1: \documentclass[preprint2]{aastex}
2: \bibliographystyle{apj}
3:
4: \newcommand{\ApJ}{ApJ}
5: \newcommand{\ApJL}{ApJ Lett.}
6: \newcommand{\ApJS}{ApJ Suppl.}
7: \newcommand{\PRL}{Phys. Rev. Lett.}
8: \newcommand{\PRD}{Phys. Rev. D}
9: \newcommand{\MNRAS}{MNRAS}
10: \newcommand{\jcph}{J.Comp.Phys.}
11: \newcommand{\ARAA}{ARA\&A}
12: \newcommand{\AsAs}{A\&A}
13:
14: \begin{document}
15: \shortauthors{KLYPIN \& PRADA }
16: \shorttitle{Testing gravity with satellites}
17:
18: \title{Testing gravity with motion of satellites around galaxies: Newtonian gravity against
19: Modified Newtonian Dynamics}
20:
21: \author
22: {Anatoly Klypin,${^1}$ and Francisco Prada$^2$}
23: \affil{$^1$Astronomy Department, New Mexico State University,
24: MSC 4500, P.O.Box 30001, Las Cruces, NM, 880003-8001, USA}
25:
26: \affil{$^2$Instituto de Astrof\'\i sica de
27: Andaluc\'\i a (CSIC), Camino Bajo de Huetor, 50, E-18008 Granada, Spain}
28:
29:
30: \date{}
31:
32: \newcommand\LCDM{$\char 3CDM$}
33: \def\simlt{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}
34: \def\simgt{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}
35: \def\ale{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$<$}}}}
36: \def\age{\mathrel{\hbox{\rlap{\hbox{\lower4pt\hbox{$\sim$}}}\hbox{$>$}}}}
37: \def\spose#1{\hbox to 0pt{#1\hss}}
38: \newcommand\lsim{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
39: \raise 2.0pt\hbox{$\mathchar"13C$}}}
40: \newcommand\gsim{\mathrel{\spose{\lower 3pt\hbox{$\mathchar"218$}}
41: \raise 2.0pt\hbox{$\mathchar"13E$}}}
42:
43: \begin{abstract}
44: The motion of satellite galaxies around normal galaxies at distances
45: 50-500~kpc provides a sensitive test for the theories. We study the
46: surface density and the velocities of satellites around isolated red
47: galaxies in the Sloan Digital Sky Survey. We find that the surface
48: number-density of satellites declines with the projected distance as
49: a power law with the slope $-1.5-2$. The rms velocities gradually
50: decline: observations exclude constant velocities at a $\sim
51: 10\sigma$ level. We show that observational data strongly favor the
52: standard model: all three major statistics of satellites -- the
53: number-density profile, the line-of-sight velocity dispersion, and
54: the distribution function of the velocities -- agree remarkably well
55: with the predictions of the standard cosmological model. Thus, that
56: the success of the standard model extends to scales (50-500)~kpc,
57: much lower than what was previously considered. MOND fails on these
58: scales for models which assume any single power-law number-density
59: profile of satellites and any constant velocity anisotropy by
60: predicting nearly constant rms velocities of satellites. Satellite
61: data can be fit by fine-tuned models, which require (1) specific
62: non-power-law density profile, (2) very radial orbits at large
63: distances (velocity anisotropy $\beta =0.6-0.7$ at $R=200-300$~kpc),
64: and (3) 2-2.5 times more stellar mass than what is found in the
65: galaxies. The external gravity force -- a necessary component for
66: MOND -- makes the situation even worse. We argue that a combination
67: of satellite data and observational constraints on stellar masses
68: make these models very problematic.
69: \end{abstract}
70: % User-supplied List of keywords.
71:
72: \keywords{cosmology: theory --- dark matter --- galaxies: halos --- galaxies: structure --- methods: numerical}
73: %------------------------------------------------------------------------------
74: %---------------------
75: \section{Introduction}
76: \label{sec:intro}
77: %---------------------
78:
79: One hundred years after Einstein, the theory of general relativity (GR) is
80: still our best theory of gravity. In the framework of GR, the standard
81: model of cosmology (\LCDM) provides a successful description of the
82: Universe. In this model, the same fluctuations which give rise to the
83: observed small variations in the temperature of the cosmic microwave
84: background (CMB) grow under the force of gravity, and eventually form
85: observed galaxies and other nonlinear structures such as filaments,
86: voids, groups and clusters of galaxies. According to the model, only
87: $\sim 4\%$ of the density in the Universe is provided by normal
88: baryonic matter \citep{WMAP06}. The \LCDM~ model requires two additional
89: components: a non-baryonic cold dark matter (CDM), which contributes
90: about 30\% of the average density of the Universe, and an even more
91: mysterious dark energy, which makes up the rest \citep{WMAP06}. The
92: model is remarkably successful on scales larger than a few
93: Megaparsecs. It predicted the amplitude and the spectrum of
94: angular fluctuations in the CMB and in the distribution of
95: galaxies \citep{Bardeen1987,Holtzman1989} that were later confirmed by
96: observations \citep{WMAP06,Boomerang2002,2dF,Tegmark2004}.
97: However, the \LCDM~ model faces challenges on
98: smaller scales. The most difficult ones are related with the
99: rotation in the inner parts of spiral galaxies. It seems
100: that the theory predicts too much dark matter inside
101: $\sim$~1kpc from the centers of galaxies \citep{Moore1994,Flores1994,deBlok2001}.
102: While there are some possible solutions of the
103: problem \citep{Rhee,Hayashi,Valenzuela}, the problems on small scales
104: are the
105: strongest challenge the standard model has encountered. When
106: compounded with the fact that there is no direct evidence of dark
107: matter or dark energy, the current problems of the standard
108: cosmological model have encouraged a small but growing community of
109: physicists to propose alternative theories of gravity to avoid the
110: need for dark matter.
111:
112: This is the case for Modified Newtonian Dynamics (MOND), proposed by
113: \citet{Milgrom} to explain the rotation of galaxies without dark
114: matter. According to MOND, the rotation curves in the outer regions
115: of galaxies do not decline because the force of gravity is
116: significantly stronger than for Newtonian gravity.
117: At early times MOND's main appeal
118: was its simplicity: there is no need to make the assumption that the Universe is
119: filled with particles that nobody has seen. Additional motivation
120: came later from difficulties with explaining anomalies in the trajectories
121: of the Pioneer 10 and 11 space missions \citep{Pioneer}.
122:
123: Yet, for a long time MOND was not more than a conjecture. Only recently,
124: Bekenstein proposed a relativistic version named tensor
125: vector scalar theory (TeVeS) \citep{Bekenstein2004}. This alternative theory of gravity
126: provides a framework to make predictions of numerous important
127: observational phenomena, which \LCDM~ has already done: the temperature
128: fluctuations seen in the CMB, gravitational
129: lensing, and the large scale structure of the universe. With maturity
130: came problems. Rotation curves of some galaxies -- the
131: initial strong argument for MOND -- cannot be explained by MOND.
132: In about 1/4 of galaxies considered by proponents of MOND the predicted velocities
133: are well above
134: the observations in the very central regions \citep{Sanders2002}.
135: RMS velocities of stars in some dwarf spheroidal galaxies \citep{Lokas2006}
136: also present problems.
137:
138: So far, the most severe challenges for MOND are coming from clusters
139: of galaxies. Dynamics of galaxies in clusters cannot be explained by
140: MOND and requires introduction of dark matter, possibly in the form of
141: a massive ($\sim 2$eV) neutrino \citep{Sanders2002}. We do not know
142: whether this modification can explain some properties of clusters of
143: galaxies such as the ``Bullet Cluster'', where the baryonic mass
144: (galaxies and gas) is clearly separated from the gravitational mass,
145: as indicated by gravitational lensing \citep{Clowe2006,Angus2006}. In
146: any case, for MOND to survive too it must invoke dark matter in the
147: form of massive neutrinos and dark energy in the form of an arbitrary
148: constant added to a combination of two scalar fields used in TeVeS
149: MOND \citep{Bekenstein2004}.
150:
151: There is no doubt that alternative theories of gravity represent a
152: challenge to the standard model of cosmology and GR. Any theory or
153: model must respond to these challenges. Here we present a number of
154: observations to test gravity and dark matter in the peripheral parts
155: of galaxies at distances 50-500~kpc from the centers of
156: galaxies. These scales can be tested by studying the motion of
157: satellites of galaxies. This is a relatively old field in
158: extragalactic astronomy and historically it was one of main arguments
159: for the presence of dark matter \citep{ZW1994,Prada2003}.
160:
161: The paper is organized as follows. In Section 2, we present the
162: observational results drawn from the SDSS and the predictions from the
163: standard model of cosmology. Predictions from MOND are computed and
164: discussed in Section 3. Finally, conclusions are given in Section 4.
165:
166: \begin{figure*}[tb!]
167: \epsscale{2.0}
168: \plotone{f1.eps}
169: \caption{{\it Left panel:} RMS of the line-of-sight velocities of
170: satellites orbiting red isolated galaxies with absolute magnitudes
171: indicated in the plot. The luminosity decreases from top to the
172: bottom curve. Satellites move faster around more luminous
173: galaxies. In all samples of galaxies the rms velocity is declining
174: with distance. {\it Right top panel:} Distribution of observed
175: line-of-sight velocities (dots with error bars) has simple structure
176: of the Gaussian distribution with a small constant background (full
177: curve). The background is due to objects, which in projection lie
178: close to the central galaxy, but that are far from it in 3D
179: space. The dashed curve is the distribution of velocities expected
180: in the standard cosmological model drawn from \LCDM~
181: simulations. {\it Right bottom panel:} Surface number-density of
182: satellites (dots with error-bars; arbitrary units) orbiting galaxies
183: with luminosity $-20.0 < M_g < -21.5$. The dashed line is the power
184: law with a slope $-2$. The full curve is the prediction of the
185: \LCDM~ model.}
186: \end{figure*}
187:
188: %---------------------
189: \section{Observational results}
190: \label{sec:results}
191: %---------------------
192:
193: We use the Sloan Digital Sky Survey (SDSS; www.sdss.org) -- the
194: largest photometric and spectroscopic astronomical survey ever
195: undertaken of the local Universe -- to study the motion of
196: satellites. As of Data Release Four (DR4) \citep{dr4}, imaging data
197: are available over 6670~deg$^2$ in five photometric bands. In addition
198: to the CCD imaging, the SDSS 2.5m telescope on Apache Point, New
199: Mexico, measured spectra of galaxies, providing distance
200: determinations. Approximately half million of galaxies brighter than
201: $r=17.77$ over 4700~deg$^2$ have been targeted for spectroscopic
202: observations as part of SDSS and are included in DR4. Redshift
203: accuracy is better than 30~km/s and the overall completeness is
204: $\sim$90\%. For our study we compute rest frame absolute magnitudes
205: in the g-band from the extinction-corrected apparent magnitudes
206: assuming a $\Lambda CDM$ cosmology with a Hubble constant $h=0.7$
207: ($H_0$ = $100 h {\rm km s}^{-1} {\rm Mpc}^{-1}$). Galaxies are split
208: into red (early-types) and blue (late-types) populations based on the
209: bimodality observed in the $u-r$ color distribution
210: \citep[e.g.]{Baldry2004}. The local minima between the peaks of the
211: color distribution occur near $u-r=2.3$. All magnitudes and colors are
212: k-corrected to $z=0$. Because calculations of MOND gravity for
213: non-spherical objects are complicated, we restrict our analysis only
214: to red galaxies, the vast majority of which are either elliptical
215: galaxies or are dominated by bulges. Our galaxy sample was selected
216: from the full redshift sample by taking all galaxies with recession
217: velocity 3000~km/s~$< cz < 25000$~km/s. The total number of selected
218: galaxies is about 215,000. The SDSS heliocentric velocities were
219: converted to the Local Group standard of rest before computing
220: distances. We select our host galaxies as galaxies with absolute
221: g-band magnitude brighter than $M_g= -19.0$ and isolated: a galaxy
222: must be at least 4 times brighter than any other galaxy within a
223: projected distance $R < 1$~Mpc and a line-of-sight velocity difference
224: $\Delta V< 1500$~km/s. We define satellites as all galaxies being at
225: least 4 times fainter than their hosts and found within a projected
226: distance $R < 1000$~kpc and velocity difference $\Delta V < 1500$~km/s
227: with respect to their hosts. Typically we find about 1.5 satellites
228: per host. In total we have 9500 satellites with a mean luminosity of
229: about $M_g= -18.0$. We bin the host galaxies
230: by luminosity and collect information about the distribution of
231: relative velocities $\Delta V$ and the number of satellites as the
232: function of projected distance $R$.
233:
234: Figure~1 presents the observational results for primaries, which are
235: at least 6 times brighter than any satellite. We also used primaries,
236: which are 4 times brighter, and primaries, which are 10 times brighter
237: than their satellites. We do not find any trend with the
238: primary-satellite magnitude gap: results are nearly the same (see
239: Section 3 for details). The only difference is the statistics, and,
240: thus, the error-bars. The distribution of line-of-sight velocities
241: (top right panel) clearly shows a two-component structure: a
242: homogeneous background of interloper galaxies (dwarfs which happen to
243: lie along the line-of-sight, with large physical distances from the
244: host galaxies but small projected and velocity differences, but which
245: are not associated with the host) and a nearly Gaussian component. The
246: surface density of the satellites also shows the same structure: at
247: large separations the number density goes to a constant due to
248: interlopers. We subtract the background and plot the surface density
249: of satellites in the bottom right panel. In order to study the
250: velocity dispersion of satellites $\Delta V_{\rm rms}^2\equiv \langle
251: \Delta V^2 \rangle$ as a function of projected distance $R$ we use a
252: maximum likelihood method to approximate the number of satellites
253: $N(R,\Delta V)$ per unit projected area and per unit velocity
254: difference $\Delta V$ using an 9-parameter function in the form of a
255: constant plus a Gaussian distribution with variable velocity
256: dispersion and normalization:
257: \begin{equation}
258: {N(R,\Delta V)} =
259: n(R)\exp\left[-{V^2}/{2\sigma^2(R)}\right]/\sqrt{ 2\pi}\sigma+ n_0,
260: \label{eq:Delta}
261: \end{equation}
262: where the surface number-density of satellites $n(R)$ and the rms
263: line-of-sight velocity $\sigma=\Delta V_{\rm rms}(R)$ are 3th order
264: shifted Chebyshev polynomials of the first kind. The parameter $n_0$
265: is a constant representing the background of interlopers. The left
266: panel in Figure~1 shows the resulting rms velocity of satellites for
267: three magnitude bins. In order to estimate statistical uncertainties,
268: we run Monte Carlo simulations using the same number of hosts and
269: satellites as in the corresponding magnitude bin in Figure~1. Note
270: that the data points are correlated. We use Monte Carlo simulations to
271: test statistical significance that the observed velocities are
272: declining with distance: data for each magnitude bin reject a constant
273: $\Delta V_{\rm rms}(R)$ at about $3\sigma$ confidence level. We also
274: studied blue galaxies and find declining velocities for them. These
275: results are in agreement with previous
276: estimates \citep{Prada2003,Conroy2007}, but we now can exclude constant
277: rms velocities much more reliably.
278:
279: \begin{figure*}[tb!]
280: \epsscale{2.0}
281: \plotone{f2.eps}
282: \caption{The line-of-sight velocities of galaxies in two luminosity
283: ranges: $-20.5 < M_g < -21.1$ (2300 satellites) and $-21.1 < M_g <
284: -21.6$ (2700 satellites). The vertical lines indicate $68\%$
285: confidence levels. The left panel shows predictions (full curves)
286: of the standard cosmological model for galaxies hosted by dark
287: matter halos with maximum circular velocities of $\sim 340$~km/s
288: (top curve) and $\sim 270$~km/s (bottom curve). The right panel
289: shows MOND predictions. The two bottom curves are for stellar masses
290: $8\times 10^{10}M_{\odot}$ estimated for galaxies with
291: $M_g=[-20.5,-21.1]$. The full curve is for models with isotropic
292: velocities $\beta=0$ and with the observed slope $\alpha=-3$ of the
293: 3D number density of the satellites. The curve was so much below the
294: observed data points that for other models we decided to use the
295: slope $\alpha=-2.5$, which is marginally compatible with the
296: observations. The dashed curve is for orbits, which are
297: preferentially radial ($\beta={\rm const}=0.5$). The two top curves
298: are for the stellar mass $2\times 10^{11}M_{\odot}$ and for $\beta=0$ (full curve) and
299: $\beta=0.6$ (dashed curve). The \LCDM~ makes quite reasonable
300: predictions, while MOND has problems.}
301: \label{fig:two}
302: \end{figure*}
303:
304: %---------------------
305: \section{Predictions of the \LCDM model}
306: \label{sec:resultsLCDM}
307: %---------------------
308:
309: In order to compare observational results with the \LCDM~ predictions,
310: we use high-resolution cosmological N-body simulations
311: \citep{Prada2006}, which we then treat as if they are the
312: observational data. The simulations were done for the standard \LCDM
313: model with parameters $sigma_8=0.9$, $\Omega_0=0.3$, $h=0.7$,
314: $n=1.$. We use two simulations. Both simulation had $512^3$
315: particles. One simulation had the computational box 120~$h^{-1}$Mpc
316: box, mass resolution $m_1=1.07\times 10^9h^{-1}M_\odot$ and force
317: resolution 1.8~$h^{-1}$kpc. The second simulation has the
318: computational box 80~$h^{-1}$Mpc box, mass resolution $m_1=3\times
319: 10^8h^{-1}M_\odot$ and force resolution 1.2~$h^{-1}$kpc. We do not
320: find any differences between the simulations and decided to present
321: results of the larger simulation, which provides better statistics of
322: halos. We select isolated halos as halos, which within projected
323: distance of 715~kpc and within relative velocity difference 1000~km/s
324: do not have halos or subhalos with maximum circular larger than 1/2 of
325: the halo's circular velocity. This corresponds to the mass ratio of a
326: factor ten for distrinct halos. Circular velocities are better
327: characteristics of subhalos, which have poorly defined
328: masses. Although we resolve subhalos, we dicided to use dark matter
329: particles as proxies for satellites. At distances larger than a
330: fraction (1/4- 1/3) of a virial radius from center of a halo
331: satellites trace the motion and the spacial distribution of the dark
332: matter \citep[e.g.][]{Sales2007}.
333:
334: Results presented in Figures 1 and 2 show that all three
335: characteristics of observed satellites are reproduced by the
336: model. Note that we actually do not make fits. Results from
337: simulations have only one parameter: the maximum circular velocity of
338: the dark matter halo. We just plot what we obtain from the
339: simulations.
340:
341: Once we fix the maximum circular velocity, the results
342: are fairly insensitive to parameters of the cosmological model. Two
343: factors may affect the shape of the theoretical V-R diagram: the halo
344: concentration and the velocity anysotropy. The halo concentrations
345: $c\equiv R_{\rm vir}/R_s$ are in the range $c=8-10$ for halos ($M_{\rm
346: vir}=(2-10)\times 10^{13}M_\odot$) and normalization $\sigma_8$
347: cosidered in this paper \citep{Bullock2001,Neto2007}. (Here $R_{\rm
348: vir}$ is the virial radius and $R_s$ is the characteristic radius of
349: the NFW profile). For virial radii in the range $300-500$~kpc, the
350: exact value of the halo concentration affects only the central region
351: $R<50-70$~kpc. The average halo concentration depends weakly on the
352: amplitude of the perturbations $\sigma_8$. Existing observational data
353: leave only very narrow range for variation of $\sigma_8=0.8-0.9$,
354: which also limits the range of concentrations. The velocity
355: anisotropy is less certain, but it is not large. N-body simulations
356: indicate that $\beta$ is slightly positive and has a tendency to
357: increase with the distance for radii smaller than the virial radius
358: \citep[e.g.][]{Cuesta2007} with typical values $\beta\approx 0.2-0.3$
359: for halo masses considered in this paper. At larger distances $\beta$
360: declines and goes to zero \citep{Cuesta2007}.
361:
362: In order to demonstrate that our results of N-body simulations are
363: robust and reliable, we use solutions of the Jeans equation, which we
364: obtain for parameters compatible with numerous previous simulations.
365: Once a solution for the radial velocity dispersion is obtained, we
366: integrate it along a line of sight with appropriate corrections for
367: the velocity anisotropy. For the density use the NFW profiles with
368: concentration $c=9-10$. For the sake of completeness, we add the
369: cosmological background density, which has only a small effect at
370: large distances. For $\beta$ we use approximation given by
371: eq.~(\ref{eq:betb}) with parameters chosen in such a way that
372: $\beta(0)=0$ and at $R=300$~kpc $\beta =0.2-0.4$. We use the same
373: range of circular velocities as in Figure~\ref{fig:two}. For more
374: massive halos we use $M_{\rm vir}=8.5\times 10^{12}h^{-1}M_\odot$ and
375: $c=9$. For less massive halos we adopted $M_{\rm vir}=3.8\times
376: 10^{12}h^{-1}M_\odot$ and $c=10$. Results presented in
377: Figure~\ref{fig:Jeans} clearly indicate that very simple stationary
378: models can accurately reproduce results of simulations.
379: \begin{figure}[tb!]
380: \epsscale{1.25}
381: \plotone{f3.eps}
382: \caption{
383: Comparison of results of N-body simulations (full curves) with
384: solutions of the Jeans equation (dashed and dot-dashed curves. In
385: the top part of the plot the vertical error bars show observational
386: constraints for galaxies in the luminosity range $-21.1 < M_g <
387: -21.6$. The corresponding full curve is for halos in simulations
388: with maximum circular velocity $\sim 340$~km/s. The two other curves
389: are solutions of stationary spherically symmetric Jeans equation
390: with the same maximum circular velocity. The dashed curve is for
391: slightly more radial velocities (velocity anisotropy $\beta=0.34$ at
392: $R=300$~kpc. The dot-dashed curve of for a model with $\beta=0.17$
393: at the same distance. The bottom curves are for halos with maximum
394: circular velocity $\sim 270$~km/s. The vertical error bars are for
395: galaxies in the luminosity range $-20.5 < M_g < -21.1$. The other
396: curves are labeled in the same way as the top curves.
397: \label{fig:Jeans}
398: }
399: \end{figure}
400:
401: \section{Predictions from MOND}
402:
403: \subsection{Analytics}
404: The situation is different for MOND because there are no predictions
405: on the same level of sophistication as for \LCDM. In principle, those
406: predictions can be made, but at this moment they have not been yet
407: made. Thus, we have only one option: solve the Jeans
408: equation for spherical systems. When doing so, we have a freedom of
409: choosing two functions: the number-density profile of satellites and
410: the velocity anisotropy $\beta(r)$. We also have two free parameters:
411: stellar mass $M_*$ and the magnitude of external force (see below for
412: details). The predictions are constrained by two functions $n(R)$ and
413: $\Delta V(R)$. The velocity distribution function also gives
414: constraints, but those are relatively weak. Only the models with a
415: large $|\beta|> 0.85$ can be excluded. The stellar luminosity and
416: colors constrain the stellar mass. Roughly speaking we have two
417: arbitrary functions to fit two observed functions. It should not be
418: difficult given that the model is viable.
419:
420: In our case the Jeans equation for the radial velocity dispersion
421: $\sigma^2(r)$ can be written in the form:
422:
423: \begin{equation}
424: \frac{d\sigma^2}{dr} + \sigma^2\frac{(2\beta+\alpha)}{r} =-g(r),
425: \alpha\equiv \ln\rho/d\ln r,
426: \label{eq:Jeans}
427: \end{equation}
428: where $\beta(r)=1-\sigma^2_{\perp}/2\sigma^2$ is the velocity
429: anisotropy and $g(r)$ is the gravitational acceleration. The formal
430: solution of the Jeans equation can be written in the form:
431:
432: \begin{equation}
433: \sigma^2 = \frac{1}{\chi (r)\rho(r)}\int_r^{\infty}\chi (r)\rho(r)g(r)dr,
434: \label{eq:JeansSol}
435: \end{equation}
436:
437: \begin{equation}
438: \chi(r) = \exp\left[2\int^r\frac{\beta dr}{r}\right].
439: \label{eq:JeansChi}
440: \end{equation}
441:
442: \begin{figure*}[tb!]
443: \epsscale{2.2}
444: \plottwo{f4a.eps}{f4b.eps}
445: \caption{
446: The line-of-sight velocities of galaxies with $-20.1 < M_g < -20.9$
447: (left panel) and $-21.1 < M_g < -21.6$ (right panel) are compared with
448: MOND models with carefully tuned parameters. We use $M_*=8\times
449: 10^{10}M_{\odot}$ and $M_*=3\times 10^{11}M_{\odot}$ for the two
450: bins. No external gravity force was assumed. Models of MOND
451: can fit the data only when very radial orbits with $\beta_0=1.0$
452: (eq.(7)) are used (dashed curves). Models with more reasonable
453: $\beta_0=0.5$ fail (full curves). Data for more luminous galaxies (right panel)
454: require twice as much of stellar mass as actually observed in the galaxies.
455: \label{fig:Tuned}
456: }
457: \end{figure*}
458:
459: It is convenient to chose forms of $\alpha$ and $\beta$ such that the
460: integral in eq.(\ref{eq:JeansChi}) is taken in elementary functions. We
461: used the following approximations:
462:
463: \begin{equation}
464: \rho = r^{-\alpha}(1+\frac{r}{r_a})^{-\alpha_1},
465: \label{eq:alpha}
466: \end{equation}
467:
468: \begin{equation}
469: \beta = \beta_0, \quad \chi = r^{2\beta_0},
470: \label{eq:beta}
471: \end{equation}
472: \begin{equation}
473: \beta = \beta_0 +\frac{\beta_1}{1+\frac{r}{r_b}}, \quad
474: \chi = r^{2\beta_0}\left[\frac{r}{r_b}\frac{1}{(1+\frac{r}{r_b})}\right]^{2\beta_1}
475: \label{eq:betb}
476: \end{equation}
477:
478: \begin{equation}
479: \beta = \beta_0 +\frac{\beta_1}{1+(\frac{r}{r_b})^2},
480: \chi = r^{2\beta_0}\left[\frac{r}{r_b}\right]^{2\beta_1}
481: \left[\frac{1}{1+(\frac{r}{r_b})^2}\right]^{\beta_1}
482: \label{eq:betc}
483: \end{equation}
484:
485: In the case of Newtonian gravity the acceleration
486: $g=g_N(r)=GM(r)/r^2$, where the mass $M(r)$ includes both normal
487: baryonic mass and dark matter. For MOND the acceleration of spherical
488: systems $g_{\rm MOND}$ is given by the solution of the non-linear
489: equation
490: \begin{equation}
491: g_{\rm MOND}\mu(|{\bf g}_{\rm MOND}+{\bf g}_{\rm ext}|/a_0)=GM_*(r)/r^2,
492: \label{eq:gmond}
493: \end{equation}
494: where $M_*$ is the mass of only baryons and $a_0=1.2\times 10^{-8}{\rm
495: cm}\: {\rm sec}^{-2}$. The term $\bf{g}_{\rm ext}$ is the external
496: constant gravitational acceleration. It formally breaks the spherical
497: symmetry. Thus, the eq.~(\ref{eq:gmond}) is only an approximation
498: valid for small external force. Here we use the same approximation as
499: eq.~(10) in \citet{Sanders2002}, which we average over angles between
500: a constant external acceleration $\bf{g}_{\rm ext}$ and internal
501: radial acceleration $g_{\rm MOND}$.
502:
503: The function $\mu(x)$ can have different shapes. We tried the originally
504: proposed form \citep{Milgrom} $\mu(x)=x/\sqrt{1+x^2}$, but accepted and used
505: for all our analysis the function $\mu(x)=x/(1+x)$, which gives slightly better results.
506: The function $\alpha(r)$ is limited by the observations presented in
507: Figure~1. The velocity anisotropy $\beta(r)$ is a free function, but
508: there are constraints. Asymptotically $\beta$ goes to zero at large
509: distances where gravitational effect of the central galaxy
510: diminishes. Ideally the velocity anisotropy at small distances also
511: should be declining. There are different arguments why this should be
512: the case. (1) The tangential velocities of few satellites of the Milky
513: Way, for which the proper motions are measured, strongly reject radial
514: orbits \citep{Kallivayalil2006, Piatek2007}. There is no reason to
515: believe that our Galaxy should be special in this respect. (2)
516: Experience with gravitational dynamical systems indicate that in
517: dynamically relaxed systems $\beta \approx 0$. Numerous simulations of
518: cosmological models illustrate this: $\beta$ is small in the central
519: region and increases to $0.2-0.3$ at the viral radius \citep{Cuesta2007}. Note that we
520: should distinguish the velocity anisotropy and the orbital
521: eccentricity. For centrally concentrated objects, which we are dealing
522: with here, already isotropic velocities imply typical peri-/apocenter
523: ratios of 1:4-1:5. If eccentricity is larger, a significant fraction
524: of satellites comes too close to the central $\sim 10$~kpc region
525: where the satellites are destroyed by tidal forces.
526:
527: The term $\bf g_{\rm ext}$ in eq.(\ref{eq:gmond}) describes the
528: effect of external gravitational field. It is specific for MOND. In
529: Newtonian dynamics a homogeneous external gravity does not affect
530: relative motions inside the object. Because the MOND gravity is
531: nonlinear, the internal force is affected by the external field (note
532: that this is not the tidal force). This external effect is quite
533: complicated. The magnitude of the
534: external force is substantial for the motion of the
535: satellites. \citet{Fam2007,Wu2007} point out that in MOND numerous
536: sources and effects generate about equal magnitude of $g_{\rm ext}$.
537: For example, 600~km/s motion of the Local Group relative to the CMB
538: implies acceleration of 600~km/s$/10$~Gyrs$=0.015a_0$. Here we assume
539: that the acceleration is constant over the whole age of the Universe.
540: If $g_{\rm ext}$ increases with time, as it may be expected, then the
541: acceleration is even larger. Infall with $250$~km/s in the direction
542: of Virgo gives about half of the value. M31 produces about the same
543: magnitude of the acceleration. While the acceleration in MOND does
544: not add linearly, it is reasonable to assume that $g_{rm
545: ext}>0.015a_0$. We will explore the effect the external field
546: later.
547:
548: \subsection{Comparison of MOND models with observations}
549: In order to make MOND predictions, we must estimate the stellar mass
550: for galaxies in our analysis. For two subsamples presented in
551: Figure~\ref{fig:Tuned} we use the magnitude bins $M_g =-20.5-21.1$
552: (2400 satellites) and $M_g =-21.1-21.6$ (2700 satellites). The
553: average luminosities of galaxies in the bins are $L_g=2.0\times
554: 10^{10}L_{\odot}$ and $L_g=4.0\times 10^{10}L_{\odot}$. When
555: estimating the luminosities, we assume $M_{g,\odot} =5.07$
556: \citep{Blanton01}. Using measured $u-g$ colors we estimate the
557: stellar masses of galaxies \citep{Bell01}: $M_*= 7.2\times
558: 10^{10}M_{\odot}$ and $M_*= 1.5\times 10^{11}M_{\odot}$ for the two
559: luminosity bins. This implies the mass-to-light ratios are nearly
560: the same for the bins: $M/L_g=3.7$. These $M/L$ estimates are close
561: to predictions of stellar population models. In our case, the
562: galaxies are red ($u-g >2.3$), old mostly ellipticals, for which we
563: expect nearly solar metallicity. Indeed, \citet{Maraston05} gives
564: $M/L_B=5$ for the Kroupa IMF for stellar population 10~Gyrs
565: old. Adjusting for the 0.4 mag difference between B and g bands, we
566: get $M/L_g=3.5$ (the Salpeter IMF gives $M/L_g=5.5$). Larger
567: metallicity is highly unlikely (the galaxies are not really massive
568: ellipticals) and it does not make much difference: 10 percent
569: increase if we take twice the solar metallicity.
570:
571:
572:
573: \begin{figure}[tb!]
574: \epsscale{1.0}
575: \plotone{f5.eps}
576: \caption{
577: RMS velocities of satellites for galaxies selected by stellar mass
578: with the average $M_*=2.2\times 10^{11}M_\odot$. The top full curve
579: with the error bars is for a sample of 2600 satellites, which are more
580: than 6 times dimmer than the primary galaxy. The dashed curve is for
581: more stringent isolation criterion of satellites 10 times less bright
582: than the primary (995 satellites). The differences are not
583: statistically significant and are less than 10~km/s for
584: $R>50$~kpc. The bottom curve is for MOND with the same stellar mass as in the observations
585: and with tuned parameters to produce the best fit. The MOND prediction
586: falls much below the observational data.
587: \label{fig:Stmass}
588: }
589: \end{figure}
590:
591: \begin{figure}[tb!]
592: \epsscale{1.0}
593: \plotone{f6.eps}
594: \caption{The ratio of mass required by MOND to the stellar mass for galaxies selected by stellar mass.
595: For small mass galaxies MOND requires masses, which are in reasonable agreement with the observed
596: stellar masses. Yet, for more massive galaxies MOND needs more mass than actually is observed.
597: \label{fig:MONDmass}
598: }
599: \end{figure}
600:
601:
602: Using different functional forms for $\beta(r)$ , we
603: solve the equations numerically and then integrate the solution along
604: the line-of-sight to get a prediction for $\Delta V_{\rm
605: rms}$. Figure~2 presents results for the two different magnitude ranges and
606: for different parameters $\beta=$const and $\alpha=$const.
607: MOND definitely has problems fitting the data because for any constant $\beta$ it predicts a
608: sharply declining rm velocity at small distances followed by nearly
609: flat velocities at large distances. The data show just the opposite
610: behavior. It is easy to understand why predictions of MOND has this
611: shape. At large distances the Newtonian acceleration is very small
612: and the acceleration is strongly dominated by the MOND correction:
613: $g_{\rm MOND}\propto 1/r$. In this case, the solution of eq.(1) gives
614: $\sigma=$const for {\it any} constant value $\beta$ and $\alpha$.
615:
616: We can try to salvage MOND, but so far there is no simple way of doing
617: it. \citet{Angus2007} find that the way to improve the situation is
618: to have a model with variable slope $\alpha$ and with extremely radial
619: orbits. We also tried different functional forms and numerous
620: combinations of parameters. While the parameters, which fit the
621: surface density and the rms velocities vary, they all give the same
622: answer: density slope in the central 40~kpc should be $\alpha >-2.0$
623: and the velocity anisotropy at large distances should be $\beta>0.7$.
624: Figure~\ref{fig:Tuned} shows (dashed curves) the results for the same set of
625: parameters as in \citet{Angus2007}: $\alpha=0$, $\alpha_1=3.1$, $r_a=40$~kpc,
626: $r_b=(20-40)$~kpc, $\beta_0 =1$ and $\beta_1=-2.57$. Here we use the
627: approximation given in eq.(\ref{eq:betb}). The solution is very
628: contrived: relatively small deviations from the best behavior (small
629: slope in the center and radial orbits in the outer radii) result in
630: failed fits. The full curves in the Figure show what happens when the
631: velocity anisotropy gets less radial: $\beta_0 =0.5$.
632:
633: Even this fine-tuned solution fails unless the mass-to-light ratio for
634: the larger magnitude bin is arificially increased by a factor of two:
635: $M/L_g=7.5$, which gives $M_*=3\times 10^{11}M_{\odot}$.
636: \citet{Angus2007} also found the same trend, but they made two
637: mistakes, which did not allow them to clearly see the problem. First,
638: the solar mass-to-light ratio was used for the B band instead of the g
639: band. Second, the width of the magnitude bin is presented as an error
640: in M/L giving impression of very large uncertainties in M/L. This is
641: not correct: the statistical uncertainty of the average luminosity of
642: galaxies in each bin is very low and can be neglected. It should be
643: noted that there is nothing special about the galaxies, which are used
644: here. The average luminosities differ by a factor 1.7. So, it is not a
645: large difference. Colors of the galaxies are practically the same,
646: which then gives the same M/L if we use stellar population models. The
647: rms velocities are also perfectly consistent with simple scaling. For
648: example, the ratio of rms velocities at the same projected distance of
649: 70~kpc is 1.3 implying simple scaling $L\propto \sigma^2R$. Roughly
650: speaking, we double the luminosity and that doubles the stellar
651: mass. This does not work for MOND: it needs twice more stellar mass.
652: This is definitely a problem because there is no justification why
653: galaxies with the same colors, with the same old population and
654: practically the same luminosity should have dramatically different
655: IMF. The differences are very large: most of the stellar mass in the
656: more luminous bin should be locked up in dwarfs with $\sim
657: 0.2M_{\odot}$, while there is relatively little of those in the lower
658: bin, which is consistent with the Kroupa IMF.
659:
660: In order to make the argument even more clear, we make analysis of
661: velocities in a different way. This time we split the sample by
662: stellar mass, but we still keep only red primaries with $u-r >2.3$. We
663: make the analysis twice: for satellites 4 times and for satellites 10
664: times less bright than the primary galaxy. There is no systematic
665: difference between the two isolation conditions: within $1\sigma$ the
666: results are the same. Figure~\ref{fig:Stmass} illustrates this
667: point. We select primary galaxies to have the stellar mass in the
668: range $M_*=(1.6-3.2)10^{11}M\odot$. The average stellar mass is
669: $\langle M_*\rangle =2.2\times 10^{11}M\odot$, and the average
670: luminosity is $\langle L_g\rangle =5.4\times 10^{10}l_\odot$. There
671: is a hint that more isolated primaries have slightly {\it larger}
672: velocities of satellites. Still, the differences are not statistically
673: significant: for radii larger than 50~kpc the differences are smaller
674: than 10~km/s. Then we take the observed stellar mass and use it for
675: MOND models and apply the best tuned parameters. We find that
676: parameters suggested by \citet{Angus2007} ($r_b=40$~kpc, $\beta_0=1$,
677: $\beta_1=-2.57$) improve the fits as compared with a constant $\beta$
678: models. Still, they are not acceptable. For example, at $R=150$~kpc
679: the MOND model is a $7\sigma$ deviation. We found a better solution
680: ($r_b=30$~kpc and other parameters the same), which places MOND
681: ``only'' at $6.2\sigma$.
682:
683: In order to envestigate what stellar mass is needed for MOND, we split
684: the sample of red primaries into 5 mass bins ranging from $M_*\sim
685: 5\times 10^{10}M_\odot$ to $M_*\sim 5\times 10^{11}M_\odot$. We chose
686: the less stringent isolation condition because it gives typically
687: 2-2.5 time more satellites resulting in smaller statistical errors. Each bin has a
688: large number of satellites: $1200-2900$. We then run a grid of MOND
689: models with different masses and select models, which make best fits.
690: Just as the MOND models in Figure~\ref{fig:Tuned}, there are models,
691: which marginally fit the data. For each model we get stellar mass
692: required by MOND and compare it with the stellar mass estimated by the
693: stellar population models. Results are presented in
694: Figure~\ref{fig:MONDmass}. For smaller masses MOND gives masses,
695: which are compatible with actual stellar mass observed in the
696: galaxies. The plot clearly shows the problem with massive galaxies:
697: MOND requires increasingly more mass than observed in the galaxies
698: ending up in large ($\sim 2.5$) disagreement with the observations.
699:
700: The external gravity force $g_{\rm ext}$ is another MOND component. It
701: must exist on the level of $g_{\rm ext}\approx
702: 0.01a_0$. Figure~\ref{fig:TunedExt} shows what happens to the models
703: when we add the external force. As the starting models we use the best
704: fits shown in the Figure~\ref{fig:Tuned} as dashed curves. The models
705: with the external force make the fits much worse. Taken at face value,
706: the models with realistic $g_{\rm ext}=0.015a_0$ can be rejected.
707: Again, we can make the model work if we increase M/L by a factor $\sim
708: 1.5$. Yet, this will make the situation with stellar masses even
709: worse, than what we already have.
710:
711: \begin{figure*}[tb!]
712: \epsscale{2.4}
713: \plottwo{f7a.eps}{f7b.eps}
714: \caption{Effects of external gravity force. Parameters are the same as in
715: Figure~3. For both low (left panel) and high (right panel) luminosity
716: bins we assume the most favorable MOND parameters: radial orbits and
717: adjusted M/L (models shown with dashed curves in
718: Figure~\ref{fig:Tuned}). The magnitude of the external field is shown
719: in the left panel. Models cannot fit the data without additional
720: increase of M/L.
721: \label{fig:TunedExt}
722: }
723: \end{figure*}
724:
725:
726: \section{Discussion and Conclusions}
727: \label{sec:conclusions}
728: %---------------------
729: Using the SDSS DR4 data we study the distribution and velocities of
730: satellites orbiting red isolated galaxies. We find that the surface
731: number-density of the satellites declines almost as a power law with
732: the slope $-2.5-3$. The distribution of the line-of-sight velocities
733: is nearly a perfect Gaussian distribution with a constant component
734: due to interlopers. The rms velocities are found to gradually decline
735: with the projected distance. The constant rms velocity (isothermal
736: solution) can be rejected at a 10~$\sigma$ level.
737:
738: Observational data strongly favor the standard cosmological model: all
739: three major statistics of satellites -- the number-density profile,
740: the line-of-sight velocity dispersion, and the distribution function
741: of the velocities -- agree remarkably well with the theoretical
742: predictions. Thus, the success of the standard model extends to scales
743: (50-500)~kpc, much lower than what was previously considered.
744:
745: MOND fails badly in cases with any acceptable power- law approximation
746: for the number-density of satellites and a constant velocity
747: anisotropy by producing sharply declining velocities at small
748: distances followed by nearly flat velocities at large distances --
749: just the opposite of what is observed in real galaxies. Models may be
750: made to fit the satellites data only when all the following conditions
751: are fulfilled:
752:
753: \begin{enumerate}
754: \item The slope of the density in the central 50~kpc is less then -2.
755: \item Satellite velocities are nearly radial in outer regions ($\beta\approx
756: 0.6-0.7$ at $R=200-300$~kpc).
757: \item Mass-to-light ratio increases from $M/L_g\approx 3.5$ at
758: $M_g=-20.9$ to $M/L_g\approx 12$ ($M/L_B\approx 17$) at $M_g=-22$.
759: \item External force of gravity is smaller than the expected value of $g_{\rm ext}$ $=0.015a_0$.
760: The negative effect of the external force can be removed by
761: additionally increasing the M/L ratios.
762: \end{enumerate}
763:
764: We find that the later three conditions are difficult to realize in
765: nature. Satellites do not fall in to their parent galaxies with zero
766: tangential velocities. The velocities are induced by other
767: neighboring galaxies and by large-scale structures such as
768: filaments. To some degree, it is similar to the tidal torque, which is
769: responsible for the origin of the angular momentum of galaxies. Yet,
770: the interactions are more efficient for providing random velocities.
771: Measurements of peculiar velocities of galaxies the Local Volume
772: (3-5~Mpc around the Milky Way) found deviations from the Hubble flow
773: that about 70-80~km/s \citep{Maccio2005}. At these distances the gravitational pull of
774: other large galaxies in the aria is larger than that of the Milky Way
775: and we expect that the same deviations exist for the perpendicular
776: velocity components. When a satellite with 70-80~km/s falls falls from
777: 1~Mpc and gets to 400~kpc from MW its tangential velocity increases
778: few times (conservation of the angular momentum). So, we expect $\beta
779: <0.5$. We have the same measurements for the satellites in SDSS. When
780: we look at distances 0.8--1~Mpc, we find that the rms velocities
781: hardly correlate with the luminosity and are about 100~km/s. Note that
782: \LCDM perfectly fits the constraint: at these distances it predicts
783: $\beta<0.4$. Thus, MOND assumption of very radial orbits does not
784: agree with the observations.
785:
786: The large M/L ratios is another problem for MOND. So far M/L was
787: treated almost as a free parameter \citep{Sanders07,Angus2007}. This
788: is not correct. It should be reminded that observationally there are
789: no indications of large variability of IMF for vastly different stellar
790: ages and metallicities. The IMF shows similar flattening for masses
791: smaller than $M_{\odot}$ in the solar neighborhood and in the spheroid
792: \citep{Chabrier2001}, in the galactic bulge \citep{Zoccali03}, and in
793: the Ursa Minor dwarf spheroidal galaxy \citep{Feltzing99}. The
794: galactic bulge has parameters similar to the red galaxies studied in
795: this paper: very old (on average) stellar population with nearly solar
796: metallicity. \citet{Zoccali03} give the $M/L_V=3.67$, which
797: corresponds to $M/L_g=3$. This is close to what we found for the red
798: galaxies in SDSS.
799:
800: In the case studied here, MOND requires that M/L changes by a factor 2-2.5 when
801: there is no change in colors and the absolute magnitude changes only
802: by $\Delta M\approx 0.5$. There is no justification for this to
803: happen. The situation is similar to what \citet{Sanders07} found for
804: early type large spirals where MOND models fail for any IMF discussed
805: in recent 20 years. Only the Salpeter IMF gives better results. Even
806: in this case the models deviate by a factor of two around the stellar
807: population predictions. The combination of all these problems makes
808: MOND a very implausible solution for the observational data on
809: satellites of galaxies.
810:
811: There is one significant difference between MOND and \LCDM. The latter
812: makes predictions for the distribution of mass and velocities for
813: isolated galaxies and those predictions match the observational
814: data. There are no theoretical predictions for
815: MOND. What we conveniently called ``MOND predictions'' were actually
816: {\it requirements}. For example, MOND must produce the radial
817: velocities of satellites or shallow slope of the number-density in the
818: central region.
819:
820: \section*{Acknowledgments}
821: We thank S.~McGaugh for challenging us to do MONDian analysis of the
822: satellite motion. We are grateful to J.~Holtzman for numerous
823: comments and suggestions and thank J. Betancort-Rijo for comments. We
824: are especially grateful to Angus et al for providing us with the draft
825: of their paper and for extensive discussions. We still have different
826: ways of looking at the situation with MOND. Still, these productive
827: discussions resulted in improvements in their and our papers. We
828: acknowledge support by the NSF grant AST-0407072 to NMSU and thank the
829: Spanish MEC under grant PNAYA 2005-07789 for their support. Computer
830: simulations used in this research were conducted on the Columbia
831: supercomputer at the NASA Advanced Supercomputing Division and on
832: Seaborg at the National Energy Research Scientific Computing Center
833: (NERSC).
834:
835:
836:
837:
838:
839: \begin{thebibliography}{10}
840: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
841:
842: \bibitem[Adelman-McCarthy~et al. (2006)]{dr4} Adelman-McCarthy, J. K., {\it et~al.}, 2006,
843: {\it Astrophys. J. Supp. Series}, 162, 38
844:
845: \bibitem[Andreson~et al. (2002)]{Pioneer}
846: Anderson {\it et~al.}, J.~A., 2002, {\it Phys. Rev. D}, {\it 65}, 082004
847:
848: \bibitem[Angus, Famaey, \& Zhao (2006)]{Angus2006} Angus, G.~W., Famaey, B.,
849: Zhao, H.~S., 2006, {\it MNRAS}, 371, 138
850:
851: \bibitem[Angus et al.(2007)]{Angus2007} Angus, G., Famaey, B., Tiret, O., Combes, F.,
852: \& Zhao, H.\ 2007, ArXiv e-prints, 709, arXiv:0709.1966
853:
854: \bibitem[Baldry~et al. (2004)]{Baldry2004}
855: Baldry, I. K., {\it et~al.}, 2004, {\it Astrophys. J.}, 600, 681
856:
857: \bibitem[Bardeen~et al. (1987)]{Bardeen1987}
858: Bardeen, J. M., Bond, J. R., Efstathiou, G., 1987, {\it Astrophys. J.}, 321, 28
859:
860: \bibitem[Bekenstein (2004)]{Bekenstein2004} Bekenstein, J.~D., 2004 {\it Phys.Rev. D}, 70, 083509
861:
862: \bibitem[Bell \& de Jong(2001)]{Bell01} Bell, E.~F., \& de Jong, R.~S.\ 2001, \apj, 550, 212
863:
864: \bibitem[Blanton et al.(2001)]{Blanton01} Blanton, M.~R., et al.\ 2001, \aj, 121, 2358
865:
866: \bibitem[Bullock et al.(2001)]{Bullock2001} Bullock, J.~S., Kolatt,
867: T.~S., Sigad, Y., Somerville, R.~S., Kravtsov, A.~V., Klypin, A.~A.,
868: Primack, J.~R., \& Dekel, A.\ 2001, \mnras, 321, 559
869:
870: \bibitem[Chabrier(2001)]{Chabrier2001} Chabrier, G.\ 2001, Dynamics
871: of Star Clusters and the Milky Way, 228, 249
872:
873:
874: \bibitem[Clowe~et al. (2006)]{Clowe2006} Clowe, D., {\it et~al.}, 2006, {\it Astrophys. J. Letter} 648, L109
875:
876: \bibitem[Cole~et al. (2005)]{2dF} Cole, S., {\it et~al.}, 2005, {\it MNRAS}, 362, 505
877:
878: \bibitem[Conroy~et al. (2007)]{Conroy2007} Conroy, C., {\it et~al.}, 2007, {\it Astrophys. J.}, 654, 153
879:
880: \bibitem[Cuesta et al.(2007)]{Cuesta2007} Cuesta, A.~J., Prada,
881: F., Klypin, A., \& Moles, M.\ 2007, ArXiv e-prints, 710, arXiv:0710.5520
882:
883: \bibitem[de Blok~et al. (2001)]{deBlok2001} {de Blok}, W.~J.~G., McGaugh, S.~S., Rubin, V.~C., 2001,
884: {\it Astron. J.}, 122, 2396
885:
886: \bibitem[Famaey \& Binney (2005)]{Famaey2005} Famaey, B., \& Binney, J., 2005, {\it MNRAS}, 363, 603
887:
888: \bibitem[Famaey et al.(2007)]{Fam2007} Famaey, B., Bruneton,
889: J.-P., \& Zhao, H.\ 2007, \mnras, 377, L79
890:
891: \bibitem[Famaey et al.(2007)]{Famaey2007} Famaey, B., Angus,
892: G.~W., Gentile, G., \& Zhao, H.~S.\ 2007, ArXiv e-prints, 706, arXiv:0706.1279
893:
894: \bibitem[Feltzing et al.(1999)]{Feltzing99} Feltzing, S., Gilmore, G., \& Wyse, R.~F.~G.\ 1999, \apjl, 516, L17
895:
896: \bibitem[Flores \& Primack (1994)]{Flores1994} Flores, R.~A., \& Primack, J.~R., 1994, {\it Astrophys. J. Letters}, 427, L1
897:
898: \bibitem[Hayashi \& Navarro (2006)]{Hayashi} {Hayashi}, E., \& {Navarro}, J.~F., 2006, {\it MNRAS}, 373, 1117
899:
900: \bibitem[Holtzman (1989)]{Holtzman1989}
901: Holtzman, J., 1989, {\it Astrophys. J. Supp. Series}, 71, 1
902:
903: \bibitem[Kallivayalil et al.(2006)]{Kallivayalil2006} Kallivayalil, N.,
904: van der Marel, R.~P., Alcock, C., Axelrod, T., Cook, K.~H., Drake, A.~J.,
905: \& Geha, M.\ 2006, \apj, 638, 772
906:
907: \bibitem[Lokas, Mamon, \& Prada (2006)]{Lokas2006}
908: {\L}okas, E., Mamon, G., Prada, F., 2006, {\it EAS Publications Series}, 20, 113
909:
910: \bibitem[Macci{\`o} et al.(2005)]{Maccio2005} Macci{\`o}, A.~V.,
911: Governato, F., \& Horellou, C.\ 2005, \mnras, 359, 941
912:
913:
914: \bibitem[Maraston(2005)]{Maraston05} Maraston, C.\ 2005, \mnras, 362, 799
915:
916: \bibitem[Milgrom (1983)]{Milgrom} Milgom, M., 1983, {\it Astrophys. J.}, 270, 365
917:
918: \bibitem[Moore (1994)]{Moore1994} Moore, B., 1994, {\it Nature} 370, 629
919:
920: \bibitem[Neto et al.(2007)]{Neto2007} Neto, A.~F., et al.\ 2007,
921: \mnras, 381, 1450
922:
923: \bibitem[Netterfield~et al. (2002)]{Boomerang2002}
924: Netterfield, C.~B., {\it et al.}, 2002, {\it Astrophys. J.}, 571, 604
925:
926: \bibitem[Piatek et al.(2007)]{Piatek2007} Piatek, S., Pryor, C.,
927: Bristow, P., Olszewski, E.~W., Harris, H.~C., Mateo, M., Minniti, D.,
928: \& Tinney, C.~G.\ 2007, \aj, 133, 818
929:
930: \bibitem[Prada~et al. (2003)]{Prada2003}
931: Prada {\it et~al.}, F., 2003, {\it Astrophys. J.}, 598, 260
932:
933: \bibitem[Prada~et al. (2006)]{Prada2006}
934: Prada, F., {it et~al.}, 2006, {\it Astrophys. J.}, 645, 1001
935:
936: \bibitem[Rhee~et al. (2004)]{Rhee} {Rhee}, G., {Valenzuela}, O., {Klypin}, A., {Holtzman}, J., {Moorthy}, B., 2004,
937: {\it Astrophys. J.}, 617, 1059
938:
939: \bibitem[Sales et al.(2007)]{Sales2007} Sales, L.~V., Navarro,
940: J.~F., Lambas, D.~G., White, S.~D.~M.,
941: \& Croton, D.~J.\ 2007, \mnras, 382, 1901
942:
943:
944:
945: \bibitem[Sanders \& McGaugh (2002)]{Sanders2002} {Sanders}, R.~H. \& McGaugh, S.~S., 2002,
946: {\it Ann.Rev.Astron. \& Astrophys} 40, 263
947:
948: \bibitem[Sanders \& Noordermeer(2007)]{Sanders07} Sanders, R.~H.,
949: \& Noordermeer, E.\ 2007, \mnras, 379, 702
950:
951: \bibitem[Spergel~et al.(2006)]{WMAP06}
952: {Spergel}, D. N., {\it et al.}, 2006,
953: {\it ArXiv Astrophysics e-prints}, arXiv:astro-ph/0603449
954:
955: \bibitem[Tegmark~et al. (2004)]{Tegmark2004} Tegmark, M., {\it et~al.}, 2004, {\it Astrophys. J.} 606, 702
956:
957: \bibitem[Valenzuela~et al. (2007)]{Valenzuela} {Valenzuela}, O., {Rhee}, G., {Klypin}, A.,
958: {Governato}, F., {Stinson}, G., {Quinn}, T., \& {Wadsley}, J., 2007,
959: {\it Astrophys. J.} 657, 773
960:
961:
962: \bibitem[Wu et al.(2007)]{Wu2007} Wu, X., Zhao, H., Famaey,
963: B., Gentile, G., Tiret, O., Combes, F., Angus, G.~W., \& Robin, A.~C.\
964: 2007, \apjl, 665, L101
965:
966: \bibitem[Zaritsky \& White (1994)]{ZW1994} Zaritsky, D., \& White, S.~D.~M., 1994, {\it Astrophys. J.}, 435, 599
967:
968: \bibitem[Zoccali et al.(2003)]{Zoccali03} Zoccali, M., et al.\ 2003, \aap, 399, 931
969: \end{thebibliography}
970:
971: \end{document}