1: \documentstyle[12pt,aaspp4]{article}
2: %\documentstyle[aasms4]{article}
3:
4: \begin{document}
5: %\baselineskip 8mm
6:
7: \newcommand{\gsim}{\mbox{\raisebox{-1.0ex}{$~\stackrel{\textstyle >}
8: {\textstyle \sim}~$ }}}
9: \newcommand{\lsim}{\mbox{\raisebox{-1.0ex}{$~\stackrel{\textstyle <}
10: {\textstyle \sim}~$ }}}
11: \newcommand{\psim}{\mbox{\raisebox{-1.0ex}{$~\stackrel{\textstyle \propto}
12: {\textstyle \sim}~$ }}}
13: \newcommand{\vect}[1]{\mbox{\boldmath${#1}$}}
14: \newcommand{\lmk}{\left(}
15: \newcommand{\rmk}{\right)}
16: \newcommand{\lnk}{\left\{ }
17: \newcommand{\nn}{\nonumber}
18: \newcommand{\rnk}{\right\} }
19: \newcommand{\lkk}{\left[}
20: \newcommand{\rkk}{\right]}
21: \newcommand{\lla}{\left\langle}
22: \newcommand{\p}{\partial}
23: \newcommand{\rra}{\right\rangle}
24: \newcommand{\vex}{{\vect x}}
25: \newcommand{\vek}{{\vect k}}
26: \newcommand{\vel}{{\vect l}}
27: \newcommand{\vem}{{\vect m}}
28: \newcommand{\ven}{{\vect n}}
29: \newcommand{\vep}{{\vect p}}
30: \newcommand{\veq}{{\vect q}}
31: \newcommand{\veX}{{\vect X}}
32: \newcommand{\veV}{{\vect V}}
33: \newcommand{\beq}{\begin{equation}}
34: \newcommand{\eeq}{\end{equation}}
35: \newcommand{\beqa}{\begin{eqnarray}}
36: \newcommand{\eeqa}{\end{eqnarray}}
37: \newcommand{\mpc}{\rm Mpc}
38: \newcommand{\hmpc}{{h^{-1}\rm Mpc}}
39: \newcommand{\ch}{{\cal H}}
40:
41: %\if0
42: \title{ Nonlinear Velocity-Density Coupling: \\Analysis by Second-Order
43: Perturbation Theory }
44: \author{\sc Naoki Seto }
45: \affil{Department of Physics, Faculty of Science, Kyoto University,
46: Kyoto 606-8502, Japan\\
47: seto@tap.scphys.kyoto-u.ac.jp }
48:
49:
50: \begin{abstract}
51: Cosmological linear perturbation theory predicts that the peculiar velocity
52: $\veV(\vex)$ and the matter overdensity $\delta(\vex)$ at a same point $\vex$
53: are statistically independent quantities, as log as the initial density
54: fluctuations are random Gaussian distributed. However nonlinear
55: gravitational effects might change the situation. Using framework of
56: second-order perturbation theory and the Edgeworth expansion method, we
57: study local density dependence of bulk velocity dispersion that is
58: coarse-grained at a weakly nonlinear scale. For a typical CDM model, the first
59: nonlinear correction of this constrained bulk velocity dispersion
60: amounts to $\sim 0.3\delta$ (Gaussian smoothing) at a weakly nonlinear scale
61: with a very weak dependence on cosmological
62: parameters. We also compare our analytical prediction with published
63: numerical results given at nonlinear regimes.
64:
65: \keywords{cosmology: theory --- large-scale structure of universe}
66: \end{abstract}
67: %\fi
68:
69:
70: \section{Introduction}
71: The peculiar velocity field is one of the most fundamental quantities to
72: analyze the
73: large-scale structure in the universe ({\it e.g.} Peebles 1980). It is
74: considered to reflect
75: dynamical nature of density fluctuations of gravitational
76: matter. The peculiar velocity field is usually observed using
77: astrophysical objects ({\it e.g.} galaxies), as determination of
78: distances is crucial for measuring peculiar velocities (Dekel 1994,
79: Strauss \& Willick 1995). There is a
80: possibility that statistical aspects of the velocity field traced by these
81: objects and that traced by dark matter particles might be different. This
82: difference is generally called ``velocity bias'' and its elucidation becomes
83: highly important in observational cosmology ({\it e.g.} Cen \& Ostriker 1992,
84: Narayanan, Berlin \& Weinberg 1998, Kaufmann et al. 1999).
85:
86: Velocity bias is often discussed numerically with making ``galaxy
87: particles" in some effective manners. But here we discussed a more basic
88: phenomenon. It is known that
89: statistics of the peculiar velocity field depend largely on local density
90: contrast. For example, both the single particle and
91: pairwise velocity dispersions of dark matter particles are
92: known to be increasing function of local density (Kepner, Summers, \&
93: Strauss 1997, Strauss, Cen \& Ostriker 1998, Narayanan et al. 1998).
94: Analysis of pairwise velocity statistics is interesting from theoretical
95: point of views, and also very important in observational cosmology
96: (Peebles 1976, Davis \& Peebles 1983, Zurek et al. 1994, Fisher et
97: al. 1994, Sheth 1996,
98: Diaferio \&
99: Geller 1996, Suto \& Jing 1997, Seto \& Yokoyama 1998a, 1999, Jing \&
100: B$\ddot{\rm o}$rner 1998,
101: Juszkiewicz, Fisher \&
102: Szapudi 1998, Seto 1999b, Juszkiewicz, Springel \&
103: Durrer 1999). But
104: we do not discuss it here and concentrate on velocity field
105: characterized by single point which is simpler to analyze
106: theoretically.
107:
108: Linear perturbation theory predicts that the peculiar velocity
109: $\veV(\vex)$ and the density contrast $\delta(\vex)$ at a given point
110: $\vex$ is statistically independent, as long as the initial density
111: fluctuations are random Gaussian distributed. Namely, the joint probability
112: distribution function $P(\veV, \delta)$ can be written in a form
113: as $P_1(\veV)P_2(\delta)$.
114:
115: It is not surprising that the peculiar velocity of each particle is largely
116: affected by nonlinear gravitational effects and shows local density
117: dependence described above. But what can we expect for the smoothed (bulk)
118: velocity that is
119: field coarse grained at some spatial scale $R$? Due to nonlinear mode
120: couplings, the relation $P_1(\veV)P_2(\delta)$ valid for linear theory must
121: be modified and bulk velocity dispersion must also depend on local
122: density contrast defined at the same smoothing scale $R$ (see
123: Bernardeau 1992,
124: Chodorowski \& {\L}okas 1997, Bernardeau et al. 1999 for the velocity
125: divergence field).
126:
127: However, Kepner, Summers \& Strauss (1997) showed from cold-dark-matter
128: (CDM) and hot-dark-matter (HDM)
129: N-body simulations that at nonlinear scales ($0.77\hmpc\le R \le
130: 4.88\hmpc$), such a local density dependence was not observed (see
131: Figs.2(a) and 3(a) of their paper).
132: This is an interesting contrast to the behavior of velocity field traced
133: by each particle, as described before (Kepner et al. 1997, Narayanan et
134: al. 1998).
135:
136:
137: In this article, we investigate local density dependence of smoothed
138: (bulk) velocity dispersion using framework of second-order perturbation
139: theory. We calculate the first-order nonlinear correction of the constrained
140: velocity
141: dispersion. Our target is weakly nonlinear scale and somewhat larger
142: than scale analyzed by Kepner et al. (1997). Since current survey depth
143: of the
144: cosmic velocity field is highly limited, our constrained statistics
145: might not be useful in observational cosmology at present ({\it e.g.}
146: Seto \& Yokoyama 1998b, see also Seto 1999a). Our interest in this
147: article is
148: theoretically motivated one about nonlinear gravitational dynamics.
149:
150:
151: As the peculiar velocity field is more weighted to large-scale fluctuations
152: (smaller wave number $k$) than the density field, perturbative treatment of
153: smoothed velocity field
154: would be reasonable at weakly nonlinear scale. Actually, Bahcall,
155: Gramann \& Cen (1994) showed that
156: smoothed unconstrained
157: velocity dispersion in N-body simulations are well predicted by
158: linear theory even at smoothing scale
159: $R=3\hmpc$ (see their Table 1). Second-order
160: analysis by Makino, Sasaki \&
161: Suto (1992) also gives consistent results to their simulations.
162:
163:
164:
165:
166:
167: \section{Formulation}
168: First we define the (unsmoothed) density contrast field $\delta(\vex)$
169: in terms
170: of the mean density of the universe $\bar{\rho}$ and the local density
171: field $\rho(\vex)$ as
172: \beq
173: \delta(\vex)=\frac{\rho(\vex)-\bar{\rho}}{\bar{\rho}}.
174: \eeq
175: Many theoretical predictions of the large-scale structure are based on
176: continuous fields, but observations as well as numerical experiments
177: (such as, N-body simulations) are usually sampled by points
178: where point-like
179: galaxies (or mass elements) exist. In comparison of theoretical
180: predictions with actual
181: observations or numerical experiments,
182: smoothing operation becomes sometimes
183: crucially important to remove sparseness of particles' system. This
184: operation is also important to reduce strong nonlinear effects which are
185: difficult to handle theoretically. Thus it is favorable to make theoretical
186: predictions of the large-scale structure including smoothing operation.
187: We can express the smoothed density
188: contrast field $\delta_R(\vex)$ and the smoothed velocity field
189: $\veV_R(\vex)$ with (spatially
190: isotropic) filter $W(x,R)$ as
191: \beq
192: \delta_R(\vex)\equiv \int d\vex'^3 \delta(\vex') W(|\vex-\vex'|,R),~~~
193: \veV_R(\vex)\equiv \int d\vex'^3 \veV(\vex') W(|\vex-\vex'|,R).
194: \eeq
195: As we discuss only the smoothed fields in this article, we hereafter
196: omit the suffix $R$ which indicates smoothing radius.
197:
198: The velocity dispersion $ \Sigma_V^2 (\delta)$ for points $\vex$ with a
199: given
200: overdensity $\delta(\vex)=\delta$ is formally written as
201: \beq
202: \Sigma_V^2 (\delta)=\frac{\lla \veV(\vex)^2\delta_{Drc}[\delta(\vex)-\delta]\rra }{\lla \delta_{Drc}[\delta(\vex)-\delta]\rra},
203: \eeq
204: where $\delta_{Drc}(\cdot)$ is Dirac's delta function and brackets
205: $\lla\cdot\rra$
206: represent to take ensemble average.
207:
208: We assume that the initial (linear) density fluctuations are isotropic
209: random Gaussian. At the linear-order we have $\veV(\vex)\propto \nabla
210: \Delta^{-1}\delta(\vex)$ and $\lla \veV(\vex)\delta(\vex)\rra=0$ due to
211: isotropy of matter fluctuations. This means that
212: $\delta$ and $\veV$ at a same point are statistically
213: independent quantities, as a multivariate probability distribution function
214: (hereafter PDF) of Gaussian variables is completely decided by their
215: covariance
216: matrix ({\it e.g.} Bardeen et al. 1986). Thus the
217: constrained velocity dispersion
218: $ \Sigma_V^2 (\delta)$ does not
219: depend on the density contrast
220: $\delta$ at linear order. However, nonlinear mode couplings would
221: change the situation. Let us
222: examine weakly non-Gaussian effects on $ \Sigma_V^2 (\delta)$. We can
223: express the first nonlinear correction of $\Sigma_V^2 (\delta)$, using
224: framework of the
225: Edgeworth expansion method (Cramer 1946, Matsubara 1994, 1995,
226: Juszkiewicz et al. 1995, Bernardeau \& Kofman 1995). This method is an
227: excellent tool to explore
228: weakly nonlinear effects of the large-scale structure induced by gravity.
229:
230: When a field $F$ is defined by weakly non-Gaussian variables
231: $\{A_\mu(\vex)\}$ with vanishing means, we can expand the expectation
232: value $\lla F \rra$ as (see appendix A)
233: \beq
234: \lla F(A_1,\cdots,A_m)\rra=\lla F \rra_G+\frac16 \sum_{\mu,\nu,\lambda}
235: \lla A_\mu A_\nu A_\lambda \rra_c \lla \frac{\p^3 F}{\p A_\mu\p A_\nu\p
236: A_\lambda } \rra_G +O(\sigma^2 F),
237: \eeq
238: where $\lla \cdot \rra_G$ is the expectation value under the assumption that
239: variables
240: $\{A_\mu(\vex)\}$ are multivariate Gaussian distributed, characterized by
241: their covariance matrix $\lla A_\mu A_\nu\rra$. The quantity
242: $\lla A_\mu A_\nu A_\lambda \rra_c$
243: is the third-order connected
244: moment of variables $\{A_\mu\}$, and we have $\lla A_\mu A_\nu A_\lambda
245: \rra_c=\lla A_\mu A_\nu A_\lambda\rra$ at third-order. The variance
246: $\sigma^2=O(A_i^2)$ is the order
247: parameter of perturbative expansion around the Gaussian distribution, and
248: we can regard $\sigma^2=\lla
249: \delta^2\rra$ in this article. The denominator of $\Sigma_V^2 (\delta)$
250: in equation (3) is nothing but the one point PDF of density contrast $\delta$.
251: From equation (4) we obtain the famous perturbative
252: formula as follows ($\nu\equiv \delta/\sigma$)
253: \beq
254: \lla
255: \delta_{Drc}[\delta(\vex)-\delta]\rra=
256: \frac{e^{-\nu^2/2}}{{\sqrt{2\pi\sigma^2}}}
257: \lmk 1+\frac{S\sigma H_3(\nu )}{6} +O(\sigma^2)\rmk,
258: \eeq
259: ({\it e.g.} Juszkiewicz et al. 1995, Bernardeau \& Kofman 1995) and this
260: is the most
261: simplified version of the
262: Edgeworth expansion.
263: Here the function
264: $H_n(\nu)\equiv (-1)^ne^{\nu^2/2}(d/d\nu)^n e^{-\nu^2/2}$
265: is $n$-th order Hermite polynomial, and $S$ is a parameter of order
266: unity and called skewness (Peebles 1980, Fry 1984, Goroff et al. 1986,
267: see also Seto 1999c),
268: \beq
269: S\equiv \frac{\lla \delta^3\rra}{\sigma^4}.
270: \eeq
271: Due to the nonlinear correction term proportional to $S\sigma$, points with
272: high-$\sigma$ overdensity are more abundant than the linear prediction
273: by a Gaussian
274: distribution.
275:
276:
277: Next the numerator of $\Sigma_V^2 (\delta)$ is given by
278: \beq
279: \lla\veV(\vex)^2
280: \delta_{Drc}[ \delta(\vex)-\delta]\rra=\sigma_V^2
281: \frac{e^{-\nu^2/2}}{{\sqrt{2\pi\sigma^2}}}
282: \lmk 1+\frac{S\sigma H_3(\nu )}{6}+C\sigma
283: H_1(\nu) +O(\sigma^2)\rmk,
284: \eeq
285: where $\sigma_V^2\equiv \lla\veV^2\rra$ is the unconstrained velocity
286: dispersion. The parameter $C=O(1)$ is defined by
287: \beq
288: C=\frac{\lla \veV(\vex)^2\delta(\vex) \rra}
289: {\sigma^2 \sigma_V^2}.
290: \eeq
291: In the studies of the large-scale structure, the Edgeworth expansion or
292: the
293: third-order moments have been mainly discussed for scalar fields, such as
294: density field $\delta(\vex)$ or velocity divergence field $\nabla\cdot
295: \veV(\vex)$ ({\it e.g.} Chodorowski \& {\L}okas 1997). Here we present
296: analytical study for couplings of
297: $\delta(\vex)$ and $\veV(\vex)$, but numerical investigation of our
298: method is also important as well as interesting.
299: From equations (5) and (7) we obtain the constrained velocity dispersion
300: $\Sigma_V^2 (\delta)$ up to the
301: first-order nonlinear correction as
302: \beqa \displaystyle
303: \Sigma_V^2(\delta) &=& \frac{\lla \veV(\vex)^2\delta_{Drc}[\delta(\vex)-\delta]\rra }{\lla \delta_{Drc}[\delta(\vex)-\delta]\rra}\nn\\
304: &=& \frac{ \displaystyle\sigma_V^2
305: \frac{e^{-\nu^2/2}}{{\sqrt{2\pi\sigma^2}}}
306: \lmk 1+\frac{S\sigma H_3(\nu )}{6}+C\sigma
307: H_1(\nu) +O(\sigma^2)\rmk}{ \displaystyle\frac{e^{-\nu^2/2}}{{\sqrt{2\pi\sigma^2}}}
308: \lmk 1+\frac{S\sigma H_3(\nu )}{6} +O(\sigma^2)\rmk}\nn\\
309: &=&\sigma_V^2\lmk1+\frac{S\sigma H_3(\nu )}6+C\delta-\frac{S\sigma H_3(\nu
310: )}6+O(\sigma^2)\rmk \nn\\
311: &=&\sigma_V^2(1+C\delta+O(\sigma^2)).
312: \eeqa
313: Note that our result $\Sigma_V^2(\delta)$ does not depend on the skewness
314: parameter $S$. Nonlinear effects appear through the quantity $C$.
315:
316: Next let us evaluate non-Gausssianity induced by gravity, using
317: higher-order perturbation theory. We perturbatively
318: expand the density and velocity fields as
319: \beqa
320: \delta(\vex)&=&\delta_1(\vex)+\delta_2(\vex)+\delta_3(\vex)+\cdots,\\
321: \veV(\vex)&=&\veV_1(\vex)+\veV_2(\vex)+\veV_3(\vex)+\cdots,
322: \eeqa
323: where $\delta_1(\vex)$ and $\veV_1(\vex)$ are the linear modes,
324: $\delta_2(\vex)$ and $\veV_2(\vex)$ are the second-order modes, and so
325: on. We solve the following three basic equations (continuity,
326: Euler and Poisson equations) order by order (Peebles 1980)
327: \beqa
328: \frac{\p}{\p
329: t}\delta(\vex)+\frac{1}{a}\nabla[\veV(\vex)\{1+\delta(\vex)\}]&=&0,\\
330: \frac{\p}{\p
331: t}\veV(\vex)+\frac1a[\veV(\vex)\cdot\nabla]\veV(\vex)+\frac{\p_t
332: a}a\veV(\vex)+\frac1a\nabla\phi(\vex)&=&0,\\
333: \nabla^2\phi(\vex)-4\pi a^2\rho(t)\delta(\vex)&=&0,
334: \eeqa
335: where $a$ represents the scale factor. In these equations we have
336: omitted explicit time dependence of fields for notational
337: simplicities. We only discuss quantities at a specific epoch and there
338: would be no confusion.
339:
340:
341: Fourier space representation is convenient to analyze the nonlinear
342: mode couplings. We denote the unsmoothed
343: linear Fourier mode by $\delta_{lin}(\vek)$.
344: Then $\delta_1(\vex)$ and $\veV_1(\vex)$ are written in terms of
345: $\delta_{lin}(\vek)$ and $W(kR)$, the Fourier transform of the filter function
346: $W(|\vex|,R)$, as
347: \beq
348: \delta_1(\vex)=\int \frac{d\vek}{(2\pi)^3}
349: \exp(i\vek\vex)\delta_{lin}(\vek)W(kR),~~~
350: \veV_1(\vex)={Hf}\int\frac{d\vek}{(2\pi)^3} \frac{i\vek}{k^2} \exp(i\vek\vex)\delta_{lin}(\vek)W(kR),
351: \eeq
352: where $H(\equiv d\ln a/dt$) is the Hubble parameter and
353: $f(\equiv d\ln D/d \ln a$, $D$: linear growth rate of density
354: fluctuation) is a function of cosmological parameters $\Omega$ and
355: $\lambda$, and well fitted by
356: \beq
357: f\simeq \Omega^{0.6}+\lambda/30,
358: \eeq
359: in the ranges $0.05\le \Omega \le 1.5 $ and $0 \le \lambda \le 1.5 $
360: (Martel 1991).
361:
362: We define the linear matter power spectrum $P(k)$ by
363: \beq
364: \lla \delta_{lin}(\vek)
365: \delta_{lin}(\vel)\rra=(2\pi)^3\delta_{Drc}^3(\vek+\vel)P(k).
366: \eeq
367: Then the dispersions
368: $\sigma^2$ and $\sigma_V^2$ are given by the following simple integrals
369: of $P(k)$ up to required order to evaluate the first nonlinear effects of
370: $\Sigma_V^2(\delta)$,
371: \beqa
372: \sigma^2&=&\int\frac{d\vek}{(2\pi)^3} P(k) W(kR)^2+O(\sigma^4),\\
373: \sigma_V^2&=&H^2f^2\int\frac{d\vek}{(2\pi)^3k^2} P(k) W(kR)^2+O(\sigma^4),
374: \eeqa
375: In this
376: article we only use the Gaussian filter defined by $W(kR)=\exp[-(kR)^2/2]$.
377:
378: As shown in equation (8), the first-order nonlinear correction of the
379: constrained dispersion
380: $\Sigma_V^2(\delta)$ is
381: characterized by the factor $C$. We need the second-order modes
382: $\delta_2(\vex)$ and $\veV_2(\vex)$ to calculate the first nonvanishing
383: contributions of
384: $\lla\veV(\vex)^2\delta(\vex) \rra$.
385: These second-order modes are given with linear mode
386: $\delta_{lin}(\vek)$ as (Fry 1984, Goroff 1986)
387: \beqa
388: \delta_2(\vex)&=&\int \frac{d\vek d\vel}{(2\pi)^6}
389: \exp[i(\vek+\vel)\vex]\delta_{lin}(\vek)
390: \delta_{lin}(\vel)\ch_{2\delta}(\vek,\vel) W(R|\vek+\vel|),\\
391: %%%%%%%%%%%%%%%%%%%%%5
392: \veV_2(\vex)&=&{Hf}\int \frac{d\vek d\vel}{(2\pi)^6}
393: \frac{i(\vek+\vel)}{|\vek+\vel|^2}
394: \exp[i(\vek+\vel)\vex]\delta_{lin}(\vek)
395: \delta_{lin}(\vel)\ch_{2V}(\vek,\vel) W(R|\vek+\vel|),
396: \eeqa
397:
398: where kernels $\ch_{2\delta}$ and $\ch_{2V}$ are defined as follows
399: \beqa
400: \ch_{2\delta}(\vek,\vel)&=&\frac12(1+K)
401: +\frac{\vek\cdot\vel}{2kl}\lmk\frac{k}{l}+\frac{l}k\rmk +\frac12(1-K)
402: \lmk \frac{\vek\cdot\vel}{kl}\rmk^2,\\
403: \ch_{2V}(\vek,\vel)&=&L
404: +\frac{\vek\cdot\vel}{2kl}\lmk\frac{k}{l}+\frac{l}k\rmk +(1-L)
405: \lmk \frac{\vek\cdot\vel}{kl}\rmk^2.
406: \eeqa
407: The factors $K$ and $L$ depend very weakly on cosmological parameters
408: $\Omega$ and $\lambda$, and are fitted as (Matsubara 1995)
409: \beqa
410: K(\Omega,\lambda)&\simeq&\frac37\Omega^{-1/30}-\frac{\lambda}{80}\lmk1
411: -\frac32\lambda\log_{10}\Omega\rmk,\\
412: L(\Omega,\lambda)&\simeq& \frac37\Omega^{-11/200}-\frac{\lambda}{70}\lmk1
413: -\frac73\lambda\log_{10}\Omega\rmk,
414: \eeqa
415: in the ranges $0.1\le\Omega\le 1$ and $0\le\lambda\le 1$. In the
416: followings we neglect these weak dependence and simply put
417: \beq
418: K=L=\frac37.
419: \eeq
420: Thus we can write down the third-order moment
421: $\lla \veV\cdot\veV \delta\rra$ in the following
422: form
423: \beqa
424: \lla \veV\cdot\veV \delta\rra&=&
425: \lla \veV_1\cdot\veV_1\delta_2\rra+2\lla
426: \veV_1\cdot\veV_2\delta_1\rra+O(\sigma^6)\\
427: &=&2H^2f^2\int\frac{d\vek d\vel}{(2\pi)^6}P(k)P(l)\lkk
428: -\frac{\vek\cdot\vel}{k^2l^2}
429: \ch_{2\delta}(\vek,\vel)+2\frac{\vek\cdot(\vek+\vel)}{k^2|\vek+\vel|^2}\ch_{2V}
430: (\vek,\vel)\rkk \nn \\
431: & &\times W(kR)W(lR)W(|\vek+\vel|R)+O(\sigma^6).
432: \eeqa
433: Due to the rotational symmetry around the origin, we can simplify the
434: six dimensional integral $d\vek d\vel$ to three dimensional integral
435: $dkdldu$. Here, $-1\le u \le 1$ is the cosine between two vectors $\vek$
436: and $\vel$ and
437: given by $u=\vek\cdot\vel/kl$. Then we obtain the first nonvanishing
438: order of $ \lla \veV\cdot\veV\delta\rra$ as
439: \beqa
440: \lla \veV\cdot\veV
441: \delta\rra&=&2H^2f^2\int_{-1}^1du\int\frac{k^2l^2dkdl}{8\pi^4}
442: P(k)P(l)\exp[-k^2-l^2-klu] \nn\\
443: & &\times \Bigg[-\frac{u}{kl}\lnk \frac57+\frac{u}2\lmk\frac{k}{l}+\frac{l}k\rmk+\frac27
444: u^2 \rnk\nn\\
445: & &~~~ +2\frac{k+lu}{k(k^2+l^2+2klu)}\lnk \frac37+\frac{u}2\lmk\frac{k}{l}+\frac{l}k\rmk+\frac47
446: u^2 \rnk \Bigg].
447: \eeqa
448: Note that the parameter $C$ does not depend on the normalization of power
449: spectrum (see eqs.[18][19] and [29]).
450: Furthermore, the factors $Hf$ cancel out between $\sigma_V^2$ and $\lla
451: \veV\cdot\veV \delta\rra$ and cosmological parameters are irrelevant for
452: the factor
453: $C$ in our treatment ($K=L=3/7$). Finally, we comment that even though
454: the constrained dispersion
455: $\Sigma_V^2(\delta)$ changes by $\sigma_V^2C\delta$ from the
456: unconstrained value $\sigma_V^2$, the
457: shape of the one-point PDF of velocity field with a given $\delta$
458: keeps Gaussian distribution at the same order of nonlinearity. We can
459: easily confirm
460: this by calculating the ratio
461: \beq
462: \frac{\lla \delta_{Drc}^3(\veV(\vex)-\veV)
463: \delta_{Drc}(\delta(\vex)-\delta)\rra}{\lla
464: \delta_{Drc}(\delta(\vex)-\delta)\rra}
465: =\frac1{(2\cdot3^{-1}\pi\Sigma_V^2(\delta))^{3/2}}\lkk
466: \exp\lmk-\frac{\veV^2}{2\cdot3^{-1}\Sigma_V^2(\delta) } \rmk +O(\sigma^2)\rkk.
467: \eeq
468: The factor $3^{-1}$ in the right-hand side arises from the dimensionality of
469: the velocity vector.
470: First-order correction is completely absorbed to the velocity dispersion
471: $\Sigma_V^2(\delta)$.
472:
473:
474:
475: \section{Results}
476: In this section we numerically evaluate the parameter $C$ for various
477: power spectra.
478: We first examine pure power-law spectra $P(k)$ given
479: by $(n >-1)$
480: \beq
481: P(k)=Ak^n.
482: \eeq
483: In this case $C$ does not depend on the smoothing radius $R$, and we can
484: simply put $R=1$. Then the dispersions $\sigma^2$ and $\sigma_V^2$ are given as
485: \beq
486: \sigma^2=\frac1{(2\pi)^2}\Gamma \lmk \frac{3+n}2\rmk,~~~
487: \sigma_V^2=\frac{(Hf)^2}{(2\pi)^2}\Gamma \lmk \frac{1+n}2\rmk,
488: \eeq
489: where $\Gamma(n)$ is the Gamma function.
490: As for the nonlinear coupling
491: $\lla\veV\cdot\veV\delta\rra=\lla\veV_1\cdot \veV_1 \delta_2 \rra+2\lla\veV_1\cdot \veV_2 \delta_1
492: \rra$, we can write down the first
493: contribution
494: $\lla \veV_1\cdot\veV_1\delta_2\rra$ explicitly in terms of Hypergeometric
495: functions as in the case of skewness parameter $S$ (Matsubara 1994,
496: {\L}okas et al 1995). However, using {\it mathematica} (Wolfram 1996),
497: we confirm that the second term $2\lla \veV_1\cdot\veV_2\delta_1\rra$
498: cannot be expressed in a closed form and numerical integration is required.
499: These two terms diverge in the limit $n\to -1$ where velocity dispersion
500: $\sigma_V^2$ also
501: diverges, but the factor $C$ approaches $0$ in this limit.
502:
503:
504: In Fig.1 we plot $C$ as a function of spectral index $n$ in the range
505: $-1<n<2$.
506: The correction $C$ is a positive and increasing function of $n$.
507: This means that the velocity dispersion of high density regions are larger
508: than that of low density regions.
509: We have $C= 0.314$ at the scale-invariant spectrum $n=1$.
510: % We can expect that difference of velocity dispersions between
511: %points with $\delta=2$ and points with $\delta=0$ is smaller than
512: %$32\%$ (for a fixed $R$).
513:
514:
515: Next we examine $C$ for a more realistic power spectrum $P(k)$. We use CDM
516: transfer function given in Bardeen, Bond, Kaiser \& Szalay (1986)
517: and assume that the primordial spectral
518: index is equal to 1. Then $P(k)$ can be written as
519: \beq
520: P(k)=Ak\lkk\frac{\ln(1+2.34q)}{2.34q}
521: \rkk^2[1+3.89q+(16.1q)^2+(5.46q)^3+(6.71q)^4 ]^{-1/2},
522: \eeq
523: where $q\equiv k/[(\Gamma h){\rm Mpc^{-1}}]$. $\Gamma$ is the shape
524: parameter of the CDM transfer function
525: and recent observational analyses of galaxy clusterings
526: support $\Gamma = 0.2\sim 0.3$ ({\it e.g.} Tadros et al. 1999, Dodelson \& Gazta$\tilde{\rm n}$aga 1999).
527: In Fig.2 we plot $C$ as a function of smoothing radius $R$
528: in units of $[(\Gamma h)^{-1}{\rm Mpc}]$.
529: For this model the factor $C$ depends weakly on the smoothing radius
530: $R$ and we have $C\sim 0.30$ at
531: a weakly nonlinear regime $R\sim 10h^{-1}\mpc$.
532: In the limit $R\to \infty$, $C$ converges to $0.314$ which is the same
533: value of
534: $C$ for the power-law model with $n=1$ presented
535: in Fig.1. This is reasonable as we have
536: \beq
537: \lim_{k\to 0}\frac{P(k)}{k}=const,
538: \eeq
539: for CDM models analyzed here.
540:
541:
542: Our results obtained so far are the velocity dispersion for points
543: constrained by the
544: matter density contrast $\delta$. One might have interest in the velocity
545: dispersion constrained by the galaxy density contrast $\delta_g$.
546: Here, let us assume deterministic but nonlinear biasing relation for the
547: smoothed galaxy
548: distribution $\delta_g(\vex)$ and the matter distribution $\delta(\vex)$ as
549: \beq
550: \delta_g(\vex)=b_1\delta(\vex)+b_2(\delta(\vex)^2-\sigma^2)+O(\sigma^3),
551: \eeq
552: where $b_1$ and $b_2$ are some constants ({\it e.g.} Fry \& Gazta$\tilde{\rm n}$aga 1993). In this case we can easily
553: show that the velocity dispersion $\Sigma_V(\delta_g)^2$ for points $\vex$
554: with $\delta_g(\vex)=\delta_g$ is given by
555: \beq
556: \Sigma_V^2(\delta_g)=\sigma_V^2(1+C\delta_g/b_1+O(\sigma^2)),
557: \eeq
558: where the factors $C$ and $\sigma_V$ are same as those appeared in
559: $\Sigma_V^2(\delta)$ (eq.[9]). Thus $\Sigma_V^2(\delta_g)$ does not
560: depend on the nonlinear coefficient $b_2$. This is also apparent when we
561: write down $\delta(\vex)$ using $ \delta_g(\vex)$ and then insert this
562: solution to equation (9). The factor proportional to $b_2$ is higher
563: effects than analyzed here. Note that in
564: equation (36), the linear bias parameter $b_1$
565: appears by itself not in the usual form $\beta\equiv \Omega^{0.6}/b_1$,
566: and the overdensity $\delta_g$ dependence becomes smaller for larger $b_1$.
567:
568: Kepner et al. (1997) numerically investigated the mean magnitude
569: $\lla|\veV(\vex)|\rra$ of smoothed
570: bulk velocity for points with given overdensity
571: $\delta$.
572: Following the fact commented in the last paragraph of section 2, we can
573: easily calculate this magnitude $ \mu_V(\delta)$ and obtain following
574: result (see appendix B)
575: \beq
576: \mu_V(\delta)=\sqrt{\frac8{3\pi}\Sigma_V^2(\delta)(1+O(\sigma^2))}=\sqrt{\frac8{3\pi}}\sigma_V \lmk 1+\frac{C\delta}2+O(\sigma^2)\rmk.
577: \eeq
578: For a typical CDM model, our analytical result predicts that the magnitude
579: $\mu_V(\delta)$ is expected to change $\sim 0.15\delta$, according to
580: the
581: local density contrast $\delta$. If we constrain points using overdensity
582: of galaxies instead of that of the gravitating matter, the combination
583: $C\delta$ is replaced by
584: $C \delta_g/b_1$ in the above equation.
585:
586: Numerical results of Kepner et al. (1997) were given for
587: CDM and HDM models
588: with $\sigma_8=0.67$ normalization. Here $\sigma_8$ is the linear rms
589: density fluctuation in a sphere of $8\hmpc$ radius.
590: They calculated the smoothed density and velocity fields with smoothing
591: radius $R$ at $0.77\hmpc\le R\le 4.88\hmpc$. Thus their results are
592: quantities at nonlinear regimes. It is true that simple application of our
593: perturbative formula to their results
594: would not be valid. However, surprisingly enough,
595: the quantity $\mu_V(\delta)$ shows almost no $\delta$ dependence in the
596: range $0<\delta\lsim 30$. \footnote{It seems that the function
597: $\mu_V(\delta)$ in their figures shows extremely weak
598: dependence of $\delta$ around $\delta\simeq 0 $.}
599:
600: If $\Sigma_V^2(\delta)$ (and thus $\mu_V(\delta)$) shows no
601: $\delta$-dependence at nonlinear scale and our second-order analysis is
602: valid for
603: weakly nonlinear regime, we confront an interesting possibility.
604: Namely, with parameterization of overdensity by normalized value $\nu\equiv
605: \delta/\sigma$, velocity dispersion does not depend on $\nu$ at linear
606: and nonlinear regime, but depends on it at (intermediate) weakly
607: nonlinear regime. Furthermore, we should notice that the
608: velocity dispersion of dark-matter
609: particles (without coarse graining) depends largely on $\delta$ (Kepner
610: et al. 1997, Narayanan et al. 1998).\footnote{Definitions of velocity
611: dispersion in these two papers are not identical. }
612:
613:
614: To make clear understanding of these transitions,
615: we need to numerically investigate the constrained dispersion
616: $\Sigma_V^2(\delta)$ in
617: detailed manner with various smoothing length $R$, from linear to
618: nonlinear scales. Performance of second-order perturbation theory for
619: the
620: velocity vector is also worth studying.
621:
622:
623:
624: \section{Summary}
625: It is commonly accepted that the large-scale structure observed today
626: is formed by
627: gravitational instability from small primordial density fluctuations
628: (Peebles 1980).
629: In this picture, the peculiar velocity and the density contrast are
630: fundamental quantities to characterize inhomogeneities in the universe.
631: Linear analysis of cosmological perturbation theory predicts that, as long as
632: initial fluctuations are random Gaussian distributed, the
633: one-point PDF of the velocity field $\veV(\vex)$ is statistically
634: independent of the local density contrast $\delta(\vex)$. This is an
635: important
636: aspect of cosmological perturbation.
637:
638: However nonlinear gravitational evolution changes the situation. Due to
639: nonlinear mode-couplings, the peculiar velocity field
640: is no longer statistically
641: independent of the local density field.
642: Here we have investigated bulk velocity dispersion
643: ($\Sigma_V^2(\delta)$) as a function
644: of the local density contrast and calculated its first nonlinear correction
645: using framework of second-order perturbation theory. Our target has
646: been set at
647: weakly nonlinear regimes where perturbative treatment must be
648: reasonable. At present, survey depth of velocity field is
649: highly limited and our constrained statistics
650: might not be directly useful for
651: observational cosmology. However, we believe that our theoretical study
652: is important to understand one interesting aspect of the cosmic velocity field
653: peculiar to its nonlinear evolution.
654:
655: We have shown that the first nonlinear correction of
656: $\Sigma_V^2(\delta)$ is proportional to the local
657: density and strongly depends on the matter power spectra, but weakly on the
658: cosmological parameters $\Omega$ and $\lambda$. For typical CDM model
659: with primordially scale-invariant fluctuations, this first-order
660: correction is about
661: $0.3\delta$. If we use overdensity of galaxies $\delta_g$, this
662: correction term becomes $0.3 \delta_g/b_1$ (the factor $b_1$ is defined in
663: eq.[35]). This dependence might be used to constrain the linear bias
664: parameter $b_1$ itself (not in the usual form $\Omega^{0.6}/b_1$) in future
665: peculiar velocity surveys.
666: We have also shown that the constrained one-point PDF of velocity field
667: keeps the Gaussian shape up to second-order of perturbation. First
668: nonlinear effects are
669: completely absorbed to
670: the velocity dispersion $\Sigma_V^2(\delta)$.
671:
672: Numerical results by Kepner et al. (1997) have been compared with our
673: analytical results.
674: Their results show almost no $\delta$-dependence,
675: contrary to ours.
676: However spatial scale of their analysis (strongly nonlinear regime) is
677: largely different from ours (weakly nonlinear regime).
678: In the forthcoming paper, detailed numerical analysis
679: will be presented for various smoothing lengths $R$ and power
680: spectra (see also Seto 2000). Validity of
681: second-order analysis as well as the Edgeworth expansion method
682: for velocity vector
683: would be also investigated numerically.
684:
685: \acknowledgments
686: The author would like to thank J. Yokoyama
687: for useful discussion and the referee R. Juszkiewicz for helpful comments.
688: This work was partially supported by the Japanese Grant
689: in Aid for Science Research Fund of the Ministry of Education, Science,
690: Sports and Culture No.\ 3161.
691:
692:
693: \newpage
694: \appendix
695:
696: \section{Weakly Non-Gaussian Averages}
697: In this appendix, we derive expression (4) for weakly non-Gaussian
698: variables $\{A_\mu\}~(\mu=1,\cdots,n)$ with $\lla A_\mu\rra=0$ ({\it
699: e.g.} Matsubara 1994).
700: The partition function $Z(J_\mu)$ for a multivariate probability distribution function
701: $P(A_\mu)$ is defined as
702: \beq
703: Z(J_\mu)\equiv \int_{-\infty}^{\infty} d^n P(A_\mu)\exp\lmk i \sum J_\nu A_\nu \rmk.
704: \eeq
705: According to the cumulant expansion theorem (Ma 1985), the function $\ln
706: Z(J_\mu)$ is a generating function of connected moments $\lla A_{\mu_1}
707: \cdots A_{\mu_N}\rra_c$. Therefore, taking the inverse Fourier transform of
708: equation (A1), the probability distribution function $P(A_\mu)$ is
709: written as
710: \beq
711: P(A_\mu)=\exp \lmk \sum_{N=3}^\infty \frac{(-1)^N}{N!}
712: \sum_{\mu_1,\cdots,\mu_N}\lla A_{\mu_1}
713: \cdots A_{\mu_N}\rra_c \frac{\p^N}{\p A_{\mu_1} A\cdots\p A_{\mu_N} }
714: \rmk P_G(A_\mu),
715: \eeq
716: where the function $ P_G(A_\mu)$ is the multivariate Gaussian
717: probability distribution function determined by a ($n\times n$)
718: correlation matrix $M_{\mu\nu}\equiv \lla
719: A_\mu A_\nu\rra$ as
720: \beq
721: P_G(A_\mu)=\frac1{\sqrt{(2\pi)^n {\rm det}M}} \exp\lmk
722: -\frac12 \sum_{\mu,\nu} A_\mu (M^{-1})_{\mu\nu} A_\nu \rmk .
723: \eeq
724: If we have relations $\lla A_{\mu_1}
725: \cdots A_{\mu_N}\rra_c=O(\sigma^{2N-2}) $ as predicted by higher-order
726: perturbation theory, equation (A2) is perturbatively expanded as
727: \beq
728: P(A_\mu)=P_G(A_\mu)-\frac16 \sum_{\mu,\nu,\lambda} \lla A_\mu A_\nu
729: A_\lambda \rra_c \frac{\p^3}{\p A_\mu\p A_\nu\p A_\lambda} P_G(A_\mu)
730: +O(\delta^2).
731: \eeq
732: Evaluating the ensemble average of a field $F(A_\mu)$ with this
733: perturbative expression and taking partial integrals, we
734: obtain expansion (4).
735:
736: One might consider that formula (3) with Dirac's delta function
737: is somewhat indirect. We can obtain same results using formula
738: for probability distribution function $P(\veV|\delta)=P(\veV,\delta)/P(\delta)$
739: and evaluating $P(\veV,\delta)$ and $P(\delta)$ with expression (A4).
740: \section{Derivation of $\mu_V(\delta)$}
741: Here we derive expression (37). As in the case of the constrained
742: velocity dispersion $\Sigma_V^2(\delta)$ (eq.[3]), the quantity
743: $\mu_V(\delta)$ is formally defined as
744: \beq
745: \mu_V(\delta)\equiv \frac{\lla|\veV(\vex)|\delta_{Drc}[\delta(\vex)-\delta]\rra}{\lla\delta_{Drc}[\delta(\vex)-\delta]\rra}.
746: \eeq
747: The r.h.s. of this equation is written as
748: \beq
749: \mu_V(\delta)=\int_{-\infty}^{\infty} d\veV \frac{\lla |\veV| \delta_{Drc}^3(\veV(\vex)-\veV)
750: \delta_{Drc}(\delta(\vex)-\delta)\rra}{\lla
751: \delta_{Drc}(\delta(\vex)-\delta)\rra}.
752: \eeq
753: With the perturbative expansion given in equation (30) the r.h.s. of this equation is evaluated as
754: \beq
755: \frac1{(2\cdot3^{-1}\pi\Sigma_V^2(\delta))^{3/2}}
756: \int_{-\infty}^{\infty} d\veV \lkk
757: \exp\lmk-\frac{\veV^2}{2\cdot3^{-1}\Sigma_V^2(\delta) } \rmk
758: +O(\sigma^2)\rkk |\veV|.
759: \eeq
760: It is straightforward to obtain expression (37) given in the main
761: text as follows
762: \beq
763: \mu_V(\delta)=\sqrt{\frac8{3\pi}\Sigma_V^2(\delta)(1+O(\sigma^2))}=\sqrt{\frac8{3\pi}}\sigma_V
764: \lmk 1+\frac{C\delta}2+O(\sigma^2)\rmk.
765: \eeq
766:
767:
768:
769:
770: %\if0%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
771: \newpage
772: %\begin{references}
773: \begin{thebibliography}{}
774:
775: \bibitem[]{} Bahcall, N. A., Gramann, M., \& Cen, R. 1994, ApJ, 436, 23
776:
777: \bibitem[]{} Bardeen, J. M., Bond, J. R., Kaiser, N. \& Szalay, A
778: S. 1986, ApJ, 305, 15
779: \bibitem[]{} Bernardeau, F. 1992, ApJ, 390, L61
780:
781: \bibitem[]{} Bernardeau, F. \& Kofman L. 1995, ApJ, 443, 479
782:
783: \bibitem[]{} Bernardeau, F., Chodorowski, M., {\L}okas, E. L., Stompor,
784: R. \& Kudlicki A., 1999, MNRAS, 309, 543
785:
786: \bibitem[]{} Cen, R. \& Ostriker, J. P. 1992, ApJ, 399, L113
787:
788: \bibitem{} Chodorowski, M. \& {\L}okas, E. L. 1997,
789: MNRAS, 287, 591
790:
791: \bibitem[]{} Cramer, H. 1946, Mathematical Methods of Statistics
792: (Princeton University Press: Princeton)
793:
794: \bibitem[]{} Davis, M. \& Peebles, P. J. E. 1983, ApJ, 267, 465
795:
796: \bibitem[]{} Dekel, A., 1994, ARA\&A, 32, 371
797:
798: \bibitem[]{} Diaferio, A. \& Geller, M. 1996, ApJ 467, 19
799:
800: \bibitem[]{} Dodelson, S., \& Gazta$\tilde{\rm n}$aga, E. 1999, preprint
801: (astro-ph/9906289)
802: \bibitem[]{} Fisher, K. B. et al. 1994, MNRAS, 267, 927
803:
804: \bibitem[]{} Fry, J. N. 1984, ApJ, 279, 499
805:
806: \bibitem[]{} Fry, J. N. \& Gazta$\tilde{\rm n}$aga, E. 1993, ApJ, 413, 447
807:
808: \bibitem[]{} Goroff, M. H., Grinstein, B., Rey, S. -J., \& Wise, M. B., 1986, ApJ,
809: 311, 6
810:
811: \bibitem[]{} Jing, Y. P., \& B$\ddot{\rm o}$rner, G. 1998, ApJ, 503, 502
812:
813: %\bibitem[]{} Juszkiewicz, R., 1981, MNRAS, 197, 931
814:
815: \bibitem[]{} Juszkiewicz, R. et al. 1995, ApJ, 442, 39
816:
817: \bibitem[]{} Juszkiewicz, R., Fisher, K. B. \& Szapudi, I. 1998, ApJ,
818: 504, L1
819: \bibitem[]{} Juszkiewicz, R., Springel, V. \& Durrer, R. 1999, ApJ,
820: 518, L25
821:
822: \bibitem[]{} Kaufmann, G., et al. 1999, MNRAS, 303, 188
823:
824: \bibitem[]{}Kepner, J. V., Summers, F. J. \& Strauss, M. A. 1997, New
825: Astron., 2, 165
826:
827: \bibitem[]{} {\L}okas, E. L., Juszkiewicz, R., Weinberg, D. H. \& Bouchet, F. R. 1995,
828: MNRAS, 274, 730
829:
830: \bibitem[]{} Ma, S.-K. 1985, {\ Statistical Mechanics} (World
831: Scientific: Philadelphia)
832:
833: \bibitem[]{} Makino, N., Sasaki, M., \& Suto, Y. 1992, Phys. Rev. D,
834: 46, 585
835: \bibitem[]{} Martel, H. 1991, ApJ, 377, 7
836:
837: \bibitem[]{} Matsubara, T. 1994, ApJ, 434, L43
838:
839: \bibitem[]{} Matsubara, T. 1995, Ph.D. Thesis, Hiroshima Univ
840:
841: \bibitem[]{} Narayanan, V. K., Berlind, A., \& Weinberg, D. H. 1998,
842: preprint (astro-ph/9812002)
843: \bibitem[]{} Peebles, P. J. E. 1976, A\&A, 53, 131
844:
845: \bibitem[]{} Peebles, P. J. E. 1980, {\ The Large Scale Structure of the Universe}
846: (Princeton University Press: Princeton)
847:
848: \bibitem[]{}Seto, N. 1999a, ApJ, 516, 527
849: \bibitem[]{}Seto, N. 1999b, ApJ, 520, 409
850: \bibitem[]{}Seto, N. 1999c, ApJ, 523, 24
851: \bibitem[]{}Seto, N. 2000, ApJ, 537 (No.2) in press
852: \bibitem[]{}Seto, N., \& Yokoyama, J. 1998a, ApJ, 492, 421
853: \bibitem[]{}Seto, N., \& Yokoyama, J. 1998b, ApJ, 496, L59
854: \bibitem[]{}Seto, N., \& Yokoyama, J. 1999, PASJ, 51, 233
855:
856: \bibitem[]{} Sheth, R. 1996, MNRAS, 2789, 1311
857:
858: \bibitem[]{} Strauss, M. A. \& Willick, J. A. 1995, Phys. Rep., 261, 271
859:
860: \bibitem[]{} Strauss, M. A., Cen, R. \& Ostriker, J. P. 1998, ApJ, 494, 20
861:
862: \bibitem[]{} Suto, Y., \& Jing, Y. P. 1997, ApJS, 110, 167
863:
864: \bibitem[]{} Tadros, H. et al. 1999, preprint (astro-ph/9901351)
865:
866: \bibitem{} Wolfram, S. 1996, {\ The Mathematica Book, 3rd ed.}
867: (Cambridge University Press: Cambridge)
868:
869: \bibitem[]{} Zurek, W. H. Quinn, P. J., Salmon, J. K. \& Warren,
870: M. S. 1994, ApJ, 431, 559
871:
872: \end{thebibliography}
873: %\fi%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
874:
875: %\if0%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
876: \newpage
877: \begin{figure}[h]
878: \begin{center}
879: \epsfxsize=7.cm
880: \begin{minipage}{\epsfxsize} \epsffile{Fig1.eps} \end{minipage}
881: \end{center}
882: \caption[]{The second-order correction $C$ for power-law matter
883: spectra with Gaussian smoothing. }
884: \end{figure}
885:
886:
887: \begin{figure}[h]
888: \begin{center}
889: \epsfxsize=7cm
890: \begin{minipage}{\epsfxsize} \epsffile{Fig2.eps} \end{minipage}
891: \end{center}
892: \caption[]{The second-order correction $C$ for the CDM
893: spectrum of Bardeen et al. (1986). We plot the factor $C$ as a
894: function of the smoothing radius $R$ in
895: units of
896: $(h\Gamma)^{-1}$Mpc. }
897: \end{figure}
898: \end{document}
899: %\fi%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
900:
901: \newpage
902: \centerline{\bf FIGURE CAPTIONS}
903: \begin{description}
904: \item[Figs.\ 1]The second-order correction $C$ for power-law matter
905: spectra with Gaussian smoothing.
906:
907: \item[Figs.\ 2]The second-order correction $C$ for the CDM
908: spectrum of Bardeen et al. (1986). We plot the factor $C$ as a
909: function of the smoothing radius $R$ in
910: units of
911: $(h\Gamma)^{-1}$Mpc.
912:
913: \end{description}
914:
915:
916:
917: \end{document}
918:
919:
920: