1:
2: \documentclass[12pt,preprint]{aastex}
3: %\documentclass[manuscript]{aastex}
4: \usepackage{graphicx,epsfig,epsf,amssymb} %for PS/EPS graphics inclusion, new
5: \newcommand{\vdag}{(v)^\dagger}
6: \newcommand{\myemail}{litp@ihep.ac.cn}
7:
8: \def \<{\langle}
9: \def \>{\rangle}
10:
11: \def \tr{$\sqrt{\<\hat{t}^2\>}$}
12: \def \t{$\mathbf{\hat{t}}$}
13: \newcommand{\degree}{^\circ}
14:
15:
16: \shorttitle{Errors in CMB maps}
17: \shortauthors{Liu \& Li}
18:
19: \begin{document}
20: \title{ Statistical and systematical errors\\ in cosmic microwave background maps}
21: \author{Hao Liu\altaffilmark{1,3} and Ti-Pei Li\altaffilmark{1,2}}
22: \altaffiltext{1}{Key Lab. of Particle Astrophys., Inst. of High Energy Phys.,
23: Chinese Academy of Sciences, Beijing}
24: \altaffiltext{2}{Center for
25: Astrophysics, Tsinghua University, Beijing, China; \myemail}
26: \altaffiltext{3}{Graduate School of Chinese Academy of Sciences, Beijing}
27:
28:
29: \begin{abstract}
30: Sky temperature map of the cosmic microwave
31: background (CMB) is one of the premier probes of cosmology.
32: To minimize instrumentally induced systematic errors,
33: CMB anisotropy experiments measure temperature differences across
34: the sky using paires of horn antennas with a fixed separation angle,
35: temperature maps are recovered from temperature
36: differences obtained in sky survey through a map-making procedure.
37: The instrument noise, inhomogeneities of the sky coverage and sky temperature
38: inevitably produce statistical and systematical errors in recovered
39: temperature maps.
40: We show in this paper that observation-dependent noise and systematic temperature
41: distortion contained in released Wilkinson Microwave Anisotropy Probe (WMAP)
42: CMB maps are remarkable. These errors
43: can contribute to large-scale anomalies detected
44: in WMAP maps and distort the angular power spectrum as well.
45: It is needed to remake temperature maps from original WMAP differential data
46: with modified map-making procedure
47: to avoid observation-dependent noise and systematic
48: distortion in recovered maps.
49: \end{abstract}
50:
51: \keywords {cosmic microwave background --- cosmology: observations ---
52: methods: statistical}
53:
54:
55: \section{Map-making}
56: The COBE and WMAP missions measure temperature differences between sky points
57: using differential radiometers consisting of plus-horn and minus-horn~\cite{smo90,ben03a}.
58: Let denote $t_i$ the temperature anisotropy at a sky pixel $i$.
59: The raw data in a certain band is a set of
60: temperature differences {\bf d} between pixels in the sky. From $N$ observations we have
61: the following observation equations
62: \begin{equation}
63: \label{dt}
64: \begin{array}{c@{\:-\:}c@{\;=\;}c}
65: t_{1^+} & t_{1^-} & d_1 \\
66: t_{2^+} & t_{2^-} & d_2 \\
67: \multicolumn{3}{c}{\dotfill}\\
68: t_{N^+} & t_{N^-} &~~d_N~.
69: \end{array}
70: \end{equation}
71: The above equation system can be expressed by matrix notation
72: \begin{equation}
73: \label{dt1}
74: \mathbf{At=d}~.
75: \end{equation}
76: Where the scan matrix of the experiment {\bf A}$ =(a(k,i)),~k=1,\cdots,N$
77: and $i=1,\cdots,L$ with $L$ being the total number
78: of sky map pixels. The most of elements $a(k,i)=0$ except for $a(k,i=k^+)=1$
79: and $a(k,i=k^-)=-1$, where $k^+$ denotes the pixel observed by the plus-horn
80: and $k^-$ the pixel observed by the minus-horn at an observation $k$.
81:
82: The normal equation of Eq.~\ref{dt} or Eq.~\ref{dt1} is
83: \begin{equation}
84: \label{ne}
85: \mathbf{Mt=A^Td}
86: \end{equation}
87: with $\mathbf{M=A^TA}$.
88: The least-squares estimate
89: of the sky map results from solving Eq.~\ref{ne}
90: \[ \mathbf{\hat{t}=M^{-1}A^Td}~. \]
91: The WMAP team \cite{hin03} uses the following approximate formula to compute the
92: iterative solution
93: \begin{equation}
94: \label{mm-w}
95: \mathbf{t^{(n+1)}=\tilde{M}^{-1}(A^Td-A^TAt^{(n)})+t^{(n)}}~,
96: \end{equation}
97: where $\mathbf{\tilde{M}^{-1}}=$ diag$(\frac{1}{N_1},\frac{1}{N_2},\cdots)$
98: is an approximate inverse of {\bf M} with $N_i$ being the total number
99: of observations for pixel $i$.
100:
101: The use of approximate inverse matrix $\mathbf{M^{-1}}$ is not necessary.
102: Here we derive an iterative formula directly from the normal equation.
103: The Eq.~\ref{ne} can be expressed as
104: \begin{eqnarray*}
105: N_i^+t_i-\sum_{k^+=i}t_{k^-}-\sum_{k^-=i}t_{k^+}+N_i^-t_i
106: =\sum_{k^+=i}d_k-\sum_{k^-=i}d_k~, \\
107: (i=1,2,\cdots,L)~.
108: \end{eqnarray*}
109: Where $\sum_{k^+=i}$ means summing over $N_i^+$ observations while the pixel
110: $i$ is observed by the plus-horn and $\sum_{k^-=i}$ means summing over $N_i^-$
111: observations while the pixel $i$ is observed by the minus-horn,
112: and the total number of observations for the pixel $i$ is $N_i=N_i^++N_i^-$.
113: From the above equations we can derive the following iterative formula
114: \begin{eqnarray}
115: \label{mm-l}
116: t_i^{(n+1)}=\frac{1}{N_i}(\sum_{k^+=i}(d_k+t_{k^-}^{(n)})-\sum_{k^-=i}
117: (d_k-t_{k^+}^{(n)}))~, \nonumber\\
118: (i=1,2,\cdots,L)~.
119: \end{eqnarray}
120:
121: With Eq.~\ref{mm-w} or Eq.~\ref{mm-l} when the number $n$ of iteration is large enough, we get
122: the final solution $\hat{t}_i=t_i^{(n)}$ for each pixel $i$.
123: The Eg.~\ref{mm-w} used by the WMAP team is an approximate formula and Eq.~\ref{mm-l}
124: is an exact one, but both has good performance for the differential data of a
125: noiseless instrument. With Eq.~\ref{mm-l} we can easily study the statistical
126: and systematical errors
127: induced by instrument noise, inhomogeneity of sky coverage, inhomogeneity of
128: sky temperature, and unbalance between two sky side measurements.
129:
130: \section{Exposure-dependent noise}
131: We can directly see from Eq.~\ref{mm-l} that their exists exposure dependent noise in
132: a recovered temperature map.
133: The WMAP differencing assembly has instrument noise {\bf n} per observation,
134: the real observation data $d_k=t_{k^+}-t_{k^-}+n_k$. The instrument noise is not negligible,
135: e.g. the mean rms noise $\sigma_0=6.7$ mK for the W4-band of WMAP \cite{lim03}.
136: The obtained temperature differences $d_k$ can be
137: taken as a random variable with a standard error $\sigma_0$.
138: The final iterative solution $\hat{t}_i$ from Eq.~\ref{mm-l} has a noise component
139: being the mean of $N_i$ variables, i.e. the temperature ${\hat t}_i$ in a WMAP
140: map for a sky pixel $i$ has an exposure-dependent error
141: \begin{equation}
142: \label{sigma}
143: \sigma_i=\frac{1}{\sqrt{N_i}}\sigma_0~.
144: \end{equation}
145:
146: By analyzing CMB maps from the first year WMAP (WMAP1) data,
147: Tegmark et al. (2003) find both the CMB quadrupole and octopole having
148: power along a particular spatial axis and more works \cite{cos04,eri04a,sch04,jaf05}
149: find that the axis of maximum asymmetry tends to lie close to the ecliptic axis.
150: A similar anomaly was also found in COBE maps \cite{cop06}.
151: The unexplained orientation of large-scale
152: patterns of CMB maps in respect to the ecliptic frame is one of the biggest surprises
153: in CMB studies \cite{sta05}.
154: A notable asymmetry of temperature fluctuation power in two opposing hemispheres is also found
155: in the WMAP1 and COBE results \cite{eri04b, han04}.
156: After the release of WMAP results in 2006 March, similar large-scale anomalies
157: are still detected
158: in the WMAP3 data \cite{abr06,jaf06,cop07,lan07,eri07,par07,vie07,sam08}.
159: The unexpected large-scale anomalies in CMB maps are extensively studied
160: with different techniques, but their reasons still remain unclear.
161: Here we show that the exposure-dependent noise should be an important
162: source of detected anomalies.
163:
164: \subsection{Large-scale non-Gaussian modulation}
165:
166: \begin{figure}[p]
167: \label{f1}
168: \vspace{2mm}
169: \begin{center}
170: \psfig{figure=f1a.ps,width=45mm,height=70mm,angle=90}\\
171: \vspace{-5mm}\psfig{figure=f1b.ps,width=45mm,height=70mm,angle=90}\\
172: \vspace{-5mm}\psfig{figure=f1c.ps,width=45mm,height=70mm,angle=90}
173: \end{center}
174: \caption{Large-scale non-Gaussian modulation features with $l_{max}=2$
175: for the WMAP3 V-band data .
176: {\sl Top panel}: large-scale feature of $1/N$ map.
177: {\sl Middle panel}: large-scale feature of $1/\sigma(N)$ map.
178: {\sl Bottom panel}: Best-fit large-scale modulation function $f(\mathbf{\hat{n}})$
179: for temperature map.}
180: \end{figure}
181:
182: Because the sky coverage of WMAP mission is inhomogeneous
183: -- the number of observations being greatest at the ecliptic poles
184: and the ecliptic plane being most sparsely observed~\cite{hin07}, from
185: Eq.~\ref{sigma} we know that the WMAP temperature maps
186: contain inhomogeneous exposure-dependent noise components.
187: The remarkable fact, the feature of large-scale anomalies
188: detected in WMAP maps being very similar with the WMAP exposure pattern,
189: strongly indicates the existence of such observational effect on WMAP maps.
190: To address the large scale anomalies, such as asymmetry,
191: alignment and low $l$ power issues detected in WMAP data
192: with different techniques, the WMAP team~\cite{sper06} describe the observed
193: temperature fluctuations, $\mathbf{{\hat{t}}}$, as a Gaussian and isotropic random
194: field, $\mathbf{t}$, modulated by a function $f(\mathbf{n})$
195: \[ \hat{t}(\mathbf{n})=t(\mathbf{n})[1+f(\mathbf{n})] \]
196: where $f(\mathbf{n})$ is an arbitrary modulation function. They expand $f(\mathbf{n})$
197: in spherical harmonics
198: \[ f(\mathbf{n})=\sum_{l=1}^{l_{max}}\sum_{l=-m}^mf_{lm}Y_{lm}(\mathbf{n}) \]
199: and use maximum likelihood technique with
200: a Markov Chain Monte Carlo solver to get the best fit values of $f_{lm}$
201: with $l_{max}=2$ for the WMAP3 V-band map.
202: The bottom panel of Fig.~1 is obtained
203: based on the best fit coefficients, showing
204: in a unifying manner the large scale anomalies in WMAP temperature fluctuations
205: which is the same feature that has been identified in a number
206: of papers on non-Gaussianity.
207: We calculate the spherical harmonic coefficients $f_{lm}$ with $l_{max}=2$
208: for the map of $1/N$ with $N$ being number of observations per sky pixel
209: from the WMAP3 V-band data. The top panel of Fig.~1 shows
210: the map of $1/N$ reconstructed based on the coefficients $f_{lm}$.
211: The middle panel shows the reconstructed result for
212: the observation fluctuation map --
213: the map of $1/\sigma(N)$ where the rms variation $\sigma(N)=\sqrt{\<(N-\<N\>)^2\>}$
214: calculated within a region of $\sim 1\degree$ side dimension for each sky pixel.
215: The large-scale non-Gaussian modulation features of WMAP temperature map
216: and scan pattern being similar for each other suggests that large scale anomalies
217: detected in WMAP maps are most probably resulted from observation effect, not
218: cosmological origin. In comparing the top and middle panels in Fig.~1,
219: the modulation pattern for the observation fluctuation map
220: is more similar to the detected anomalies shown in the bottom panel,
221: indicating that the fluctuation of observation numbers could produce additional
222: noise component to the recovered temperature map.
223:
224: \subsection{Alignment and planarity}
225: \begin{figure}[t]
226: \label{f2}
227: % \vspace{-50mm}
228: \begin{center}
229: \psfig{figure=f2a.ps,width=45mm,height=60mm,angle=270}\\
230: \psfig{figure=f2b.ps,width=45mm,height=60mm,angle=270}
231: \caption{Latitude distributions from the WMAP3 foreground-cleaned Q1-band
232: map in ecliptic coordinates.
233: {\sl Black line}: rms variation in the map, $\sigma_0=2.27$ mK.
234: {\sl Green line}: residual rms after excluding the exposure-dependent anisotropy
235: component by Eq.~\ref{rms}.
236: {\sl Red line}: average number of observations per sky pixel.
237: {\sl Blue line}: average map temperature (in mK, times by a factor of
238: $2.76\times 10^4$).
239: }
240: \end{center}
241: \end{figure}
242:
243: To show the anisotropy noise, we compute the average rms variation $\sqrt{\<{\hat{t}}^2\>}$
244: for the WMAP3 foreground-cleaned Q1-band
245: temperature map at different latitudes; the results are shown by the black
246: line in Fig.~2. The red line in Fig.~2
247: shows that of average number of observations per sky pixel.
248: To modify the exposure-induced rms anisotropy,
249: we use the following formula to get the residual variation
250: \begin{equation}
251: \label{rms}
252: rms=\sqrt{\<({\hat{t}}-\<{\hat{t}}\>)^2\>-\<\sigma_0^2/N\> }~.
253: \end{equation}
254: The residual rms of WMAP3 Q1-band , the green line in Fig.~2,
255: shows that the latitude dependences are well excluded
256: by using Eq.~\ref{rms}, or, in other words, the noise per pixel in
257: the WMAP map is really exposure-dependent in a way described by Eq.~\ref{sigma}.
258:
259: \begin{figure}[t]
260: \label{f3}
261: % \vspace{-10mm}
262: \begin{center}
263: \psfig{figure=f3a.ps,width=45mm,height=55mm,angle=270}\hspace{8mm}
264: \psfig{figure=f3b.ps,width=45mm,height=55mm,angle=270}
265: \caption{Latitude distributions of temperature rms variation
266: from WMAP3 foreground-cleaned maps in ecliptic coordinates.
267: {\sl left}: V1 band, $\sigma_0=3.29$ mK.
268: {\sl right}: W1 band, $\sigma_0=5.83$ mK.
269: }
270: \end{center}
271: \end{figure}
272: \begin{figure}[t]
273: \label{f4}
274: % \vspace{-10mm}
275: \begin{center}
276: \psfig{figure=f4a.ps,width=45mm,height=60mm,angle=270}\hspace{4mm}
277: \psfig{figure=f4b.ps,width=45mm,height=60mm,angle=270}
278: \caption{Latitude distributions of temperature rms variation
279: from WMAP foreground-cleaned maps in ecliptic coordinates.
280: {\sl left}: WMAP1 W4 band, $\sigma_0=6.7$ mK.
281: {\sl right}: WMAP3 W4 band, $\sigma_0=6.7$ mK.
282: }
283: \end{center}
284: \end{figure}
285:
286:
287: Similar features are also found
288: in other waveband maps and the combined frequency maps (TOH and ILC maps)
289: as well.
290: For example, Fig.~3 shows the results from WMAP3 V1 and W1 bands.
291: The similarity of structures of latitude distribution of rms variation for different bands shown
292: in Fig.~2 and Fig.~3 indicate that they are commonly originated
293: from the ununiformity of sky exposure.
294: The average pixel rms is 0.20 mK from $N=1.68\times 10^6$
295: observations for W1 band, 0.15 mK from $1.12\times 10^6$ observations for V1 band and
296: 0.13 mK from $0.84\times 10^6$ observations for Q1 band.
297: The instrument noise $\sigma_0$ for W1, V1 and Q1 band are 5.85, 3.29 and
298: 2.27 mK respectively \cite{lim03}. We find that the ratios
299: between rms$\times \sqrt{N}$ of different bands
300: are approximately equal to that between $\sigma_0$ of corresponding bands,
301: which is what expected by Eq.~\ref{sigma}.
302: Fig.~4 shows that the noise anisotropy for the three year WMAP W4-band map is
303: almost the same as that for the first year map, not suppressed with
304: data accumulation.
305:
306: As demonstrated above, the released WMAP temperature maps
307: contain considerable exposure-dependent noise.
308: The remarkable similarity
309: between the large scale non-Gaussian modulation feature of WMAP temperature fluctuation
310: and exposure-pattern of WMAP observation, as shown in Fig.~1, suggests that
311: large scale anomalies detected in WMAP maps are most probably resulted from
312: observation effect, not astrophysical or cosmological origin.
313:
314: For foreground removal the WMAP team produce the internal linear
315: combination (ILC) map~${\mathbf t}$ from the five frequency sky maps
316: ${\mathbf t_i}~(i=1,\cdots,5)$ by
317: \(
318: {\mathbf t}=\sum_iw_i{\mathbf t}_i \)
319: where the weights minimize the variance of final map, Var({\bf t}), under the constraint
320: $\sum_iw_i=1$.
321: For the region outside of the inner Galactic plane and by a nonlinear search, the weights $w$
322: are found to be 0.109, -0.684, -0.096, 1.921, -0.250 for
323: K, Ka, Q, V, and W bands, respectively \cite{ben03c}.
324: Eriksen et al. (2004b) make minimization of the variance
325: under the constraint by means of Lagrange multipliers and obtain
326: the solutions are the inverse covariance weights
327: \(
328: w_i=\frac{\sum_jC_{ij}^{-1}}{\sum_{j,l}C_{jl}^{-1}}
329: \)
330: with $C_{ij}$ being the map-to-map covariance matrix.
331:
332: It is needed to inspect the effect of exposure dependent noise in the ILC map,
333: which is extensively used
334: in cosmological analysis, although the WMAP team warns against its use for
335: CMB studies because of the complex noise properties of this map \cite{lim03}.
336: The left panel of Fig.~5 shows the latitude distribution of rms deviation from
337: the WMAP3 ILC map. There exists a variation component monotonically decreasing from the
338: southern pole to the northern pole, expressed roughly by the hand-drawn
339: dotted line in the left panel of Fig.~5. The right panel shows the residual
340: rms deviations after subtracting the north-south asymmetry component.
341: From Fig.~5 we see that the procedure of linear combination
342: with respect to the covariance between different frequencies can effectively
343: depress the instrument noise. The latitude dependence of rms variation
344: in the ILC map is weaker, but still visible with a structure similar
345: to what observed in W, V and Q band.
346: We can not make modification for the ILC map by sky exposure like what we do
347: for Q, V and W data sets, because the ILC map is reconstructed by re-adding the data
348: at different frequencies, it is difficult to assign an observation number $N_i$ to a pixel $i$.
349: \begin{figure}[tb]
350: \label{f5}
351: % \vspace{-10mm}
352: \begin{center}
353: \psfig{figure=f5a.ps,width=45mm,height=60mm,angle=270}\hspace{4mm}
354: \psfig{figure=f5b.ps,width=45mm,height=60mm,angle=270}
355: \end{center}
356: %\vspace{-5mm}
357: \caption{{\sl Left}: RMS deviation of temperatures vs.
358: ecliptic latitude for the three-year ILC map.
359: The vertical coordinate is \tr, where \t~denotes temperature anisotropies (in mK)
360: at a certain latitude.
361: {\sl Right}: Residual rms deviations obtained from the curve shown in the left graph
362: by subtracting the north-south asymmetry component
363: (the dotted line in the left graph)}
364: \end{figure}
365:
366: \subsection{Pseudo sources}
367: \begin{figure}[tb]
368: \label{f6}
369: \vspace{-5.5cm}
370: \begin{center}
371: \includegraphics[width=14cm,angle=0]{f6.ps}
372: \end{center}
373: \vspace{-9cm}
374: \caption{Cross-correlation map in ecliptic coordinates.
375: Points show pixels with $c\ge 3\sigma_c$,
376: $\mathbf c$ -- cross-correlation of beam profile and sky temperature,
377: $\sigma_c$ -- standard deviation of $\mathbf c$. {\it Left panel}, from W4-band map.
378: {\it Right panel}, from modified W4-band map
379: with multiplying temperatures by $\frac{\sqrt{M_i}}{\sqrt{\<M\>}}$.
380: {\it Top panel}, from WMAP1 data. {\it Bottom panel}, from WMAP3 data.}
381: \end{figure}
382:
383: In studying the nature of discrete $\gamma$-ray sources discovered by the COS-B satellite
384: experiment, we found that quite a lot apparent discrete sources
385: revealed in the cross-correlation map with high significances (about half of sources in 2CG catalog)
386: are pseudo-sources produced by the fluctuation of
387: structured diffuse background of the galactic plane \cite{lit82}.
388: Here we use cross-correlation as another indicator to inspect
389: the possible large-scale anomaly caused by the unevenly distributed noise in WMAP maps.
390: We compute the cross-correlation function $\mathbf{c}$ of the
391: W4-band map \t~and beam profile $\mathbf{B}$ by $\mathbf{c=B^T\hat{t}}$,
392: and the standard deviation $\sigma_c$ of the correlation map. The pixels with
393: $|c|\ge 3\sigma_c$ in ecliptic
394: coordinates are shown in the left panel of Fig.~6. That the apparent features of point-like
395: sources concentrated along the ecliptic plane are just generated by observation effect
396: as this feature disappears after modifying WMAP temperatures by a factor
397: of $\frac{\sqrt{M_i}}{\sqrt{\<M\>}}$ for both one-year and three-year data,
398: as shown by the right column of Fig.~6. The result shown by Fig.~6 remind us that
399: one should take a care before to claim a finding of CMB anomaly from observation maps
400: before carefully considering the effect of exposure dependent noise.
401:
402: \subsection{North-south asymmetry}
403:
404: \begin{figure}[htb]
405: \label{f7}
406: %\vspace{0.5cm}
407: \begin{center}
408: \includegraphics[width=5cm,angle=270]{f7.ps}
409: \end{center}
410: \vspace{-0.1cm}
411: \caption{North-south asymmetry of rms deviation of temperatures in ILC map.
412: $\beta$ -- ecliptic latitude.
413: rms($\beta$)/rms($-\beta$) -- ratio between rms deviations
414: at the latitude $\beta$ in the northern hemisphere and at that in the southern hemisphere.}
415: \end{figure}
416:
417: \begin{figure}[htb]
418: \label{f8}
419: \vspace{-8mm}
420: \begin{center}
421: \psfig{figure=f8.ps,width=75mm,height=70mm,angle=0}
422: \vspace{-5mm}
423: \caption{Ecliptic latitude dependence of residual temperature rms fluctuations
424: of the foreground-cleaned WMAP3 V-band map.
425: The rms fluctuations are modified by using Eq.~\ref{rms}. The best-fit line
426: $y=-4.46\times10^{-5}x+0.085$.
427: }
428: \end{center}
429: \end{figure}
430:
431: A remarkable character in Fig.~5 is
432: the monotonically decline trend of rms variation from the south pole to the north pole
433: in the ILC map. To show the north-south asymmetry quantitatively, we calculate the asymmetry ratio
434: rms$(\beta)$/rms$(-\beta)$ for different latitude $\beta$, where rms$(\beta)$
435: is the average rms of sky pixels with latitude within $\beta\pm1.5^o$ in the northern hemisphere
436: and rms$(-\beta)$ is that in the south. The result is shown in Fig.~7.
437: From Fig.~7 we see that there exists a
438: systematic decrease of rms deviation from south to north with maximum asymmetry amplitude
439: $\sim 10\%-20\%$.
440: For different ecliptic latitudes we calculate residual rms from the WMAP3
441: V-band foreground-cleaned
442: map after modifying the exposure ununiformity by using Eq.~\ref{rms},
443: the result is shown in Fig.~8, where we also find a clear north-south asymmetry.
444:
445: \begin{figure}[tb]
446: \label{f9}
447: \begin{center}
448: \vspace{-5mm}
449: \includegraphics[width=4.5cm,angle=270]{f9a.ps}
450: \hspace{3mm}\includegraphics[width=4.5cm,angle=270]{f9b.ps}
451: \hspace{3mm}\includegraphics[width=4.5cm,angle=270]{f9c.ps}
452: \end{center}
453: \vspace{-0.1cm}
454: \caption{North-south asymmetry of WMAP3 exposure.
455: $N(\beta)$ is the total number of observations contributing to
456: the pixels with ecliptic latitude from $(\beta-1.5)$ deg to $(\beta+1.5)$ deg
457: in the northern hemisphere,
458: $N(-\beta)$ is that in the southern hemisphere.
459: {\it Left panel}, W1 band. {\it Middle panel}, V1 band. {\it Right panel}, Q1 band.}
460: \end{figure}
461:
462: To see if the north-south asymmetry also exists in WMAP exposure,
463: we calculate
464: the exposure asymmetry ratio $N(\beta)/N(-\beta)$ for W1, V1 and Q1 band separately, where
465: $N(\beta)$ is the three-year observation number contributing to the pixels
466: with latitude within $\beta\pm1.5^o$
467: in the northern hemisphere and $N(-\beta)$ is that in the south,
468: the results are shown in Fig.~9.
469: From Fig.~9 we see that there exists north-south asymmetry in WMAP sky survey:
470: exposure increases
471: from south to north with maximum amplitude $\sim 2\%-4\%$. It is natural to explain the rms asymmetry
472: by the exposure asymmetry as $\sim 10\%-20\%$ decreasing of rms deviation
473: is just expected by $\sim 2\%-4\%$ increasing of exposure, provided that the linear combination
474: can not suppress the common and systematic north-south asymmetry in multi frequencies
475: and the exposure dependent
476: noise follows Eq.~\ref{sigma}, rms $\propto 1/\sqrt{(N_i)}$.
477:
478: An apparent asymmetry in the distribution of fluctuation power in two opposing hemispheres
479: with a remarkable absence of power in the vicinity of the northern ecliptic pole is observed
480: from the WMAP1 Q, V, and W band sky maps \cite{eri04b, han04}.
481: By analyzing the issue of power asymmetry with the WMAP3 ILC map and a model
482: of an isotropic CMB sky modulated by a dipole field, Eriksen et al. (2007) find that
483: the modulation amplitude is $11.4\%$ and that the results on hemispherical power asymmetry
484: are not sensitive to data set or sky cut.
485: All of these features can be explained naturally by the effect of exposure dependent noise
486: as we discuss above in this section.
487:
488: \clearpage
489: \section{Foreground contamination}
490: Besides the statistical noise, another kind of error in recovered WMAP CMB maps
491: is systematic distortion.
492: From Eq.~\ref{mm-l} we can see that in an iteration the temperature estimation at a sky pixel $i$
493: is evaluated based on temperatures of many pixels on a circle in the sky sphere
494: $\theta_{beam}$ (beam separation) apart from the pixel $i$, including $N_i^-$ pixels
495: pointed by minus-horn
496: when plus-horn pointing to $i$ and $N_i^+$ pixels pointed by plus-horn
497: when minus-horn pointing to $i$. The sky temperature $t_{k^+}$
498: observed by plus-horn and $t_{k^-}$ by minus-horn are differently placed
499: in the right side of the iterative formula Eq.~\ref{mm-l}.
500: The iterative solution $\hat{t}_i$ could be deviated from
501: the true temperature $t_i$ due to inhomogeneity of temperature sky
502: and unbalance between two sky side beam measurements.
503:
504: \subsection{Map distortion by a hot source}
505: The foreground induced systematic effect on a recovered map is not
506: only limited in the region containing foreground sources, but spreated
507: over the whole sky.
508: A hot foreground source pointed by the side beam A of a radiometer can
509: distort the recovered temperatures of sky pixels pointed by the side beam B
510: with separation angle $\theta_{beam}$ to the source
511: through map-making iterations. From Eq.~\ref{mm-l} (or Eq.~\ref{mm-w}) the first
512: iterative solution of temperature of a sky pixel $i$ can be expressed as
513: \begin{equation}
514: \label{t1}
515: t^{(1)}_i=t_i-\<t-t^{(0)}\>_{ring} \end{equation}
516: where $\<~ \>_{ring}$ denotes averaging on the scan ring
517: with separation angle $\theta_{beam}$ to the pixel $i$.
518: For zero initials $t^{(0)}=0$
519: \begin{equation}
520: \label{t10}
521: t^{(1)}_i=t_i-\<t\>_{ring}~, \end{equation}
522: a hot source contained on the scan ring will let $\<t\>_{ring}>0$ and
523: the recovered temperature $t_i^{(1)}<0$, indicating that a hot foreground
524: source might systematically make the recovered temperatures on its scan ring lower.
525:
526: A one-dimensional simulation is made to show such effect.
527: A true temperature ``sky'' containing 500 pixels is produced
528: as a white noise series of zero mean and 0.2 mK standard deviation
529: with a hot source in the pixel interval 240-260, as shown in the upper panel of Fig.~10.
530: To produce differential data, we simulate a scan process of a differential
531: radiometer with instrument noise $\sigma_0=4.0$ mK and beam separation
532: $\theta_{beam}=100$ pixel.
533: From $i=1$ to 500, for each pixel $i$ we produce 2500 temperature differences
534: $t_i-t_j+n$ where $j=i+100$ for $i\le400$ or otherwise $j=i-400$,
535: $n$ is sampled from the normal distribution with zero mean and standard
536: deviation $\sigma_0$.
537: We use Eq.~\ref{mm-w} to reconstruct temperature map from the differential data.
538: For pixel $k=350$ the recovered temperatures from first fifteen iterations,
539: $t_k^{(1)}, t_k^{(2)}, \cdots, t_k^{(15)}$
540: are shown in Fig.~11, where we see
541: that the recovered temperature of first iteration for the pixel
542: $\theta_{beam}$ away from the hot pixel really drops down dramatically
543: as expected by Eq.~\ref{t1} and after a few iterations converged to a stable
544: value which still significantly lower than its real temperature.
545: The lower panel of Fig.~10 shows the recovered temperature map
546: after 50 iterations.
547: There exist two significant cold regions in the recovered map
548: with pixels $340-360$ and $140-160$,
549: each have a distance of $\theta_{beam}$ to the hot source interval
550: and their temperature variation negatively related to that of the hot source.
551: Two more regions distorted in the similar way
552: can be found in the recovered map, each
553: placed 100 pixel away from the above two cold regions respectively,
554: pixel intervals $440-460$ and $40-60$.
555: The above simulation result indicates that a hot source can distort
556: temperatures in the map recovered from
557: differential data at pixels away from the source with distance of
558: the beam separation and such distortion can propagate to more far away pixels.
559:
560: \begin{figure}[htb]
561: \label{f10}
562: % \vspace{-10mm}
563: \begin{center}
564: \psfig{figure=f10a.ps,width=45mm,height=70mm,angle=270}\\
565: \vspace{3mm}\psfig{figure=f10b.ps,width=45mm,height=70mm,angle=270}
566: \caption{True and recovered temperatures. {\sl Upper panel}:
567: True temperatures.
568: {\sl Lower panel}: Recovered temperatures.
569: }
570: \end{center}
571: \end{figure}
572:
573: \begin{figure}[htb]
574: \label{f11}
575: % \vspace{-10mm}
576: \begin{center}
577: \psfig{figure=f11.ps,width=45mm,height=70mm,angle=270}
578: \caption{Recovered temperature at pixel 350 vs. step number of iteration by
579: map-making Eq.~\ref{mm-w}. The one-dimensional true temperature map,
580: shown in the upper panel of Fig.~10, has a hot source with
581: a peak at pixel 250, and the differential data is simulated
582: for a radiometer with beam separation of 100 pixel.
583: }
584: \end{center}
585: \end{figure}
586:
587: \subsection{$141\degree$ rings in WMAP maps}
588: The beam separation angle $\theta_{beam}$ of WMAP radiometers is $141\degree$.
589: We predict from the above simulation that there should exist strongest negative
590: correlation in WMAP temperature maps between temperatures of two sky pixels
591: separated $141\degree$ each other. This prediction is confirmed
592: by analyzing temperature maps
593: in Q, V and W bands released by WMAP team.
594: For the WMAP3 Q-, V- and W-band maps with HEALPix resolution parameter
595: $N_{side}=128$ \cite{gor05}, we calculate cross-correlation coefficients
596: $C_{t_a\<T_b\>}(\theta)$ between $t_a$ and $\<t_b\>$ for different
597: separation angle $\theta$,
598: where $t_a$ denotes the temperature of a sky pixel $a$ and
599: $\<t_b\>$ the average temperature of the ring with separation angle $\theta$
600: to the pixel $a$, the obtained distributions are shown in Fig.~12.
601:
602: \begin{figure}[tb]
603: \label{f12}
604: %\vspace{-2cm}
605: \begin{center}
606: \vspace{-4mm}\psfig{figure=f12a.ps,width=65mm,height=42mm,angle=0}\\
607: \vspace{-4mm}
608: \psfig{figure=f12b.ps,width=65mm,height=42mm,angle=0}\\
609: \vspace{-4mm}
610: \psfig{figure=f12c.ps,width=65mm,height=42mm,angle=0}
611: \vspace{-5mm} \caption{Correlation coefficient $C_{t_a\<t_b\>}(\theta)$ between
612: temperatures $t_a$ and $\<t_b\>$ vs. separation
613: angle $\theta$. $t_a$ denotes the temperature of sky
614: pixel $a$,
615: $\<t_b\>$ the average temperature of the ring with separation angle $\theta$ to $a$.
616: {\it Filled square}: pixel $a$ only within the foreground mask Kp12.
617: {\it Square}: for whole sky.
618: {\it Triangle}: only for the sky region out of the mask Kp0.
619: {\sl Top panel}: WMAP3 Q-band.
620: {\sl Middle panel}: WMAP3 V-band.
621: {\sl Bottom panel}: WMAP3 W-band. }
622: \end{center}
623: \end{figure}
624:
625: From Fig.~12 we find that the strongest negative correlation appear
626: indeed around $141\degree$ separation in the correlation distribution
627: for each band.
628: If pixel $a$ is limited within the
629: sky region of the foreground mask Kp12 \cite{ben03b} which consists mainly of hot pixels,
630: around $141\degree$ separation show more strong negative correlation
631: just as expected from assuming that
632: hot foreground emission induce the correlation.
633:
634: The program synfast in HEALPix software package (available at
635: http://healpix.jpl.nasa.gov)
636: can create temperature maps computed as realizations
637: of random Gaussian fields on a sphere characterized by the user provided
638: spherical harmonic coefficients of a angular power spectrum.
639: To test the significance of the correlation $C(141\degree)=-0.238$
640: in Q-band WMAP map, simulated temperature maps are created with
641: the synfast program
642: from the best fit $\Lambda$CDM model power spectrum with Q-band
643: beam function and noise property. For each simulated CMB map,
644: we compute the correlation coefficient between temperatures of the pixels in Kp12 region
645: and average temperatures on their $141\degree$ rings. From 1000 simulated maps
646: we get $C^\prime(141\degree)=0.108\pm0.080$, indicating that the $141\degree$
647: negative correlation detected in WMAP Q-band map has a significance
648: of $4.3\sigma$. Similarly, for V-band we get $C(141\degree)=-0.231$
649: with significance $4.4\sigma$ evaluated by
650: $C^\prime(141\degree)=0.099\pm0.075$ from simulated CMB maps,
651: and for W-band $C(141\degree)=-0.268$ with significance $5.0\sigma$ from
652: $C^\prime(141\degree)=0.094\pm0.072$.
653:
654: \begin{figure}[p]
655: \label{f13}
656: \begin{center}
657: \vspace{-10mm}\includegraphics[width=40mm,angle=90]{f13a.ps}\\
658: \vspace{-2mm}\includegraphics[width=40mm,angle=90]{f13b.ps}\\
659: \vspace{-2mm}\includegraphics[width=40mm,angle=90]{f13c.ps}\\
660: \vspace{-2mm}\includegraphics[width=40mm,angle=90]{f13d.ps}
661: \caption{Average temperature maps of $141\degree$ ring in Galactic coordinates
662: from WMAP3 Q-band, V-band, W-band and ILC map (from top to bottom panel).
663: }
664: \end{center}
665: \end{figure}
666: It is expected from our simulation shown in Fig.~10 that foreground hot sources can
667: produce cold rings of $141\degree$ separation in WMAP temperatures.
668: To test this effect, for each sky pixel we calculate the average temperature
669: of its $141\degree$ ring from the WMAP3 Q-band, V-band, W-band and ILC temperature
670: maps with $N_{side}=128$ and Kp0 mask \cite{ben03b} for foreground clean.
671: The result is shown in Fig.~13. There is a low temperature region
672: near Galaxy center in Fig.~13 for each band
673: indicating that most of the $141\degree$ rings corresponding to the galaxy center region
674: are cold. This is consistent with our expectation.
675:
676: To roughly estimate the magnitude of temperature distortion by foreground emission,
677: we pick up 2000 hottest pixels in the 3-year WMAP Q-band map and then
678: find out their $141\degree$ rings. The average temperature of all pixels on these rings
679: and out of the Kp0 mask (cover about $14\%$ of the sky, see Fig.~14) is calculated to be
680: $\<t\>=-0.0117$ mK.
681: The correspondent value from 1000 simulated CMB maps is $\<t^\prime\>=3.4\times10^{-5}\pm0.006$ mK.
682: Similarly, for V-band we get $\<t\>=-0.0125$ mK and $\<t^\prime\>=-4.7\times10^{-5}\pm0.006$ mK,
683: for W-band $\<t\>=-0.0141$ mK and $\<t^\prime\>=8.8\times10^{-5}\pm0.006$ mK.
684: Therefore, in foreground-cleaned WMAP maps, considerable foreground-induced deviations
685: still exist with amplitude comparable to the fluctuation of CMB signal and
686: over broad sky region.
687:
688: \begin{figure}[htb]
689: \label{f14}
690: \begin{center}
691: \includegraphics[width=48mm,angle=90]{f14.ps}
692: \caption{$141\degree$ rings of 2000 hottest pixels in
693: 3-year Q-band WMAP map, after using the Kp0 mask, in Galactic coordinates.
694: }
695: \end{center}
696: \end{figure}
697:
698: The foreground-induced deviation in WMAP temperatures should also distort the temperature
699: angular power spectrum. We use simulated temperature maps created with
700: the synfast program to roughly estimate the magnitude of foreground-induced distortion
701: for the WMAP power spectrum on large-scale (low-$l$) region.
702: For each simulated map, we subtract 0.01 mK from each temperature of pixels
703: on the $141\degree$ rings corresponding to the 2000 hottest pixels (shown in Fig.~14)
704: to get a distorted map. We compute the power spectrum for simulated and
705: distorted map respectively and find that, on average, the deviation between the original
706: and distorted spectrum densities at $l=2, 3$ is $\sim 10\%$,
707: indicating that the foreground-induced distortion on WMAP power spectrum
708: cannot be ignored for a precision cosmology study.
709: Strongest hot sources may produce cold spots by the foreground induced distortion
710: out of the foreground region.
711: Temperature distortions in a CMB map caused by foreground sources with different
712: scales and observation dependent noise could distort the angular power spectrum
713: on wide range of angular scale.
714:
715: \subsection{Large non-Gaussian spots}
716: A large cold spots with a radius of $\sim 5\degree$
717: centered at $(l,b)=(209\degree, -57\degree)$ and the lowest
718: temperature $\sim -0.15$ mK has been
719: detected in WMAP1 and WMAP3 maps with the wavelet and other techniques
720: \cite{vie04,cru05,cru06,cru07a,vie07},
721: and was explained as a cosmic texture, a remnant of symmetry-breaking
722: at energies close to the Planck scale in the very early universe~\cite{cru07b}.
723: We find that the $141\degree$ rings of the pixels in the spot region
724: across through the hot Galactic plane with temperature up to $\sim 5$ mK,
725: therefore, the foreground induced systematic effect on recovered maps
726: discussed in \S3.1 should be a more plausible explanation to the detected large cold spot
727: than the texture hypothesis. The foreground contamination and other systematic
728: effect may also produce other more
729: detected large spots in WMAP maps.
730:
731: \section{Discussion}
732: \subsection{Can the exposure induced anisotropy be corrected ?}
733: It has to be pointed out that Eq.~\ref{rms} can be used to modify the effect of
734: exposure dependent noise, like what we do for the latitude distribution
735: of rms variation shown in Fig.~2, only for the case that the map rms
736: fluctuation $\<{\hat{t}}^2\>$ itself is the directly analyzed quantity.
737: However, almost all analysis works are based on CMB maps of temperature.
738: We know from Figs.~2-5 that the released WMAP temperature maps
739: contain considerable exposure dependent noise.
740: It is no way from a recovered temperature map to produce a corrected map
741: in which
742: the instrument-induced and exposure-dependent noise can be eliminated.
743: Therefore, the anisotropy noise should contribute to large scale anomalies
744: detected in existent CMB maps.
745:
746: The exposure inhomogeneity of WMAP comes from its scan strategy, which can't
747: be suppressed through accumulating observation time.
748: To avoid the observation effect, it is needed to remake temperature maps
749: from a uniform differential data set obtained by giving up partial observation data
750: for pixels of high exposure. Comparing released WMAP maps and new maps from uniform
751: data will help us to judge their origin of detected large scale anomalies,
752: e.g. the low $l$ power issues detected in WMAP data,
753: the unexplained orientation of large-scale patterns of CMB maps in respect
754: to the ecliptic frame, the north-south asymmetry of temperature fluctuation power
755: etc., and to see if the observational effect can also influence the angular
756: power spectrum as well.
757:
758: \subsection{Avoiding the foreground-induced distortion}
759: \begin{figure}[p]
760: \label{f15}
761: \vspace{-18mm}
762: \begin{center}
763: \psfig{figure=f15a.ps,width=45mm,height=70mm,angle=270}\\
764: \vspace{2mm}\psfig{figure=f15b.ps,width=45mm,height=70mm,angle=270}\\
765: \vspace{2mm}\psfig{figure=f15c.ps,width=45mm,height=70mm,angle=270}
766: \caption{Differences between recovered $t$ and true temperature $t_0$.
767: The one-dimensional temperature distribution $t_0$ consists of white
768: noise and hot source between 240 - 260 pixel (mask region), as shown
769: in the upper panel of Fig.~10. Differential data is obtained by simulation
770: of one-dimensional scan for $t_0$ by a differential radiometer
771: with beam-separation of 100 pixel. Recovered $t$ are calculated with map-making
772: Eq.~\ref{mm-w} after 50 iterations.
773: {\sl Upper panel}: $t$ recovered from all differential data (the lower panel of Fig.~10).
774: {\sl Middle panel}: $t$ recovered by iterations from all data for mask region
775: and with excluding differences that contain hot source temperature for pixels out of mask.
776: {\sl Bottom panel}: $t$ recovered by iterations with initials estimated
777: from the differential data by Eq.~\ref{tj}. }
778: \end{center}
779: \end{figure}
780:
781: The distortion by hot foreground sources on their $141\degree$ rings
782: in a WMAP map can not be removed with a foreground mask on the recovered temperature map.
783: What's needed is to use the mask on the original differential data before map-making
784: to avoid the foreground-induced error in the recovered map.
785: The top panel of Fig.~15 shows the difference between the recovered and
786: true temperature distributions (shown in the lower and upper panel of Fig.~10 respectively),
787: where the distortion structure caused by the hot source on pixel 240 - 260 and
788: beam separation of 100 pixel is clearly shown. We redo the temperature reconstruction
789: with excluding the temperature differences that contain a
790: beam side pointing to a pixel between 240 - 260 (``mask region'')
791: during iterations for the pixels out of mask, the result
792: is shown in the middle panel of Fig.~14, where the distortion structures
793: are really suppressed.
794:
795: A weakness of using mask in map-making process is decreasing the number of
796: useful differential data. Another approach to avoid the distortion in recovered map
797: by foreground emission is to properly set the initials of iteration for
798: the foreground region. From Eq.~\ref{t1} we see that the temperature deviation
799: of the first iterative solution, $t_i^{(1)}-t_i$, will be suppressed if the
800: temperature initials at pixels of hot source are set to be close to their true values
801: to let $\<t-t^{(0)}\>_{ring}\approx~0$.
802: The initial $t_i^{(0)}$ of pixel $i$ can be taken as
803: \begin{equation}
804: \label{tj}
805: t_i^{(0)}=\frac{1}{N_i^\prime}(\sum_{k^+=i}d_k-\sum_{k^-=i}d_k)~.
806: \end{equation}
807: Where $\sum_{k^+=i}$ means summing over the observations while the pixel
808: $i$ is observed by the plus-horn and the pixel pointed by the minus-horn
809: is out of mask, $\sum_{k^-=i}$ means summing over the
810: observations while the pixel $i$ is observed by the minus-horn and
811: the pixel pointed by the plus-horn is out of mask,
812: $N_i^\prime$ is the total number of used observations.
813: For the simulated differential data from the true temperatures shown in the upper
814: panel of Fig.~10, we make 50
815: iterations with Eq.~\ref{mm-w} starting from initials calculated by Eq.~\ref{tj},
816: the distortion structures are satisfactory suppressed
817: in the resultant solution, as shown in the bottom panel of Fig.~15.
818:
819: \subsection{Remaking WMAP maps}
820: We demonstrate in this paper that for existent CMB maps
821: both the observation dependent noise and
822: systematic error induced by foreground emission can not be
823: neglected and both can produce large-scale anomalies
824: and distort the angular power spectrum.
825: These errors can not be completely excluded by performing noise suppressing or
826: using foreground mask on temperature maps.
827: We suggest to remake temperature maps from the original WMAP time-order-data
828: by a modified algorithm with applying foreground mask in map-making
829: to exclude mask pixels from use
830: in iterations for CMB dominated region (or properly set
831: temperature initials before iteration),
832: and/or keeping
833: used differential data uniform by giving up partial observation data
834: for pixels of high exposure.
835: New maps from modified map-making algorithm
836: will help us to judge the origin of large scale anomalies detected in released
837: WMAP maps, e.g. the low $l$ power issues,
838: the unexplained orientation of large-scale patterns in respect
839: to the ecliptic frame, the north-south asymmetry of temperature fluctuation power
840: and the large non-Gaussian spots,
841: and to see to what extent the statistical and systematical errors
842: influence the angular power spectrum and the derived cosmological parameters.
843: Believable conclusions on CMB anisotropy anomalies and precise temperature
844: angular power spectrum from differential measurement
845: should be based on temperature maps with homogeneous sky exposure
846: and should avoid foreground-induced distortion during map-making.
847:
848: This study is supported by the National Natural
849: Science Foundation of China and the CAS project KJCX2-YW-T03.
850: The data analysis in this work made use of the
851: WMAP data archive and the HEALPIX software package.
852:
853: \begin{thebibliography}{}
854: \bibitem[Abramo et al. 2006]{abr06} Abramo L.R., Bernui A., Ferreira I.S.,
855: Villela T. \& Wuensche C.A., 2006, Phys. Rev. D, 74, 063560
856: \bibitem[Bennett et al. 2003a]{ben03a} Bennett, C.L. et~al., 2003c,
857: \apj, 583, 1
858: \bibitem[Bennett et al. 2003b]{ben03b} Bennett, C.L. et~al., 2003b,
859: \apjs, 148, 1
860: \bibitem[Bennett et al. 2003c]{ben03c} Bennett, C.L. et~al., 2003c,
861: \apjs, 148, 97
862: \bibitem[Copi et al. 2006]{cop06} Copi, C.J., Huterer, D.,
863: Schwarz, D.J. \& Starkman, G.D., 2006, \mnras, 367, 79
864: \bibitem[Copi et al. 2007]{cop07} Copi, C.J., Huterer, D.,
865: Schwarz, D.J. \& Starkman, G.D., 2007, Phys. Rev. D, 75,023507
866: \bibitem[Cruz et al. 2005]{cru05} Cruz M., Martinez-Gonzalez E.,
867: Vielva P. \& Cayon L., 2005, \mnras, 356, 29
868: \bibitem[Cruz et al. 2006]{cru06} Cruz M., Tucci M., Martinez-Gonzalez E.
869: \& Vielva P., 2006, \mnras, 369, 57
870: \bibitem[Cruz et al. 2007a]{cru07a}Cruz M., Cayon L., Martinez-Gonzalez E.
871: \& Vielva P., 2007a, \apj, 655, 11
872: \bibitem[Cruz et al. 2007b]{cru07b} Cruz, M., Turok, N., Vielva,P., Martinez-Gonzalez E.
873: \& Hobson M., {\sl Science}, 2007b, 318, 1612
874: \bibitem[de Oliveira-Costa et~al. 2004]{cos04}de Oliveira-Costa, A.,
875: Tegmark, M., Zaldarriaga. \& Hamilton A., 2004, Phys. Rev. D, 69, 063516
876: \bibitem[Eriksen et~al. 2004a]{eri04a} Eriksen H.K., Hansen, F.K,
877: Banday, A.J., Grski, K.M. \& Lilje, P.B., 2004a, \apj, 605, 14
878: \bibitem[Eriksen et~al. 2004b]{eri04b} Eriksen H.K., Banday A.J.,
879: Gorski K.M. \& Lilje, P.B., 2004b, \apj, 612, 633
880: \bibitem[Eriksen et~al. 2007]{eri07} Eriksen H.K.,
881: Banday, A.J., Grski, K.M. Hansen, F.K \& Lilje, P.B., 2007, \apj, 660, L81
882: \bibitem[Gorski et al. 2005]{gor05} Gorski, K. M., Hivon, E., Banday, A. J.,
883: Wandelt, B. D., Hansen, F. K., Reinecke, M., Bartelmann, M., 2005, \apj, 622, 759
884: \bibitem[Hansen et~al. 2004]{han04} Hansen F.K., Banday A.J.
885: \& Gorski K.M., 2004, \mnras, 354, 641
886: \bibitem[Hinshaw et al. 2003]{hin03} Hinshaw, G. et~al., 2003, \apjs, 148, 63
887: \bibitem[Hinshaw et al. 2007]{hin07} Hinshaw, G. et~al., 2007, \apjs, 170, 288
888: \bibitem[Jaffe et al. 2005]{jaf05} Jaffe T.R., Banday A.J., Eriksen H.K.,Gorski K.M.
889: \& Hansen F.K., 2005, \apj, 629, L1
890: \bibitem[Jaffe et al. 2006]{jaf06} Jaffe T.R., Banday A.J., Eriksen H.K.,Gorski K.M.
891: \& Hansen F.K., 2006, A\&A, 460, 393
892: \bibitem[Land \& Magueijo 2007]{lan07} Land K. \& Magueijo J., 2007, \mnras, 378, 153
893: \bibitem[Li \& Wolfendale 1982]{lit82} Li T.P. \& Wolfendale A.W., 1982, A\&A, 116, 95
894: \bibitem[Limon et al. 2003]{lim03} Limon, M. et~al., 2003,
895: {\sl Wilkinson Microwave Anisotropy Probe (WMAP) : Explanatory Supplement},
896: Greenbelt, MD: NASA/GSFC;
897: Available in electronic form at http://lambda.gsfc.nasa.gov/
898: \bibitem[Park et al. 2007]{par07} Park C.G., Park C. \& Gott J.R., 2007, \apj, 660, 959
899: \bibitem[Samal et al. 2008]{sam08} Samal, P.K., Saha, R.,
900: Jian, P. \& Ralston, J.P., 2008, \mnras, 385, 1718
901: \bibitem[Schwarz et al. 2004]{sch04} Schwarz, D.J., Starkman, G.D.,
902: Huterer, D. \& Copi, C.J., 2004, Phys. Rev. Lett., 93, 221301
903: \bibitem[Spergel et al. 2006]{sper06} Spergel D. N. et al., 2006, arXiv:astro-ph/0603449v1
904: \bibitem[Starkman and Schwarz 2005]{sta05} Starkman, G.D. \& Schwarz, D.J.,
905: 2005, Sci. Am., 293(2), 48
906: \bibitem[Smoot et al. 1990]{smo90} Smoot, G.F. et~al., 1990,
907: \apj, 360, 685
908: \bibitem[Smoot et al. 1991]{smo91} Smoot, G.F. et~al., 1991,
909: \apj, 371, L1
910: \bibitem[Tegmark et al. 2003]{teg03} Tegmark, M., de Oliveira-Costa, A. \& Hamilton, A.J.,
911: 2003, Phys. Rev. D, 68, 123523
912: \bibitem[Vielva et al. 2004]{vie04} Vielva P., Martinez-Gonzalez E.,
913: Barreiro R. B., Sanz J. L., Cayon L., 2004, \apj, 609, 22
914: \bibitem[Vielva et al. 2007]{vie07} Vielva P., Wiaux Y., Martinez-Gonzalez E. \&
915: Vandergheynst P., 2007, \mnras, 381, 932
916:
917: \end{thebibliography}
918: \end{document}
919: