1: \documentclass[12pt,preprint]{aastex}
2:
3: %\documentclass{article}
4: %\usepackage{emulateapj5,pstricks,apjfonts}
5: %\usepackage{emulateapj5,pstricks}
6:
7: \newcommand{\be}{\begin{equation}}
8: \newcommand{\ee}{\end{equation}}
9: \newcommand{\ba}{\begin{eqnarray}}
10: \newcommand{\ea}{\end{eqnarray}}
11:
12: \shorttitle{The Primordial Power Spectrum from WMAP Data}
13: \shortauthors{Mukherjee \& Wang}
14:
15: %\def\baselinestretch{2}
16: \begin{document}
17:
18: \title{Model-Independent Reconstruction of the Primordial Power Spectrum
19: from WMAP Data}
20: \author{Pia~Mukherjee, Yun~Wang}
21: \affil{Department of Physics \& Astronomy, Univ. of Oklahoma,
22: 440 W Brooks St., Norman, OK 73019;
23: email: pia,wang@nhn.ou.edu}
24:
25: \begin{abstract}
26:
27: Reconstructing the shape of the primordial power spectrum in a model
28: independent way from cosmological data is a useful consistency check
29: on what is usually assumed regarding early universe physics. It is also
30: our primary window to unknown physics during the inflationary era.
31: Using a power-law form for the primordial power spectrum
32: $P_{in}(k)$
33: and constraining the scalar spectral index and its running,
34: \cite{Peiris03} found that the first year WMAP data
35: seem to indicate a preferred scale in $P_{in}(k)$.
36: We use two complementary methods:
37: the wavelet band powers method of \cite{pia1},
38: and the top-hat binning method of \cite{Wang99}
39: to reconstruct $P_{in}(k)$ as a free function
40: from CMB data alone (WMAP, CBI, and ACBAR), or from CMB data
41: together with large scale structure data
42: (2dFGRS and PCSZ). The shape of the reconstructed $P_{in}(k)$ is consistent
43: with scale-invariance, although it allows some indication of a preferred
44: scale at $k \sim 0.01$Mpc$^{-1}$.
45: While consistent with the possible evidence for a running of the
46: scalar spectral index found by the
47: WMAP team, our results
48: highlight the need of more stringent and independent constraints
49: on cosmological
50: parameters (the Hubble constant in particular) in order
51: to more definitively constrain deviations of $P_{in}(k)$ from scale-invariance
52: without making assumptions about the inflationary model.
53:
54:
55: \end{abstract}
56:
57: %\end{document}
58:
59: %\keywords{Cosmology}
60:
61:
62:
63: \section{Introduction}
64:
65: The long anticipated first year WMAP data \citep{Bennett03,Spergel03}
66: have very interesting implications for the state of cosmology today.
67: On one hand, WMAP results on cosmological parameters are consistent
68: with and refine previous constraints from various independent and
69: complementary observations.
70: On the other hand, the data seem to indicate a preferred scale in
71: the primordial scalar power spectrum $P_{in}(k)$ \citep{Peiris03},
72: with or without complementary large scale structure data,
73: though not at high significance. If true,
74: this would contradict the simple assumption of scale-invariance
75: of $P_{in}(k)$ made
76: by most researchers in cosmology and the prediction of the
77: simplest inflationary models, but would be
78: consistent with earlier lower significance findings of
79: \cite{WangMathews02}, \cite{pia1} (hereafter MW03a), and \cite{pia2}.
80:
81:
82: With WMAP, CMB data continues to be fully consistent with inflation
83: \citep{Guth81,KolbTurner90,HuDodelson02,PeeblesRatra03}.
84: WMAP reveals new evidence for inflation from the anti-correlation
85: between CMB temperature and polarization fluctuations near $l$ of 150.
86: Focus is now shifting towards distinguishing between the different
87: inflationary models. The simplest models of inflation predict a
88: power-law primordial matter power spectrum (for example,
89: \cite{Linde83,naturalinf,extendedinf}). Thus some efforts
90: have been focused on constraining slow roll parameters
91: \citep{Liddle92,Leach02,Barger03},
92: evaluated at a certain epoch during inflation, or
93: at the Hubble crossing time of a certain scale usually chosen
94: to be at the center of the scales probed by observations. The
95: primordial power spectrum that results from slow-roll
96: inflation can be computed to high accuracy in terms of these
97: parameters. The WMAP team fits to observables
98: which can be written as derivatives of the slow roll parameters.
99: However, there are also viable models of inflation which
100: predict primordial power spectra which cannot be parametrized
101: by a simple power-law
102: (for example,
103: \cite{Holman91ab,Linde94,Wang94,Randall96,Adams97,Les97}).
104: In such models, features in
105: $P_{in}(k)$ can result
106: from unusual physics during inflation
107: \citep{Chung00,Enqvist00,Lyth02}.
108: The assumption of a power-law $P_{in}(k)$ could then lead to
109: our missing the discovery of the possible features
110: in the primordial matter power spectrum
111: and erroneous estimates of cosmological parameters
112: \citep{Kinney01}.
113:
114: WMAP results underscore the importance of model-independent
115: measurements of the shape of the primordial power spectrum \citep{Wang99}.
116: In this paper we reconstruct $P_{in}(k)$ as a free function using
117: two different methods,
118: the wavelet band powers method of MW03a,
119: and the top-hat binning method of \cite{Wang99}.
120: We briefly describe our methods in Sec.2. Sec.3 contains our results.
121: Sec.4 contains a summary and discussions.
122:
123:
124: \section{Methods}
125:
126: Both the methods that we have used to reconstruct $P_{in}(k)$ are
127: essentially binning methods in which we can write $P_{in}(k)$ as
128: \be
129: P_{in}(k) = \sum_i \alpha_i f_i(k),
130: \label{eq:Pink}
131: \ee
132: where $f_i(k)$ are functions of wavenumber $k$, and $\alpha_i$ are constants.
133: We use Eq.(\ref{eq:Pink}), instead of
134: $P_{in}(k) = A\, k^{n_S-1}$, to parametrize $P_{in}(k)$
135: as an arbitrary function.
136: Our $P_{in}(k)$ parameters are the coefficients $\alpha_i$'s,
137: instead of the normalization parameter $A$,
138: the power-law spectral index, $n_S$, and its running,
139: $\mbox{d} n_S/\mbox{d}\ln k$ \citep{Kosow95}.
140:
141: This allows us to expand the CMB temperature angular
142: power spectrum as follows:
143: \begin{eqnarray}
144: C_l(\{\alpha_i\}, \mbox{\bf s}) &=&(4\pi)^2 \int \frac{dk}{k} P_{in}(k)
145: \left|\Delta_{Tl}(k, \tau=\tau_0)\right|^2
146: \nonumber\\
147: &=& \sum_{i} \alpha_i \int \frac{dk}{k} f_i(k)
148: \left|\Delta_{Tl}(k, \tau=\tau_0)\right|^2\nonumber\\
149: &\equiv & \sum_i \alpha_i \, C_l^i (\mbox{\bf s}),
150: \label{Clwaveletproj}
151: \end{eqnarray}
152: where the cosmological model dependent transfer function
153: $\Delta_{Tl}(k,\tau=\tau_0)$ is an integral over
154: conformal time $\tau$ of the sources which generate CMB
155: temperature fluctuations,
156: $\tau_0$ being the conformal time today,
157: and $\mbox{\bf s}$ represents cosmological parameters other than
158: the $\alpha_i$'s. We use CAMB\footnote{Similar to
159: CMBFAST, CAMB can be used to compute the CMB and
160: matter power spectra from a given set of cosmological parameters and primordial
161: power spectrum. For details, see http://camb.info/.} to compute the
162: CMB angular power spectra,
163: in a form such that for given cosmological
164: parameters other than the $\alpha_i$'s, the $C_l^i(\mbox{\bf s})$ are computed,
165: so that there is no need to call CAMB when we vary only the $\alpha_i$'s.
166:
167: The choice of the basis functions $f_i(k)$ in Eq.(\ref{eq:Pink})
168: differ in the two methods as described below.
169:
170: \subsection{The wavelet band powers method}
171:
172: In this method, we are using wavelets essentially as band pass filters.
173: The wavelet band powers of the primordial power spectrum are given by
174: (MW03a)
175: \begin{equation}
176: P_j = \frac{1}{2^j} \sum_{n=-\infty}^{\infty} \left| \hat{\psi}
177: \left(\frac{n}{2^j}\right) \right|^2 P_{in}(k_n).
178: \label{impeqn1}
179: \end{equation}
180: The wavelet band power window functions in $k$ space,
181: $\left| \hat{\psi}\left(\frac{k}{2^j}\right)\right|^2$,
182: are the modulus squared
183: of the Fourier transforms of the wavelet basis functions of different $j$
184: (dilation index). The translation index has been integrated out as we
185: Fourier transformed the basis functions. The resulting band powers $P_j$'s
186: are thus band averaged Fourier power spectrum.
187:
188: The wavelet band power window functions,
189: $\left| \hat{\psi}\left(\frac{k}{2^j}\right)\right|^2$,
190: are plotted in Fig.1 of MW03a.
191: Fig.2 of MW03a shows the window functions in CMB
192: multipole $l$ space that these functions map on to.
193:
194:
195: If the primordial density field is a Gaussian random field, the
196: $P_j$'s, which represent the variance of wavelet coefficients of scale $j$,
197: are uncorrelated (e.g. \cite{pia00}):
198: \be
199: \frac{ \langle P_j\, P_{j'} \rangle} {P_j \, P_{j'}}
200: = 1.
201: \ee
202:
203: Furthermore, the primordial power spectrum $P_{in}(k)$ can be reconstructed as
204: a smooth function from
205: the wavelet band powers $P_j$'s as follows (\cite{Fang00}, MW03a)
206: \begin{equation}
207: \hat{P}_{in}(k) = \sum_{j} P_j \left| \hat{\psi}\left(\frac{k}{2^j}
208: \right)\right|^2,
209: \label{impeqn2}
210: \end{equation}
211: i.e., $\alpha_i=P_i$, and $f_i(k)=\left| \hat{\psi}\left(\frac{k}{2^i}
212: \right)\right|^2$ in Eq.(1).
213: MW03a has shown that Eq.(\ref{impeqn2}) gives excellent estimates
214: of $P_{in}(k)$ at the centers of the wavelet window functions.
215: Smooth wavelets work best in this method. We have chosen
216: the wavelet Daubachies 20 \citep{Daub92}. Our results are insensitive
217: to the choice of the particular wavelet among smooth wavelets.
218:
219: The wavelet band powers method is an optimal binning method,
220: in which the locations of the bands is not arbitrary.
221: Here the position
222: and momentum spaces are decomposed into elements that satisfy
223: $\Delta x\, \Delta k \sim 1$, with $\Delta x \propto 1/k$,
224: and $\Delta k/k = \log_{10}\,2$. Thus on small length scales
225: (large $k$), $\Delta x$ is small, and on large length scales,
226: $\Delta x$ is large, and since the wavelet bases are complete,
227: one cannot have more independent bands than used here \citep{Fang00}.
228: We choose to estimate 11 $P_j$'s
229: that cover the $k$ range probed by current data. The $P_j$'s outside
230: of this $k$ range are set equal to their adjacent $P_j$'s.
231:
232: Note that although the $P_j$'s are mutually uncorrelated by construction,
233: the $P_j$'s estimated from CMB data (the
234: measured CMB temperature anisotropy angular power spectrum $C_l$ bands)
235: will be somewhat correlated
236: because of the cosmological model dependent nonlinear mapping between the wavenumber $k$ and the
237: CMB multipole number $l$, and due to correlations with
238: cosmological parameters.
239:
240: We have chosen to use the wavelet band powers method in this paper,
241: since the first year WMAP data seem to be relatively well fit by a
242: smooth $P_{in}(k)$ and are not yet sensitive to very sharp features
243: (see Fig. 7 of \cite{Peiris03}). The direct wavelet expansion method
244: of \cite{pia2} is more suited to reconstructing an unknown function with
245: possible sharp features on scales smaller than $log_{10}$$\,2$.
246:
247: \subsection{The top-hat binning method}
248:
249: We also use the top-hat binning method of \cite{Wang99} to
250: parametrize $P_{in}(k)$. We write
251: \be
252: P_{in}(k) = \left\{ \begin{array}{lll}
253: \alpha_1, \hskip 1cm k < k_{0}; \\
254: \alpha_i, \hskip 1cm k_{i-1}< k < k_i; \\
255: \alpha_{n}, \hskip 1cm k > k_{n}.
256: \end{array} \right.
257: \ee
258: The $k_i$'s are chosen to be uniformly spaced in $\log\,k$ as in
259: \cite{Wang99}.
260: In this method, the basis functions $f_i(k)$ in
261: Eq.(\ref{eq:Pink}) are top-hat window functions:
262: $f_i(k) = 1$, for $ k_{i-1} < k < k_i$, and $f_i(k) = 0$ elsewhere.
263: The boundary conditions at the minimum and maximum $k$ values
264: are similar to what we imposed in the wavelet band powers method.
265:
266: We choose the centers of the top-hat window functions to
267: coincide with the central $k$ values of the wavelet band power window
268: functions, so that the results from the two methods can be compared
269: for consistency.
270:
271: The advantage of this binning is its simplicity.
272: The disadvantage is that the reconstructed $P_{in}(k)$
273: is a {\it discontinuous} step function, which might introduce
274: additional degeneracies with cosmological parameters.
275:
276: We have included this binning method primarily to cross-check
277: the results of the wavelet band powers method.
278:
279: \section{Results}
280:
281: We work with CMB temperature anisotropy data from WMAP
282: \citep{Bennett03}, complemented at $l>800$, and upto an $l_{max}$ of 2000,
283: by data from
284: CBI \citep{Pearson02} and
285: ACBAR \citep{Kuo02}, and large scale
286: structure (LSS) power spectrum data from
287: the 2dFGRS \citep{Percival02} and PSCZ \citep{Hamilton02}
288: galaxy redshift surveys.
289: We use the data covariance matrices and window functions provided by
290: the different experimental teams. For CMB data we marginalize analytically
291: over known beanwidth and calibration uncertainties (Bridle et al. 2002).
292: For LSS data we assume that the galaxy power spectrum
293: is a multiple of the underlying matter power spectrum and marginalize
294: analytically over a linear bias \citep{Lewis02}. Both CMB and LSS data
295: depend on $P_{in}(k)$ and on the cosmological parameters.
296: Note however that as pointed out by \cite{Elgaroy02}, it is hard to detect
297: features in the primordial power spectrum at $k<0.03 h$Mpc$^{-1}$
298: using LSS data.
299:
300: We estimate the $P_{in}(k)$ parameters $\alpha_i$'s (see
301: Eq.(\ref{eq:Pink})), together with
302: the Hubble constant $H_0$, baryon density $\Omega_b\,h^2$,
303: cold dark matter density $\Omega_c\,h^2$, and reionization
304: optical depth $\tau_{ri}$.
305: We assume Gaussian adiabatic scalar perturbations in a flat universe
306: with a cosmological constant. We do not use tensor modes in this paper,
307: since current data are not sensitive to tensor contributions.
308: We make use of the WMAP constraint on $\tau_{ri}$, derived
309: for a $\Lambda$CDM cosmology from WMAP's TE polarization data
310: \citep{Kogut03}, by
311: imposing a Gaussian prior on $\tau_{ri}$, $p(\tau_{ri}) \propto
312: \exp\left[ - (\tau_{ri}-0.17)^2/(2 \sigma_{\tau_{ri}}^2)\right]$,
313: with $\sigma_{\tau_{ri}}=0.04$ (this error estimate includes systematic
314: and foreground uncertainties).
315:
316: For a power-law primordial fluctuation spectrum,
317: $P_{\cal R}= 2.95 \times 10^{-9} \, A\,(k/k_0)^{n_S-1}$
318: \citep{Spergel03}. For an arbitrary primordial power spectrum,
319: we define
320: \be
321: P_{in}(k)=\frac{P_{\cal R}}{2.95 \times 10^{-9}}.
322: \ee
323: We use $k_0 = 0.05\,$Mpc$^{-1}$ when quoting
324: parameter constraints for a power law model.
325:
326:
327: We use the Markov Chain Monte Carlo (MCMC) technique to estimate
328: the likelihood functions of the parameters
329: \citep{neil,Knox01,Kosowsky02,Lewis02,Verde03}.
330: The use of MCMC is necessitated by the large number of parameters,
331: and it is free of the interpolation errors expected in the conventional
332: and much slower grid method.
333: At its best, the MCMC method scales approximately linearly with the number of
334: parameters. The method samples from the full posterior distribution of the
335: parameters, and from these samples the marginalized posterior distributions
336: of the parameters can be estimated.
337:
338: Table 1 lists the mean values and marginalized 1 $\sigma$ confidence limits
339: of the cosmological parameters estimated using four different models
340: for $P_{in}(k)$, as well as $\chi^2_{eff}= -2 \ln {\cal L}$ where ${\cal L}$ is
341: the likelihood of each model.
342: Both the wavelet bandpowers model and the top-hat binning model allow
343: $P_{in}(k)$ to be an arbitrary non-negative function.
344: The scale invariant model assumes $P_{in}(k)=A$ ($A$ is a constant),
345: while the powerlaw model assumes $P_{in}(k)$ to be a powerlaw,
346: $P_{in}(k)=A k^{n_S-1}$ ($A$ and $n_S$ are constants).
347: The only priors used
348: are a Gaussian prior on $\tau_{ri}$, $p(\tau_{ri}) \propto
349: \exp\left[ - (\tau_{ri}-0.17)^2/(2 \sigma_{\tau_{ri}}^2)\right]$,
350: with $\sigma_{\tau_{ri}}=0.04$ as discussed earlier, and a weak age
351: prior of age of the universe $t_0 > 10$ Gyrs.
352: Also, we do not use
353: tensor modes
354: in this paper,
355: since current data are not sensitive to tensor contributions.
356:
357:
358: Fig.1 shows the reconstructed $P_{in}(k)$ from the two different
359: methods discussed in Sec.2, using CMB temperature anisotropy data.
360: The dotted line in each panel indicates the scale-invariant model that fits
361: these data\footnote{The uncertainty in $P_{in}(k)$ in the wavelet band
362: powers method is calculated using $\sigma^2_{P_{in}(k)} = \frac{1}{N}
363: \sum_N \left [ \sum_j (P_j-\bar{P}_j) \left| \hat{\psi}
364: \left(\frac{k}{2^j}\right) \right|^2 \right ] ^2$, where the average
365: is over the MCMC samples and $\bar{P}_j$ denotes the mean or expectation
366: value of the wavelet band power.} .
367:
368: From Table 1 we see that the power-law model differs by
369: $\Delta\chi^2_{eff}=8$ from the
370: the wavelet band powers model, and by $\Delta\chi^2_{eff}=10$
371: from the top-hat binning model. Since the difference in the
372: number of degrees of
373: freedom is approximately 9, the power-law model is disfavored
374: at approximately $\sim 0.7\,\sigma$ and $\sim 1\, \sigma$
375: compared to the wavelet band powers model and
376: the top-hat binning model respectively.
377: Also, note that the powerlaw model is favored over the
378: scale-invariant model at less than $1\,\sigma$.
379:
380: Given the estimated wavelet band powers $P_j$'s with their full covariance matrix
381: $C_{P_j}$,
we can treat the $P_j$'s as data
382: and compute $\chi^2 \equiv \Delta^T C_{P_j}^{-1} \Delta$,
383: $\Delta$ being the difference between the estimated $P_j$'s
384: and the $P_j$'s corresponding to the fitted power-law and
385: scale-invariant spectra.
386: The $\chi^2$'s turn out to be 8 and 7
387: for the power-law and scale-invariant parametrization
388: respectively. In the top-hat binning case, we can treat the estimated
389: top-hat bin amplitudes as data
390: and compute a similarly defined $\chi^2$.
391: We find that the corresponding $\chi^2$ is 9 for both power-law and scale-invariant
392: parametrizations.
393: Thus similar significances for
deviation of the reconstructed $P_{in}(k)$
394: from the simpler parametrizations are indicated in this way also.
395: In general, note that the low
396: levels of significance are also due to the large number of degrees of
397: freedom that we are allowing for in the analysis here, and it is clear
398: from the figure which points do not contribute much to the $\chi^2$.
399:
400:
401:
402: %\pspicture(0,0.2)(5.5,12.4)
403: %\rput[tl]{0}(-0.2,12.2){\epsfxsize=8.5cm \epsfclipon
404: %\epsffile{f1.eps}}
405: %\rput[tl]{0}(0,3.3){
406: %\begin{minipage}{8cm}
407: %\small\parindent=3.5mm
408: %{\sc Fig.}~1.---
409:
410: \clearpage
411:
412: \begin{figure}
413: \plotone{f1.eps}
414: \figcaption{
415: the reconstructed $P_{in}(k)$ with 1$\sigma$ error bars from the two different
416: methods discussed in Sec.2, using only CMB data.
417: The dotted line indicates the scale-invariant model that best fits the data.
418: }
419: \end{figure}
420:
421: %\end{minipage}
422: % }
423: % \endpspicture
424: % \vskip -1cm
425:
426: \clearpage
427:
428:
429:
430: Fig.2 shows the reconstructed $P_{in}(k)$ from the two different
431: methods discussed in Sec.2, using CMB temperature anisotropy data
432: as above together with
433: LSS data from the 2dFGRS and PSCZ galaxy redshift surveys.
434: The constraints on cosmological parameters are listed in Table 1.
435:
436: The results of fitting the same CMB and LSS data to a scale-invariant model
437: and a power-law model are shown in Table 1.
438: The power-law model differs by
439: $\Delta\chi^2_{eff}=6$ from the
440: the wavelet band powers model\footnote{Since the two
441: methods give very similar estimates of $P_{in}(k)$,
442: the smaller $\chi^2_{eff}$ of the top-hat binning method
443: seems to indicate that the wavelet band power method
444: has not yet sampled the parameter values
445: with the smallest possible $\chi^2_{eff}$.
446: }, and by $\Delta\chi^2_{eff}=9$
447: from the top-hat binning model. Since the difference in the
448: number of degrees of
449: freedom is approximately 9, the power-law model is disfavored
450: at approximately $\sim 0.4\,\sigma$
451: and $\sim 0.9\, \sigma$
452: compared to the wavelet band powers model and
453: the the top-hat binning model respectively.
454: The powerlaw model is again favored over the
455: scale-invariant model at less than $1\,\sigma$.
456:
457: As discussed previously, the estimated parameters that describe the
458: power spectrum as an arbitrary function (wavelet band powers or
459: the tophat bin amplitudes) can be treated as data,
460: and compared with the fitted power-law and scale-invariant
spectra
461: by computing a $\chi^2$ using the full covariance matrix of
462: the estimated parameters (wavelet band powers or
463: the tophat bin amplitudes).
464: In the wavelet band powers method,
465: the $\chi^2$ for both power-law and scale-invariant parametrizations is 6.
466: Similarly in the top-hat case the $\chi^2$'s
467: are 10 and 12 for the power-law and scale-invariant parametrization
468: respectively.
469: These indicate similar significances for deviation of the
470: reconstructed spectrum from the simpler parametrizations.
471:
472: %\pspicture(0,0.2)(5.5,12.4)
473: %\rput[tl]{0}(-0.2,12.2){\epsfxsize=8.5cm \epsfclipon
474: %\epsffile{f2.eps}}
475: %\rput[tl]{0}(0,3.3){
476: %\begin{minipage}{8.75cm}
477: %\small\parindent=3.5mm
478: %{\sc Fig.}~2.---
479:
480: \clearpage
481:
482: \begin{figure}
483: \plotone{f2.eps}
484: \figcaption{
485: The reconstructed $P_{in}(k)$ with 1$\sigma$ error bars from the two different
486: methods discussed in Sec.2, using CMB data and
487: LSS data.
488: The dotted line indicates the scale-invariant model that best fits the data.}
489: \end{figure}
490:
491: \clearpage
492:
493: %\end{minipage}
494: %}
495: % \endpspicture
496:
497: %\vskip -1cm
498:
499:
500: Fig.3 shows the reconstructed $P_{in}(k)$ using the wavelet band powers
501: method, compared with the $P_{in}(k)$ constraints derived using
502: the WMAP team's constraints on $A$, $n_S$ and
503: $\mbox{d} n_S/\mbox{d}\ln k$ for $P_{in}(k)=A\,(k/k_0)^{n_S-1}$
504: \citep{Peiris03,Spergel03} (shaded region).\footnote{
505: The WMAP constraints in Fig.3 is similar to Fig.2
506: of \cite{Peiris03} (which considers tensor contributions), but the
507: $P_{in}(k)$ parameter constraints are taken from Table 8 of
508: \cite{Spergel03}, since we do not consider
509: tensor contributions in this paper.}
510: The shaded region in Fig.3 are only
511: meant to illustrate roughly the WMAP team's constraints, since we
512: have not included the covariances among $A$, $n_S$ and
513: $\mbox{d} n_S/\mbox{d}\ln k$ estimated by them (these are not publicly
514: available). Clearly, our results are consistent with the WMAP results within
515: 1$\,\sigma$.
516:
517:
518: \clearpage
519:
520: %\pspicture(0,0.2)(5.5,12.4)
521: %\rput[tl]{0}(-0.2,12.2){\epsfxsize=8.5cm \epsfclipon
522: %\epsffile{f3.eps}}
523: %\rput[tl]{0}(0,3.3){
524: %\begin{minipage}{8.75cm}
525: %\small\parindent=3.5mm
526: %{\sc Fig.}~3.---
527:
528: \begin{figure}
529: \plotone{f3.eps}
530: \figcaption{
531: The reconstructed $P_{in}(k)$ using the wavelet bandpower
532: method (with 1$\sigma$ error bars), compared with the $P_{in}(k)$ constraints
533: derived using the WMAP team's constraints on $A$, $n_S$ and
534: $\mbox{d} n_S/\mbox{d}\ln k$ for $P_{in}(k)=A\,(k/k_0)^{n_S-1}$
535: \citep{Peiris03,Spergel03} (shaded region). We have not included the
536: covariances among
537: $A$, $n_S$ and $\mbox{d} n_S/\mbox{d}\ln k$.
538: }
539: \end{figure}
540:
541: \clearpage
542:
543: %\end{minipage}
544: % }
545: %\endpspicture
546: % \vskip -1cm
547:
548: We note that the cosmological parameters are relatively
549: well constrained even when the primordial power spectrum is
550: reconstructed as a free function. In the MCMC method, the
551: parameters are allowed to vary within wide limits. We have monitored
552: convergence and mixing as advocated in Verde et al. (2003).
553: The amplitude of the band
554: on the smallest scale (centered at $k\sim 0.2\,$Mpc$^{-1}$) is
555: essentially unconstrained in both the wavelet band power and top hat binning
556: methods when using just CMB data.
557: Using CMB and LSS data,
558: the amplitude of this band gets constrained.
559: The amplitude in the band centered at $k\sim 0.0003\,$Mpc$^{-1}$
560: is single tailed towards larger values, but well constrained within the prior.
561: Besides these bands, the power in all the other
562: bands and the cosmological parameters are all well constrained, and have
563: close to Gaussian 1d marginalized distributions. Parameter constraints
564: from the full $n$-D distribution are somewhat weaker,
565: as expected, but consistent with the 1d marginalized distributions.
566:
567:
568:
569: We find only slight evidence for a preferred scale at $k \sim 0.01\,$Mpc$^{-1}$
570: in the primordial power spectrum from current data (Figs.1-2).
571: This apparent deviation from scale-invariance of $P_{in}(k)$ accompanies
572: a slightly low Hubble constant, and non-vanishing
573: reionization optical depth $\tau_{ri}$ (see Table 1).
574: Inclusion of tensor contributions in the analysis would also
575: increase the effect
576: (as the data would then be consistent with reduced power on large
577: scales which can be filled in by tensor contributions, as also noted
578: in \cite{Seljak03}). The data do not require this deviation however, and
579: within parameter degeneracies appear consistent with scale invariance,
580: as well as with a slight red tilt (see Table 1).
581: Note that the current data are consistent with the tensor to scalar
582: ratio $T/S=0$, and the fit is not improved by including $T/S$ as a parameter
583: \citep{Spergel03}. Therefore, current data
584: do not require a tensor contribution. However, this
585: does not imply that a non-zero tensor contribution
586: is ruled out.
587: Similarly, deviation of the primordial power spectrum
588: from scale invariance is not ruled out at present, though limits
589: can be placed on such deviations (see Figures).
590: We would be able to better distinguish between these models if
591: cosmological parameters could be constrained to better accuracy.
592: We have not included Lyman $\alpha$,
593: weak lensing and supernovae data, since these have larger uncertainties
594: at present.
595:
596:
597: \section{Summary and Discussion}
598:
599: Reconstructing the shape of the primordial power
600: spectrum $P_{in}(k)$ in a model
601: independent way from cosmological data is a useful consistency check
602: on what is usually assumed regarding early universe physics. It is also
603: our primary window to unknown physics during inflation. We have used two
604: methods to reconstruct
605: $P_{in}(k)$ as a free function from CMB temperature
606: anisotropy and LSS data (Figs.1 and 2).
607: The two methods are complementary to each other, and
608: give consistent results.
609: We find that $P_{in}(k)$ reconstructed from CMB data alone (WMAP,
610: CBI, and ACBAR), or from CMB data together with LSS data
611: (2dFGRS and
612: PCSZ), seems to
613: indicate excess power for $ 0.002\,$Mpc$^{-1} \la k \la 0.03\,$Mpc$^{-1}$,
614: consistent with that found by the WMAP team \citep{Peiris03}
615: but at a lower significance of $\sim 1\,\sigma$ (Fig.3).
616: Note that the significance level deduced here is also low because we
617: are reconstructing $P_{in}(k)$ in a large number of bins.
618: Neither a scale-invariant $P_{in}(k)$ nor a power-law $P_{in}(k)$
619: is ruled out by the current data.
620:
621:
622: We find that this apparent deviation of $P_{in}(k)$ accompanies
623: a slightly low Hubble constant (and correspondingly
624: a slightly high $\Omega_m$), and a non-vanishing $\tau_{ri}$ (Table 1).
625: However, the 1$\sigma$ error bars on our derived $H_0$
626: values
627: overlap with the 1$\sigma$ error bar obtained by the
628: HST Key project \citep{Freedman01} and matches well the $H_0$
629: determined using supernovae \citep{Branch98}.
630: Note that because of parameter degeneracies, the
631: $H_0$ values derived from CMB data represent {\it indirect} measurements,
632: while the $H_0$ derived from Cepheid distances \citep{Freedman01}
633: or supernova data \citep{Branch98} are {\it direct} measurements.
634: It is also important to include the systematic uncertainty
635: in local direct measurements of $H_0$ due to matter inhomogeneity
636: in the universe \citep{Wang98}, as included in the error estimate
637: of $H_0$ by \cite{Freedman01}.
638: Clearly, more stringent independent measurements of $H_0$
639: can help tighten the constraints on $P_{in}(k)$.
640:
641: We have not included tensor contributions in our analysis, because
642: the WMAP data are not yet constraining on the tensor perturbations
643: from inflation. The inclusion of tensor contributions are expected
644: to increase the deviation of $P_{in}(k)$ from scale-invariance.
645:
646: Our results are consistent with that of \cite{Bridle03} and \cite{Barger03};
647: both find that the $P_{in}(k)$ derived from current data from WMAP (\cite{Bridle03}
648: included LSS data as well) are consistent with
649: scale-invariance.
650: There are two basic differences between the analysis presented in \S4 of \cite{Bridle03} and this paper.
651: The sensitivity of their method to the
652: feature around $k \sim 0.01$Mpc$^{-1}$ is reduced because their banding
653: oversamples this region by a factor of three.
654: Also, \cite{Bridle03} reconstructed $P_{in}(k)$ using linear interpolation
655: of amplitudes of $P_{in}(k)$ at discrete $k$ points, an approach
656: pursued in \cite{WangMathews02} and MW03a. This method is expected to lead to
657: stronger correlations between adjacent $P_{in}(k)$ amplitudes estimated
658: from data. Since the different binning choices made by us and \cite{Bridle03}
659: lead to different correlations between the
660: estimated parameters, they provide complementary and somewhat
661: different information. \cite{Barger03} used WMAP temperature data to constrain slow-roll
662: inflationary models.
663: %The difference in results between \cite{Barger03}
664: %and this paper can be explained as follows.
665: They find that $\tau_{ri}=0$ is preferred based on
666: WMAP temperature data.
667: We have found that taking $\tau_{ri}=0$
668: greatly diminishes the significance of any deviations of $P_{in}(k)$
669: from scale-invariance.
670: We have chosen to take into consideration the implications of the
671: WMAP polarization data
672: by applying a Gaussian prior on $\tau_{ri}$
673: based on the results of \cite{Kogut03}.
674:
675: We note that Miller et al. (2002) have
676: examined the pre-WMAP CMB temperature data in a non-parametric way to check
677: whether the data can be better fit by breaking away from our assumed
678: cosmological model, and to deduce the
679: significance levels of the acoustic peaks in the $C_l$ spectrum
680: non-parametrically. This also helps test the robustness of the
681: cosmological model.
682:
683: We conclude that without making assumptions about the form of $P_{in}(k)$,
684: the $P_{in}(k)$ derived from first year WMAP data deviates from
685: scale-invariance (with a preferred scale at
686: $k \sim 0.01\,$Mpc$^{-1}$) only at a significance level
687: of approximately $1\,\sigma$ (Fig.1-2).
688: Simplest forms of $P_{in}(k)$ (scale-invarant, or power-law)
689: are thus consistent with the data at present.
690: The WMAP data in subsequent years, together with improved constraints
691: from other independent cosmological probes, will allow us to place firmer
692: constraints on very early universe physics.
693:
694: \acknowledgements
695: It is a pleasure for us to thank Hiranya Peiris and Dipak Munshi
696: for helpful discussion,
697: the referee for useful comments,
698: and Henry Neeman for computational assistance.
699: We acknowledge the use of CAMB and CosmoMC. This work is
700: supported in part by NSF CAREER grant AST-0094335.
701:
702:
703:
704: \begin{thebibliography}{}
705:
706: \bibitem[Adams, Ross, \& Sarkar(1997)]{Adams97}
707: Adams, J.A., Ross, G.G., \& Sarkar, S. 1997, Nuclear Physics B, 503, 405
708:
709: \bibitem[Barger, Lee, \& Marfatia(2003)]{Barger03}
710: Barger, V., Lee, H.S., \& Marfatia, D. 2003, astro-ph/0302150
711:
712: \bibitem[Bennett et al.(2003)]{Bennett03}
713: Bennett, C., et al. 2003, astro-ph/0302207
714:
715: \bibitem[Branch(1998)]{Branch98}
716: Branch, D. 1998, ARA\&A, 36, 17
717:
718: \bibitem[Bridle et al.(2003)]{Bridle03}
719: Bridle, S.L., Lewis, A.M., Weller, J., \& Efstathiou, G. 2003,
720: astro-ph/0302306
721:
722: \bibitem[Bridle et al.(2002)]{Bridle02}
723: Bridle, S. L., Crittenden, R., Melchiorri, A., Hobson, M. P.,
724: Kneissl, R., \& Lasenby, A. N. 2002, MNRAS, 335, 1193
725:
726: \bibitem[Chung et al.(2000)]{Chung00}
727: Chung, D.J.H., Kolb, E.W., Riotto, A., \& Tkachev, I.I. 2000,
728: Phys. Rev. D, 62, 043508
729:
730: \bibitem[Daubechies(1992)]{Daub92}
731: Daubechies, I. 1992, Ten Lectures on Wavelets, S.I.A.M., Philadelphia.
732:
733: \bibitem[Elgaroy, Gramann, \& Lahav(2002)]{Elgaroy02}
734: Elgaroy, O., Gramann, M., \& Lahav, O. 2002, MNRAS, 333, 93
735:
736: \bibitem[Enqvist \& Kurki-Suonio(2000)]{Enqvist00}
737: Enqvist, K., \& Kurki-Suonio, H. 2000, Phys. Rev. D, 61, 043002
738:
739: \bibitem[Fang \& Feng(2000)]{Fang00}
740: Fang, L.Z., \& Feng, L.L. 2000, ApJ, 539, 5
741:
742: \bibitem[Freedman et al.(2001)]{Freedman01}
743: Freedman, W.L., et al. 2001, ApJ, 553, 47
744:
745: \bibitem[Freese, Frieman, \& Olinto(1990)]{naturalinf}
746: Freese, K., Frieman, J.A., \& Olinto, A.V. 1990, Phys. Rev. Lett., 65, 3233
747:
748: \bibitem[Guth(1981)]{Guth81}
749: Guth, A.H. 1981, Phys. Rev. D, 23, 347
750:
751: \bibitem[Hamilton \& Tegmark(2002)]{Hamilton02}
752: Hamilton, A.J.S., Tegmark, M. 2002, MNRAS, 330, 506
753:
754:
755: \bibitem[Holman et al.(1991ab)]{Holman91ab}
756: Holman, R., Kolb, E.W., Vadas, S.L., \& Wang, Y. 1991a, Phys.\ Rev., D43, 3833
757:
758: \bibitem[Holman et al.(1991b)]{Holman91b}
759: Holman, R., Kolb, E.W., Vadas, S.L., \& Wang, Y. 1991b, Phys.\ Lett., B269, 252
760:
761: \bibitem[Hu \& Dodelson(2002)]{HuDodelson02}
762: Hu, W., \& Dodelson, S. 2002, ARA\&A, 40, 171
763:
764: \bibitem[Kinney(2001)]{Kinney01}
765: Kinney, W. H. 2001, Phys.Rev. D63, 043001
766:
767: \bibitem[Knox et al.(2001)]{Knox01}
768: Knox, L., Christensen, N., \& Skordis, C. 2001, ApJ, 563, L95
769:
770: \bibitem[Kogut et al.(2003)]{Kogut03}
771: Kogut, A., et al. 2003, astro-ph/0302213
772:
773: \bibitem[Kolb \& Turner(1990)]{KolbTurner90}
774: Kolb, E.W., \& Turner, M.S. 1990, {\it The Early Universe} (Addison-Wesley
775: Publishing Company)
776:
777: \bibitem[Kosowsky, Milosavljevic, \& Jimenez(2002)]{Kosowsky02}
778: Kosowsky, A., Milosavljevic, M., \& Jimenez, R. 2002, Phys.Rev. D66, 063007
779:
780: \bibitem[Kosowsky \& Turner(1995)]{Kosow95}
781: Kosowsky, A.; Turner, M.S. 1995, Phys.Rev. D52, 1739
782:
783: \bibitem[Kuo et al.(2002)]{Kuo02}
784: Kuo, C.L., et al. 2002, submitted to ApJ, astro-ph/0212289
785:
786: \bibitem[La \& Steinhardt(1991)]{extendedinf}
787: La, D., \& Steinhardt, P.J. 1991, \prl, 62, 376
788:
789: \bibitem[Leach et al.(2002)]{Leach02}
790: Leach, S., Liddle, A.R., Jerome, M., \& Schwarz, D.J. 2002, Phys.Rev. D66, 3515
791:
792: \bibitem[Lesgourgues, Polarski, \& Starobinsky(1997)]{Les97}
793: Lesgourgues, J., Polarski, D., \& Starobinsky, A.A. 1997, Nuclear
794: Physics B, 497, 479
795:
796: \bibitem[Lewis \& Bridle(2002)]{Lewis02}
797: Lewis, A., \& Bridle, S.L. 2002, Phys.Rev. D66, 103511
798:
799: \bibitem[Liddle \& Lyth(1992)]{Liddle92}
800: Liddle, A.R. \& Lyth, D.H. 1992, Phys. Lett., B291, 391
801:
802: \bibitem[Linde(1994)]{Linde94}
803: Linde, A. 1994, Phys. Rev. D49, 748
804:
805: \bibitem[Linde(1983)]{Linde83}
806: Linde, A.D. 1983, Phys.~Lett., 129B, 177
807:
808: \bibitem[Lyth, Ungarelli, \& Wands(2002)]{Lyth02}
809: Lyth, D. H., Ungarelli, C., \& Wands, D. 2002, Phys. Rev. D, in press, astro-ph
810: /0208055
811:
812: \bibitem[Miller et al.(2002)]{miller}
813: Miller, C.J., Nichol, R.C., Genovese, C., \& Wasserman, L. 2002, ApJ, 565, 67
814:
815: \bibitem[Mukherjee, Hobson \& Lasenby(2000)]{pia00}
816: Mukherjee, P., Hobson, M., \& Lasenby, A. 2000, MNRAS, 318, 1157
817:
818: \bibitem[Mukherjee \& Wang(2003a)]{pia1}
819: Mukherjee, P., \& Wang, Y. 2003a, ApJ, in press, astro-ph/0301058 (MW03a)
820:
821: \bibitem[Mukherjee \& Wang(2003b)]{pia2}
822: Mukherjee, P., Wang, Y. 2003b, ApJ submitted, astro-ph/0301562
823:
824: \bibitem[Neil(1993)]{neil}
825: Neil, R.M. 1993, ftp://ftp.cs.utoronto.ca/pub/~radford/review.ps.gz
826:
827: \bibitem[Pearson et al.(2002)]{Pearson02}
828: Pearson, T.J., et al. 2002, submitted to ApJ, astro-ph/0205388
829:
830: \bibitem[Peebles \& Ratra(2003)]{PeeblesRatra03}
831: Peebles, P.J.E., \& Ratra, B. 2003, RMP, in press, astro-ph/0207347
832:
833: \bibitem[Percival et al.(2002)]{Percival02}
834: Percival, W.J. et al. 2002, MNRAS, in press, astro-ph/0206256
835:
836: \bibitem[Peiris et al.(2003)]{Peiris03}
837: Peiris, H. V., et al. 2003, astro-ph/0302225
838:
839: \bibitem[Randall, Soljacic, \& Guth(1996)]{Randall96}
840: Randall, L., Soljacic, M., \& Guth, A. 1996,
841: Nucl. Phys. B472, 377
842:
843:
844:
845:
846: \bibitem[Seljak, McDonald \& Makarov(2003)]{Seljak03}
847: Seljak, U., McDonald, P., \& Makarov, A. 2003, astro-ph/0302571
848:
849: \bibitem[Spergel et al.(2003)]{Spergel03}
850: Spergel, D. N.; et al. 2003, astro-ph/0302209
851:
852: \bibitem[Verde et al.(2003)]{Verde03}
853: Verde, L. et al. 2003, astro-ph/0302218
854:
855: \bibitem[Wang(1994)]{Wang94}
856: Wang, Y. 1994, Phys.\ Rev., D50, 6135
857:
858:
859: \bibitem[Wang, Spergel, \& Strauss(1999)]{Wang99}
860: Wang, Y., Spergel, D.N., \& Strauss, M.A. 1999,
861: ApJ, 510, 20
862:
863: \bibitem[Wang, Spergel, \& Turner(1998)]{Wang98}
864: Wang, Y.; Spergel, D.N.; \& Turner, E.L. 1998,
865: ApJ 498, 1
866:
867: \bibitem[Wang \& Mathews(2002)]{WangMathews02}
868: Wang, Y., \& Mathews, G.J. 2001, ApJ, 573, 1
869:
870: \end{thebibliography}
871:
872: \begin{deluxetable}{lccccccccc}
873: \tablecolumns{8}
874: \tabletypesize{\scriptsize\tiny}
875: %\tablewidth{0pt}
876: \tablecaption{Parameters Estimated from CMB\tablenotemark{a} and
877: LSS\tablenotemark{b} data\tablenotemark{c}}
878: \tablehead{
879: $P_{in}(k)$ model & data used & $P_{in}(k)$ parameters &
880: $\Omega_b\,h^2$ & $\Omega_m\,h^2$
881: & $h $ & $\tau_{ri}$ & $\chi^2_{eff}$\\
882: }
883: \startdata
884: wavelet band powers & CMB only & see Fig.1 &
885: $.0180 \pm .0038$ &
886: $.143 \pm .029$ & $ .575 \pm .082$ & $.185 \pm .045$ & 980.04 \\
887: top-hat binning & CMB only & see Fig.1 &
888: $.0185 \pm .0031$ &
889: $.129 \pm .029$ & $ .617 \pm .080$ & $ .175 \pm .044$ & 977.81\\
890: scale-invariant & CMB only & $A=.893 \pm .050 $ &
891: $.0237 \pm .0006$ & $.123 \pm .015$
892: & $.710\pm .044 $& $.173\pm .036$ & 988.29\\
893: power-law & CMB only & $A=.799\pm .117$ &
894: $.0228\pm .0012$&
895: $.116\pm .016$ & $.713\pm .044 $& $.136\pm .054$ & 987.92\\
896: & & $n_S=.974\pm .028$ & & & & & \\
897: \hline
898: wavelet band powers & CMB \& LSS & see Fig.2 &
899: $.0187\pm .0031$ &
900: $.136 \pm .021$ & $ .601 \pm .069 $ & $.191 \pm .047$ & 1037.86 \\
901: top-hat binning & CMB \& LSS & see Fig.2 &
902: $.0189 \pm .0019$ &
903: $.134 \pm .016$ & $ .597 \pm .049 $ & $ .164 \pm .047 $ & 1034.80\\
904: scale-invariant & CMB \& LSS & $A=.883\pm .050$ &
905: $.0238\pm .0006$ &
906: $.121\pm .007$ & $.714\pm .022$ & $.170\pm .032$ & 1044.33\\
907: power-law & CMB \& LSS & $A=.836\pm .107$ &
908: $.0233 \pm .0010$ &
909: $.120\pm .007$ & $.707\pm .026$ & $.147\pm .055$ & 1043.68\\
910: & & $n_S=.985\pm .028$ & & & & & \\
911: \enddata
912: \tablenotetext{a}{CMB
913: temperature anisotropy data from WMAP, CBI and ACBAR}
914: \tablenotetext{b}{LSS power spectrum data from the 2dFGRS and PSCZ galaxy redshift surveys}
915: \tablenotetext{c}{The number of data points in the different data sets used are 899 (WMAP), 4 (CBI), 7 (ACBAR), 32 (2dFGRS) and 22 (PSCZ).}
916: \end{deluxetable}
917:
918:
919:
920:
921:
922: \end{document}
923:
924:
925:
926: