1: \documentclass[useAMS,usenatbib]{mn2e}

2: \bibliographystyle{mn2e}

3: %\def\baselinestretch{1}

4: %\usepackage[dvips]{graphicx}

5: \usepackage{graphicx}

6: \usepackage{times}

7: \usepackage{amssymb}

8: \usepackage{natbib}

9: %%%%%%%%%% user-defined commands %%%%%%%%%%%%%%%%%

10: \def\gsim { \lower .75ex \hbox{$\sim$} \llap{\raise .27ex \hbox{$>$}} }

11: \def\lsim { \lower .75ex \hbox{$\sim$} \llap{\raise .27ex \hbox{$<$}} }

12: %%%%%%%%%%%%%%%%%%%%%%%%%%%

13: 

14: \newcommand{\apj}{ApJ}

15: \newcommand{\apjl}{ApJL}

16: \newcommand{\apjs}{ApJS}

17: \newcommand{\aj}{AJ}

18: \newcommand{\mnras}{MNRAS}

19: \newcommand{\nat}{Nature}

20: \newcommand{\pasj}{PASJ}

21: \newcommand{\pasp}{PASP}

22: \newcommand{\araa}{ARA\&A}

23: \newcommand{\aapr}{A\&ARv}

24: \newcommand{\aap}{A\&A}

25: \newcommand{\prd}{PhRvD}

26: \newcommand{\physrep}{PhysRep}

27: 

28: %%%%%%%%%% end user-defined commands %%%%%%%%%%%%%%%%%

29: 

30: 

31: \begin{document}

32: 

33: \title[Satellites and Fossil Groups in the Millennium Simulation]

34: {Satellite Galaxies and Fossil Groups in the Millennium Simulation}

35: 

36: \author[Sales et al.]{

37: \parbox[t]{\textwidth}{

38: Laura V. Sales$^{1,2}$,

39: Julio F. Navarro$^{3,4}$\thanks{Fellow of the Canadian Institute for Advanced Research},

40: Diego G. Lambas$^{1,2}$,

41: Simon D. M. White$^{4}$ \&

42: Darren J. Croton$^{5}$

43: }

44: \vspace*{6pt} \\

45: \\

46: $^{1}$ Observatorio Astron\'omico, Universidad Nacional de C\'ordoba, Laprida

47: 854, 5000 C\'ordoba, Argentina.

48: \\

49: $^{2}$ Instituto de Astronom\'{\i}a Te\'orica y Experimental, Conicet, Argentina.\\

50: $^3$Department of Physics and Astronomy, University of Victoria, B.C., Canada.\\

51: $^4$Max-Planck-Institut f\"ur Astrophysik, D-85740 Garching, Germany.\\

52: $^5$Department of Astronomy, University of California, Berkeley, CA, 94720, USA.}

53: 

54: \maketitle

55: 

56: 

57: \begin{abstract}

58: We use a semianalytic galaxy catalogue constructed from the {\it

59: Millennium Simulation} ({\it MS}) to study the satellites of isolated

60: galaxies in the $\Lambda$CDM cosmogony. The large volume surveyed by

61: the {\it MS} ($500^3 \, h^{-3}{\rm Mpc}^3$), together with its

62: unprecedented numerical resolution, enable the compilation of a large

63: sample of $\sim 80,000$ bright ($M_r<-20.5$) primaries, surrounded by

64: $\sim 178,000$ satellites down to the faint magnitude limit

65: ($M_r=-17$) of our catalogue. This sample allows the characterization,

66: with minimal statistical uncertainty, of the dynamical properties of

67: satellite/primary galaxy systems in a $\Lambda$CDM universe. The

68: details of this characterization are sensitive to the details of the

69: modeling, such as its assumptions on galaxy merging and dynamical

70: friction timescales, but many of its general predictions should be

71: applicable to hierarchical formation models such as $\Lambda$CDM. We

72: find that, overall, the satellite population traces the dark matter

73: rather well: its spatial distribution and kinematics may be

74: approximated by an NFW profile with a mildly anisotropic velocity

75: distribution.  Their spatial distribution is also mildly anisotropic,

76: with a well-defined ``anti-Holmberg'' effect that reflects the

77: misalignment between the major axis and angular momentum of the host

78: halo. Our analysis also highlights a number of difficulties afflicting

79: studies that rely on satellite velocities to constrain the primary

80: halo mass. These arise from variations in the star formation

81: efficiency and assembly history of isolated galaxies, which result in

82: a scatter of up to $\sim 2$ decades in halo mass at fixed primary

83: luminosity. Our isolation criterion (primaries may only have

84: companions at least 2 mag fainter within $1 \, h^{-1}$ Mpc)

85: contributes somewhat to the scatter, since it picks not only galaxies

86: in sparse environments, but also a number of primaries at the centre

87: of ``fossil'' groups. We find that the abundance and luminosity

88: function of these unusual systems are in reasonable agreement with the

89: few available observational constraints. Much tighter halo

90: mass-luminosity relations are found when splitting the sample by

91: colour: red primaries inhabit halos more than twice as massive as

92: those surrounding blue primaries, a difference that vanishes, however,

93: when considering stellar mass instead of luminosity. The large scatter

94: in the halo mass-luminosity relation hinders the interpretation of the

95: velocity dispersion of satellites stacked according to the luminosity

96: of the primary. We find $L\propto \sigma^3$ (the natural scaling

97: expected for $\Lambda$CDM) for truly-isolated primaries, i.e., systems

98: where the central galaxy contributes more than $85\%$ of the total

99: luminosity within its virial radius. Less strict primary selection,

100: however, leads to substantial modification of the scaling relation:

101: blindly stacking satellites of all primaries results in a much

102: shallower $L$-$\sigma$ relation that is only poorly approximated by a

103: power law.

104: \end{abstract}

105: 

106: \date{}

107: 

108: \begin{keywords}

109: galaxies: haloes - galaxies: kinematics and dynamics - dark matter -

110: methods: statistical

111: \end{keywords}

112: 

113: 

114: \section{INTRODUCTION}

115: \label{sec:intro}

116: 

117: Satellite galaxies may be thought of as the visible fossil relics of

118: hierarchical galaxy formation, where the mass of a galaxy is

119: envisioned to be assembled in a sequence of accretion events. As

120: surviving witnesses of the accretion process, satellites bear an

121: invaluable record of the assembly history of the primary galaxy they

122: orbit, and provide at the same time prime information about the mass

123: and extent of the dark matter halo they inhabit.

124: 

125: Satellites are a particularly valuable tool for studying the outer

126: regions of dark matter halos, where few other tracers exist that can

127: provide effective constraints. This is especially true in the case of

128: the Local Group, where the dynamics of the outer satellites has played

129: an important role in mass estimates of the Milky Way and M31

130: \citep{littleandtremaine87,zaritsky89,kochanek96,wilkinson99,

131: evans00,battaglia05}. Local Group satellites can be found even if they

132: are extremely faint; Draco, for example, is $\sim 8 \times 10^{4}$

133: times fainter than its primary, the Milky Way, and some of the new

134: satellites discovered over the past few years have luminosities

135: rivaling that of ordinary star clusters

136: \citep{zucker04,zucker06,willman05b,martin06,belokurov06,

137: belokurov07,irwin07,majewski07,ibata07,chapman07}.

138: 

139: On the other hand, extragalactic satellite studies have been

140: traditionally limited by the scarcity of satellite/primary systems

141: easily accessible to observation. Partly as a result of the strict

142: isolation criteria that are imposed on the primaries in order to

143: minimize interlopers and to avoid complications that may arise from

144: having more than one dominant object, it is rare for primaries to have

145: more than a few satellites bright enough for spectroscopic

146: confirmation and follow up.

147: 

148: The scarcity of satellites surrounding one given primary has prompted

149: the adoption of stacking techniques in order to overcome small number

150: statistics. For example, the satellites of all primaries of given

151: luminosity, $L$, might be combined to yield estimates of the {\it

152: average}, rather than individual, halo properties as a function of

153: $L$.  These techniques are clearly vulnerable to the presence of

154: luminosity-dependent biases in the satellite distribution and of

155: systematic trends between halo mass and primary luminosity that may be

156: difficult to detect and evaluate in observational datasets.

157: 

158: Such techniques, nevertheless, seem appropriate to address a number of

159: important issues in galaxy formation studies.  How does the average

160: dark halo mass depend on the luminosity of the primary? What is the

161: mass profile of the dark halo and how does it vary with luminosity?

162: What is the three-dimensional shape of the dark halo and how does it

163: relate to the primary and to the spatial distribution of satellites?

164: These are some of the questions traditionally addressed by extragalactic

165: satellite studies, and there is a rich literature documenting prior

166: attempts at exploiting the dynamics of satellites to constrain the

167: mass and extent of dark matter halos.

168: 

169: The pioneering work of \citet{zaritsky93,zaritsky97a}, for example,

170: confirmed the presence of massive halos around isolated spirals, 

171: but also hinted at a few odd results that are difficult to 

172: reconcile with current galaxy formation models. For example, these 

173: authors noted that satellite velocities seemed to be independent 

174: of the luminosity of the primary, contrary to what might be naively

175: expected from the Tully-Fisher relation. They also remarked that 

176: satellites seemed to be distributed anisotropically around the 

177: primary, in agreement with the early suggestion of \citet{holmberg69}.

178: 

179: These issues have been revisited using the much larger datasets

180: compiled by the 2dfGRS \citep{colless01} and the SDSS

181: \citep{york00,strauss02} surveys, and broad consensus seems to be

182: gradually emerging.  In particular, \citet{mckay02} and

183: \citet{prada03} find that a well-defined trend between satellite

184: velocities and primary luminosity {\it does} appear when considering

185: samples substantially larger than the ones considered by Zaritsky et

186: al. Better statistics have also clarified the anisotropic distribution

187: of satellites around spirals, which, contrary to Holmberg's

188: suggestion, apparently tend to avoid the rotation axis of the disk

189: \citep{brainerd05,azzaro06,yang06,agustsson07}.  Finally, the

190: availability of larger samples have also enabled studies of the

191: satellite velocity dispersion profile, which may be used to probe the

192: outer dark mass distribution and to compare it with cosmological

193: N-body simulations (see \citealt{prada03}, also the review of

194: \citealt{brainerd04b}, and van den Bosch et al. 2005a,b).

195: \nocite{vandenbosch05a,vandenbosch05b}

196: 

197: The success of these studies depends crucially on identifying and

198: understanding possibly subtle biases between dark matter and the

199: satellite population. This is best accomplished through direct

200: numerical simulations, where full 3D dynamical information is

201: available and from which mock observational datasets may be created to

202: assess the ability of analysis techniques to identify and correct for

203: such biases. Progress, however, has been slow, mainly because

204: simulations with enough dynamic range to resolve simultaneously a

205: galaxy and its satellites have only recently become possible

206: \citep{klypin99a,moore99,kravtsov04,gao04b,diemand04,maccio06,

207: diemand07,libeskind07,sales07a,sales07b}. Further, these simulations

208: have typically followed small volumes (single halos, in many cases)

209: and many of them have focussed on the dark matter component only,

210: hindering direct comparisons between theory and observation. 

211: 

212: We investigate these issues here using the galaxy catalogue created

213: by \citet{croton06} from the {\it Millennium Simulation} 

214: (\citealt{springel05}, hereafter {\it MS} for short).  The catalogue is the

215: result of a sophisticated semianalytic galaxy formation model applied

216: to an N-body simulation of unprecedented dynamic range. This model has

217: been calibrated to reproduce many of the large-scale properties of the

218: observed galaxy population, such as the luminosity and correlation

219: functions, as well as their color dependence. However, no specific

220: information about satellite systems has been taken into account and,

221: therefore, the results we present here may be considered true

222: theoretical ``predictions'' that can be contrasted fruitfully with

223: observation.

224: 

225: Our principal aim is to provide a detailed characterization of the 3D

226: properties of the satellite population of bright, isolated galaxies in

227: the $\Lambda$CDM cosmogony. This characterization is intended (i) to

228: guide the interpretation of observational datasets, (ii) to improve

229: the identification of primary/satellite systems so as to minimize the

230: contamination by interlopers, and (iii) to provide a framework within

231: which the questions posed above may be profitably addressed.  The

232: sheer size of the catalogue, which lists $\sim 10^7$ galaxies brighter than

233: $M_r=-17$ in the $500^3 h^{-3}$ Mpc$^3$ volume of the {\it MS}

234: {\footnote{We express the present-day value of the Hubble constant as

235: $H_0=100\, h$ km/s/Mpc.  Throughout this paper, masses and

236: luminosities will assume $h=0.73$ unless otherwise specified.}, ensures

237: that our results have negligible statistical uncertainty.

238: 

239: Our plan for the paper is as follows. After a brief description of the

240: {\it MS} (\S~\ref{ssec:MS}) and of the semianalytic galaxy formation

241: model (\S~\ref{ssec:sam}), we discuss our selection criteria for

242: primaries and for satellites in \S~\ref{ssec:gxcat}. Our main results

243: are presented and discussed in \S~\ref{sec:results}. We begin in

244: \S~\ref{ssec:lmc} with a discussion of the relation between halo mass,

245: primary luminosity and color. Noting that our isolation criteria for

246: primaries allows for the inclusion of ``fossil groups'' in our sample,

247: we compare their abundance and luminosity function to observations in

248: \S~\ref{ssec:fossil}. The satellites' spatial distribution is

249: discussed next (\S~\ref{ssec:satprof}), together with their kinematics

250: (\S~\ref{ssec:satvel}). We end the discussion of our results by

251: analyzing the relation between primary luminosity and satellite

252: velocity dispersion (\S~\ref{ssec:lsg}), as well as the presence of

253: anisotropies (\S~\ref{ssec:holm}) in the satellite spatial

254: distribution. We conclude with a brief summary in \S~\ref{sec:conc}.

255: 

256: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

257: \begin{figure*}

258: \begin{center}

259: \includegraphics[width=\linewidth,clip]{figures/fig1_lr.ps}

260: \end{center}

261: \caption{Virial mass vs r-band luminosity relation for primary

262: galaxies.  Panel $(a)$ shows data for all primaries, other panels are

263: subsamples after applying a color cut (see labels). The total number

264: of primaries plotted in each panel is quoted.  Colors indicate the

265: density of primaries at various locations in each panel, in

266: logarithmic units; each color step corresponds to a variation of about

267: 10 in number. The same applies to other panels, although the color

268: coding has been renormalized in each case to make use of the whole

269: color palette.  Dotted, solid and dashed lines indicate loci of

270: constant virial mass-to-light ratio, and are chosen to ease the

271: comparison between panels and to outline the presence of three

272: ``sequences'' in the data (see text for further discussion).}

273: \label{figs:lmc}

274: \end{figure*}

275: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

276: 

277: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

278: \begin{figure*}

279: \begin{center}

280: \includegraphics[width=0.475\linewidth,clip]{figures/mvir_lum_seqhost.ps}

281: \hspace{0.4cm}

282: \includegraphics[width=0.475\linewidth,clip]{figures/mvir_lum_seqiso.ps}

283: \caption{Virial mass distribution in luminosity bins for {\it host}

284: (left panel) and {\it single} (right panel) primaries, split according

285: to color (see Figure~\ref{figs:lmc}). Histograms have been normalized

286: to the total number of galaxies in each luminosity bin. For {\it

287: hosts} we have: $(12,747)$, $(23,585)$, $(26,039)$, $(13,817)$, $(2,436)$ and

288: $(454)$ galaxies in each of the $6$ bins of increasing $L$,

289: respectively. For {\it single} primaries, the corresponding numbers

290: are : $(91,060)$, $(58,364)$, $(23,893)$, $(4,742)$, $(361)$ and $(66)$,

291: respectively.  Percentages quoted in the right column of each panel indicate 

292: the fraction of primaries in each color sequence for a given luminosity bin.

293: Left columns show the rms values $\sigma(\rm log(M_{200}/M_\odot))$ 

294: corresponding to each subsample.}

295: \label{figs:lmcdist}

296: \end{center}

297: \end{figure*}

298: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

299: 

300: \section{The Catalogue}

301: 

302: \subsection{The Millennium Simulation}

303: \label{ssec:MS}

304: 

305: The {\it{Millennium Simulation}} ({\it MS}), one of the projects of

306: the Virgo Consortium \footnote{http://www.virgo.dur.ac.uk}, is a

307: cosmological N-body simulation of the $\Lambda$CDM universe that

308: follows the evolution of more than 10 billion particles in a box of

309: 500 $h^{-1}$ Mpc (comoving) on a side. The particle mass is $8.6

310: \times 10^9 \, h^{-1} M_\odot$, and particle-particle gravitational

311: interactions are softened on scales smaller than 5 $h^{-1}$ kpc.  The

312: simulation adopts parameters consistent with the WMAP1 results

313: \citep{spergel03,seljak05}: $\Omega_{\rm m}=0.25$,

314: $\Omega_{\Lambda}=0.75$, $h=0.73$, $n=1$, and $\sigma_8=0.9$ (for

315: details see Springel et al. 2005). Data from the simulation output

316: at 64 times spaced logarithmically in expansion factor before $z=1$,

317: and at approximately $200$ Myr intervals thereafter. These data are

318: used to build merger trees which encode the assembly history of each

319: halo and its resolved substructure. These form the basis for the

320: galaxy formation modeling. Complete data for these halo merging trees

321: as well as for several galaxy formation models can be found at {\small \tt

322: http://www.mpa-garching.mpg.de/millennium}. The $z=0$ data from the

323: Croton et al (2006) model used here can be found at 

324: {\small \tt http://www.mpa-garching.mpg.de/galform/agnpaper/}.

325: 

326: \subsection{The Semianalytic Model}

327: \label{ssec:sam}

328: 

329: The information contained in the {\it MS} is harvested using

330: semianalytic galaxy formation models that follow, with simple, but

331: physically motivated, laws the formation of galaxies within the

332: evolving dark matter halos. The actual prescriptions used derive from

333: the work of \citet{kauffmann99}, \citet{springel01} and

334: \citet{delucia04b}, but have been reformulated in the model of Croton

335: et al (2006) to take into account newer observational constraints.

336: % and is able to reproduce, within a

337: %hierarchical cosmogony and with the help of various forms of feedback,

338: %the ``down-sizing'' of galaxies with redshift.

339: 

340: In brief, the model tracks the build-up of the luminous component of

341: each dark matter halo by prescribing how gas cools and transforms into

342: stars, as well as how enriched gas is devolved to the interstellar

343: medium in halos of various masses and at different stages of the

344: hierarchy. A novel feature of the implementation of Croton et al

345: (2006) is the inclusion of AGN feedback to curtail star formation in

346: massive objects and to prevent the formation of overluminous galaxies

347: at the center of galaxy clusters. The star formation and enrichment

348: history of galaxies is then processed with standard spectrophotometric

349: models to provide estimates of galaxy luminosities and colors. The

350: main data we use here is the luminosity in the 5 Sloan bands,

351: {\it{ugriz}}, and we concentrate our analysis in the $g$ and $r$

352: bands. Unless otherwise specified, luminosities and magnitudes will

353: refer to the $r$ band in what follows.

354: 

355: The semianalytic approach follows explicitly the evolution of galaxies

356: within dark matter halos, even when they are accreted into (and become

357: satellites of) larger structures. After accretion, satellite galaxies

358: are followed consistently until their parent subhalo is destroyed by

359: the tidal field of the more massive system, at which point the

360: satellite orbit is subsequently identified as that of the most bound

361: particle of the parent subhalo before disruption. The satellite is

362: subsequently assumed to survive for some residual time consistent with

363: dynamical friction estimates. As we shall see below, this careful

364: treatment of the N-body simulation is crucial to characterize the

365: dynamics of satellites of isolated galaxies in the {\it MS}. Some

366: aspects of the true evolution are nevertheless neglected; in

367: particular, the effects of dynamical friction on the orbit of a

368: satellite after its dark halo is disrupted but before it merges with

369: the central galaxy of its host halo, as well as a necessarily rough

370: estimation of its merging time. This neglect will affect some aspects

371: of the results presented below, notably the overall abundance and

372: the radial distribution of satellites.

373: 

374: \subsection{The Primary/Satellite Galaxy Catalogue}

375: \label{ssec:gxcat}

376: 

377: \subsubsection{Primary galaxies}

378: \label{sssec:primgx}

379: 

380: We follow standard practice and identify a sample of primary galaxies

381: from the {\it MS} galaxy catalogue through (i) an {\it isolation}

382: criterion, imposed to ensure that a single object dominates the local dynamics

383: traced by the satellites, and (ii) a brightness cutoff, imposed to

384: ensure that most primaries have a fair chance of having detectable

385: satellites. Hereafter, we shall refer to systems brighter than

386: $M_r=-20.5$ and surrounded, within 1 $h^{-1}$ Mpc, {\it only} by

387: companions at least 2 magnitudes fainter, as {\it primary

388: galaxies}. (Note that neither the Milky Way nor M31 satisfy this

389: strict isolation criterion.)

390: 

391: \subsubsection{Satellites}

392: \label{sssec:satgx}

393: 

394: We identify as satellites any galaxy brighter than $M_r=-17$ (the

395: limiting magnitude of the catalogue) that lies within the virial

396: radius

397: %

398: \footnote{We define the {\it virial} radius, $r_{\rm 200}$, of a

399: system as the radius of a sphere of mean density $200$ times the

400: critical density for closure, 

401: $\rho_{\rm crit}=3H_0^2/8\pi G\sim 277.5 \, h^{2}\, M_\odot/\, \rm kpc^3$.

402: This implicitly defines the virial mass of a halo, $M_{\rm 200}$,

403: as that enclosed within $r_{200}$, and the virial velocity, $V_{\rm 200}$, 

404: as the circular velocity measured at $r_{\rm 200}$.

405: Quantities characterizing a system will be referred to as ``virial''

406: and measured within $r_{\rm 200}$, unless otherwise specified.}

407: %

408: of a primary. Primaries with {\it no} satellites brighter than

409: $M_r=-17$ within their virial radius will be referred to as {\it

410: singles}. Primaries with at least one satellite will be referred to as

411: {\it hosts}.

412: 

413: \subsubsection{Further nomenclature}

414: \label{sssec:nomenc}

415: 

416: One complication arises from the fact that, despite our isolation

417: criterion, some primaries might themselves be satellites of larger

418: systems. This tends to happen in the rarefied outskirts of massive

419: clusters. In this case, the virial radius would refer to the cluster,

420: rather than to the parent halo of the primary, and the velocities of

421: nearby satellites would be contaminated by high-speed cluster

422: members. To take this into account, we shall distinguish, if

423: pertinent, two classes of primaries: {\it central} and {\it

424: non-central}. However, we emphasize that the fraction of non-central

425: primaries is rather small, and that excluding them from the

426: statistical analysis has no major influence on our main conclusions.

427: 

428: A second note refers to satellites, some of which, as described in

429: \S~\ref{ssec:sam}, are still surrounded at the present time by their

430: parent halos (hereafter {\it WSUB} satellites, for short), while

431: others are identified with the most bound dark matter particle of the

432: parent subhalo immediately prior to disruption ({\it NOSUB}

433: satellites). Clearly, results concerning {\it NOSUB} satellites are

434: likely to be more model-dependent, since the identification of the

435: satellite as a distinct object depends in this case heavily on the

436: assumptions made in the model about dynamical friction timescales. It

437: is important to keep this distinction in mind as we try and interpret

438: the results.

439: 

440: \subsubsection{Statistics}

441: \label{sssec:stats}

442: 

443: It is interesting to assess how our primary selection criteria select

444: galaxies from the general population. We find that the fraction of

445: galaxies that are designated as primaries depends only weakly on

446: luminosity: for example, $23\%$ of galaxies brighter than $M_r=-23$

447: are {\it primaries}, the vast majority of which ($90\%$) are {\it

448: hosts}; $94.6\%$ of them are of the {\it central} type. Fainter galaxies

449: ($-21<M_r<-20.5$) have a $19\%$ chance of being classified as {\it

450: primaries}, but the majority of them ($85\%$) are actually {\it

451: singles} and therefore will not contribute to satellite studies. Of

452: these fainter primaries, $99.5\%$ are classified as {\it

453: central}. Overall, only $1\%$ of primaries are {\it non-central}, and

454: therefore our results are unlikely to be affected by the presence of

455: these unusual systems.

456: 

457: The final sample contains $258,321$ {\it primaries}, $79,000$ of which

458: are {\it hosts} to at least one satellite. We find more than $178,000$

459: satellites orbiting within the virial radius of the primaries, a

460: number that rises to $508,000$ if all satellites within 1 $h^{-1}$ Mpc

461: of the primaries are considered. We list the fraction of primaries as

462: a function of $r$-band luminosity and halo mass, along with other

463: useful data, in Tables~\ref{tab:mfrac} and ~\ref{tab:lfrac}.

464: 

465: 

466: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

467: \begin{figure}

468: \begin{center}

469: \includegraphics[width=84mm]{figures/lumgap_n10.ps}

470: \caption{ {\it Top panel:} Magnitude gap between the primary and the

471: 10th brightest satellite within $r_{200}$ as a function of

472: $L_{\rm host}$ (expressed in units of $h_{70}^{-2}\, L_\odot$).  The

473: dashed line indicates the $M_r=-17$ faint magnitude cutoff of our

474: catalogue. For reference, we show also the approximate location of a

475: few well-known systems, such as the Milky Way and M31 (both of which

476: fall outside the plot), the Virgo cluster \citep{trentham02}, and three ``fossil''

477: groups: RX J1340.6+4018 \citep{jones00}, RX J1552.2+2013

478: \citep{mendez_oliveira06} and RX J1416.4+2315 \citep{cypriano06}.

479: Magnitudes have been converted to the r-band using \citep{fukugita95}

480: when necessary.

481: {\it Bottom panel:} Luminosity

482: distribution of all red ($(g-r)>0.95$) primaries (black solid

483: histogram).The shaded magenta histogram indicate the subsample shown

484: in the upper panel. These are the red {\it host} primaries having at

485: least $10$ satellites within $r_{200}$.}

486: \label{figs:lgap}

487: \end{center}

488: \end{figure}

489: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

490: 

491: 

492: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

493: \begin{center}

494: \begin{figure}

495: \includegraphics[width=84mm]{figures/radial_dist3d.ps}

496: \caption{Number density profile of satellites within $\sim 2$ virial

497: radii, computed after scaling the position of all satellites to the

498: virial radius of the primary halo, and normalized at $r=r_{200}$. The

499: contribution of {\it WSUB} and {\it NOSUB} satellites to the density

500: profile is shown as well. Note that the majority of satellites near

501: the center have had their parent halo stripped. The shape of the

502: density profile is well approximated by an NFW profile with

503: $c_{200}\sim 5.6$. This is slightly less concentrated than the

504: ``average'' halo around these primaries, for which we estimate

505: $\langle c_{200}\rangle \sim 8.1$. Both of these curves are shown in

506: the figure, using the same normalization as for the satellites. Arrows

507: indicate the radius containing half of the objects in each profile.}

508: \label{figs:dprof}

509: \end{figure}

510: \end{center}

511: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

512: 

513: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

514: \begin{center}

515: \begin{figure}

516: \includegraphics[width=84mm]{figures/radial_dist3d_colsat.ps}

517: \caption{The color dependence of the satellite number density

518: profile. The thick dotted line shows the profile corresponding to {\it

519: all} satellites, as in Figure~\ref{figs:dprof}. The dashed (red) and

520: solid (blue) curves show the contribution to the total profile of the

521: reddest ($(g-r)>1.01$) and bluest ($(g-r)<0.95$) one-third of the

522: satellites. Note that the reddest satellites are significantly more

523: concentrated than the bluest ones. Arrows indicate the radius

524: enclosing half of the satellites in each subsample.}

525: \label{figs:dprofcol}

526: \end{figure}

527: \end{center}

528: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

529: 

530: 

531: \section {Results and Discussion}

532: \label{sec:results}

533: 

534: \subsection {Mass-Luminosity-Color relation for primaries}

535: \label{ssec:lmc}

536: 

537: We begin by exploring the virial mass-luminosity relation for

538: primaries in our catalogue, one of the main topics addressed by

539: studies of satellite dynamics. This is shown in the top left panel of

540: Figure~\ref{figs:lmc}, and illustrates the expected trend for brighter

541: primaries to inhabit more massive halos. The most striking aspect of

542: this figure, however, is the large scatter: at given luminosity, halos

543: span over a decade in mass. Conversely, halos of given mass may host

544: primaries spanning a factor of $\sim 5$ in luminosity.  The rms

545: deviation at given $L$ is substantially smaller (see table \ref{tab:lfrac})

546: but the broad tails

547: in the mass-luminosity distribution are important, and we shall see

548: below.  The large scatter in the mass-luminosity relation reflects the

549: variety of mass assembly and star formation histories of isolated

550: galaxies, and must be borne in mind when ``stacking'' galaxies

551: according to luminosity in order to study their average halo

552: properties.

553: 

554: Different colors in Figure~\ref{figs:lmc} code the number density of

555: primaries at each location in the $M$-$L$ plane, varying by a factor

556: of $\sim 10$ for each step in color, and decreasing from red to

557: blue. Most of the galaxies are in a relatively tight ``main sequence''

558: (outlined by the red dots in Figure~\ref{figs:lmc} and delineated

559: roughly by the solid line), but are surrounded by a broad cloud

560: covering a large fraction of the panel. Closer inspection reveals the

561: presence of two further relatively well-defined ``sequences'': one

562: consisting of galaxies that are systematically brighter than the

563: ``main sequence'', indicated roughly by the dotted line, and another

564: one consisting of ``underluminous'' galaxies inhabiting fairly massive

565: halos (see dashed line in Figure~\ref{figs:lmc}).

566: 

567: We note that the ``main sequence'' follows roughly a constant virial

568: mass-to-light ratio. This would be the natural expectation for halos

569: where a similar fraction of their baryonic content has been

570: transformed into stars of roughly similar mass-to-light ratio. 

571: 

572: The ``bright sequence'' (dotted line) is caused by the transient

573: brightening of the luminous component resulting from a starburst

574: triggered by a major merger.

575: 

576: The ``underluminous sequence'' (dashed line), on the other hand, is

577: linked to the stunted growth of the stellar component of central

578: galaxies in massive halos that results from the ``radio-mode'' AGN

579: feedback (see for details Croton et al 2006). In such systems, halos

580: can increase their mass substantially while their central galaxies

581: grow only through mergers. This is indeed the way in which the

582: semianalytic model is able to reconcile the hierarchical growth of

583: structure with the observed down-sizing of star formation in galaxies

584: and the presence of ``red and dead'' galaxies at early epochs (Croton

585: et al 2006, De Lucia et al 2006, Bower et al 2006). \nocite{bower06,

586: delucia06}

587: 

588: The above interpretation suggests that the three ``sequences'' may be

589: best appreciated by applying color cuts to the sample, since starburst

590: galaxies will be bluer than average, while the opposite will be true

591: for galaxies that have not formed stars recently. This expectation is

592: largely borne out, as shown by the other three panels in

593: Figure~\ref{figs:lmc}: galaxies bluer than $g-r=0.65$ trace

594: predominantly the ``bright'' sequence, whereas those redder than

595: $g-r=0.95$ largely trace the ``underluminous sequence''. The ``main

596: sequence'' is nicely traced by galaxies of intermediate, less extreme,

597: colors. We hasten to add that the color discrimination is not perfect,

598: and that residual evidence for the ``main sequence'' is clearly

599: apparent in both the red and blue panels of Figure~\ref{figs:lmc}.

600: 

601: The main conclusion from this discussion is that, at fixed luminosity,

602: the virial mass depends strongly on color, and that a strict color

603: selection criterion applied to primaries helps to tighten

604: substantially the scaling between virial mass and luminosity 

605: \citep[see e.g.,][]{klypin_prada07}. As we

606: discuss below, the large scatter that would result from blindly

607: stacking primaries in luminosity bins would most likely obscure many

608: of the underlying trends.

609: 

610: The prevalence of each of these sequences is a sensitive function of

611: halo mass and galaxy luminosity. This is illustrated in

612: Figure~\ref{figs:lmcdist}, where each panel shows, for several

613: luminosity bins, the distribution of virial masses of galaxies in each

614: of the three sequences. The sample is further split into {\it host}

615: and {\it single} primaries, since, by definition, only the halos of

616: {\it host} primaries are amenable to satellite dynamical studies.

617: 

618: Blue primaries are clearly in the minority amongst host galaxies,

619: except at the brightest magnitudes, where they make a substantial

620: ($\sim 29\%$) fraction of the brightest ($L>10^{11}\, L_\odot$)

621: primaries. Red galaxies, on the other hand, are reasonably well

622: represented at all magnitudes, making up about $\sim 30$-$40\%$ of all

623: primaries fainter than $10^{11} \, L_\odot$. The ``main sequence'' of

624: primaries dominates at all but the brightest luminosities:

625: intriguingly, only blue and red primaries make up the population of

626: $L>10^{11} \, L_\odot$ hosts. These are clearly very unusual objects,

627: either the result of a recent starburst in a relatively low mass halo,

628: or the result of growth by mergers but without star formation at the

629: center of a very massive halo.

630: 

631: {\it Single} primaries show similar mass-color-luminosity trends as

632: {\it hosts}, although galaxies with extremely red colors are less well

633: represented, and the ``main sequence'' is more prevalent than amongst

634: {\it hosts}. Interestingly, only blue {\it single} primaries in low

635: mass halos populate the brightest luminosity bin ($L>10^{11} \,

636: L_\odot$): this is clearly the result of a major 

637: merger fueling a starburst of extreme but short-lived brightness.

638: 

639: At given luminosity, {\it single} primaries tend to have lower virial

640: masses than {\it hosts}. This is expected, since more massive halos

641: tend to have more of everything, including satellites, implying a

642: small bias toward more massive halos in a sample selected to have at

643: least one satellite. The effect is noticeable, as witnessed by the

644: shift in the position of the green histogram in corresponding panels

645: of the {\it single} and {\it host} distributions shown in

646: Figure~\ref{figs:lmcdist}. As expected from the argument below, it is

647: largest in the lowest luminosity bins (i.e., $L<10^{10.4} \,

648: L_\odot$), where it amounts to a shift by a factor of $\sim 2$ in the

649: average mass of galaxies in the ``main'' sequence.

650: 

651: \subsection {Application to ``fossil groups''}

652: \label{ssec:fossil}

653: 

654: As is clear from Figures~\ref{figs:lmc} and ~\ref{figs:lmcdist}, our

655: primary selection criteria select a number of systems in fairly

656: massive halos, $10^{13}$-$10^{15} \, M_\odot$, a range usually

657: associated with galaxy groups and poor clusters rather than isolated

658: galaxies. These are, indeed, systems of galaxies which, by chance,

659: have evolved a gap of at least two magnitudes between the brightest

660: and second brightest galaxy, and therefore are included in our

661: sample. These systems are not rare; according to

662: Table~\ref{tab:mfrac}, the brightest galaxy of about $\sim 10\%$ of

663: all halos in the range $10^{13}$-$10^{15} \, M_\odot$ is a primary

664: according to our definition.

665: 

666: The presence of these systems in our sample has the potential of

667: biasing our satellite velocities toward large values, and therefore

668: one must be careful to take this into account in the analysis. We

669: shall return to this issue below (\S~\ref{ssec:lsg}); but we note here

670: that these are the analogs of ``fossil groups'', systems of galaxies

671: which are unusually X-ray bright (and massive) for their optical

672: richness. ``Fossil'' groups or clusters are defined, just like our

673: primaries, as systems where a large ($>2$) magnitude gap exists

674: between the brightest and second-brightest galaxy.

675: 

676: Although the statistics of these objects are still fairly poorly

677: known, their unusual properties have attracted attention from

678: observers and theorists alike, and there has been speculation that the

679: abundance of these systems might be difficult to reproduce in the 

680: $\Lambda$CDM cosmogony \citep{donghia04,donghia05,milosavljevic06,

681: sommerlarsen06}. Subsequent work has shown, however, that the predicted

682: abundance of "fossil" systems is in reasonable agreement with the still

683: rather uncertain observational constraints \citep[see, e.g.][]{donghia07,

684: vandenbosch07}. Worries remain that the atypical luminosity function of 

685: known 'fossil' systems may be difficult to reconcile with the substructure

686: function of $\Lambda$CDM halos \citep{donghia07}, but we emphasize that 

687: {\it only three} "fossil" groups have published luminosity functions. This,

688: however, should change soon, as available X-ray data are systematically 

689: surveyed for the presence of ``fossil'' groups, and as imaging and 

690: spectroscopic surveys provide secure luminosity functions for a growing 

691: number of ``fossil'' groups.

692: 

693: Our intention here is to characterize the abundance and luminosity

694: function of ``fossil''-like systems in our galaxy catalogue in order

695: to guide the interpretation of future observational constraints. As

696: stated above, these objects are not rare, since up to one in ten of

697: the most massive halos may be regarded as a ``fossil'' system

698: according to the above definitions (see Table~\ref{tab:mfrac}).  These

699: systems are not rare amongst bright galaxies either, as shown in

700: Table~\ref{tab:lfrac}; approximately $22\%$ of galaxies brighter than

701: $5\times 10^{10} L_\odot$ live in a ``fossil''-like

702: environment. ``fossil''-like systems are more prevalent amongst

703: fainter galaxies, which tend to inhabit

704: lower mass halos, where large magnitude gaps between the two brightest

705: galaxy members are common (e.g., the Milky Way is the ``ultimate

706: fossil group'').

707: 

708: This gives a total number density of $8.1 \times 10^{-6}\, h^{3}$

709: Mpc$^{-3}$ for ``fossil'' systems in halos exceeding $10^{13} \,

710: h^{-1} \, M_\odot$, or $8.3 \times 10^{-5}\, h^{3}$ Mpc$^{-3}$ for

711: fossil groups whose brightest galaxy exceeds $5\times 10^{10}

712: h^{-2}\, L_\odot$. This actually exceeds the (still rather

713: uncertain) observational estimates of $\sim 5 \times 10^{-7}-2 \times

714: 10^{-6} \, h^{3}$ Mpc$^{-3}$ \citep{jones03, vikhlinin99,romer00} for

715: groups with X-ray luminosities $\geq 10^{43} h_{50}^{-2}$ erg ${\rm

716: s}^{-1}$. We note, however, that only about $10$ fossil groups are

717: actually known, and that this may very well be an underestimate. Our

718: preliminary conclusion is that the semianalytic approach has no

719: difficulty accounting for the abundance of groups with such a strong

720: distinction between brightest and second-brightest galaxy.

721: 

722: A further test concerns the actual luminosity distribution of galaxies

723: in the groups. In a recent paper, D'Onghia et al (2007) argue, on the

724: basis of N-body simulations, that ``fossil'' groups present a

725: challenge to the $\Lambda$CDM scenario, since few cold dark matter

726: halos are as deficient in substructure as ``fossil'' groups. This

727: suggestion is based on the analysis of simulations that follow solely

728: the dark matter component, so it is important to check the predictions

729: of models that actually follow the formation of the galaxies.

730: 

731: We adopt a simple measure of the shape of the luminosity distribution

732: of galaxies in a ``fossil'' group, namely the magnitude difference

733: between the brightest and tenth brightest galaxy member, $\Delta

734: M_{10}$. This is shown as a function of primary luminosity in the top

735: panel of Figure~\ref{figs:lgap}. A typical cluster with a well

736: populated Schechter-like luminosity distribution, such as Virgo, has

737: $\Delta M_{10} \sim 2$ \citep{trentham02}. Individual galaxies, such as the

738: Milky Way or M31, have a much larger percentage of the total light

739: concentrated in a single object, and as a result their $\Delta M_{10}$

740: differ markedly from Virgo; we find $\Delta M_{10}=9$ and $11$ for the

741: Milky Way and M31, respectively \citep{mateo98}.

742: Fossil groups have intermediate values of $\Delta M_{10}$; for 

743: the three groups with published luminosity functions, the values 

744: span the range from $\sim 3$ to $\sim 5$ \citep{jones00,

745: mendez_oliveira06,cypriano06}.

746: 

747: These values are not unusual in our catalogue of isolated primaries,

748: where $\Delta M_{10}$ is a strong function of the primary luminosity,

749: approaching $\sim 2$ (the minimum possible value given our isolation

750: criterion) for the brightest primaries, but increasing rapidly with

751: decreasing luminosity. As shown in the top panel of

752: Figure~\ref{figs:lgap}, there is also a large scatter in $\Delta

753: M_{10}$; at $L\sim 10^{11} L_\odot$ we find values that go from $\sim

754: 3$ to $\sim 6$, spanning the range observed in fossil groups. It would

755: clearly be difficult to argue on the basis of this evidence that there

756: is a substantial discrepancy between $\Lambda$CDM predictions and the

757: observations of ``fossil'' groups. Further data are needed to clarify

758: this issue further.

759: 

760: 

761: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

762: \begin{center}

763: \begin{figure}

764: \includegraphics[width=84mm]{figures/mean_satlum_cutlumsat.ps}

765: \caption{The average satellite luminosity (expressed in units of

766: $L_{\rm host}$) is shown as a function of radius for various primary

767: luminosity bins. Because of our faint magnitude cutoff ($M_r=-17$),

768: we consider a different magnitude range for satellites, depending on

769: the brightness of the host. The three dot-dashed lines correspond to

770: primaries brighter than $10^{10.8}\, L_\odot$ ($M_r\lsim-22.5$).

771: Note that when considering the full magnitude range resolved in the {\it MS},

772: ($2<M_{\rm sat}-M_{\rm host}<5.5$) a strong luminosity segregation is

773: clearly present: the average brightness of satellites decreases by a

774: factor of $\sim 2$ from the center out to the virial radius. As

775: expected, the magnitude of the effect decreases as the magnitude range

776: narrows when considering fainter hosts. Once this effect is taken into

777: account, the satellite luminosity segregation seems to be roughly

778: independent of host brightness.}

779: \label{figs:dprofl}

780: \end{figure}

781: \end{center}

782: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

783: 

784: \subsection {Satellite density profile}

785: \label{ssec:satprof}

786: 

787: The solid circles in Figure~\ref{figs:dprof} show the number density

788: profile of satellites, computed after rescaling the position of each

789: satellite to the virial radius of the host and stacking the full

790: sample. The large number of satellites in our catalogue makes the

791: statistical error in the profile negligible; bootstrap error bars are

792: smaller than the size of each symbol.

793: 

794: The shape of the satellite density profile is well described by the

795: NFW (Navarro, Frenk \& White 1996, 1997) formula, with

796: $c_{200}=5.6$. This result is rather insensitive to halo mass or

797: luminosity, for example, splitting the sample of primaries in two by

798: halo mass results in $c=5.6\pm0.3$ and $6.2\pm0.5$ for the high and 

799: low mass sample, respectively. A similar split in luminosity yields 

800: $c=6.2 \pm 0.6$ and $c=5.5 \pm 0.4$, respectively. 

801: \nocite{nfw96,nfw97}

802: 

803: Satellites are slightly less concentrated than the host dark matter

804: halos.  The average host dark halo concentration may be found by averaging

805: the concentrations of host halos, estimated from the

806: mass-concentration relation of Fausti-Neto et al (in preparation):

807: %,

808: \begin{equation}

809: \log(c_{200})= 2.1-0.1 \, \log(M_{200}/(M_\odot/h)^{-1}),

810: \end{equation}

811: %

812: and taking into account the lognormal dispersion

813: $\sigma(\log(c_{200}))=0.10$. We find an average concentration of

814: $\langle c_{200} \rangle =8.1$.

815: 

816: These two NFW profiles are plotted in Figure~\ref{figs:dprof}, showing

817:  that the difference is not large. For example, the half mass radius

818:  of the average dark halo is $0.39 \, r_{200}$, which is very similar

819:  to the radius that contains half of the stacked satellites, $0.42 \,

820:  r_{200}$. We conclude that satellites are a relatively unbiased

821:  tracer of the dark mass distribution within the virial radius of a

822:  halo, at least for this model of galaxy formation and evolution.

823: 

824: In a recent paper, \citet{chen06} study the radial projected

825: distribution of satellite galaxies in the SDSS. These authors find

826: that the projected satellite number density profile is well fitted by

827: a power law: $\Sigma(R) \propto R^\alpha$, with

828: $\alpha=-1.7\pm0.1$. In order to approximately mimic the selection

829: criteria applied by Chen et al., we have projected all galaxies in our

830: catalogue within 1 $h^{-1}$Mpc around each isolated primary,

831: considering as "satellites" those within projected distances $\Delta

832: R<500\; h^{-1}$ kpc and line-of-sight velocities $\Delta V_{los}<500$

833: km/s (notice that from our definition of isolated galaxies, all

834: "satellites" are at least two magnitude fainter than the primary). We

835: find $\alpha=-1.55 \pm 0.08$ in the distance range $26<R<500 \,

836: h^{-1}$kpc, which is consistent with the results of Chen et al within

837: the quoted errors.

838: 

839: \subsubsection{Dependence on satellite color}

840: \label{sssec:dprofcol}

841: 

842: Although the satellite population as a whole traces the dark mass

843: reasonably well, there is a strong dependence on satellite color. This

844: is shown in Figure~\ref{figs:dprofcol}, where the profiles of the

845: reddest ($(g-r)>1.01$, red dashed line) and bluest ($(g-r)<0.95$, solid

846: blue line) one-third of satellites is compared with the overall

847: profile presented in Figure~\ref{figs:dprof}. Red satellites are

848: clearly much more centrally concentrated than blue ones: half of the

849: red sample is contained within $0.32 \, r_{200}$, a radius that climbs

850: to $0.71 \,r_{200}$ for the bluest one-third of satellites.

851: 

852: This difference in concentration may be traced to the assumption in

853: the semianalytic treatment that a satellite loses its reservoir of hot

854: gas (the future fuel for star formation) once it is accreted into a

855: larger structure. Thus star formation in satellites declines quickly

856: after accretion: the earlier a satellite was accreted the older (and

857: redder) its stellar population will be. Early-accreting satellites

858: have smaller turnaround radii and are likely to orbit closer to the

859: center than late accreting ones, resulting in the trend shown in

860: Figure~\ref{figs:dprofcol}. 

861: 

862: A trend of similar origin is shown also by the bottom two curves in

863: Figure~\ref{figs:dprof}, which show the contribution to the satellite

864: density profile from satellites that have preserved ({\it WSUB}) or

865: lost ({\it NOSUB}) their parent dark halos. Early accreting satellites

866: are more likely to have been more affected by tides and to have lost

867: their parent halos, leading to the spatial segregation between {\it

868: WSUB} and {\it NOSUB} satellites seen here. This figure also shows

869: clearly the importance of tracking satellites in dark matter-only

870: simulations even after their parent halos have been disrupted in the

871: tidal field of the primary: {\it NOSUB} satellites are crucial to the

872: satellite profile in the inner regions of the primary.

873: 

874: 

875: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

876: \begin{center}

877: \begin{figure}

878: \includegraphics[width=84mm]{figures/radialveloc_lr.ps}

879: \caption{The radial velocity of all satellites in our sample as a

880: function of radius, both scaled to the virial values of the host

881: halo. The solid curve shows the median of the distribution, while

882: dashed lines outline the 25\% and 75\% percentiles of the distribution

883: as a function of radius. The velocity dispersion decreases out to the

884: virial radius beyond which it remains approximately constant. A

885: ``first infall'' sequence of negative radial velocities is clearly

886: defined outside $\sim 0.5 \, r_{200}$. Satellites with very high

887: radial velocities (sometimes exceeding $\sim 3\, V_{200}$) are present,

888: especially at large radii. This is a result of some primaries lying in

889: the periphery of much larger structures, such as galaxy clusters; the

890: large velocities are those of cluster members typically unrelated to

891: the primary.}

892: \label{figs:rvr}

893: \end{figure}

894: \end{center}

895: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

896: 

897: \subsubsection{Dependence on satellite brightness}

898: \label{sssec:dprofl}

899: 

900: Figure~\ref{figs:dprofl} explores the luminosity segregation of

901: satellites and its dependence on primary luminosity. The faint

902: magnitude cutoff of our catalogue implies that, in order to compare

903: meaningfully primaries of different brightness, satellites must be

904: selected within a definite magnitude range. Satellites can only have

905: absolute magnitudes between the cutoff at $M_r=-17$ and $M_{\rm

906: host}+2$, so that fainter primaries have, by construction, satellites

907: that span a narrower magnitude range.

908: 

909: The three (red) dot-dashed curves in Figure~\ref{figs:dprofl} indicate

910: the radial dependence of the average satellite luminosity (in units of

911: the host's) for the brightest hosts ($L_r>10^{10.8} \, L_\odot$;

912: $M_r<-22.5$). When all

913: satellites are considered (i.e., in the magnitude range $2<M_{\rm

914: sat}-M_{\rm host}<5.5$; bottom curve in Figure~\ref{figs:dprofl}) a

915: significant radial trend is seen: the average satellite luminosity

916: drops by a factor of $\sim 2$ from the center out to the virial

917: radius. The trend is, as expected, more difficult to detect when a

918: narrower range in satellite brightnesses is imposed (upper two

919: curves). Such a narrower range is needed when considering fainter

920: primaries, for which results are shown by the dashed and dotted

921: curves in Figure~\ref{figs:dprofl}. Once this is taken into account

922: the luminosity segregation we find seems to be independent of the

923: brightness of the primary.

924: 

925: The luminosity segregation shown in Figure~\ref{figs:dprofl} is most

926: likely due to the effects of dynamical friction, which operate faster

927: on more massive/more luminous satellites, bringing them closer to the

928: center on a shorter timescale. It should be possible to contrast this

929: result against observation.  However, a note of caution about this

930: interpretation is in order, recalling the results of

931: Figure~\ref{figs:dprof}. Since most satellites, especially those near

932: the center, have lost their parent halo, their present-day location is

933: being traced by a single dark matter particle, and their survival

934: depends directly on the semianalytic model assumptions about dynamical

935: friction. As such, the luminosity segregation shown in

936: Figure~\ref{figs:dprofl} is likely to be model-dependent.

937: 

938: \subsection {Satellite Velocities}

939: \label{ssec:satvel}

940: 

941: \subsubsection{Radial velocities}

942: \label{sssec:radv}

943: 

944: Figure~\ref{figs:rvr} shows the radial velocities of all satellites in

945: our sample as a function of their distance to the primary, after

946: rescaling to the virial quantities of the host halo. The solid line

947: traces the median satellite velocity in radial bins; the dashed lines

948: the 25\% and 75\% of the distribution, respectively. Note that the

949: median velocity within $r_{200}$ is constant and consistent with zero,

950: as expected from a relaxed population in equilibrium.

951: 

952: Outside $r_{200}$, negative velocities are more prevalent, as

953: satellites on their first approach to the primary start to

954: dominate. These ``first-infall'' satellites delineate the negative

955: velocity boundary at all radii, forming a sequence that becomes fairly

956: obvious outside $\sim 0.5\, r_{200}$ in Figure~\ref{figs:rvr}. The

957: velocity along this sequence decreases outward, and approaches zero at

958: $r \sim 3\, r_{200}$, the approximate location of the

959: turnaround radius according to the simple secondary infall model

960: \citep[see][]{bertschinger85,white93,navarroandwhite93}.

961: 

962: The ``first-infall'' sequence is an interesting feature of the radial

963: velocity distribution, one whose detection may be used to yield a

964: direct estimate of the mass of the host halo. This is the rationale of

965: a number of studies that attempt to pin down the location of the

966: turnaround radius in the outskirts of galaxy groups and clusters by

967: looking at ``caustics'' in the velocity distribution

968: \citep{diaferio97,diaferio99,geller99,biviano03,rines03,diaferio05}.

969: Although some progress has been made

970: on this issue, the observational evidence remains elusive and its

971: interpretation controversial \citep{reisenegger00,drinkwater01,

972: mahdavi05,mohayaee06a,gavazzi06}.

973: 

974: Figure~\ref{figs:rvr} offers a possible explanation for these findings:

975: satellites on first approach make up a relatively small fraction of

976: systems populating the outskirts of the halo. We investigate this

977: quantitatively in Figure~\ref{figs:rvdist}, where we show the

978: satellite radial velocity distribution as a function of radius,

979: $x=r/r_{200}$, normalized to the virial radius. 

980: 

981: Satellites within the

982: virial radius have an approximately Gaussian velocity distribution

983: with dispersion that declines with radius. Interestingly, the radial

984: velocity dispersion in the inner regions is comparable to the virial

985: velocity of the halo, $\sigma_r\sim 0.96 \, V_{200}$. This result has

986: been used by Sales et al (2007a) to argue that the relatively low

987: velocity dispersion of Galactic satellites implies a fairly low mass

988: for the Milky Way halo ($V_{200}^{MW}\sim 110$ km/s, see that paper for

989: further details). The velocity dispersion declines outward: the best

990: fitting Gaussian for satellites with $0.5<x<1$ is $\sigma_r\sim 0.7\,

991: V_{200}$.

992: 

993: Beyond the virial radius, the radial velocity distribution becomes

994: clearly asymmetric, with an excess of satellites with negative radial

995: velocities. As mentioned above, this is a result of the increasing

996: importance of satellites on their first approach to the host halo. We

997: have chosen to quantify this with a double Gaussian fit: one of zero

998: mean velocity and dispersion fit to the distribution of {\it positive}

999: radial velocities, and a second one whose dispersion and mean are

1000: chosen so that the sum of the two Gaussians match best the whole

1001: distribution. The two Gaussian fits are shown with dashed lines in

1002: Figure~\ref{figs:rvdist}, and we shall hereafter refer to them, for

1003: short, as the ``zero mean velocity'' component ({\it zmv}) and the

1004: ``infall'' component ({\it inf}). Values quoted in each panel of

1005: figure \ref{figs:rvdist} indicate the best-fit Gaussian parameters to

1006: the {\it total} distribution\footnote{We will use $\bar V_i$ and

1007: $\sigma_i$ to indicate the mean and dispersion of the $i$-component

1008: velocity distribution, where $i=r,\theta,\phi$. The skewness $\xi$ is

1009: defined as $\frac{\Sigma (V_i-\bar V_i)^3}{(N-1)\sigma_i^3}$ and the

1010: kurtosis $\kappa=\frac{\Sigma (V_i-\bar V_i)^4}{(N-1)\sigma_i^4} -3$}.

1011: In contrast with the results of \cite{prada06} and \cite{power06}, the

1012: infall patern is insensitive to halo mass; we find a non-negligible

1013: population of first-infall satellites around low and high-mass halos

1014: in our catalogue.

1015: 

1016: As expected, the mean infall velocity decreases outward, from ${\bar

1017: V}_r \sim -0.85 \, V_{200}$ at $1<x<1.5$ to ${\bar V_r}\sim

1018: -0.4\,V_{200}$ at $2.5<x<3$, and consistent with a turnaround radius

1019: located just outside $\sim 3\, r_{200}$. Interestingly, the velocity

1020: dispersion of infalling satellites amounts to about $15\%$ and $25\%$

1021: of the virial velocity of the host halo, about a factor of two to

1022: three colder than the rest of the population. The prevalence of the

1023: first-infall population increases outward: it makes up $15\%$, $25\%$,

1024: $36\%$, and $40\%$ of all satellites in each of the four $x>1$ bins

1025: shown in Figure~\ref{figs:rvdist}, respectively. These results may be

1026: used to improve algorithms intended to detect infalling galaxies in

1027: the regions surrounding groups and clusters.

1028: 

1029: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1030: \begin{center}

1031: \begin{figure}

1032: \includegraphics[width=84mm]{figures/gauss_fit_allr.ps}

1033: \caption{Distribution of satellite radial velocities, in bins of

1034: different distance to the primary. We consider only {\it central}

1035: hosts here in order to minimize the contribution of velocity

1036: interlopers. The radial range considered in each panel is labeled by

1037: the range in $x=r/r_{200}$ used. Dashed lines show Gaussian fits to

1038: the profiles: {\it two} Gaussians are used when the distribution shows

1039: a strong asymmetry between negative and positive radial

1040: velocities. One of the Gaussians (the {\it zmv} component) is assumed

1041: to have zero mean velocity and to match the distribution of positive

1042: radial velocities; the parameters of the second one (the {\it inf}

1043: component) are then fit so as to match the whole distribution. The fit

1044: parameters (mean $V_r$, dispersion $\sigma_r$, skewness $\xi$ and kurtosis

1045: $\kappa$) are given in Table~\ref{tab:Gaussfits} as well as quoted 

1046: in each panel.}

1047: \label{figs:rvdist}

1048: \end{figure}

1049: \end{center}

1050: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1051: 

1052: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1053: \begin{center}

1054: \begin{figure*}

1055: \includegraphics[width=0.475\linewidth,clip]{figures/vthe_histo_gauss_allr.ps}

1056: \hspace{0.4cm}

1057: \includegraphics[width=0.475\linewidth,clip]{figures/vphi_histo_gauss_allr.ps}

1058: \caption{Distribution of the tangential spherical components of the

1059: satellite velocities, $V_\theta$ and $V_\phi$, shown in various

1060: distance bins, $x=r/r_{200}$, as labelled in each panel. We have split

1061: the sample in two components, an ``infall'' component ({\it inf},

1062: magenta long-dashed lines) and a ``zero-mean velocity'' component

1063: ({\it zmv}, solid black line), using the double Gaussian fits to the

1064: radial velocity distribution of Figure~\ref{figs:rvdist} to assign

1065: probabilistically satellites to each. Dotted curves are Gaussian fits

1066: to each of these distributions. Parameters (mean, dispersion $\sigma$,

1067: skewness $\xi$ and kurtosis $\kappa$, are quoted in each panel 

1068: and in Table~\ref{tab:Gaussfits}.}

1069: \label{figs:vtandist}

1070: \end{figure*}

1071: \end{center}

1072: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1073: \subsubsection{Tangential velocities}

1074: \label{sssec:vtan}

1075: 

1076: In order to complete our characterization of satellite velocities we

1077: show in Figure~\ref{figs:vtandist} the distribution of the tangential

1078: velocity components, binned by distance just as in

1079: Figure~\ref{figs:rvdist}. The spherical components are measured in a

1080: reference frame where the $z$-axis is chosen to coincide with the

1081: angular momentum of the host halo, so that $V_{\phi}$ carries

1082: information about a satellite's sense of rotation relative to the

1083: host.

1084: 

1085: Within the virial radius the dispersion in $V_{\phi}$ and $V_{\theta}$

1086: is substantially lower than the radial dispersion, and there are also

1087: hints of significant departures from Gaussianity. There is, for

1088: example, an excess of satellites that co-rotate with the host (i.e.,

1089: $V_{\phi}>0$); also, the $V_{\theta}$ distribution is platykurtic;

1090: i.e., it is more centrally peaked and has broader wings than a

1091: Gaussian. Such departures from Gaussianity are more pronounced for

1092: satellites outside the virial radius.

1093: 

1094: The long- and short-dashed curves in Figure~\ref{figs:vtandist} show

1095: the contribution to the total velocity distribution of the ``infall''

1096: and ``zero-mean velocity'' components discussed in

1097: \S~\ref{sssec:radv}. This is done by assigning satellites

1098: probabilistically to each of the two components, according to the

1099: Gaussian decomposition shown in

1100: Figure~\ref{figs:rvdist}. Interestingly, the $V_{\theta}$

1101: distribution, which is {\it not} well approximated by a Gaussian, is

1102: found to be the result of adding the nearly Gaussian distributions

1103: corresponding to the ``infall'' and the ``zero-mean velocity''

1104: components. Note, as well, that even {\it inside} $r_{200}$ the

1105: $V_{\theta}$ distribution differs slightly from a Gaussian. This is

1106: most likely due to the presence of infalling satellites within the

1107: virial radius. Although they are difficult to pick up in radial

1108: velocity, there are apparently enough of them to modify the

1109: $V_{\theta}$ distribution from Gaussian to platykurtic.

1110: 

1111: As may be seen from the right-hand panel of

1112: Figure~\ref{figs:vtandist}, the $V_{\phi}$ distribution of the

1113: ``infall'' component is approximately Gaussian as well, and shows only

1114: a weak excess of satellites co-rotating with the host. The ``zero-mean

1115: velocity'' component, on the other hand, shows a much more pronounced

1116: asymmetry and a broader dispersion. It is not clear at this point what

1117: causes this difference, but we plan to follow it up in future work. In

1118: all cases, the dispersion in the ``infall'' component is significantly

1119: lower than in the ZMV component. Table~\ref{tab:Gaussfits} lists a

1120: summary of the fit parameters for all these velocity components.

1121: 

1122: 

1123: \subsubsection{Velocity anisotropy}

1124: \label{sssec:vanis}

1125: 

1126: The velocity dispersion of the various spherical components declines

1127: gradually with radius, as shown in Figure~\ref{figs:vanis}. The

1128: biggest decline is seen for the radial velocity dispersion, which

1129: drops by almost a factor of two from the center out to the virial

1130: radius. The dispersion in the other components drops with radius at a

1131: different rate, leading to an anisotropy profile that increases from

1132: the center outwards, reaches a maximum of $\beta_{\rm max}\sim 0.5$ at

1133: $r\sim 0.2\, r_{200}$ and declines to become almost isotropic

1134: $\beta\sim 0$ just outside the virial radius.

1135: 

1136: The radial dependence of the anisotropy is a reflection of the

1137: increasing importance of the first-infall population at larger

1138: radii. Because it is on its first approach, the {\it inf} population

1139: is ``stretched'' along the radial direction and has therefore a

1140: smaller radial velocity dispersion at any given radius, compared with

1141: its tangential dispersion (compare, e.g., the radial and tangential

1142: dispersions for the {\it inf} population in Figures~\ref{figs:rvdist}

1143: and ~\ref{figs:vtandist}). As the prevalence of this component

1144: increases outwards, it brings down the radial bias characteristic of

1145: the inner regions, leading to a decline of the anisotropy in the

1146: outskirts of the system. In support of this interpretation, we note

1147: that the anisotropy of {\it WSUB} satellites differs strongly from

1148: that of {\it NOSUB} satellites (Figure~\ref{figs:vanis}). The former

1149: still retain their parent halos; are therefore more likely to have

1150: been accreted into the system more recently; and have velocities 

1151: that are more isotropic than the rest.

1152: 

1153: Figure \ref{figs:vanis} shows that the satellite velocity

1154: dispersion at the virial radius is about $40\%$ its maximum value

1155: in the inner regions. This decline is in agreement with that expected

1156: for NFW halos and can be successfully recovered from observational

1157: samples when contamination from interlopers is properly accounted for.

1158: Although early studies suggested a nearly flat satellite velocity

1159: dispersion profiles favouring isothermal models for host halos

1160: \citep{mckay02,brainerd03}, subsequent analysis suggested that it might

1161: be due to poor removal of background and foreground interlopers. \citet{prada03} studied

1162: the dynamic of satellite

1163: galaxies in the SDSS removing interlopers from their samples fitting

1164: a ''Gaussian + constant'' function to the satellite velocity distributions.

1165: These authors find that the projected velocity dispersion of {\it true}

1166: satellites drops to $\sim 40-60\%$ its maximum value at a projected

1167: distance of $\Delta R\sim 300 h^{-1}$kpc from the primary. The analysis

1168: of satellites in the 2dFGRS catalogue also show a declining $\sigma$

1169: profile after interloper remotion, although the measured drop in the

1170: velocity dispersion with distance is somewhat weaker \citep{brainerd04b}.

1171: The satellite galaxy kinematics in the Millennium Simulation appears broadly consistent

1172: with these observational results.

1173: 

1174: \subsection{Halo mass profile from satellite dynamics}

1175: \label{ssec:mprof}

1176: 

1177: Once the spatial distribution and the kinematics of the satellite

1178: population have been characterized, we may use them to constrain the

1179: shape of the host halo mass profile. Assuming spherical symmetry,

1180: equilibrium, and that satellites are massless tracers of the

1181: potential, Jeans' equations link the potential with the velocity

1182: dispersion and density profiles of the satellites. Expressed in terms

1183: of a circular velocity, we have:

1184: 

1185: \begin{equation}

1186: V_c^2(r)=-\sigma_r^2 \,

1187: (\frac{d\ln{\rho}}{d\ln{r}}+\frac{d\ln{\sigma_r}^2}{d\ln{r}} + 2\beta),

1188: \label{eqs:jeans}

1189: \end{equation}

1190: where the terms in the right hand side may be estimated from the

1191: results in the preceding discussion, and are summarized in the top

1192: panel of Figure~\ref{figs:mprof}.

1193: 

1194: The implied circular velocity profile for the {\it average} halo

1195: populated by the satellites is shown in the bottom panel of

1196: Figure~\ref{figs:mprof}. This $V_c$ profile has several of the same

1197: characteristics of the NFW profile: it rises to a maximum and then

1198: drops near the virial radius. The maximum circular velocity implied is

1199: $V_{\rm max}\sim 1.25 \, V_{200}$, and occurs at $r_{\rm max}\sim 0.3

1200: \, r_{200}$, which corresponds to a concentration of $c_{200}\sim

1201: 7$. We note that this is higher than the concentration ($\sim 5.6$)

1202: derived from fitting an NFW profile to the number density of

1203: satellites (Figure~\ref{figs:dprof}), and is closer to the {\it

1204: average} concentration of $\langle c \rangle \sim 8.1$ found for the

1205: host halos in the {\it MS} (see \S~\ref{ssec:satprof}). In spite of

1206: these differences, it is remarkable that satellites are overall

1207: reasonably good tracers of the dark mass profile. These similarities

1208: have also been reported in N-body/gasdynamical simulations by Sales et

1209: al (2007a), and augur well for studies of the mass profile of galactic

1210: halos based on satellite data.

1211: 

1212: 

1213: 

1214: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1215: \begin{center}

1216: \begin{figure}

1217: \includegraphics[width=84mm]{figures/sigmas_ty0host.ps}

1218: \caption{{\it Upper panel:} Velocity dispersion profile of central

1219: host satellites in spherical coordinates, where the polar $z$ axis is

1220: chosen to coincide with the direction of the host angular momentum.

1221: {\it Lower panel:} The corresponding anisotropy parameter,

1222: $\beta=1-{\sigma_t}^2/2{\sigma_r}^2$, where

1223: ${\sigma_t}^2=\sigma_\theta^2 + \sigma_\phi^2$.  We distinguish in the

1224: bottom panel the contribution of satellites that have preserved their

1225: parent dark matter halo ({\it WSUB}, filled triangles) and those that

1226: have not ({\it NOSUB}, open circles).}

1227: \label{figs:vanis}

1228: \end{figure}

1229: \end{center}

1230: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1231: 

1232: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1233: \begin{center}

1234: \begin{figure}

1235: \includegraphics[width=84mm]{figures/circveloc_jeans.ps}

1236: \caption{{\it Top:} Terms in the right-hand side of Jeans' equation

1237: relating satellite dynamics and the halo mass profile

1238: (eq.~\ref{eqs:jeans}). The dotted line shows a fit to $\sigma_r(r)$ of

1239: the form: $\sigma_r/V_{200}=\sigma_0+ (x/x_0)\, \exp(-(x/x_0)^\alpha)$, with

1240: $x=r/r_{200}$. Best-fitting dimensionless parameters are: $x_0=0.068$,

1241: $\alpha=3/4$ and $\sigma_0=0.6$. 

1242: {\it Bottom:} Solid line shows the average circular

1243: velocity profile of the potential sampled by the satellites, as

1244: derived from applying Jeans' equation. Note that the circular velocity

1245: implied by the satellite population rises to a maximum before dropping

1246: near the virial radius. The maximum circular velocity implied is

1247: consistent with an NFW profile with $c_{200}\approx 7$. This is in

1248: reasonable agreement with the average concentration ($\langle c

1249: \rangle \approx 8.1$) of host halos inhabited by satellites, and is

1250: higher than the concentration derived from the satellite density

1251: profile in Figure~\ref{figs:dprof}.}

1252: \label{figs:mprof}

1253: \end{figure}

1254: \end{center}

1255: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1256: 

1257: \subsection {Primary luminosity vs satellite velocity dispersion}

1258: \label{ssec:lsg}

1259: 

1260: The velocity dispersion of satellites is one of the prime tools used

1261: to investigate the mass of dark halos surrounding isolated galaxies

1262: and its dependence on luminosity.  In hierarchical formation scenarios

1263: like $\Lambda$CDM, velocities vary with mass according to the virial

1264: definitions discussed in the footnote to \S~\ref{sssec:satgx}, which

1265: imply that mass scales with velocity like $M_{200}\propto V_{200}^3$.

1266: 

1267: Due to the small number of satellites surrounding each primary, it is

1268: necessary to combine them in some way in order to beat small-number

1269: statistics. One obvious choice is to bin primaries in a narrow

1270: luminosity range, stack their satellites, compute their velocity

1271: dispersion, and see how the dispersion varies as a function of the

1272: luminosity of the primary.

1273: 

1274: If satellite velocities are an unbiased tracer of the virial velocity,

1275: and luminosity is a reasonable proxy for mass, then one may expect

1276: $L\propto \sigma^3$.  The former assumption appears to be borne out by

1277: the analysis presented above, but the latter one is afflicted by the

1278: large scatter in the mass-luminosity relation discussed in

1279: \S~\ref{ssec:lmc}. How does this affect the scaling between primary

1280: luminosity and satellite velocity dispersion?

1281: 

1282: We show this in Figure~\ref{figs:lsg}, where hosts are binned

1283: according to their $r$-band luminosity, and the one-dimensional

1284: velocity dispersion is computed for each bin after stacking all of

1285: their satellites. In each panel, the solid circles show the result of

1286: this procedure including all satellites of hosts satisfying the

1287: condition expressed in the label. The short-dashed (blue), long

1288: dashed (green) and dot-dashed (red) curves correspond to

1289: selecting primaries according to the color cuts adopted in

1290: Figure~\ref{figs:lmc}. The dotted lines indicate, for reference, the

1291: $L\propto \sigma^3$ and $L\propto \sigma^2$ scalings.

1292: 

1293: When considering all primaries (top-left panel), the $L$-$\sigma$

1294: relation is poorly fit by a power law, bending from approximately

1295: $L\propto \sigma^2$ at faint luminosities to a substantially shallower

1296: scaling for $L>5\times 10^{10} L_\odot$.

1297: A similar departure of the $L-\sigma$ relation from a simple power-law scaling has been

1298: reported and discussed previously by \citet{vandenbosch04}.

1299:  This is quite 

1300: different from the naive scaling mentioned above for hierarchical 

1301: models, and is largely due to the large scatter in the halo 

1302: mass-luminosity relation shown in Figure~\ref{figs:lmc}. The presence

1303: of some very massive halos at all luminosities, together with their

1304: growing importance with increasing luminosity bend the relation off the natural $L\propto

1305: \sigma^3$ scaling. At $\sim 5 \times 10^{10} L_\odot$ the prevalence

1306: of such massive objects increases and the relation becomes even

1307: shallower.

1308: 

1309: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1310: \begin{center}

1311: \begin{figure}

1312: \includegraphics[width=84mm]{figures/tf_3d_seq_f.ps}

1313: \caption{Velocity dispersion of satellites around isolated galaxies

1314: binned by $r$-band luminosity. Dispersions are computed after stacking

1315: all satellites within the virial radius of the host of each galaxy in

1316: the bin. Solid connected circles show the results for all satellites

1317: of primaries satisfying the criterion indicated by the label in each

1318: panel. Short-dashed (blue), long-dashed (green) and dot-dashed (red)

1319: curves correspond to primaries selected according to the color cuts

1320: chosen in Figure~\ref{figs:lmc}. Dotted lines indicate $L\propto

1321: \sigma^3$ and $L\propto \sigma^2$ scalings, to ease the comparison

1322: from panel to panel. Panels differ in the sample of primaries

1323: used. The importance of the primary within its own virial radius

1324: distinguishes these samples, as given by $L_{\rm host}/L_{\rm sys}$,

1325: where $L_{\rm sys}$ is the total luminosity of galaxies within the

1326: virial radius. }

1327: \label{figs:lsg}

1328: \end{figure}

1329: \end{center}

1330: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1331: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1332: \begin{center}

1333: \begin{figure}

1334: \includegraphics[width=84mm]{figures/tf_3d_mstr_seq_f.ps}

1335: \caption{Same as figure \ref{figs:lsg}, but considering stellar

1336: masses rather than $r$-band luminosities. Cuts applied on 

1337: $M_{\rm host}/M_{\rm sys}$ refer also to the mass in stars.}

1338: \label{figs:mstsg}

1339: \end{figure}

1340: \end{center}

1341: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1342: 

1343: Restricting primaries by color helps; for example, for blue hosts

1344: ($(g-r)<0.65$) we obtain a $L\propto \sigma^3$ relation that holds

1345: over the whole luminosity range.  However, color cuts work less well

1346: for redder primaries: for $0.65<(g-r)<0.95$ hosts we recover the

1347: $L\propto \sigma^3$ relation at faint luminosities, although for

1348: galaxies brighter than $\sim 6 \times 10^{10} L_\odot$ the scaling

1349: becomes again much shallower as a result of the growing importance of

1350: massive halos with underluminous central galaxies. The same aplies to

1351: the primaries with the reddest colors, although the change in the

1352: slope occurs at fainter luminosities ($\sim 4 \times 10^{10}

1353: L_\odot$).  Selecting by color alone is thus not enough to ensure a

1354: sample of primaries with a well-defined power-law scaling between

1355: luminosity and velocity dispersion.

1356: 

1357: One way of eliminating underluminous primaries within very massive

1358: halos from our sample is to consider the richness of their

1359: surroundings. Massive halos will typically host a large number of

1360: galaxies and in such systems, despite our isolation criteria, which

1361: ensures the dominance of the central galaxy, the other members may

1362: contribute a significant fraction of the total luminosity. This is

1363: shown in the other three panels of Figure~\ref{figs:lsg}, where we

1364: consider only primaries making up more than, respectively, $70\%$,

1365: $85\%$, and $95\%$ of the total combined luminosity of galaxies within

1366: the host halo, $L_{\rm sys}$. The stricter the criteria for selecting

1367: the primaries the nearer the scaling is to the ``natural'' $L\propto

1368: \sigma^3$ relation. Selecting truly isolated galaxies thus requires

1369: more than just imposing a magnitude gap, but also a color cut and a

1370: conscientious survey of the surroundings to weed out fossil groups and

1371: poor clusters from the sample which may unduly bias the $L$-$\sigma$

1372: scaling. Alternatively, one may try and eliminate from the

1373: analysis the brightest galaxies, where the contamination by ``fossil''

1374: systems is worse. This is the approach adopted by \cite{prada03}, who

1375: show that the ``natural'' scaling may also be recovered in that case.

1376: 

1377: Despite the pruning of the sample, the dependence of the satellite

1378: velocity dispersion on the color of the primary remains. For our

1379: strictest isolation criterion, $L_{\rm host} > 0.95 \, L_{\rm sys}$,

1380: the velocity dispersion of satellites of red galaxies is $\sim 55\%$ higher

1381: than that of blue galaxies of given luminosity. This is reminiscent of

1382: the observations of Brainerd T. (2004a,b) in the 2dFGRS

1383: who found larger velocity dispersion in satellites associated to 

1384: early-type (red) primaries than satellites of late-type (blue)

1385: hosts.\nocite{brainerd04a,brainerd04b}

1386: 

1387: This effect is due largely to the different stellar mass-to-light

1388: ratios of galaxies of different colors. Indeed, as shown in

1389: Figure~\ref{figs:mstsg}, the shift in velocity dispersion between

1390: primaries of different color basically disappears when {\it stellar

1391: masses} are considered instead of $r$-band luminosities. Using,

1392: whenever possible, stellar mass estimates rather than luminosity in

1393: order to bin galaxies is likely to give more robust results. Combining

1394: this with a strict isolation criterion that evaluates not only the

1395: luminosity gap between brightest and second-brightest galaxy but also

1396: the richness of the surrounding field appears essential in order to

1397: avoid biases and to recover the natural scaling expected from

1398: hierarchical structure formation scenarios.

1399: 

1400: 

1401: \subsection {Spatial anisotropies and the Holmberg effect}

1402: \label{ssec:holm}

1403: 

1404: An issue that has drawn recurring attention over time is whether the

1405: spatial distribution of satellites around primaries has anisotropies

1406: of particular significance. For example, the brightest satellites

1407: around the Milky Way seem to align on a plane perpendicular to the

1408: disk of the Galaxy \citep{lyndenbell82,majewski94,kroupa05,libeskind05},

1409: and a similar result seems to apply to

1410: at least some of the satellites of the Andromeda galaxy

1411: \citep{kochandgrebel06,metz07}. The small number of

1412: satellites involved in these analyses precludes robust conclusions to

1413: be drawn on the basis of the Local Group \citep{zentner05}, but

1414: it is intriguing that both Holmberg (1969) and \citet{zaritsky97b}

1415: find a similar effect in their samples of satellites of spiral

1416: galaxies.

1417: 

1418: The advent of large datasets, mostly from the SDSS, has allowed the

1419: issue to be revisited, and the latest work suggests that satellites of

1420: isolated spirals tend to distribute themselves preferentially along

1421: the direction of the disk: the {\it opposite} of the effect claimed by

1422: Holmberg and present in the Milky Way

1423: \citep{brainerd05,azzaro06,yang06, agustsson07}. One recent paper

1424: \citep{sales04} argued that satellites in the 2dfGRS actually

1425: follow Holmberg's suggestion, but this was in error due apparently to

1426: ambiguities in the way position angles are defined in the 2dfGRS

1427: database, and has now been resolved \citep{yang06}.

1428: 

1429: The anti-Holmberg effect is due to misalignment between the angular

1430: momentum and the triaxial structure of the host halos. A long

1431: literature has now established dark matter halos to be triaxial

1432: objects, with a preference for nearly prolate shapes, and whose

1433: angular momentum is perpendicular to the major axis of the halo

1434: \citep{frenk88,bullock01a,bullock02,jing02,bailin05,hopkins05,bett07}.

1435: If satellites trace the shape of the dark matter halo and if the spin of

1436: the central galaxy disks preserves the direction of the halo angular

1437: momentum, then this would explain the scarcity of satellites near the

1438: rotation axis of the disk.

1439: 

1440: We examine this in Figure~\ref{figs:Holmb}, where the anisotropy in

1441: the spatial distribution of satellites around primaries is measured by

1442: the distribution of the cosine of the polar angle (measured from the

1443: angular momentum axis of the SUBFIND subhalo centred on the host galaxy).

1444: An isotropic distribution

1445: would be horizontal in this plot, and it is clear that the spatial

1446: distribution of satellites is significantly anisotropic. Simulated

1447: satellites clearly show an ``anti-Holmberg'' effect, aligning

1448: themselves preferentially along the plane perpendicular to the angular

1449: momentum axis of the halo.

1450: 

1451: The effect is significant but relatively weak; satellites along the

1452: plane outnumber those closer to the rotation axis by roughly

1453: $2$:$1$. The effect depends only very weakly on the luminosity of the

1454: primary, as shown in Figure~\ref{figs:Holmb}, or on the relative

1455: brightness of the satellites. We have also checked that the effect

1456: is essentially independent of the color of the

1457: satellites. Quantitatively, we show this in Table~\ref{tab:costheta},

1458: where we report the {\it average} value of $|cos(\theta)|$ for various

1459: combinations of primary/satellite luminosity/color.

1460: 

1461: A theoretical study of these alignment issues based on an

1462: N-body/semi-analytic model similar to but substantially smaller than

1463: the Millennium Simulation has recently been carried out by \citet{kang07}.

1464: These authors did not require their primary systems to be

1465: isolated, nor did they require a substantial magnitude difference

1466: between host galaxy and satellites.  As a result it is difficult to

1467: compare their results quantitatively with our own.  Nevertheless,

1468: there are a number of results in common between our two studies.  They

1469: found that the observed alignments between satellite and central

1470: galaxies were best explained by assuming the minor central galaxy to

1471: align with the spin of its host subhalo (as assumed here) rather than

1472: assuming the major axis of the galaxy to align with that of its

1473: subhalo, and they found alignments of similar strength to those we show

1474: here. Their model gave dependences of the strength of the alignment

1475: signal both on the colour of the host galaxy and on the colour of the

1476: satellites. Such trends are weak or absent for the MS samples we study

1477: here. This difference is most likely due to the much wider range of

1478: systems included in the Kang et al study.

1479: 

1480: The weak dependence we find of the anisotropy on galaxy properties such as

1481: luminosity and color seems at odds with a number of observational

1482: studies. Some of them suggest that only satellites of red primaries

1483: are anisotropically distributed (Azzaro et al. 2006, Yang et al. 2006,

1484: Agustsson \& Brainerd 2007), and that the effect is enhanced when

1485: considering red companions to red primaries (Yang et al 2006). We

1486: note, however, that this result may be affected by the large scatter

1487: in the mass-luminosity-color relation discussed in

1488: \S~\ref{ssec:lmc}. Satellites are typically searched within a fixed

1489: radius (of order $500$ kpc), a choice that would lead to the inclusion

1490: of a larger fraction of interlopers around blue primaries, which tend

1491: to inhabit halos of lower mass. This may dilute the anisotropy in blue

1492: subsamples, explaining the observational results.

1493: 

1494: A similar comment applies to other potential correlations reported in

1495: the literature, such as the trend for spatial anisotropies to decline

1496: at large radius (Brainerd 2005; Augustsson \& Brainerd 2007) or to

1497: increase with halo mass (Yang et al 2006). As this paper was nearing

1498: submission, \citet{bailin07} argued in a recent preprint that the

1499: alignment of satellites may significantly depend on the isolation

1500: criteria applied. These authors find that satellites of ``truly''

1501: isolated SDSS primaries {\it do} show a polar excess (the original

1502: ``Holmberg effect''). We have checked for this in our catalogue, but

1503: find little dependence of our conclusions on $L_{\rm host}/L_{\rm

1504: sys}$. A detailed assessment of the correlations claimed by Bailin et

1505: al (2007) in light of our results is beyond the scope of this paper,

1506: but it is clearly an issue of interest to which we plan to return in

1507: future work.

1508: 

1509: 

1510: 

1511: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1512: \begin{center}

1513: \begin{figure}

1514: \includegraphics[width=84mm]{figures/alignment_lumhost_cutlumsat.ps}

1515: \caption{Spatial anisotropy of the satellite spatial distribution, as

1516: measured by the distribution of the cosine of the polar angle,

1517: $|cos(\theta)|$, measured from the rotation axis of the host

1518: halo. Different colors correspond to different primary luminosity, as

1519: labelled. Different line types correspond to varying the magnitude

1520: range used to select satellites, also as labelled (see also

1521: Figure~\ref{figs:dprofl}). An isotropic distribution would be

1522: horizontal in this plot. Satellites show a well-defined

1523: ``anti-Holmberg'' effect; i.e., they tend to populate preferentially

1524: the plane {\it perpendicular} to the angular momentum axis of the host

1525: halo. }

1526: \label{figs:Holmb}

1527: \end{figure}

1528: \end{center}

1529: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1530: 

1531: %\subsection {Orbital alignment}

1532: %\label{ssec:orbal}

1533: 

1534: \section{Summary and Conclusions}

1535: \label{sec:conc}

1536: 

1537: We analyse a large sample of isolated galaxies and their satellites

1538: selected from the semianalytic galaxy catalogue constructed by Croton

1539: et al (2006)  from the {\it Millennium Simulation} ({\it MS}). The large

1540: number of galaxies in the catalogue, together with the large volume

1541: surveyed by the {\it MS}, allow us to characterize in detail the 3D

1542: dynamical properties of the satellite population of bright isolated

1543: galaxies. Our isolation criterion stipulates that all galaxies within

1544: $1\, h^{-1}$ Mpc should be at least 2 mag fainter than the

1545: primary. This criterion typically selects galaxies in sparse

1546: environments, but also picks systems of galaxies with peculiar

1547: luminosity gaps between the two brightest galaxies, such as ``fossil

1548: groups''.

1549: 

1550: Our main conclusions may be summarized as follows:

1551: 

1552: \begin{itemize}

1553: 

1554: \item The relation between the halo mass of isolated galaxies and

1555: their luminosity shows very large scatter (halo masses span over a

1556: decade in mass at given luminosity), compromising the ability of

1557: studies that rely on stacking satellites of galaxies of similar

1558: luminosity to probe their dark matter halos. Selecting primaries by

1559: color in order to eliminate the reddest and bluest primaries helps to

1560: tighten the mass-luminosity relation, but still a number of

1561: ``underluminous'' central galaxies of massive groups remain in the

1562: sample. These may, however, be excluded by surveying the environment

1563: of the primary and rejecting those in regions of anomalously high

1564: richness.

1565: 

1566: \item One corollary of the above conclusion is that the relation

1567: between primary luminosity and satellite velocity dispersion is rather

1568: sensitive to the primary selection criteria. Stacking all satellites

1569: of all primaries leads to an $L$-$\sigma$ relation that is poorly

1570: approximated by a power law, and much shallower that the $L\propto

1571: \sigma^3$ scaling expected for hierarchical models. Only after weeding

1572: out massive halos with underluminous central galaxies do we recover

1573: the expected $L \propto \sigma^3$ scaling.

1574: 

1575: \item Since our isolation criterion readily selects the central

1576: galaxies of ``fossil'' groups, our analysis may be used to predict the

1577: abundance of groups with unusual gaps in the luminosity of the two

1578: brightest galaxies. We find that about $8$ to $10\%$ of halos

1579: exceeding $10^{13} \, M_\odot/h$ would qualify as ``fossil'' systems,

1580: a result that seems consistent with the (so far rather uncertain)

1581: observational constraints. We examine recent claims that the

1582: luminosity function of ``fossil'' groups may be difficult to reproduce

1583: in the $\Lambda$CDM cosmogony but find no obvious discrepancy with

1584: observational constraints for the three fossil group luminosity

1585: functions in the literature. Further data are needed to settle this

1586: issue.

1587: 

1588: \item The density profile of satellites around primaries may be well

1589: approximated by an NFW profile that is slightly less concentrated than

1590: the average dark matter profile. This conclusion is sensitive to the

1591: color of the satellites; red satellites are significantly {\it more}

1592: concentrated than the dark matter; the opposite is true for {\it blue

1593: } satellites. We also find evidence for luminosity segregation in the

1594: satellite population; i.e., a weak tendency for satellites near the

1595: primary to be brighter than those further away.

1596: 

1597: \item The velocity distribution of satellites is, like the dark

1598: matter, dominated by radial motions within the virial radius. The

1599: anisotropy is maximal at intermediate radii, becoming gradually more

1600: isotropic near the virial radius. This is a result of the

1601: radially-increasing contribution of satellites on their first infall

1602: onto the primary, a population of objects with rather small dispersion

1603: in radial velocity whose contribution raises the importance of

1604: tangential motions in the outskirts of the host halo.

1605: 

1606: \item Satellites are distributed anisotropically around primaries,

1607: with a well-defined but relatively weak preference for the plane

1608: perpendicular to the angular momentum of the halo (an

1609: ``anti-Holmberg'' effect). This is consistent with the latest

1610: observational studies, and is a direct result of misalignment between

1611: the angular momentum axis and the triaxial structure of dark matter

1612: halos.

1613: 

1614: \end{itemize}

1615: 

1616: The characterization of the satellite population of isolated, bright

1617: galaxies we present here has several goals: (i) to guide the

1618: compilation of primary/satellite systems that minimize the presence of

1619: interlopers, (ii) to facilitate the interpretation of observational

1620: results, and (iii) to provide predictions that may be used to validate

1621: the semianalytic model of galaxy formation applied to the {\it

1622: Millennum Simulation}, as well as, more generally, the hierarchical

1623: nature of galaxy assembly. Many of our results are amenable to direct

1624: confrontation with observation, and it is to be hoped that such a

1625: comparison will provide a number of insights into galaxy formation

1626: physics, and perhaps even some challenges to the $\Lambda$CDM

1627: paradigm.

1628: 

1629: 

1630: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1631: %------- TABLE WITH HALO MASS STATISTICS ---------

1632: \begin{table*}

1633: \caption{Abundance of {\it primary} galaxies in the {\it Millennium

1634: Simulation} as a function of halo mass. Columns 2 and 3 indicate the

1635: median and the rms dispersion of the luminosity that correspond to

1636: primaries residing in halos within each mass bin. Columns 4 and 5 show

1637: the density ($n_{\rm prim}$) and the fraction ($f_{\rm prim}$) of

1638: halos in the Millennium Simulation that host a {\it primary} galaxy in

1639: our catalogue respectively. The last three columns list the

1640: contribution to $f_{\rm prim}$ in each mass bin after applying the

1641: color cuts shown in Figure~\ref{figs:lmc}.}

1642: \begin{tabular}{|c|c|c|c|c|c|c|c|}

1643: \hline

1644: %

1645: {$M_{200}$} & ${\log(L_r/L_\odot)}$ & $\sigma(\log(L_r/L_\odot))$ & $n_{\rm prim}$  

1646: & $f_{\rm prim}$ & $f_{g-r}$ (\%) & $f_{g-r}$ (\%) & $f_{g-r}$ (\%) \\

1647: $[M_\odot/h]$ & [median] & & $[h^3$Mpc$^{-3}]$& \% & $< 0.65$ & $<0.95 \ \& >0.65$ & $>0.95$ \\

1648: \hline

1649: %

1650: $10^{11}-10^{12}$ & 10.17 & 0.15 & $15.04 \times 10^{-4}$& 24.7 & 1.6 & 21.2 & 1.9 \\

1651: %

1652: $10^{12}-10^{13}$ & 10.51 & 0.18 & $5.50 \times 10^{-4}$& 72.1 & 0.6 & 37.0 & 34.5 \\

1653: %

1654: $10^{13}-10^{14}$ & 10.85 & 0.18 & $8.37 \times 10^{-6}$& 10.1 & 0.1 & 0.3 & 9.7 \\

1655: %

1656: $10^{14}-10^{15}$ & 11.34 & 0.16 & $0.34 \times 10^{-6}$& 8.3 & 0.2 & 0.6 & 7.5 \\

1657: \hline

1658: \end{tabular}

1659: \label{tab:mfrac}

1660: \end{table*}

1661: %

1662: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1663: %------- TABLE WITH HALO MASS STATISTICS ---------

1664: \begin{table*}

1665: \caption{Abundance of {\it primary} galaxies in the {\it Millennium

1666: Simulation} as a function of $r$-band luminosity. Columns 2 and 3

1667: indicate the median and the rms dispersion of the halo mass where

1668: primaries of a given luminosity reside. Columns 4 and 5 show the

1669: density ($n_{\rm prim}$) and the fraction ($f_{\rm prim}$) of galaxies

1670: in the Millennium Simulation that are classified as {\it primaries} in

1671: our catalogue respectively. The last three columns list the

1672: contribution to each luminosity bin after applying the color cuts

1673: shown in Figure~\ref{figs:lmc}.  }

1674: \begin{tabular}{|c|c|c|c|c|c|c|c|}

1675: \hline

1676: %

1677: {$\log{L_r}/L_\odot$} & ${\log(M_{200}/M_\odot)}$ & $\sigma(\log({M_{200}/M_\odot}))$ & $n_{\rm prim}$ 

1678: & $f_{\rm prim}$ & $f_{g-r}$ (\%) & $f_{g-r}$ (\%) & $f_{g-r}$ (\%) \\

1679:  & [median] & & $[h^3$Mpc$^{-3}]$ & \% & $< 0.65$ & $<0.95 \ \& >0.65$ & $>0.95$ \\

1680: \hline

1681: %

1682: $10.0-10.2$& 11.75 & 0.17 & $9.00 \times 10^{-4}$ & 18.8 & 1.3  & 15.03 & 2.5  \\

1683: $10.2-10.4$& 11.95 & 0.20 & $6.22 \times 10{-4}$  & 22.1 & 0.7  & 16.5  & 4.9 \\

1684: $10.4-10.6$& 12.19 & 0.23 & $3.82 \times 10^{-4}$ & 24.8  & 0.5  & 16.9  & 6.1 \\

1685: $10.6-10.8$& 12.44 & 0.25 & $1.39 \times 10^{-4}$ & 18.4  & 0.9  & 15.1  & 7.3  \\

1686: $10.8-11.0$& 12.64 & 0.37 & $2.03 \times 10^{-5}$ & 25.3  & 5.0  & 11.7  & 7.8 \\

1687: $>11$      & 13.34 & 0.76 & $3.70 \times 10^{-6}$ & 25.1  & 15.6  & 1.92  & 7.8  \\

1688: %

1689: \hline

1690: \end{tabular}

1691: \label{tab:lfrac}

1692: \end{table*}

1693: %

1694: 

1695: %------- TABLE OF THE GAUSSIAN FITS TO THE SATELLITE VELOCITY DISTRIBUTIONS ---------

1696: \begin{table*}

1697: \caption{Moments of the satellite velocity distributions shown in

1698: Figures~\ref{figs:rvdist} and ~\ref{figs:vtandist} in several distance

1699: bins. The last column show $f_{\rm inf}$, the fraction of satellites in

1700: each radial shell that are in their first approach to the host.}

1701: \begin{tabular}{|c|c|c|c|c|c|c|}

1702: \hline

1703: Component & $r/r_{200}$ & mean & sigma ($\sigma$) & skewness ($\xi$) & kurtosis ($\kappa$) & $f_{\rm inf}$\\

1704: \hline

1705: $V_r$ & 0.0-0.5 & -0.014 & 0.96 & 0.025 & -0.68 & --\\

1706: & 0.5-1.0 & -0.072 & 0.70 & -0.002 & -0.45 & --\\

1707: & 1.0-1.5 & -0.133 & 0.54 & 0.014 & 0.23 & 15\%\\

1708: & 1.5-2.0 & -0.187 & 0.51 & 0.257 & 2.39 & 25\% \\

1709: & 2.0-2.5 & -0.208 & 0.52 & 0.687 & 5.97 & 36\% \\

1710: & 2.5-3.0 & -0.179 & 0.55 & 0.644 & 8.22 & 40\% \\

1711: \hline

1712: $V_\theta$ & 0.0-0.5 & 0.0 & 0.45 & 0.004 & 0.99&--\\

1713: & 0.5-1.0 & 0.0 & 0.39 & 0.058 & 1.44&--\\

1714: & 1.0-1.5 & 0.0 & 0.49 & -0.087 & 2.30& 15\%\\

1715: & 1.5-2.0 & 0.0 &  0.51 & 0.143 & 8.82& 25\%\\

1716: & 2.0-2.5 & 0.0 & 0.55 & 0.002 & 16.58& 36\%\\

1717: & 2.5-3.0 & 0.0 & 0.60 & 0.240 & 20.35& 40\%\\

1718: \hline

1719: $V_\phi$ & 0.0-0.5 & 0.20 & 0.x55 & -0.020 & 0.48&--\\

1720: & 0.5-1.0 & 0.25 & 0.50 & -0.240 & 1.15 & --\\

1721: & 1.0-1.5 & 0.22 & 0.51 & -0.456 & 2.23 &  15\%\\

1722: & 1.5-2.0 & 0.15 & 0.52 & -0.387 & 4.00 &  25\%\\

1723: & 2.0-2.5 & 0.10 & 0.57 & -0.785 & 18.23 & 36\%\\

1724: & 2.5-3.0 & 0.05 & 0.60 & -0.015 & 15.00 & 40\%\\

1725: \hline

1726: \end{tabular}

1727: \label{tab:Gaussfits}

1728: \end{table*}

1729: %-------------------------------------------------

1730: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1731: \begin{table*}

1732: \caption{Average $|cos(\theta)|$ (where $\theta$ is the polar angle

1733: measured from the halo angular momentum axis) of all satellites for

1734: different bins in host luminosity and color.  The last three columns

1735: show the values corresponding to three equal-number subsamples: blue

1736: satellites ($(g-r)_{\rm sat} < 0.95$), intermediate color satellites

1737: ($0.95<(g-r)_{\rm sat}< 1.01$) and red satellites ($(g-r)_{\rm sat}

1738: >1.01$).}

1739: \begin{tabular}{|c|c|c|c|c|}

1740: \hline

1741: % & $[h^3$Mpc$^{-3}$]  & \% & $< 0.65$ & $<0.95 \ \& >0.65$ & $>0.95$ \\%

1742: $M_{\rm host}$ & $\langle{|cos(\theta)|}\rangle$ & $\langle{|cos(\theta)|}\rangle$ & $\langle{|cos(\theta)|}\rangle$ & $\langle{|cos(\theta)|}\rangle$\\

1743: %

1744: & all & $(g-r)_{\rm sat}<0.95$ & $0.95<(g-r)_{\rm sat}<1.01$ & $(g-r)_{\rm sat}>1.01$ \\ 

1745: \hline

1746: %

1747: $-21.5<M_{\rm host}<-20.5$ & 0.410 & 0.404 & 0.408 & 0.422\\

1748: $-22.5<M_{\rm host}<-21.5$ & 0.410 & 0.447 & 0.441 & 0.439\\

1749: $M_{\rm host}<-22.5$ & 0.410 & 0.476 & 0.463 & 0.452\\

1750: %

1751: \hline

1752: $(g-r)_{\rm host}$ & & & & \\

1753: \hline

1754: %

1755: $(g-r)_{\rm host}\leq0.65$ & 0.426 & 0.437 & 0.421 & 0.417\\

1756: $0.65<(g-r)_{\rm host}\leq0.95$ & 0.428 & 0.447 & 0.430 & 0.436\\

1757: $(g-r)_{\rm host}>0.95$ & 0.426 & 0.497 & 0.458 & 0.450\\

1758: %

1759: \hline

1760: \end{tabular}

1761: \label{tab:costheta}

1762: \end{table*}

1763: %

1764: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

1765: 

1766: \section*{Acknowledgments}

1767: 

1768: LVS would like to thank Dr. Mario Abadi for many helpful suggestions

1769: and discussions. This work was partially supported by the Latin

1770: American European Network on Astrophysics and Cosmology of the

1771: European Union's ALFA Programme and the Consejo Nacional de

1772: Investigaciones Cient\'{\i}ficas y T\'ecnicas (CONICET),

1773: Argentina. LVS is grateful for the hospitality of the Max-Planck

1774: Institute for Astrophysics in Garching, Germany, where much of the

1775: work reported here was carried out. Data for the galaxy formation

1776: model on which this study is based are available at 

1777: {\small \tt http://www.mpa-garching.mpg.de/galform/agnpaper/}

1778: Data for other galaxy formation models and

1779: for the halo/subhalo populations of the {\it Millennium Simulation}

1780: are available for all redshifts at {\small \tt

1781: http://www.mpa-garching.mpg.de/millennium}.

1782: We acknowledge Eduardo Cypriano for sending us electronic

1783: data for the fossil groups RX1416 and RX1552. We also thank

1784: the anonymous referee for useful suggestions and comments

1785: that helped to improve the previous version of the paper.

1786: 

1787: \bibliography{references}

1788: 

1789: \end{document}

1790: