measurement accuracies}

329: %\label{tab:dist}

330: %\end{table}

331: 

332: {\it Parameter Estimation.---} To estimate the sensitivity of this

333: technique for measurement of dark energy parameters, we follow the

334: procedure in Seo and Eisenstein (2007) \cite{Seo:2007} and carry out

335: the standard figure-of-merit calculation used in the DETF

336: report\cite{detf:2006}.  Our calculations include the effects of

337: redshift space distortion and a simplified model of non-linearity

338: using a smoothing window.  This type of analysis, the conversion from

339: survey data to cosmological parameters, is an area of active research,

340: and several effects such as mode-mode coupling are still being

341: investigated.  Our use of shells of 21 cm surface brightness

342: introduces two new types of unknowns into the analysis: the HI

343: density, and the bias of 21 cm sources.  To accommodate these

344: uncertainties we make one sensitivity estimate using conservative

345: values for the unknown quantities (IM) as our baseline model, and one

346: with a worst case assumption that each of these effects works to

347: increase the error (IMW). The IM case estimate assumes that

348: $\Omega_{\rm HI}$ remains at the present day value of 0.0005,

349: independent of redshift, and we adopt a bias $b=1$.  The shot noise

350: from present day HI galaxies \cite{Zwaan:2005} is negligible.

351: Motivated by the reconstruction of BAO degradation due to

352: non-linearity in \cite{Eisenstein:2007} the Seo and Eisenstein damping

353: scale is shortened by a factor of two.  For the worst case (IMW)

354: estimate we let $\Omega_{\rm HI}$ to be 0.0004.  We adopt a bias

355: $b=0.8$, which is the current estimate of the low $z$ value. This

356: increases the noise, and seems securely to be a worst case because

357: galaxies are unlikely to be both antibiased and rare.  We assume that

358: 21 cm sources are rare and massive at high redshift, and we use the

359: Lyman-Break Galaxy luminosity function (with a density of 0.004

360: h$^3$ Mpc$^{-3}$) for the shot noise calculation.  The Seo and Eisenstein

361: damping scale is assumed.

362: 

363: The time variable dark energy density is

364: parametrized as $\rho_{DE} \propto (1+z)^{3(1+w)}$; an ordinary

365: cosmological constant corresponds to $w=-1$, and the principal goal of

366: dark energy experiments is to measure deviations of $w$ from $-1$.  We

367: first determine the Fisher matrix for the tangential and radial

368: distances errors.  From these, we compute the errors on dark energy

369: parameters.  The resulting projected errors to the equation of state

370: are shown in Figure~\ref{fig:planck}.  Table \ref{tab:par} quantifies

371: the errors, which can be compared to other dark energy surveys.

372: 

373: \begin{figure}%[h]

374: \centerline{\epsfig{file=w.eps, width=\columnwidth}}

375: \caption{The 1-$\sigma$ contour for IM combined with Planck (inner thick solid for

376: baseline model, outer thin solid for worst case), the Dark Energy Task Force Stage II projects with Planck (outer dotted), the Stage II and III projects with Planck (intermediate dotted), the Stage II, III and IV projects with Planck (inner dotted), and all above experiments combined (dashed, again thick for baseline, thin for worst case;  the two contours are nearly indistinguishable).

377: } 

378: \label{fig:planck}

379: \end{figure}

380: 

381: The mapping of cosmological large scale structure by three dimensional 21 cm intensity mapping

382: is an approach that can be tested with existing  

383: telescopes today, which appears to offer very competitive constraints on dark energy parameters.

384: This new tool may soon allow astronomers to economically map huge volumes of the

385: universe.  

386: 

387: {\it Acknowledgments.---}

388: We acknowledge financial support by NSERC.

389: 

390: \bibliographystyle{apsrev}

391: \bibliography{bao}

392: \end{document}

393: 

394: 

395: 

396: 

397: 

398: