astro-ph0311269/ms.tex
1: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
2: % 
3: % 
4: %
5: %        Jounghun Lee \& Naoki Yoshida
6: %
7: %                              Time-stamp: <03/11/12 16:17:05 lee>
8: %
9: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
10: \documentclass[12pt,preprint]{aastex}
11: %\usepackage{emulateapj}
12: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
13: \newcommand{\beq}{\begin{equation}}
14: \newcommand{\eeq}{\end{equation}}
15: \newcommand{\ben}{\begin{eqnarray}}
16: \newcommand{\een}{\end{eqnarray}}
17: \newcommand{\bth}{\vec{\theta}}
18: \newcommand{\hbth}{\hat{\theta}}
19: \newcommand{\bu}{{\bf u}}
20: \def\comment#1{\par\noindent\llap{$\Rightarrow$\enskip}{\bf #1}\par}
21: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
22: \begin{document}
23: \title{Reconstructing the Peak Distribution in the Sunyaev-Zeldovich
24: Effect Surveys} 
25: \author{\sc Jounghun Lee\altaffilmark{1} and
26: Naoki Yoshida\altaffilmark{2}}
27: \altaffiltext{1}{Department of Physics, The University of Tokyo, Tokyo 
28: 113-0033, Japan ; lee@utap.phys.s.u-tokyo.ac.jp}
29: \altaffiltext{2}{National Astronomical Observatory Japan, Mitaka, 
30: Tokyo 181-8588, Japan ; naoki@th.nao.ac.jp}
31: 
32: %
33: \received{2003 Nov. 1}
34: \accepted{2003 ???}
35: %
36: \begin{abstract} 
37: We examine the ability of the Sunyaev-Zel'dovich effect (SZE) 
38: statistics proposed by Lee (2002) using numerical simulations.
39: The statistics describe the distribution of the peak heights
40: in the noise-contaminated SZE sky maps, and provide an analysis
41: technique to sort out the noise-contamination effect and estimate 
42: the number density of real clusters. 
43: The method is devised to be suitable for the interferometric 
44: SZE observations in drift-scanning mode like AMiBA experiment. 
45: We apply the proposed method to a set of realistic SZE sky maps 
46: constructed from large-scale cosmological simulations, and show that 
47: the method indeed allows us to estimate the number density of clusters 
48: efficiently. The efficiency of the method is demonstrated in two 
49: aspects: (i) it can reconstruct the number density of 
50: clusters even though the clusters contribute only a small fraction 
51: ($\sim 10 \%$) of the total number of peaks in each SZE map; 
52: (ii) it can count even those clusters whose amplitudes are low enough 
53: to be compatible with the noise level while conventional method of 
54: cluster identification using the threshold cutoff cannot count them 
55: properly within the same observation time.  
56: Thus, the proposed method may be useful for the study of 
57: cluster abundance exploiting future interferometric SZE surveys 
58: such as AMiBA.
59: \end{abstract} 
60: \keywords{cosmology: theory  --- SZ effect: clusters}
61: 
62: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
63: \section{INTRODUCTION}
64: 
65: The thermal Sunyaev-Zel'dovich effect (SZE) by galaxy clusters represents 
66: the spectral distortion of the cosmic microwave background radiation 
67: (CMBR) through the inverse Compton scattering by the hot intra-cluster 
68: electrons \citep{sun-zel72}.  Since the SZE is independent of the redshift, 
69: it provides a unique tool to detect high-redshift clusters. Thus, the 
70: blank-field SZE survey can measure the formation epoch and the abundance
71: evolution of galaxy clusters, which can be used as a probe to put constraints 
72: on the cosmological parameters 
73: \citep{bar-etal96,hen97,bah-fan98,via-lid99,bor-etal99,fan-chi01,
74: gre-etal01,mol-etal02}. 
75: 
76: The SZE flux diminution is directly proportional to the cluster 
77: Comptonization parameter $y$, defined as 
78: \beq
79: y(\bth) = \frac{\sigma_{T}k_{B}}{m_{e}c^{2}}
80: \int n_{e}(l\hbth)T_{e}(l\hbth)dl, 
81: \label{eqn:compton} 
82: \eeq 
83: where $\bth$ represents the location vector along the line-of-sight 
84: ($\hbth \equiv \bth/\vert \bth\vert$) and the integral is over the 
85: CMBR path in the line-of-sight direction. Hence the SZE survey at a fixed 
86: frequency basically provides a map of the $y$-parameter. The number
87: density of galaxy clusters, however, cannot be directly measured from
88: the observed SZE maps because of the presence of noise. 
89: 
90: Conventionally, signals from the noise-contaminated maps are 
91: identified as the local peaks whose flux amplitudes (for the SZE
92: observation, proportional to the peak height of the Comptonization
93: parameter, $y$) 
94: significantly exceed the noise level. Therefore, to detect lower-amplitude 
95: signals using the conventional approach, one has to decrease the noise 
96: level by increasing the observational time. Unfortunately, the necessary 
97: observational time increases inversely proportional to the squares of the 
98: imposed noise level. It is clearly inefficient to use the conventional 
99: signal-selection method for the estimate of the cluster number counts,  
100: because of the high costs of the interferometric observations.  
101: 
102: Recently \citet{lee02} proposed a statistical strategy to estimate efficiently
103: the number density of galaxy clusters from the noise-contaminated SZE 
104: maps constructed from the interferometric observation in drift-scanning
105: mode, with the AMiBA experiment as a target model.  
106: By {\it the efficient estimate},  we mean that the method is
107: capable of estimating the number density of clusters which contribute 
108: only a small fraction of the total number density of peaks in the 
109: SZE maps, counting properly even those clusters whose peak heights 
110: are low enough to be compatible with the noise level without increasing 
111: the observational time. 
112: The key feature of the method is to exploit the property that the 
113: noise-contamination is described by a Gaussian statistic.
114: Then the cluster number density can be estimated by simply deconvolving 
115: a one-dimensional Gaussian function and the zero-th order approximation 
116: to the cluster number density.  Although the method has been shown to 
117: render reasonable results for idealized Monte-Carlo simulations 
118: \citep{lee02}, it is necessary to test its ability by applying 
119: it to more realistic simulations before using it in real SZE surveys.  
120: 
121: In this {\it Letter}, we test the ability of the SZE statistical strategy 
122: using the outputs of a large cosmological simulation. 
123: In $\S 2$, we provide a concise overview of the model experiment AMiBA
124: and the SZE statistics. In $\S 3$, we describe the numerical simulation, 
125: explain how to apply the statistical strategy to the simulated SZE maps, 
126: and compare the reconstructed cluster number densities with the original 
127: ones. In $\S 4$, we summarize the results, and give concluding remarks.
128: 
129: 
130: \section{The SZE CLUSTER SURVEY}
131: 
132: \subsection{AMiBA in a Drift-Scan Mode}
133: 
134: For the blank-field SZE survey,  we have two types of observations: 
135: the single dish bolometric observations such as BOLOCAM 
136: \citep{gle-etal98} and the multiple dish interferometric 
137: observations such as our model experiment, AMiBA \citep{lo-etal00}. 
138: AMiBA (Array for Microwave Background Anisotropy) is an interferometer 
139: telescope with $19$ dishes of $1.2$ meter,  dedicated to SZE 
140: observation at operating frequency of $90 \pm 16$ GHz. 
141: It is designed to be most suitable for a deep survey of low-mass
142: clusters at high redshifts. For this purpose, it has relatively small 
143: beam of $2$ arcminute, and plans to survey $5$ square degree for 20
144: hours per field  on average (the total period is over $7$ months), 
145: expected to be able to detect $1$ cluster per every $9$ hour.  
146: 
147: An observational strategy of AMiBA is to use the drift-scan method,  
148: which will help keep the ground fringes constant with time, improve the 
149: Fourier-space resolution, and most importantly simplify the noise
150: analysis \citep{pen-etal02}. A drift-scanned sky map from AMiBA experiment 
151: will be a combination of a small fraction of cluster signals 
152: ($\sim 10\%$) and a large fraction of dominant instrumental noise.  
153: The distribution of cluster signals is non-Gaussian, while that of the
154: instrumental noise is Gaussian with no spatial correlation. 
155: Since the statistical property of a Gaussian field is well known, 
156: the noise analysis for the SZE sky maps from AMiBA experiment 
157: in drift-scanning mode may be analytically tractable 
158: \citep{zha-etal02,lee02}.  
159: 
160: \subsection{The SZE Statistics : An Overview}
161:  
162: The peak number density of a Gaussian random field is analytically given by 
163: \citep{lon57,bon-efs87}: 
164: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
165: \beq
166: n_{g}(\nu) = \frac{1}{(2\pi)^{3/2}}R_{*}^{-2}e^{-\frac{\nu^2}{2}}
167: \int^{\infty}_{0} [x^2+e^{-x^2}-1]
168: \frac{\exp[-\frac{1}{2}(x-\gamma\nu)^{2}/(1-\gamma^{2})]}
169: {[2\pi(1-\gamma^2)]^{1/2}}dx,   
170: \label{eqn:ng}
171: \eeq 
172: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
173: where $\nu$ is the peak height rescaled by the noise 
174: dispersion, $\sigma_{0}$, 
175: $\sigma^{2}_{i} = \frac{1}{(2\pi)^{2}}\int \vert 
176: W(u)\vert^{2}u^{2i}d^{2}{\bf u}$ with a  noise filter $W(u)$,  
177: $R_{*} \equiv \sqrt{2}(\sigma_{1}/\sigma_{2})$, and 
178: $\gamma \equiv \sigma_{1}^{2}/(\sigma_{0}\sigma_{2})$. 
179: Equation (\ref{eqn:ng}) implies that the number density of the Gaussian
180: noise is a sharply decreasing function of $\nu$, which in turn suggests 
181: that the noise effect will be almost completely negligible for very 
182: high signals $\nu > 5$. 
183: 
184: The presence of noise contaminates the SZE sky map, making it obscure 
185: to count the number density of true signal peaks in many different ways: 
186: it changes the amplitudes and locations of the signal peaks, and also it 
187: creates or compromises signal peaks. To make matters worse, the 
188: fraction of the signal peaks is usually only a few ten percent of the noise
189: peaks.  \citet{lee02} suggested, however, that all the various 
190: noise-contamination effects on the number density of real clusters 
191: $n_{cl}(y)$ could be quantified by a one-dimensional Gaussian
192: distribution with the unit dispersion $p(x) =
193: (1/\sqrt{2\pi})e^{-x^2/2}$, and that $n_{cl}(y)$ be statistically
194: estimated by deconvolving $p(x)$ and the zeroth-order approximation to 
195: the cluster number density $n^{(0)}_{cl}(y) \equiv n_{sz}(y) - n_{g}(y)$ 
196: where $n_{sz}(y)$ is the number density of local peaks in a SZE map:  
197: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
198: \beq 
199: n_{cl}^{(0)}(y) = \frac{N_{cl}^{\rm tot}}{N_{sz}^{\rm tot}}
200: \int p(x)n_{cl}(y-x)dx,
201: \label{eqn:dec}
202: \eeq  
203: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
204: where $N_{sz}^{\rm tot}$ and $N_{cl}^{\rm tot}$ are the total number   
205: of local peaks in a SZE map and that of real clusters, respectively. 
206: 
207: The stabilized deconvolution of $p(x)$ and $n^{(0)}_{cl}(y)$ can be 
208: performed in the Fourier space as 
209: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
210: \beq
211: \tilde{n}_{cl}(y_{k}) = 
212: \frac{W_{p}(y_{k})\tilde{n}^{(0)}_{cl}(y_{k})}{\tilde{p}(y_{k})},
213: \eeq
214: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
215: where $y_{k}$, $\tilde{n}_{cl}$, $\tilde{n}^{(0)}_{cl}$,
216: and $\tilde{p}$ are the Fourier counter parts of $y$,  
217: $n_{cl}$, $n^{(0)}_{cl}$, and $p$, respectively. Here, 
218: $W_{p}(y_{k})$ represents the Wiener optimal filter \citep{pre-etal92}, 
219: a key quantity to the deconvolution process. It is introduced 
220: to stabilize the process of the deconvolution on a discretized mesh. 
221: The functional form of $W_{p}(y_{k})$ is determined from the 
222: power spectrum of the $y$-peaks; the ratio of the true power spectrum 
223: ($P_{-}$) to the measured one ($P_{+}$). The measured power spectrum
224: $P_{+}$ is usually a corrupted one due to the numerical noise effect. 
225: The true power spectrum $P_{-}$ is supposed to be well estimated by 
226: subtracting the obvious noise tail from the measured power spectrum 
227: $P_{+}$ (see Fig. \ref{fig:power}). Finally, the inverse Fourier
228: transform of $\tilde{n}_{cl}(y_{k})$ gives the desired estimate of the 
229: cluster number density $n_{cl}(y)$.  
230: 
231: The validity and stability of the above statistical method depends 
232: on the heights of $y$-peaks considering the definitions of $n^{(0)}_{cl}(y)$ 
233: and $p(x)$.  In the very low $y$ range where the noise peaks are dominant 
234: (say, $y < \sigma^{N}$, where $\sigma_{N}$ is the noise level), 
235: $n^{(0)}_{cl}(y)$ is close to zero, and hence the deconvolution likely fails. 
236: In other words, in the low $y$ range, most of the local peaks can
237: be safely regarded as just noise. Also, in the very high $y$ range (say, 
238: $y > 5\sigma_{N}$) where the signal peaks are dominant, we have 
239: $n^{(0)}_{cl}(y) \approx n_{sz}(y) \approx n_{cl}(y)$. 
240: In this range the noise contamination effect is almost completely
241: negligible so that the deconvolution of $p(x)$ is not meaningful any 
242: longer. Hence it is naturally expected that the above statistical 
243: method works best in the range of medium $y$ peaks 
244: (say, $\sigma_{N} < y \le 5\sigma_{N}$) where $n^{(0)}_{cl}(y)$ 
245: has non-zero values while containing a substantial 
246: noise-contamination that can be quantified as $p(x)$. 
247: We emphasize that it is this range of $\sigma_{N} < y \le 5\sigma_{N}$ 
248: where we hope to estimate the cluster number density more efficiently 
249: than the conventional method. 
250: 
251: \section{APPLICATION TO NUMERICAL SIMULATIONS}
252: 
253: We use the SZE cluster catalogues produced from a large cosmological
254: simulation carried out by the Virgo consortium
255: \citep{Jenkins01,yosh-etal01}. 
256: We make maps of $2048^{2}$ pixels in a field of view of 3.14 degree 
257: on a side. Details of the generation of the SZE cluster maps
258: are found in \citet{yosh02}.
259: For our purpose, we use a total of 8 realizations of the SZE cluster
260: maps,  each of which corresponds to a patch of the SZE sky map. 
261: We first smooth the SZE cluster maps by a Gaussian filter of 
262: scale radius $R_{g}$ in order to mimic the diffuse and extended clusters 
263: observed in real surveys. The scale radius $R_{g}$ corresponds to the
264: FWHM (full-width at half maximum) of the clusters which varies with 
265: different observations. Here we choose $R_{g} = 2$ arcminute, given that
266:  the natural beam is expected to have a FWHM of $2$ arcminute in AMiBA 
267: experiment. Next, we construct a Gaussian random field on the same 
268: $2048^{2}$ pixels using the Monte-Carlo method, assuming the flat-sky 
269: approximation. To simulate the instrumental noise of the real SZE 
270: observation, we adopt a white-noise power spectrum as the random field, 
271: and then convolve it with the noise-cleaning filter $W_{C}$ which is
272: given in Fourier $\bu$-space ($\bu$: the Fourier counterpart of $\bth$, 
273: $u \equiv \vert \bu \vert$) as \citep{pen-etal02,lee02}:
274: \begin{equation} 
275: W_{C}(u) = \frac{1}{u}\left[\exp\left(-\frac{u^{2}\theta^{2}_{A}}{2}\right) 
276: - \exp\left(-\frac{u^{2}\theta^{2}_{B}}{2}\right)\right], 
277: \end{equation}  
278: where $\theta_{B} = 6\theta_{A}$ with $\theta_{A} = 0.5$ arcminute 
279: suitable for AMiBA (Pen 2002, private communication). 
280: For each of the 8 realizations of SZE cluster maps, we construct an
281: independent noise field, and combine the cluster
282: sources with the noise field to construct a realistic, noise-added SZE map. 
283: In combining, we rescale each field by the y-dispersion of the noise, 
284: $\sigma_{y}^{N}$. Figure \ref{fig:omap} shows one realization of the sky 
285: maps of original SZE clusters (the top panel) and its smoothed version 
286: combined with the noise field (the bottom panel), respectively.  
287: From each realization of the total SZE map, we determine the number
288: density of the peaks $n_{sz}$ as a function of the rescaled peak
289: heights. Using the measured $n_{sz}$ along with equation 
290: (\ref{eqn:ng}), we compute the zero-th approximation to the cluster 
291: number density, $n^{(0)}_{cl}$, and measure the power spectrum of 
292: the peaks from the total SZE map in order to find a Wiener optimal 
293: filter $W_{p}$. Figure \ref{fig:power} shows an example of the peak 
294: power spectrum computed from one realization. 
295: The solid line corresponds to the measured power spectrum $P_{+}$ from 
296: the total SZE map realization, revealing the noise tail. 
297: The dashed line represents the estimate of the true power spectrum
298: $P_{-}$,  obtained by subtracting the noise tail from $P_{+}$. 
299: The Wiener optimal filter for the deconvolution is then approximated 
300: as $P_{-}/P_{+}$.  
301: 
302: Finally, by using this Wiener optimal filter, we perform the 
303: stabilized deconvolution of the zero-th order signal distribution 
304: $n_{cl}^{(0)}$ and the one-dimensional Gaussian distribution $p(x)$ 
305: to estimate the true number density of the signal peaks. 
306: We found that for all realizations our method works quite well 
307: in the range of $\sigma_{y}^{N} < y \le 5\sigma_{y}^{N}$ 
308: within a $\sim 25 \%$ error. Figure \ref{fig:recon} plots the number 
309: counts of local peaks per degree square as a cumulative function of 
310: the rescaled Comptonization parameter 
311: $N_{cl}(\ge y) \equiv \int _{y}n_{cl}(y^{\prime}) dy^{\prime}$ for 
312: the case of each realization separately. 
313: The long-dashed, the dotted, and the solid lines represent 
314: the cumulative number densities of local peaks from the total SZE map, 
315: noise peaks from equation (\ref{eqn:ng}), and the signal peaks from 
316: the reconstruction, respectively, while the square dots represent the 
317: cumulative number densities of signals from the original SZE simulations 
318: before the noise contamination effect added.
319: Figure \ref{fig:trecon} plots the same as Figure \ref{fig:recon} but 
320: averaged over all 8 realizations. Our method reconstructs
321: also the {\it averaged} peak distribution fairly well.
322: 
323: \section{SUMMARY AND CONCLUSIONS}
324: 
325: We have tested the validity of the SZE statistics proposed by 
326: \citet{lee02} using numerical simulations. The statistical method is 
327: devised for the interferometric SZE observations in drift scanning mode, 
328: targeting at AMiBA as a model experiment, and applied to estimate 
329: the number density of cluster signals that occupy only 
330: a small fraction ($\sim 10 \%$) of the noise-contaminated 
331: SZE maps more efficiently than the conventional method based on 
332: the threshold cut-off. The method is developed to model the dominant 
333: noise contamination effect on the SZE peak distribution by a Gaussian 
334: scatter and to deconvolve the noise scatter and the SZE peak 
335: distribution. The stability of the deconvolution process has been 
336: achieved by convolution with a Wiener optimal filter estimated from the 
337: measurable power spectrum of the SZE peaks. 
338: 
339: We have applied the method to a large number of realistic SZE sky maps, 
340: and found that it indeed works for all realizations, giving a fairly 
341: good estimate of the cluster number density, including even those
342: clusters whose peak heights are compatible with noise-levels.   
343: In the conventional approach, those low-amplitude peaks are usually 
344: all disregarded as noise, so that a large fraction of low-amplitude cluster 
345: signals are missing. Our method allows us to count those missing low-amplitude
346: clusters as quickly and accurately as possible. Hence, we conclude that 
347: it will provide an efficient method to determine the number density
348: of galaxy clusters in the future interferometric SZE surveys. 
349: 
350: \acknowledgments
351: 
352: We thank K. Yoshikawa, U.L. Pen, and Y. Suto for helpful discussions 
353: and useful comments.  We both acknowledge gratefully the financial 
354: support of the JSPS (Japan Society of Promotion of Science)
355: fellowships.  This research was funded by the Grant-in-Aid for 
356: Scientific Research of JSPS (02674 and 12640231). 
357: The simulation data used in this paper are publicly available at 
358: www.mpa-garching.mpg.de/galform/virgo/vls
359: 
360: \clearpage
361: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
362: \begin{thebibliography}{}
363: \bibitem[Bahcall \& Fan(1998)]{bah-fan98}
364: Bahcall, N., \& Fan, X. 1998, \apj, 504, 1
365: \bibitem[Barbosa et al.(1996)]{bar-etal96}
366: Barbosa, D. Bartlett, J. G., Blanchard, A., \& Oukbir, J. 1996, \aap, 314, 13
367: \bibitem[Bond \& Efstathiou(1987)]{bon-efs87}
368: Bond, J. R., \& Efstathious, G. 1987, \mnras, 226, 655
369: \bibitem[Borgani et al.(1999)]{bor-etal99}
370: Borgani, S., Rosati, P., Tozzi, P., \& Colin, N. 1999, \apj, 517, 40
371: \bibitem[Diaferio et al.(2003)]{dia-etal03}
372: Diaferio, A., Nusser, A., Yoshida, N. \& Sunyaev, R. A. 2003, MNRAS, 338, 433
373: \bibitem[Fan \& Chiueh(2001)]{fan-chi01}
374: Fan, Z., \& Chiueh, T. 2001, \apj, 550, 547 
375: \bibitem[Glenn et al.(1998)]{gle-etal98}
376: Glenn, J., Bock, J. J., Chattopadhyay, G., Edgington, S. F., 
377: Lange, A. E., et al. 1998, Proc. SPIE 3357, 326 
378: \bibitem[Grego et al.(2001)]{gre-etal01}
379: Grego, L., Carlstrom, J. E., Reese, E. D., Holder, G. P., 
380: Holzapfel, W. L., Joy, M. K., Mohr, J. J., \& Patel, S. 2001, 552, 2
381: \bibitem[Henry(1997)]{hen97}
382: Henry, J. P. 1997, \apj, 489, L1
383: \bibitem[Jenkins et al.(2001)]{Jenkins01}
384: Jenkins, A., Frenk, C. S., White, S. D. M., Colberg, J. M., Cole, S., 
385: Evrard, A. E., Couchman, H. M. P., Yoshida, N., 2001,\mnras, 321, 372
386: \bibitem[Lee(2002)]{lee02}
387: Lee, J. 2002, \apj, 578, L27 
388: \bibitem[Longuet-Higgins(1957)]{lon57}
389: Longuet-Higgins, M. S. 1957, Phil Trans Roy Soc London A, 249, 321 
390: \bibitem[Lo et al.(2000)]{lo-etal00}
391: Lo, K. H., Chiueh, T. H., Martin, R. N., Ng, K. W., Liang, H., 
392: Pen, U. L, \&  Ma, C. P. 2000, in IAU symposium, Vol. 201, E31 
393: \bibitem[Molnar et al.(2002)]{mol-etal02}
394: Molnar, S. M., Birkinshaw, M.,\&  Mushotzky, R. F. 2002, \apj, 570, 1
395: \bibitem[Pen et al.(2002)]{pen-etal02}
396: Pen, U. L., Ng, K. W., Kesteven, M. J., \& Sault, B. 2002, preprint 
397: (http://www.cita.utoronto.ca/~pen/download/Amiba/)
398: \bibitem[Press et al.(1992)]{pre-etal92}
399: Press, W. H., Teukolsky, S. A., Vetterling, W. T., \& Flannery, B. P. 
400: 1992, Numerical Recipes in Fortran (Univ. of Cambridge: New York)
401: \bibitem[Sunyaev \& Zel'dovich(1972)]{sun-zel72}
402: Sunyaev, R. A., \& Zel'dovich, Y. B. 1972, Comm. Astrophys. Sp. Phys., 
403: 4, 173 
404: \bibitem[Viana \& Liddle(1999)]{via-lid99}
405: Viana, P. T. P., \& Liddle, A. R. 1999, MNRAS, 303, 535
406: \bibitem[Yoshida et al.(2001)]{yosh-etal01}
407: Yoshida, N., Sheth, R. K. \& Diaferio, A. 2001, MNRAS, 328, 669
408: \bibitem[Yoshida(2002)]{yosh02}
409: Yoshida, N., 2002, PhD thesis, Ludwig-Maximilians-Univ., Munich
410: \bibitem[Zhang et al.(2002)]{zha-etal02}
411: Zhang, P., Pen, U.L., \& Wang, B. 2002, \apj, 577, 555
412: \end{thebibliography}
413: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
414: \clearpage
415: \begin{figure}
416: \begin{center}
417: \epsscale{0.6} 
418: \plotone{f1.eps}
419: \caption{Simulated thermal SZ effect maps. The maps shows 
420: the same field of view of 3.14 degree on a side. {\it Top}: a pure
421: thermal SZ effect; {\it Bottom}: a smoothed SZE map combined with 
422: a Gaussian noise field (see $\S 3$ in text).
423: \label{fig:omap}}
424: \end{center}
425: \end{figure}
426: 
427: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
428: \clearpage
429: \begin{figure}
430: \begin{center}\epsscale{1.0} 
431: %\leavevmode\epsfxsize=11.0cm \epsfbox{f2.eps} 
432: \plotone{f2.eps}
433: \caption{The power spectrum of the rescaled $y$-peaks measured from one 
434: realization of the total SZE map. The solid line is the measured power 
435: spectrum, while the dashed line represents our estimate of the true power 
436: spectrum. 
437: \label{fig:power}}
438: \end{center}
439: \end{figure}
440: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
441: 
442: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
443: \clearpage
444: \begin{figure}
445: \begin{center}\epsscale{1.0} 
446: %\leavevmode\epsfxsize=11.0cm \epsfbox{f3.eps} 
447: \plotone{f3.eps}
448: \caption{The number counts of local peaks per degree square as a cumulative 
449: function of the rescaled Comptonization parameter in each realization.  
450: The long-dashed, the dotted, and the solid lines correspond 
451: to the number counts of the total peaks of the SZE maps, the noise,  
452: and the reconstructed signals, respectively, while the square dots 
453: correspond to the number counts of the true signals from the original 
454: SZE map from simulations before the combination with the noise field. 
455: \label{fig:recon}}
456: \end{center}
457: \end{figure}
458: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
459: 
460: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
461: \clearpage
462: \begin{figure}
463: \begin{center}\epsscale{1.0} 
464: %\leavevmode\epsfxsize=11.0cm \epsfbox{f4.eps} 
465: \plotone{f4.eps}
466: \caption{Same as Figure \ref{fig:recon} but the averaged over the 
467: 8 realizations.
468: \label{fig:trecon}}
469: \end{center}
470: \end{figure}
471: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
472: 
473: \end{document}
474: