1: % It also requires running BibTeX. The commands are as follows:
2: %
3: % 1) latex apssamp.tex
4: % 2) bibtex apssamp
5: % 3) latex apssamp.tex
6: % 4) latex apssamp.tex
7: %
8: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
9: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
10:
11: \usepackage{graphicx}% Include figure files
12: \usepackage{dcolumn}% Align table columns on decimal point
13: \usepackage{bm}% bold math
14:
15: \newcommand{\fpr}[1]{\left( #1\right)}
16: \newcommand{\fbr}[1]{\left\{ #1\right\}}
17: \newcommand{\fsq}[1]{\left[ #1\right]}
18: \newcommand{\fang}[1]{\langle #1\rangle}
19: \newcommand{\fabs}[1]{\left| #1\right|}
20: \newcommand{\bfx}{{\bf{x}}}
21: \newcommand{\bfk}{{\bf{k}}}
22: \newcommand{\WMAP}{{\it{WMAP }}}
23: \newcommand{\fnl}{f_{\mathrm{NL}}}
24: \newcommand{\side}{{\mathrm{side}}}
25: \begin{document}
26:
27: \title{Probing Non-Gaussianity In The Cosmic Microwave Background Anisotropies: One Point Distribution Function}
28: \author{E. Jeong$^1$ and G. F. Smoot$^{1,2}$}
29: \email{ehjeong@berkeley.edu, gfsmoot@lbl.gov}
30: \affiliation{$^1$Department of Physics, University of California, Berkeley,
31: CA, 94720\\
32: $^2$Lawrence Berkeley National Laboratory, 1 Cyclotron Road, Berkeley, CA,
33: 94720}
34:
35: \date{\today}% It is always \today, today,
36:
37: \begin{abstract}
38: We analyze \WMAP 3 year data using the one-point distribution functions to
39: probe the non-Gaussianity in the Cosmic Microwave Background (CMB) Anisotropy
40: data. Computer simulations are performed to determine the uncertainties of the
41: results. We report the non-Gaussianity parameter $\fnl$ is constrained to
42: $26<\fnl<82$ for Q-band, $12<\fnl<67$ for V-band, $7<\fnl<64$ for W-band
43: and $23<\fnl<75$ for Q+V+W combined data at 95\% confidence level (CL).
44: \end{abstract}
45:
46: \pacs{Valid PACS appear here}
47: \maketitle
48:
49: \section{introduction}
50: Non-Gaussianity is one of the most important tests of models of the inflation.
51: Among the various theoretical models on the inflation, slow-roll inflation is
52: currently most lively being studied. There are various predictions on the
53: magnitude of non-Gaussianity based on the simple model of slow-roll
54: inflation and its extensions, ranging from undetectably tiny values to large
55: enough values to be detectable with currently available data
56: \cite{barnabycline,battefeldeasther,calcagni,creminelli,bartolo.et.al}. On the
57: other hand, observational works have claimed both detection and non-detection
58: of non-Gaussianity (for reviews on recent works, see
59: \cite{komatsu.et.al,Spergel.et.al,troia.et.al,creminelli.et.al,gaztanagawagg}).
60: Among the popular techniques for detecting non-Gaussianity are one-point
61: distribution function fitting, bispectrum, trispectrum and Minkowski
62: functionals. Here, we investigate the one-point distribution functions to
63: probe primordial non-Gaussianity in the CMB anisotropy data. An observed CMB
64: anisotropy at a
65: direction ($\delta T_{obs}$) can be regarded as the superposition of three
66: parts: physical fluctuation of cosmic origin ($\delta T_p$), instrumental noise
67: ($\delta T_n$), and foreground emissions ($T_{fg}$). Since the foreground
68: templates are separately prepared, we start with foreground-removed data of
69: which the CMB anisotropy can be decomposed into two uncorrelated components,
70: \begin{equation}\label{eq1}
71: \delta T=\delta T_{obs}-T_{fg}=\delta T_{p}+\delta T_n.
72: \end{equation}
73: The primary source for the cosmic fluctuation of CMB at the large scale is
74: attributed to the Sachs-Wolfe effect which is again triggered by the
75: primordial curvature perturbation. The curvature perturbation $\Phi$ by
76: primordial seed during the inflation is transferred to CMB anisotropy with the
77: relation
78: \begin{equation}\label{eq2}
79: \frac{\delta T_p\fpr{\bfx}}{T_0}=\eta_{t}\Phi\fpr{\bfx}
80: \end{equation}
81: where $T_0=2.725$ K, the thermodynamic temperature of the CMB today, and
82: $\eta_{t}$ is the radiation transfer function. For the super-horizon scale,
83: we take $\eta_{t}=-1/3$ from the Sachs-Wolfe effects. At the first-order of
84: perturbation, we may replace $\Phi =\Phi_g$, where $\Phi_g$ is an auxiliary
85: random Gaussian field with its mean $\fang{\Phi_g}=0$ and its variance denoted
86: by $\fang{\Phi_g^2}$. When the second-order perturbation is considered, it is
87: conventional to prescribe the nonlinear coupling of the curvature perturbation
88: as \cite{komatsuspergel}
89: \begin{equation}\label{eq3}
90: \Phi (\bfx )\simeq\Phi_g(\bfx )+\fnl\fpr{\Phi_g^2(\bfx )-\fang{\Phi_g^2}}
91: \end{equation}
92: where $\fnl$ is the non-Gaussianity parameter. The second term in
93: \eqref{eq3} is responsible for the non-Gaussianity of the primordial
94: fluctuation. Then, the probability distribution function of the non-Gaussian
95: field $\Phi$ can be derived as
96: \begin{eqnarray}
97: f_{\Phi}(\Phi)&=&\int f_G(\Phi_g)\delta_D
98: \fsq{\Phi-\Phi_g-\fnl\fpr{\Phi_g^2-\fang{\Phi_g^2}}}d\Phi_g\nonumber\\
99: &=&\frac{1}{\sqrt{2\pi\fang{\Phi_g^2}\fnl^2\fpr{\Phi_+-\Phi_-}^2}}\nonumber\\
100: & &\times\fsq{\exp\fpr{-\frac{\Phi_+^2}{2\fang{\Phi_g^2}}}
101: +\exp\fpr{-\frac{\Phi_-^2}{2\fang{\Phi_g^2}}}}\label{eq4}
102: \end{eqnarray}
103: where $\Phi_{\pm}$ are defined by
104: \begin{equation}\label{eq5}
105: \Phi_{\pm}=\frac{1}{2\fnl}
106: \fsq{-1\pm\sqrt{1+4\fnl\Phi+4\fnl^2\fang{\Phi_g^2}}}
107: \end{equation}
108: and $\Phi$ has to be limited by the reality of $\Phi_{\pm}$ as
109: \begin{equation}\label{eq6}
110: \fnl\Phi > -\frac{1}{4}-\fnl^2\fang{\Phi_g^2}.
111: \end{equation}
112: $\fang{\Phi_g^2}$ can be expressed in terms of $\eta_t$, $T_0$ and
113: $\sigma_{CMB}$,
114: \begin{equation}\label{eq7}
115: \fang{\Phi_g^2}=\frac{1}{4\fnl^2}\fsq{-1+\sqrt{1+8
116: \fpr{\frac{\fnl\sigma_{CMB}}{\eta_t T_0}}^2}}.
117: \end{equation}
118: For a pixelized CMB anisotropy data set, the probability distribution function
119: for Gaussian instrumental noise becomes
120: \begin{equation}\label{eq8}
121: f_N(\delta T_n)=\frac{1}{N_{pix}}\sum_{i=1}^{N_{pix}}
122: \frac{1}{\sqrt{2\pi\sigma_0^2/n_i}}
123: \exp\fsq{-\frac{\delta T_n^2}{2\sigma_0^2/n_i}}
124: \end{equation}
125: where $n_i$ is the effective number of measurements at the $i_{th}$ pixel and
126: $\sigma_0$ represents the dispersion of the instrumental noise per observation
127: ($\sigma_0$=2.1898, 3.1249, 6.5112 mK for Q, V, W-band, respectively
128: \cite{Limon.et.al}). Now, it is straightforward to express the probability
129: density function for $\delta T$ in an integral form,
130: \begin{eqnarray}
131: f(\delta T)&=&\int f_{\delta T_p}(\delta T_p)f_N(\delta T_n)\nonumber\\
132: & &\times\delta_D\fpr{\delta T-\delta T_p-\delta T_n}
133: d\delta T_p d\delta T_n\nonumber\\
134: &=&\int f_{\Phi}(\Phi )f_N(\delta T_n)\nonumber\\
135: & &\times\delta_D\fpr{\delta T-\eta_t T_0\Phi-\delta T_n}d\Phi d\delta T_n.
136: \label{eq9}
137: \end{eqnarray}
138: The probability density function derived in \eqref{eq9} explicitly contains
139: the non-Gaussianity parameter $\fnl$, and it can serve as the prediction of
140: one-point distribution function with a given $\fnl$ for a (ideally)
141: foreground-removed CMB anisotropy data set to estimate the magnitude of
142: deviation from Gaussian distribution in a quantitative manner.
143: \section{Application to \WMAP data}
144: We use the three channels of \WMAP 3 year CMB anisotropy data sets (Q (33GHz)-,
145: V (61GHz)-, W (94GHz)-band) which contain dominant signal over contaminations
146: to investigate the non-Gaussianity of the CMB anisotropy data. To remove the
147: foreground emissions, the Maximum Entropy Method (MEM) maps of the
148: synchrotron, free-free and thermal dust are
149: used\footnote[1]{http://lambda.gsfc.nasa.gov/}. The Kp0-mask is
150: applied to the sky maps to remove the intense Galactic emissions and scattered
151: bright point sources, which leaves 76.5\% of the sky. We also prepare a
152: combined map (Q+V+W) by taking a weighted sum for a pixel temperature,
153: \begin{equation}\label{eq10}
154: \delta T\fpr{\bfx}=\frac{\sum_i\delta T_i\fpr{\bfx}n_i\fpr{\bfx}/\sigma_{0i}^2}
155: {\sum_in_i\fpr{\bfx}/\sigma_{0i}^2},\quad i={\mathrm{Q,\: V,\: W}}
156: \end{equation}
157: where $n_i\fpr{\bfx}$ is the effective number of measurements at the pixelized
158: position $\bfx$ and $\sigma_{0i}$ is the dispersion of the instrumental noise
159: of the $i_{\mathrm{th}}$ channel. We can trace the effective variance of
160: the instrumental noise as a result of weighted sum defined in \eqref{eq10} as
161: \begin{equation}\label{eq11}
162: \sigma^2\fpr{\bfx}=\fsq{\sum_in_i\fpr{\bfx}/\sigma_{0i}^2}^{-1}
163: ,\quad i={\mathrm{Q,\: V,\: W}}.
164: \end{equation}
165: The sky map data are degraded from $N_{\mathrm{side}}=512$
166: ($6.87^{\prime}$ pixel) to $N_{\mathrm{side}}=128$ ($27.48^{\prime}$ pixel)
167: where the number of pixels in a full sky map is given
168: by $12\times N_{\mathrm{side}}^2$. The purpose of demotion of the resolution is
169: to suppress the small scale fluctuation which is dominated by the instrumental
170: noise. We perform the $\chi^2$-test for the goodness of fit for the probability
171: density function given in \eqref{eq9} as a prediction to the observed
172: probability density function which is directly
173: calculated from the \WMAP data. Figure \ref{fig1} and Table \ref{t1} show the
174: results of $\chi^2$ fitting of \WMAP data sets with varying $\fnl$ as a free
175: parameter. All data sets are best fitted at positive $\fnl$
176: (dubbed $\fnl^{\fpr{\mathrm{opt}}}$) which are consistent with one another as
177: well as the results with previous work \cite{Spergel.et.al} within the
178: statistical errors.
179: \begin{figure*}[ht]
180: \begin{center}
181: \includegraphics[width=6.2cm,angle=0]{fig1a.ps}
182: \includegraphics[width=6.2cm,angle=0]{fig1b.ps}
183: \includegraphics[width=6.2cm,angle=0]{fig1c.ps}
184: \includegraphics[width=6.2cm,angle=0]{fig1d.ps}
185: \caption[$\fnl$ vs. $\chi^2$]
186: {\label{fig1}$\chi^2$-test for goodness of fit is used to find the optimal
187: value of the non-Gaussianity parameter. The anisotropy maps are demoted to
188: $N_{\side}=2^7$. Top left: Q-band, top right: V-band,
189: bottom left: W-band, bottom right: Q+V+W combined. The color (shaded) bands
190: indicate the bounds on $\fnl$ in which the probability that the statistic
191: $\chi^2$ is smaller than that of the observed with respect to the prediction
192: curve with $\fnl$.}
193: \end{center}
194: \end{figure*}
195: \begin{table}[ht]
196: \begin{center}
197: \begin{tabular}{c|r@{$<\!\fnl\!<$}l|r@{$<\!\fnl\!<$}l|r@{$<\!\fnl\!<$}l
198: |r@{$<\!\fnl\!<$}l}
199: \hline\hline
200: Map&\multicolumn{2}{c|}{Q-band}&\multicolumn{2}{c|}{V-band}
201: &\multicolumn{2}{c|}{W-band}&\multicolumn{2}{c}{Q+V+W}\\
202: \hline
203: $\fnl^{\fpr{\mathrm{opt}}}$&\multicolumn{2}{c|}{53}&\multicolumn{2}{c|}{39}
204: &\multicolumn{2}{c|}{35}&\multicolumn{2}{c}{48}\\
205: \hline
206: 68\%&\multicolumn{2}{c|}{N/A}&-9&86&\multicolumn{2}{c|}{N/A}&-6&101\\
207: 95\%&34&71&-37&113&-20&89&-29&123\\
208: 99\%&3&102&-49&126&-37&106&-40&134\\
209: \hline
210: DOF&\multicolumn{2}{c|}{119}&\multicolumn{2}{c|}{119}
211: &\multicolumn{2}{c|}{119}&\multicolumn{2}{c}{119}\\
212: \hline\hline
213: \end{tabular}
214: \end{center}
215: \caption{\label{t1}The results of $\chi^2$-tests for Goodness-of-fit for \WMAP
216: 3 year data. The percentages on the first column represent the tail
217: probabilities at which the statistic $\chi^2$ would be smaller than the
218: observed. DOF on the bottom row stands for {\it{Degrees of freedom}}.}
219: \end{table}
220:
221: \section{Simulation And Statistical Uncertainties}
222: As is shown in Figure \ref{fig1}, \WMAP data fit well with finite range of the
223: non-Gaussianity parameter $\fnl$. We pick $\fnl^{\fpr{\mathrm{opt}}}$ as the
224: representative magnitude of non-Gaussianity for a data set, and carry out the
225: Monte-Carlo simulations to test the pertinence of $\fnl^{\fpr{\mathrm{opt}}}$
226: as a proper measure of non-Gaussianity for a data set in a quantitative manner.
227: A simulated data set is prepared as follows: First, a Gaussian field $\Phi_g$
228: with its variance equal to \eqref{eq7} is generated and we use it to generate
229: $\Phi$-field of which the deviation from Gaussianity is denoted by $\fnl$.
230: Second, we prepare a noise map in which each pixel contains a random value
231: picked from a normal distribution with variance $\sigma_0^2/n_i$ as defined in
232: \eqref{eq8} and add this to $\Phi$-field. A simulated map generated in this
233: way has the same noise structure and the dispersion of physical fluctuation
234: ($\sigma_{CMB}$) as a real data set. Since a data set contains nontrivial
235: instrumental noise and the number of pixels is finite, the returned magnitude
236: of non-Gaussianity ($\fnl^{\fpr{\mathrm{opt}}}$) would have some uncertainty.
237: For a given value of $\fnl$, we repeat the simulation and find that the
238: algorithm returned an unbiased, normal distribution of
239: $\fnl^{\fpr{\mathrm{opt}}}$ which is centered at the input value of $\fnl$.
240: Thus, from the results of the simulations, we are able to set the bounds on
241: $\fnl$ for the real data. The simulation results and error bands are plotted
242: in Figure \ref{fig2} and the deduced uncertainties for $\fnl$ are summarized
243: in Table \ref{t2}. It is very remarkable that this analysis strongly disfavors
244: the null hypothesis ($\fnl=0$) and all the data sets show consistent results
245: within the statistical errors. The results are also consistent with the works
246: by the \WMAP team ($-54<\fnl<114$ at 95\% CL from bispectrum
247: \cite{Spergel.et.al}) but with much tighter limits and more importantly, it
248: excludes $\fnl=0$ at 95\% CL.
249: \begin{figure*}[ht]
250: \begin{center}
251: \includegraphics[width=6.2cm,angle=0]{fig2a.ps}
252: \includegraphics[width=6.2cm,angle=0]{fig2b.ps}
253: \includegraphics[width=6.2cm,angle=0]{fig2c.ps}
254: \includegraphics[width=6.2cm,angle=0]{fig2d.ps}
255: \caption[Simulation]
256: {\label{fig2}Error bands found through the simulations with the same
257: profiles with \WMAP 3 year data (blue-green slanted bands). Here the
258: ``same profile'' means that the simulated data set has exactly same
259: instrumental noise structure and $\sigma_{CMB}$ as a real \WMAP data set has.
260: The vertical bands indicate the bounds for $\fnl^{\fpr{\mathrm{opt}}}$ found
261: from the analysis of the \WMAP data. Top left: Q-band, top right: V-band,
262: bottom left: W-band, bottom right: Q+V+W. On the vertical axis,
263: $\fnl^{\fpr{\mathrm{out}}}$ is the measured $\fnl^{\fpr{\mathrm{opt}}}$ from
264: the simulated data.}
265: \end{center}
266: \end{figure*}
267:
268: \begin{table}[ht]
269: \begin{center}
270: \begin{tabular}{c|r@{$<\fnl<$}l|r@{$<\fnl<$}l|r@{$<\fnl<$}l
271: |r@{$<\fnl<$}l}
272: \hline\hline
273: Map&\multicolumn{2}{c|}{Q-band}&\multicolumn{2}{c|}{V-band}
274: &\multicolumn{2}{c|}{W-band}&\multicolumn{2}{c}{Q+V+W}\\
275: \hline
276: $N_{\mathrm{sim}}$\footnote[2]{Number of simulations for each input $\fnl$.}&\multicolumn{2}{c|}{100}&\multicolumn{2}{c|}{100}
277: &\multicolumn{2}{c|}{100}&\multicolumn{2}{c}{100}\\
278: \hline
279: 68\%&40&68&26&53&21&50&36&62\\
280: 95\%&26&82&12&67&7&64&23&75\\
281: 99\%&12&96&-1&80&-7&78&9&88\\
282: \hline\hline
283: \end{tabular}
284: \end{center}
285: \caption{\label{t2}Summary of results from Simulations with \WMAP 3 year data
286: profiles and the bounds at three confidence levels from the simulations.}
287: \end{table}
288:
289: \section{Conclusion}
290: We developed an algorithm that uses the one-point distribution function
291: to investigate the non-Gaussianity of CMB anisotropy data, and applied
292: it to \WMAP 3 year data. We found that the null result ($\fnl$=0) is
293: manifestly excluded at 95\% CL. The estimated magnitude of non-Gaussianity
294: parameter is $23<\fnl<75$ at 95\% CL and $9<\fnl<88$ at 99\% CL for the
295: (Q+V+W)-combined map. Since the quadratic term in \eqref{eq3} takes a generic
296: form of Taylor series for a perturbative expansion, it is a good
297: possibility that the observed non-Gaussianity in this work is a combined
298: effects of various physical processes, while the primordial seeds are very
299: likely to be the leading one. There are two premises we have taken in
300: developing the algorithm, which, provided they are not precise enough, could
301: cause non-Gaussianity of not cosmic but systematic origin: (1) the probability
302: distribution function of the instrumental noise for each pixel is centered at
303: zero, and (2) the foreground emissions are removed efficiently enough in the
304: foreground-removed maps. The first condition can be broken when the thermal
305: and radiation environments of the \WMAP satellite in its orbit are taken into
306: account, while the \WMAP team assessed they are insufficient to influence the
307: science data \cite{Limon.et.al}. So, we tested the effects of the alternative
308: noise distributions with a random mean in each of the Gaussian
309: distribution in \eqref{eq8} and the algorithm was not misled to show
310: non-Gaussianity within the statistical error. It is difficult to directly
311: estimate how much residual foreground emissions after foreground subtraction
312: would affect the one-point distribution function. We solely rely on the
313: quality of foreground templates and it is remarkably successful, showing that
314: the observed total Galactic emission matches the model to less than 1\%
315: \cite{bennett.et.al2,Hinshaw.et.al}. We also analyzed simulated maps which
316: are (Gaussian map + foreground templates), and all the templates for Q, V and
317: W-channel showed negative values of the non-Gaussianity parameter with
318: $\fabs{\fnl}\sim\mathcal{O}\fpr{10^1}$ at the resolution
319: $N_{\mathrm{side}}=512$.
320:
321: \begin{acknowledgments}
322: We would like to thank Dr. Sara Ricciardi and Dr. Oliver Zahn for valuable
323: discussion and comments on the foreground emission and other topics.
324: Computer simulation and data analysis with \WMAP data set were done using the
325: HEALPix\cite{Gorski.et.al}. This work was supported by LBNL and the Department
326: of Physics at University of California, Berkeley.
327: \end{acknowledgments}
328: \bibliography{ms}
329: \end{document}
330: