astro-ph0610520/ms.tex
1: \documentclass[aps,prl,showpacs,nobibnotes,twocolumn,preprintnumbers,
2: nobalancelastpage]{revtex4}        
3: \usepackage{graphics}
4: \usepackage{graphicx}
5: 
6: \begin{document} 
7: \title{Void Ellipticity Distribution as a Probe of Cosmology}
8: \author{Daeseong Park and Jounghun Lee}
9: \email{pds2001@astro.snu.ac.kr, jounghun@astro.snu.ac.kr}
10: \affiliation{Department of Physics and Astronomy, FPRD, Seoul National 
11: University, Seoul 151-747, Korea}
12: 
13: \begin{abstract} 
14: Cosmic voids refer to the large empty regions in the universe with a very low 
15: number density of galaxies. Voids are likely to be severely disturbed by the 
16: tidal effect from the surrounding dark matter. We derive a completely analytic 
17: model for the void ellipticity distribution from physical principles. 
18: We use the spatial distribution of galaxies in a void as a measure of its 
19: shape, tracking the trajectory of the void galaxies under the influence of 
20: the tidal field using the Lagrangian perturbation theory. Our model implies 
21: that the void ellipticity distribution depends sensitively on the cosmological 
22: parameters. Testing our model against the high-resolution Millennium Run 
23: simulation, we find excellent quantitative agreements of the analytic 
24: predictions with the numerical results.
25: \end{abstract}
26: 
27: \pacs{98.65.Dx, 95.75.-z, 98.80.Es}
28: \maketitle
29: 
30: {\it Introduction.---}One of the most fundamental goals in physical cosmology 
31: is to understand the large-scale structure in the universe. The organization 
32: of the large-scale structure begins at the scale of galaxies and appears to 
33: follow a hierarchical pattern up to the scale of superclusters 
34: \cite{col-etal01}.   
35: 
36: The study of the large-scale structure has been so far focused mostly on 
37: the bound objects like clusters and superclusters. However, more recent 
38: observations indicate that the universe is in fact a collection of bubble-like 
39: voids separated by sheets and filaments, at the dense nodes of which the 
40: clusters and superclusters are located \cite{hoy-vog04}. Hence, to address 
41: the large scale structure in the universe, it seems essential to account 
42: for the presence and characterization of voids.
43: 
44: According to the standard gravitational instability theory, voids 
45: originate from the local minima of the initial density field and expand 
46: faster than the rest of the universe. The extremely low-density of the 
47: observed voids ($\delta_{v} \approx -0.9$) supports this scenario 
48: \cite{hoy-vog04}. A common expectation based on this standard 
49: scenario is that voids are likely to have quite spherical shapes 
50: \cite{ick84,dub-etal93,van-van93,she-van04}. However, a recent systematic 
51: analysis of simulation data has revealed that the shapes of voids are 
52: in fact far from spherical symmetry \cite{sha-etal04,sha-etal06}. 
53: 
54: Confronted with this somewhat unexpected result, Ref.\cite{sha-etal06} 
55: claimed that the gravitational tidal field be responsible for the 
56: nonspherical shapes of voids. They explained that since the voids have 
57: very low-density, they must be more easily disturbed by the tidal 
58: effect from the surrounding matter. 
59: 
60: Reference \cite{lee-par06} has investigated the tidal effect on voids 
61: analytically using the linear tidal torque theory, which was inspired by 
62: \cite{sha-etal06}. It is shown by them that the tidal field indeed has a 
63: significant influence on voids, generating non-radial motions of matter that 
64: make up voids. They tested their analytic model against high-resolution 
65: N-body simulation and found remarkably good agreements between the analytical 
66: and the numerical results. 
67: 
68: The success of the analytic model of \cite{lee-par06} has two crucial 
69: implications. First, the non-radial motions of galaxies in voids  
70: generated by the tidal effect would cause the nonsphericity in void shapes, 
71: as hinted first by \cite{sha-etal06}.  Second, since the Lagrangian theory 
72: has turn out to work well in predicting the tidal effect on voids, 
73: it may be also possible to model the nonspherical shapes using the 
74: Lagrangian theory that depends only on the initial conditions.
75: 
76: One difficulty in modeling the nonspherical shapes of voids, however, 
77: lies in the fact that there is no well defined way to determine the shape 
78: of a void  since a void is an unbound system having no clear-cut boundary.
79: Nevertheless, given that our purpose is not to construct the full geometrical 
80: shape of a void but to quantify the degree of the deviation of its shape 
81: from spherical symmetry, a practical way that we choose here is to use the 
82: spatial distribution of the void galaxies as a measure of the nonspherical 
83: shape of a void. If a void has a spherical shape, then the spatial 
84: distribution of its galaxies will show a more or less isotropic pattern. 
85: If it has a nonspherical shape, the void galaxy spatial distribution will 
86: also deviate from the isotropic pattern. In this practical way, the 
87: nonsphericity of the shape of a given void may be quantified by the 
88: {\it ellipticity} of the spatial distribution of the galaxies that it 
89: contains. 
90: 
91: In this Letter we aim for constructing a complete analytic model for the 
92: void ellipticity using the Zel'dovich approximation (the first order 
93: Lagrangian perturbation theory) and investigate if the void ellipticity 
94: distribution can put constraints on the cosmological parameters. 
95: 
96: {\it The Analytic Model---}Assuming that the density field is smoothed on the 
97: void scale, we use the Zel'dovich approximation \cite{zel70} to describe the 
98: trajectory of a galaxy located in a void region as 
99: ${\bf x}={\bf q}-D(t)\nabla\Psi({\bf q})$. Here ${\bf x}$ and ${\bf q}$ 
100: represent the Eulerian and the Lagrangian positions of a void galaxy, 
101: respectively, $D(t)$ is the linear growth factor and $\Psi({\bf q})$ is 
102: the perturbation potential smoothed on the Lagrangian void scale, 
103: $R_{L}$. Note that here a galaxy is treated as a particle moving under 
104: the influence of $\Psi$.
105: 
106: Applying the mass conservation law, 
107: $\rho d^{3}{\bf x} =\bar{\rho}d^{3}{\bf q}$ to the Zel'dovich approximation, 
108: the mass density of a void region, $\rho_{v}$, can be written as 
109: \begin{equation}
110: \rho_{v} = \frac{\bar{\rho}}{(1-\lambda_1)(1-\lambda_2)(1-\lambda_3)}, 
111: \label{eqn:rho}
112: \end{equation}
113: where $\lambda_{1},\lambda_{2},\lambda_{3}$ 
114: (with $\lambda_{1}>\lambda_{2}>\lambda_{3}$) are the three eigenvalues of 
115: the tidal tensor, $T_{ij}$, defined as the second derivative of $\Psi$. 
116: The dimensionless residual overdensity of the void region, $\delta_{v}$, 
117: equals the sum of the three eigenvalues: 
118: $\delta_{v}=\sum_{i=1}^{3}\lambda_{i}$. 
119: 
120: Eq.~(\ref{eqn:rho}) implies that under the influence of the tidal field, 
121: $T_{ij}$, the void region will experience a triaxial expansion. 
122: Let $p_{1},p_{2},p_{3}$ (with $p_{1}>p_{2}>p_{3}$) represent the three 
123: semi-axis lengths of the inertia momentum tensor of the galaxies that make 
124: up void , and let $\mu$ and $\nu$ be the two axial ratios defined as 
125: $\mu\equiv p_{2}/p_{1}$ and $\nu\equiv p_{3}/p_{1}$. The ellipticity 
126: of a given void region in terms of the minor-to-major axis ratio as 
127: $\varepsilon \equiv 1 -\nu$. 
128: 
129: According to Eq.~(\ref{eqn:rho}), the two axial ratios are related to the 
130: three eigenvalues of $(T_{ij})$ as 
131: \begin{equation}
132: \label{eqn:axis_lambda}
133: \mu = \left(\frac{1-\lambda_2}{1-\lambda_3}\right)^{1/2},\qquad 
134: \nu = \left(\frac{1-\lambda_1}{1-\lambda_3}\right)^{1/2}.
135: \end{equation}
136: The distribution of $\varepsilon$ can be found from the distribution of $\nu$, 
137: which in turn can be derived from the conditional distribution of the three 
138: eigenvalues of the tidal tensor under the constraint of 
139: $\delta_{v}=\sum_{i=1}^{3}\lambda_{i}$.
140: 
141: In Ref.\cite{dor70}, the unconditional joint probability density distribution 
142: of $\{\lambda_1,\lambda_2,\lambda_3\}$ was derived. On the void scale 
143: $R_{L}$, it is written as 
144: \begin{eqnarray}
145: p(\lambda_1,\lambda_2,\lambda_3;\sigma_{R_L})&=&
146: \frac{3375}{8\sqrt{5}\pi\sigma^6_{R_L}}
147: \exp\bigg{(}-\frac{3K_{1}^2}{\sigma^2_{R_L}}
148: +\frac{15K_{2}}{2\sigma^2_{R_L}}\bigg{)}\cr
149: &&\times(\lambda_{1}-\lambda_{2})(\lambda_{2}-\lambda_{3})
150: (\lambda_{1}-\lambda_{3}),
151: \label{eqn:dor}
152: \end{eqnarray}
153: where $K_{1} \equiv \lambda_{1}+\lambda_{2}+\lambda_{3}$, 
154: $K_{2} \equiv \lambda_{1}\lambda_{2} + \lambda_{2}\lambda_{3} +
155: \lambda_{3}\lambda_{1}$, and $\sigma_{R_L}$ is the rms density fluctuation 
156: related to the dimensionless linear power spectrum $\Delta^{2}(k)$ as 
157: \begin{equation}
158: \sigma^{2}_{R_L} \equiv \int_{-\infty}^{\infty}\Delta^{2}(k)
159: W^{2}(kR_{L})d\ln k, 
160: \end{equation}
161: where $W(kR_L)$ represents a smoothing function with the Lagrangian filtering 
162: radius of $R_L$. Here, we adopt the approximation formula given by 
163: \cite{bar-etal86} for $\Delta^{2}(k)$ and the top-hat spherical filter 
164: for $W(kR_{L})$. 
165: 
166: By Eq.~(\ref{eqn:dor}), the conditional joint probability 
167: density distribution of the void axial ratios is found as  
168: \begin{eqnarray}
169: p(\mu,\nu|\delta_{v};\sigma_{R_L})&=& 
170: A~p[\lambda_{1}(\mu,\nu),\lambda_{2}(\mu,\nu)|\delta_{v};\sigma_{R_L}]
171: \cr
172: && \times\frac{4(\delta_{v}-3)^2\mu\nu}{(\mu^{2}+\nu^{2}+1)^{3}}.
173: \label{eqn:munu}
174: \end{eqnarray}
175: Here $p(\lambda_{1},\lambda_{2}|\delta_{v};\sigma_{R_L})$ is the conditional 
176: probability density of $\lambda_{1}$ and $\lambda_{2}$ provided that a given 
177: void has an average density $\delta_{v}$ on the scale of $R_{L}$.  This 
178: conditional probability has the following form:
179: \begin{eqnarray} 
180: &&p(\lambda_{1},\lambda_{2}|\delta_{v};\sigma_{R_L}) \cr
181: &&=\frac{3375\sqrt{2}}{\sqrt{10\pi}\sigma^{5}_{R_L}}
182: \exp\left[-\frac{5\delta^{2}_{v}}{2\sigma^{2}_{R_L}} + 
183: \frac{15\delta_{v}(\lambda_{1}+\lambda_{2})}{2\sigma^{2}_{R_L}}\right]\cr
184: &&\times\exp\left[-\frac{15(\lambda^{2}_{1}+\lambda_{1}\lambda_{2}+
185: \lambda^{2}_{2})}{2\sigma^{2}_{R_L}}\right] \cr
186: &&\times(2\lambda_{1}+\lambda_{2}-\delta_{v})(\lambda_{1}-\lambda_{2})
187: (\lambda_{1}+2\lambda_{2}-\delta_{v}).
188: \label{eqn:con}
189: \end{eqnarray}
190: where the functional forms of $\lambda_{1}(\mu,\nu)$ and 
191: $\lambda_{2}(\mu,\nu)$ are given as \cite{lee-etal05}
192: \begin{eqnarray}
193: \label{eqn:lamu1}
194: \lambda_{1}(\mu,\nu) &=& \frac{1 + (\delta_{v}- 2)\nu^{2} + 
195: \mu^{2}}{(\mu^{2} + \nu^{2} + 1)},\\
196: \label{eqn:lamu2} 
197: \lambda_{2}(\mu,\nu) &=& \frac{1 + (\delta_{v}- 2)\mu^{2} + \nu^{2}}
198: {(\mu^{2} + \nu^{2} + 1)},
199: \end{eqnarray}
200: and the normalization constant, $A$, satisfies the constraint of 
201: $A\int p(\mu,\nu | \delta=\delta_v ; \sigma_{R_L})d\mu d\nu = 1$
202: 
203: The distribution, $p(\nu)$ is now straightforwardly evaluated by integrating 
204: Eq.~(\ref{eqn:munu}) over 
205: $\mu$ as 
206: \begin{equation}
207: p(\nu; R_{L}) = \int_{\nu}^{1}
208: p(\mu,\nu|\delta =\delta_{v};\sigma_{R_L})d\mu.
209: \label{eqn:nu}
210: \end{equation}
211: Then, the distribution of the void ellipticity, $p(\varepsilon)$, is nothing 
212: but $p(\epsilon)=p(1-\nu)$.
213: %%%%%%%%%%%%%%%%%%%%%%%%%
214: \begin{figure}
215: \includegraphics[scale=0.45]{fig1.ps}
216: \caption
217: {Probability density distributions of the void ellipticity  predicted by 
218: our analytic model for the four different sets of cosmological parameters 
219: at two different smoothing scales: $R_{L}=5$ and $8h^{-1}$Mpc in the upper 
220: and the lower panel, respectively. For each case, the void density contrast 
221: is set at $\delta_{v}=-0.9$.}
222: \label{fig:analytic}
223: \end{figure}
224: %%%%%%%%%%%%%%%%%%%%%%%%%
225: Fig.~\ref{fig:analytic} plots $p(\varepsilon)$ at two different Lagrangian 
226: scales: $R_{L}=5$ and $8h^{-1}$Mpc in the upper and the lower panel, 
227: respectively. Here we assume a $\Lambda$-dominated cold dark matter cosmology 
228: ($\Omega_{\Lambda}=1-\Omega_{m}$) and consider four different sets of 
229: the key cosmological parameters: $\Omega_{m}$ 
230: (matter density parameter), $h$ (dimensionless Hubble parameter), 
231: $\sigma_{8}$ (amplitude), and $n_{s}$ (power spectrum slope). 
232: 
233: As can be seen, the ellipticity distribution depends quite sensitively on 
234: the cosmological parameters. Note that for the cases of low $\Omega_{m}$ 
235: and high $h$, voids tend to have more spherical shapes, which in fact can 
236: be understood as follows. The void ellipticity distribution is an outcome 
237: of the counterbalance between the tidal effect of the dark matter and 
238: the expansion of the universe: The tidal effect disturbs the void shapes 
239: from spherical symmetry, while the expansion of the universe resists such 
240: disturbance.It is also worth noting that the void ellipticity distribution 
241: moves toward the low ellipticity as the smoothing scale increases. It 
242: indicates that the larger a void is, the more spherical it is.
243: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
244: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
245: 
246: {\it Numerical Test.---}To examine the validity of our analytic model, 
247: we test its prediction against the Millennium Run simulation for which the 
248: cosmological parameters are specified as 
249: $\Omega_{m}=0.25$, $n_{s}=1$, $\sigma_{8}=0.9$ and $h=0.73$ \citep{spr-etal05}.
250: 
251: In our previous work \cite{lee-par06} we have already identified voids 
252: in the Millennium Run catalog using the Hoyle-Vogeley (HV02) void-finder 
253: algorithm \cite{hoy-vog02}. For the detailed description of the void-finding  
254: scheme and the properties of the identified voids, see \cite{lee-par06}. 
255: Among these voids, we select only those voids which contain 
256: more than $5$ galaxies. A total of $23652$ voids are found to contain more 
257: than $5$ galaxies.
258: 
259: For each selected void, we first calculate the pseudo inertia momentum 
260: tensor, $(I_{ij})$, defined as 
261: \begin{equation}
262: I_{ij} = \frac{1}{N_{vg}}\sum_{\alpha}x^{g}_{\alpha i}x^{g}_{\alpha j},
263: \label{eqn:iner}
264: \end{equation}
265: where $N_{vg}$ is the number of galaxies belonging to a given void, and 
266: $(x_{\alpha i}^{g})$ is the position vector of the $\alpha$-th void galaxy. 
267: As mentioned in \cite{sha-etal06}, the inertia momentum tensor is computed 
268: without weighting the position vectors with mass since we are interested 
269: in the geometry of voids.
270: 
271: Diagonalizing $(I_{ij})$, we find the three eigenvalues, $I_{1}$, $I_{2}$, 
272: and $I_{3}$ (with $I_{1}>I_{2}>I_{3}$), which are related to the semi 
273: axis-lengths of a given void as 
274: $I_{1}\propto (p_{1}^{2}+p_{2}^{2})$, $I_{2}\propto (p_{1}^{2}+p_{3}^{2})$, 
275: $I_{3}\propto (p_{2}^{2}+p_{3}^{2})$ \cite{sha-etal06}. Now, the void 
276: ellipticity, $\varepsilon \equiv 1-p_{3}/p_{1}$, can be written in terms of 
277: $I_{1},I_{2},I_{3}$ as 
278: %%%%%%%%%%%%%%%%%%%%%%%%%
279: \begin{equation}
280: \varepsilon = 1 -\nu = 1 -
281: \left(\frac{I_{2}+I_{3}-I_{1}}{I_{1}+I_{2}-I_{3}}\right)^{1/2}.
282: \label{eqn:ell}
283: \end{equation}
284: %%%%%%%%%%%%%%%%%%%%%%%%%%
285: Measuring $\varepsilon$ of each selected void using Eq.~(\ref{eqn:ell}), 
286: we determine the void ellipticity distribution. Since the distribution 
287: depends on the size of void, we divide the whole sample of voids into the 
288: four bins according to their effective radius, $R_{E}$, which is defined as 
289: $R_{E}\equiv (3\Gamma_{V}/4\pi)^{1/3}$ where $\Gamma_{V}$ is the total volume 
290: of a void measured by the Monte-Carlo method devised by \cite{hoy-vog02}. 
291: Each bin is chosen to include equal number of voids. The range of $R_{E}$, 
292: the mean values of $R_{E}$, $\delta_{v}$, and $\varepsilon$ for the four 
293: bins are listed in Table ~\ref{tab:list}.
294: %%%%%%%%%%%%%%%%%%%%%%%%%%%
295: \begin{table}
296: \caption{The range of the effective radius ($R_{\textrm{E}}^{\textrm{min}}$ 
297: and $R_{\textrm{E}}^{\textrm{max}}$), the mean effective radius 
298: ($\bar{R}_{\textrm{E}}$), the mean density ($\bar{\delta}_{\textrm{v}}$), 
299: the mean ellipticity ($\bar{\varepsilon})$, and the mean number of the 
300: galaxies ($\bar{N}_{\textrm{vg}}$) of the voids belonging to the four bins.
301: \label{tab:list}}
302: \begin{ruledtabular}
303: \begin{tabular}{lcccccc}
304: & $R_{\textrm{E}}^{\textrm{min}}$ &
305: $R_{\textrm{E}}^{\textrm{max}}$& $\bar{R}_{\textrm{E}}$&
306: $\bar{\delta}_{\textrm{v}}$
307: & $\bar{\varepsilon}$ & $\bar{N}_{\textrm{vg}}$ \\
308: \hline
309: $(a)$ & $6.42$  & $9.04$  & $8.37$& $-0.89$ & $0.48$ & $18$ \\
310: $(b)$ & $9.04$  & $10.28$ & $9.65$& $-0.89$ & $0.43$ & $28$ \\
311: $(c)$ & $10.28$ & $11.66$ & $10.92$ & $-0.89$ & $0.41$ & $39$ \\
312: $(d)$ & $11.66$ & $21.52$ & $13.00$& $-0.90$ & $0.36$ & $63$ \\
313: %\caption{ \label{table1}}
314: \end{tabular}
315: \end{ruledtabular}
316: \end{table}
317: %%%%%%%%%%%%%%%%%%%%%%%%%
318: %%%%%%%%%%%%%%%%%%%%%%%%%
319: \begin{figure} 
320: \includegraphics[scale=0.45]{fig2.ps}
321: \caption
322: {Illustration of the four examples of the voids identified in the Millennium 
323: Run simulation. Each panel corresponds to the four bins. The shaded regions 
324: and the dots represent the areas of the voids and the positions of the void 
325: galaxies in them projected onto the two dimensional plane.} 
326: \label{fig:exa}
327: \end{figure}
328: %%%%%%%%%%%%%%%%%%%%%%%%%
329: %%%%%%%%%%%%%%%%%%%%%%%%%
330: \begin{figure}
331: %\hbox to \hsize{\hss \epsfxsize=8.6cm \epsffile{rhom.ps}\hss} 
332: \includegraphics[scale=0.45]{fig3.ps}
333: \caption{The ellipticity distributions of the voids belonging to the 
334: four bins. In each panel the histogram represents the numerical data points 
335: from the Millennium Run simulation while the solid line corresponds to the 
336: analytic prediction. For the analytic results, the values of the 
337: Lagrangian scale radius, $R_{L}$, are evaluated by the relation of 
338: $R_{L}=(1+\bar{\delta}_{v})^{1/3}\bar{R}_{E}$. }
339: \label{fig:com}
340: \end{figure}
341: %%%%%%%%%%%%%%%%%%%%%%%%%
342: %%%%%%%%%%%%%%%%%%%%%%%%%
343: %\begin{equation}
344: %R_{L} = (1+\delta_{v})^{1/3}R_{E}.
345: %\label{eqn:cov}
346: %\end{equation}
347: %%%%%%%%%%%%%%%%%%%%%%%%%%
348: As shown, $\bar{\varepsilon}$ decreases as the mean $\bar{R}_{E}$ increases. 
349: That is, the larger a void is, the more spherical shape it tends to have.
350: 
351: Fig.~\ref{fig:exa} illustrates the projected images of the four sample voids 
352: selected from the four bins, respectively. In each panel the dots represent 
353: the positions of the void galaxies and the shaded regions correspond to 
354: the areas of the voids. Fig.~\ref{fig:com} plots the void ellipticity 
355: distributions for the four bins. In each panel, the histogram and the solid 
356: line represent the numerical and the analytical results, respectively. 
357: For the analytic distribution,  the Lagrangian scale radius, $R_{L}$, is 
358: given analytically as $R_{L} = (1+\delta_{v})^{1/3}\bar{R}_{E}$. 
359: As can be seen the analytic and the numerical results are in excellent 
360: quantitative  agreements for all cases, proving the validity of our model.
361: It is worth emphasizing here that our model is completely analytical without 
362: having any fitting parameter derived from principles. It depends only on the 
363: initial cosmological parameters.
364: 
365: {\it Discussion.---}The ellipticity distribution of galaxy clusters has been 
366: used recently to constrain cosmology \cite{jin-sut02,lee06,ho-etal06}. 
367: For the case of clusters, however, the ellipticity distribution cannot be 
368: evaluated analytically from the initial conditions because the nonlinear 
369: clustering process tends to modify the cluster ellipticities significantly 
370: in the subsequent evolution. Although \cite{lee-etal05} attempted to derive 
371: an analytic fitting model for the cluster ellipticity distribution, their 
372: model was shown to be not in good quantitative agreements with the numerical 
373: results.
374: 
375: In contrast, we have for the first time shown here that the void ellipticity 
376: distribution can be evaluated fully analytically using physical principles and 
377: thus be a better indicator of the initial conditions of the universe since 
378: the void ellipticities are less vulnerable to nonlinear modifications.  
379: By applying our analytic model to real data that will be available from 
380: future large galaxy surveys , it may be possible to constrain the cosmological 
381: parameters precisely in an independently way. 
382: 
383: It is, however, worth nothing that the Hoyle-Vogeley algorithm works very 
384: well if a void is found as a spherical low-density region. Nevertheless, we 
385: expect that our numerical result would not change sensitively with respect 
386: to the void-finding algorithm, given the fact that the ellipticity of a void 
387: here is defined in terms of the spatial distribution of galaxies that make 
388: up the void but not by the boundary shape of a void.
389: 
390: The Millennium Run simulation used in this paper was carried out by the Virgo 
391: Supercomputing Consortium at the Computing Centre of the Max-Planck Society 
392: in Garching. We thank the referee and V. Springel for useful comment. 
393: This work was supported by the New Faculty Settlement fund of the Seoul 
394: National University.
395: 
396: \begin{thebibliography}{}
397: \bibitem{col-etal01}
398: Colless {\it et al.},  Mon. Not. R. Astron. Soc. {\bf 328}, 1039
399: (2001)
400: \bibitem{hoy-vog04}
401: F. Hoyle, and M. S. Vogeley, Astrophys. J. {\bf 607}, 751 (2004)
402: \bibitem{ick84}
403: V. Icke, Mon. Not. R. Astron. Soc. {\bf 206}, 1P (1984)
404: \bibitem{dub-etal93}
405: J. Dubinski, L. N. da Costa, D. S. Goldwirth, M. Lecar, and T. Piran, 
406: Astrophys. J. {\bf 410}, 458 (1993)
407: \bibitem{van-van93}
408: Van de Weygaert, R., and Van Kampen, E., Mon. Not. R. Astron. Soc.
409: {\bf 263}, 481 (1993)
410: \bibitem{she-van04}
411: R. K. Sheth, and R. van de Weygaert, Mon. Not. R. Astron. Soc. {\bf
412: 350}, 517 (2004)
413: \bibitem{sha-etal04}
414: S. F. Shandarin, J. V. Sheth, and V. Sahni, Mon. Not. R.
415: Astron. Soc. {\bf 353}, 162 (2004)
416: \bibitem{sha-etal06}
417: S. F. Shandarin, H. A. Feldman, K. Heitmann, and S. Habib, Mon. Not. R.
418: Astron. Soc. {\bf 367}, 1629 (2006)
419: \bibitem{lee-par06}
420: J. Lee, and D. Park, Astrophys. J. {\bf 652}, 1 (2006)
421: \bibitem{zel70}
422: Ya. B. Zel'dovich, Astrophys. \& Astron. {\bf 5}, 84 (1970)
423: \bibitem{dor70}
424: A. G. Doroshkevich, Astrofizika, {\bf 3}, 175 (1970)
425: \bibitem{bar-etal86}
426: J. M. Bardeen, J. R. Bond, N. K. Kaiser, and  A. S. Szalay,  
427: Astrophys. J. {\bf 304}, 15 (1986)
428: \bibitem{lee-etal05} 
429: J. Lee, Y. Jing, and Y. Suto, Astrophys. J. {\bf 632}, 706 (2005)
430: \bibitem{spr-etal05}
431: V. Springel {\it et al.}, Nature, {\bf 435}, 629 (2005). 
432: The semianalytic galaxy catalogue is publicly available at 
433: http://www.mpa-garching.mpg.de/galform/agnpaper
434: \bibitem{hoy-vog02}
435: F. Hoyle, and M. S. Vogeley, Astrophys. J. {\bf 566}, 641 (2002)
436: \bibitem{jin-sut02}
437: Y. Jing, and Y. Suto, Astrophys. J. {\bf 574}, 538 (2002)
438: \bibitem{lee06}
439: J. Lee, Astrophys. J. {\bf 643}, 724 (2006)
440: \bibitem{ho-etal06}
441: Ho. Shirley, Bahcall, N. A., and P. Bode, Astrophys. J. {\bf 647}, 8 (2006)
442: \end{thebibliography}
443: 
444: \end{document}
445: