1: \documentclass[12pt,preprint]{aastex}
2: %\documentclass[manuscript]{aastex}
3: \usepackage{graphicx,epsfig,epsf,amssymb} %for PS/EPS graphics inclusion, new
4: \newcommand{\vdag}{(v)^\dagger}
5: \newcommand{\myemail}{litp@tsinghua.edu.cn}
6: \newcommand{\liuemail}{liuhao@ihep.ac.cn}
7: \def \<{\langle}
8: \def \>{\rangle}
9:
10: \def \tr{$\sqrt{\<\hat{t}^2\>}$}
11: \def \t{$\mathbf{\hat{t}}$}
12: \newcommand{\degree}{^\circ}
13:
14:
15: %\shorttitle{Errors in CMB maps}
16: %\shortauthors{Liu \& Li}
17:
18: \begin{document}
19:
20: \title{Systematic Distortion in Cosmic Microwave Background Maps}
21: \author{Hao LIU\altaffilmark{1,3} \& 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; \liuemail}
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: \begin{abstract}
29: To minimize instrumentally induced systematic errors,
30: cosmic microwave background (CMB) anisotropy experiments measure
31: temperature differences across the sky using paires of horn antennas,
32: temperature map is recovered from temperature
33: differences obtained in sky survey through a map-making procedure.
34: To inspect and calibrate residual systematic errors in recovered temperature maps
35: is important as
36: most previous studies of cosmology are based on these maps.
37: By analyzing pixel-ring couping and latitude dependence of CMB temperatures,
38: we find notable systematic deviation from CMB Gaussianity
39: in released Wilkinson Microwave Anisotropy Probe (WMAP) maps.
40: The detected deviation can not be explained by the best-fit $\Lambda$CDM
41: cosmological model at a confidence level above $99\%$
42: and can not be ignored for a precision cosmology study.
43: \end{abstract}
44:
45: \keywords{Cosmology: cosmic microwave background - Methods: data analysis}
46:
47:
48: \section{Introduction: differential observation and scan-ring}
49: To minimize instrumentally induced systematic errors,
50: the WMAP mission measures temperature differences across
51: the sky using paires of horn antennas with a fixed separation angle $\theta$,
52: temperature maps are recovered from raw time-ordered temperature
53: differences obtained in sky survey with a map-making algorithm$^{[1]}$.
54: When an antenna points to a sky pixel $i$, the scan path of the other one will
55: draw a ring $R_\theta(i)$ in the sky with angular radius $\theta$
56: to the center pixel $i$.
57: The map temperature $t(i)$ of a pixel $i$
58: should be in some extent correlated to measured temperatures in its scan-ring
59: $R_\theta(i)$ through the map-making procedure.
60: The beam separation angle of WMAP radiometers is $\theta\sim141\degree$.
61: The off-diagonal terms of noise covariance matrixes have been inspected
62: by the WMAP team with two-point correlation functions and small positive blips
63: of order $0.3\%$ of the diagonal elements have been found
64: at $141\degree$ pixel separation angle$^{[2,3]}$.
65:
66: Instead of the pixel-pixel noise coupling, in this work we inspect the
67: pixel-ring temperature coupling,
68: and find in released WMAP CMB maps that notable temperature distortions
69: exist on scan-rings of hot foreground sources: scan-rings of hot sources
70: are significantly cooled (presented in \S2) and strongest anti-correlations
71: between pixel and scan-ring temperatures appear at a separation angle
72: $\theta\sim141\degree$ (\S3).
73: In \S4 we find a systematic deviation from CMB isotropy in
74: Galactic latitude distributions of released WMAP maps, which
75: can be understood by pixel-ring temperature coupling found in \S2 and \S3.
76: Finally we give a brief discussion in \S5.
77:
78: \section{Cold scan-rings of hot sources}
79: We choose 2000 brightest pixels that are close to the galactic plane
80: from five year WMAP (WMAP5) intensity maps$^{[4]}$ with HEALPix resolution parameter
81: $N_{side}=512$$^{[5]}$ and almost the same for different bands,
82: and draw their scan-rings of $0.5\degree$ width, shown
83: in Fig.~\ref{rings}. The Kp0 mask is used to exclude bright galactic sources.
84: All the scan-rings in Fig. \ref{rings} form a sample (shortened as
85: ``sample 1'' henceforth) that covers about $15\%$
86: of the entire celestial sphere. Generally speaking, sample 1 is large enough
87: to merge most CMB anisotropy and the expectation of average temperature should be very close
88: to zero.
89: \begin{figure}
90: \label{rings}
91: \begin{center}
92: \includegraphics[width=46mm,angle=90]{f1.ps}
93: \vspace{-3mm} \caption{\footnotesize The scan-rings of 2000 hottest pixels in
94: the Q-band WMAP5 map in Galactic coordinates after using the Kp0 mask.
95: The ring width is $0.5\degree$.
96: }
97: \end{center}
98: \vspace{-4mm}
99: \end{figure}
100:
101: For the WMAP5 Q-, V- and W- band foreground-reduced maps and ILC map$^{[6]}$
102: we calculate the average temperature of sample 1 (``ring temperature'' hereafter)
103: respectively.
104: The results are listed in Table~1.
105: To estimate their significances we simulate observed CMB temperature maps
106: with the synfast program in HEALPix software package (available at
107: http://healpix.jpl.nasa.gov) from the best fit $\Lambda$CDM model
108: power spectrum with proper beam function
109: and noise property. For each input map, 1000 simulated maps are created and their
110: ring temperatures calculated. The average ring temperature and its standard deviation
111: is shown in the bottom row of Table~1 for each input map.
112: From this table we can see that, for the four maps the ring temperatures
113: are all notably lower than the expectation with a confidence level higher than $99\%$.
114: With these results, it is obvious that map temperatures on scan-rings of hot sources
115: suffer systematical distortion, and the observed violence to CMB Gaussianity
116: should come, at least partially, from the combined effect of hot foreground sources
117: and WMAP's differential nature.
118:
119: We have also rotated WMAP5 ILC temperature map (by rotating the north pole to other pixels)
120: to place sample 1 to other locations and compute the new average temperatures.
121: The original sample 1 is known to be abnormal; therefore it is excluded together
122: with the KP0 foreground mask. The objective of this test is to check the
123: variances given in Table~1 with the original map, and the result should be close
124: to simulation; otherwise the simulation is suspicious. The north pole of WMAP5 ILC
125: map is rotated to all 3072 sky pixels in HEALPix
126: resolution $N_{side}=16$ and the expectation and RMS fluctuation
127: of the average temperature of sample 1 are -0.5 $\mu$K and 4.3 $\mu$K respectively.
128: This is consistent with Table~1; therefore, the simulation and
129: Table~1 are reliable.
130:
131:
132: \begin{table}
133: \begin{center}
134: \caption{Temperatures ($\mu$K) averaged over scan-rings of hot sources}
135: \vspace{1mm}
136: \begin{tabular}{ l l l l l } \hline
137: & Q-band & V-band & W-band & ILC map \\ \cline{2-5}
138: WMAP5 & -11.67 & -12.62 & -12.92 & -11.34 \\
139: Simulation & $0.09\pm 4.7$ & $-0.09\pm4.7$ & $0.09\pm4.8$ & $0.04\pm4.4$\\
140: Significance& $2.5\sigma$ &$2.7\sigma$& $2.7\sigma$&$2.7\sigma$\\
141: \hline
142: \end{tabular}
143: \end{center}
144: \end{table}
145:
146: \section{Pixel-ring temperature coupling}
147: We use $t_R(i,\theta)$ to denote the average temperature over the scan ring
148: of a pixel $i$ with a separation angle $\theta$, i.e.
149: $t_R(i,\theta)=\<t(k)\>$ where $k\in R_\theta(i)$.
150: For a temperature map with
151: $N_{side}=128$, we calculate the cross-correlation coefficient
152: $C(\theta)$ between $t(i)$ and $t_R(i,\theta)$
153: for a designed separation angle $\theta$
154: \[ C(\theta)=\frac{\sum_i(t(i)-\overline{t})(t_R(i,\theta)-\overline{t}_R)}
155: {\sqrt{\sum_i(t(i)-\overline{t})^2(t_R(i,\theta)-\overline{t}_R)^2}}~.\]
156:
157: \begin{figure}
158: \label{correlation}
159: %\vspace{2cm}
160: \begin{center}
161: \vspace{-1mm}\includegraphics[height=4.5cm, width=6.5cm, angle=0]{f2a.ps}\\
162: \vspace{0mm}
163: \includegraphics[height=4.5cm, width=6.5cm, angle=0]{f2b.ps}\\
164: \vspace{0mm}
165: \hspace{1mm}\includegraphics[height=4.5cm, width=6.5cm, angle=0]{f2c.ps}
166: \vspace{-3mm} \caption{\footnotesize Separation angle dependence of pixel-ring
167: temperature correlation. The vertical coordinate shows the correlation coefficient
168: between temperature $t(i)$ of pixel $i$ and average temperature $t_R(i,\theta)$
169: of the ring with separation angle $\theta$ to $i$. The abscissa marks
170: the separation angle $\theta$ in degree.
171: {\it Filled square}: pixel $i$ only within the foreground mask Kp12.
172: {\it Square}: for whole sky.
173: {\it Triangle}: for the sky region out of the mask Kp0.
174: {\sl Top panel}: WMAP5 Q-band.
175: {\sl Middle panel}: WMAP5 V-band.
176: {\sl Bottom panel}: WMAP5 W-band. }
177: \end{center}
178: \vspace{-4mm}
179: \end{figure}
180:
181: First, we limit pixel $i$ only within the region of foreground
182: mask Kp12 where contain hottest foreground sources and calculate
183: $C(\theta)$ at different $\theta$ and for
184: WMAP5 Q, V and W band maps separately. The obtained correlation distributions
185: are shown by filled squares in Fig.~2, where the strongest negative correlation
186: appear around $141\degree$ separation for each band.
187: The separation angle dependence of pixel-ring
188: temperature correlation is also obtained from the WMAP5 ILC map, which is
189: similar with what from the WMAP5 Q, V and W band maps shown in Fig.~2.
190: The correlation coefficients at $141\degree$ separation, $C(\theta=141\degree)$,
191: from WMAP5 Q-, V- and W-band maps
192: are listed in Table~2. The bottom row of Table~2 comes from 1000 simulations
193: for each band. The foreground emission maps$^{[6]}$
194: have to be used in the simulation calculation because we need the original recovered
195: temperatures (without foreground reduction) for $t(i)$, which is the reason that we do not
196: give the result for ILC map in Table~2 (ILC
197: map does not have corresponding foreground emission map).
198: From Table~2 we can see that the negative correlation
199: at $\theta=141\degree$ has a significance from $2.2\sigma$ to $3.9\sigma$.
200:
201: \begin{table}
202: \begin{center}
203: \caption{Correlation coefficients
204: between temperatures of pixels
205: within Kp12 mask and average temperatures on their $141\degree$ scan-rings}
206: %\begin{minipage}{1cm}
207: \vspace{2mm}
208: \begin{tabular}{ l l l l } \hline
209: & Q-band & V-band & W-band \\ \cline{2-4}
210: WMAP5 & -0.234 & -0.223 & -0.262 \\
211: Simulation & $0.001\pm 0.106$ & $-0.001\pm0.101$ & $-0.002\pm 0.066$ \\
212: Significance & $2.2\sigma$ & $2.2\sigma$ &$3.9\sigma$\\
213: \hline
214: \end{tabular}
215: \end{center}
216: %\end{minipage}
217: \end{table}
218:
219: In the sky region out of the Galaxy plane (out of Kp0 mask) the correlation dip
220: around the $141\degree$ separation almost completely disappear (shown by triangles
221: in Fig.~2), indicating again that the detected correlation structure
222: is most possibly a combined effect of WMAP differential observation and Galactic hot emission.
223: The broad feature of the correlation dip around the $141\degree$ separation
224: can be caused by finite width and noncircular response of instrument beam,
225: structure of hot emission regions, and diffusion of temperature distortion
226: in map-making process.
227: The anti-correlation between temperatures of Galactic plane and its
228: $141\degree$ scan-rings is consistent with the detected fact that
229: $141\degree$ scan-rings of hot sources being cold (shown in \S2).
230:
231: Since the noise in WMAP temperature maps is known to be correlated
232: at $141\degree$$^{[2,3]}$, it is necessary to check
233: pixel-ring correlation with the same noise maps they used
234: (e.g., Q1-Q2, V1-V2). The pixel-ring correlation inside KP12 mask
235: is computed for both bands V1 \& V2, and compared with the result
236: from noise map (V1-V2). As we expect, the V1 and V2 results are very
237: close to the V-band result in Fig. 2;
238: however, in the noise map result, the pixel-ring correlation coefficients
239: go to nearly 0. Therefore, the pixel-ring correlation is unaffected
240: by known $141\degree$ noise correlation.
241:
242: \section{Latitude dependence of CMB temperatures}
243: The four graphs in Fig.~3 show the Galactic latitude dependence of average temperature $\<t\>$
244: from the foreground cleaned WMAP5 Q-, V-, W-band maps and ILC map respectively,
245: where for $|b|=85\degree$,
246: $\<t\>$ is calculated over the two regions $\pm(80\degree, 90\degree)$,
247: and for $|b|<80\degree$ over $\pm(|b|-2.5\degree, |b|+2.5\degree)$.
248: From Fig.~3 we can see systematic distortions existing in WMAP foreground cleaned
249: maps evidently.
250: The relationships of average temperature vs. Galactic latitude
251: in the four WMAP maps are highly consistent: all maps have $\<t\><0$ for the nine
252: latitude intervals in $10\degree<|b|<55\degree$
253: (listed in Table~3), and, contrarily, $\<t\>>0$
254: for the region of $|b|>55\degree$.
255: That the turning at $|b|\sim 55\degree$ and
256: the negativity of WMAP CMB maps in $|b|<55\degree$ can be understood
257: by the detected cooling effect of hot sources on their scan-rings
258: (shown in \S2 ans \S3)
259: with the fact that most of $141\degree$ rings of Galactic hot sources
260: are contained in the region of $|b|<55\degree$ as shown by Fig.~1.
261:
262: \begin{figure}
263: \label{f3}
264: \begin{center}
265: \includegraphics[width=35mm,angle=270]{f3a.ps}
266: \hspace{2mm} \includegraphics[width=35mm,angle=270]{f3b.ps}\\
267: \vspace{3mm}
268: \includegraphics[width=35mm,angle=270]{f3c.ps}
269: \hspace{2mm} \includegraphics[width=35mm,angle=270]{f3d.ps}
270: \caption{\footnotesize Average temperature vs. absolute Galactic latitude
271: in WMAP5 maps with the Kp0 mask, from top to bottom for Q, V, W band
272: and ILC map respectively. The standard deviations of average temperatures are shown
273: in the graphs, they are small and hard to be recognized.
274: }
275: \end{center}
276: \end{figure}
277:
278: \begin{table}
279: %\onecolumn
280: \caption{Average temperatures $\<t\>$ ($\mu$K) over
281: Galactic latitude regions $|b|\pm2.5\degree$}
282: \begin{center}
283: \begin{tabular}{llllllllll}
284: \hline
285: $|b|$ (deg) & 12.5 & 17.5& 22.5& 27.5& 32.5& 37.5& 42.5& 47.5&52.5\\
286: \hline
287: Q band & -3.4 & -1.2& -4.2& -4.0 & -5.7& -6.3& -7.6& -4.2& -3.8\\
288: V band & -4.4& -1.9& -5.2& -4.3& -5.2& -6.3& -7.8& -4.1& -3.5\\
289: W band & -4.9& -2.6& -5.7 & -4.5& -6.7& -6.1& -7.7& -4.0& -3.1\\
290: ILC map & -3.8 & -1.8 & -4.9 & -3.8 & -5.0 & -5.0 & -6.7 & -3.0 & -1.7 \\
291: \hline
292: \end{tabular}
293: \end{center}
294: \end{table}
295:
296: \begin{table}
297: \begin{center}
298: \caption{Average temperatures of foreground-cleaned
299: WMAP5 maps over $10\degree < |b| < 55\degree$} %\\[1ex]
300: %\begin{minipage}{7.5cm}
301: \vspace{3mm}
302: \begin{tabular}{ l l l l l} \hline
303: Map & Q-band & V-band & W-band & ILC \\ \cline{1-5}
304: $\<t\>$ ($\mu$K) & -4.6 & -5.1 & -5.3 & -4.28\\
305: Simulation& $0.05\pm2.1$ &$0.03\pm2.2$ &$-0.02\pm2.4$ & $0.01\pm2.0$\\
306: Significance & $ 2.2\sigma$ &$ 2.3\sigma$ & $ 2.2\sigma$ & $ 2.1\sigma$ \\
307: \hline
308: \end{tabular}
309: %\end{minipage}
310: \end{center}
311: \vspace{-2mm}
312: \end{table}
313:
314: \section{Discussion}
315: The magnitude of temperature distortion of the hottest Galactic sources upon their scan-rings can
316: be estimated to be $>10\mu$K from Table~1 (the corresponding region covers about $15\%$ of the
317: entire celestial sphere after removing KP0 mask region), and the average distortion for the region
318: of $|b|<55\degree$ to be $\sim 5\mu$K from Table~4 (the corresponding region covers $>80\%$ of the
319: entire celestial sphere or $>60\%$ after removing KP0 mask region). Such a wide-spread effect
320: with considerable strength can not be ignored for a precise cosmology study and need to be further
321: investigated.
322:
323: The detected large scale distortion from CMB Gaussianity is closely connected with
324: the Galactic hot sources and the WMAP beam separation angle, which is hard to explain by
325: the best-fit $\Lambda$CDM cosmological model
326: and most possibly comes from foreground effect and map-making.
327: To mitigate errors produced by effects from the hot foreground sources, the WMAP team used the Kp8
328: mask as a "processing mask" during the map-making process$^{[2,7]}$. However, the
329: detected temperature coupling between hot sources and their scan-rings indicates that the
330: Kp8 mask might be not wide enough to exclude all unwanted effect in the map-making process,
331: a broader mask, e.g. Kp0, might be better.
332: The future CMB observation project Planck is
333: designed to measure the CMB anisotropy with completely different mode to WMAP,
334: therefore, it is expected to be totally unaffected by such distortion.
335:
336: \begin{acknowledgements}
337: This study is supported by the National Natural
338: Science Foundation of China (Grant No. 10533020),
339: the National Basic Research Program of China
340: (Grant No. 2009CB-824800), and the Directional Research Project
341: of the Chinese Academy of Sciences (Grant No. KJCX2-YW-T03).
342: The data analysis made use of the WMAP data archive and the
343: HEALPix software packages.
344: \end{acknowledgements}
345:
346: \begin{thebibliography} {}
347: \bibitem{ben03}1. Bennett, C. et~al. 2003, ApJ, 583, 1
348: \bibitem{hin03}2. Hinshaw, G. et~al. 2003, \apjs, 148, 63
349: \bibitem{eri05}3. Eriksen, H. et al. 2005, \apj, 622, 58
350: \bibitem{hin08}4. Hinshaw, G. et~al. 2008, arXiv: 0803.0732
351: \bibitem{gor05}5. Gorski, K. et~al. 2005, \apj, 622, 759
352: \bibitem{gol08}6. Gold, B. et~al. 2008, arXiv: 0803.0715
353: \bibitem{jar07}7. Jarosik, N. et~al. 2007, \apjs, 170, 263
354: \end{thebibliography}
355: \end{document}
356: