1: 

2: \documentclass[12pt,preprint]{aastex}

3: 

4: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%

5: 

6: \slugcomment{The Astrophysical Journal, in press}

7: \shorttitle{Anisotropic Locations of Satellite Galaxies}

8: \shortauthors{Agustsson \& Brainerd}

9: 

10: \begin{document}

11: 

12: \title{Anisotropic Locations of Satellite Galaxies: Clues to

13: the Orientations of Galaxies within their Dark Matter Halos}

14: 

15: \author{Ingolfur Agustsson \& Tereasa G. Brainerd}

16: \affil{Boston University, Institute for Astrophysical Research, 725

17: Commonwealth Ave., Boston, MA 02215}

18: \email{ingolfur@bu.edu, brainerd@bu.edu}

19: 

20: \begin{abstract}

21: We investigate the locations of the satellites of relatively isolated

22: host galaxies in the Sloan Digital Sky Survey and the Millennium

23: Run simulation.  Provided we use two distinct prescriptions to embed luminous

24: galaxies within the simulated dark matter halos (ellipticals share the

25: shapes of their halos, while disks have angular momenta that are aligned

26: with the net angular momenta of their halos),

27: we find a fair agreement between observation and theory.

28: Averaged over scales $r_p \le 500$~kpc,

29: the satellites of red, high--mass hosts with low

30: star formation rates are found preferentially near the major axes of

31: their hosts.  In contrast, the satellites of blue, low--mass hosts with

32: low star formation rates show little to no anisotropy when averaged over the

33: same scale.  The difference between the locations of the satellites of

34: red and blue hosts cannot be explained by the effects of interlopers

35: in the data.

36: Instead, it is caused primarily by marked differences in the

37: dependence of the mean satellite

38: location, $\left< \phi \right>$, on the projected distance at which the

39: satellites are found.  We also find that

40: the locations of red, high--mass satellites with low star formation rates

41: show considerably more anisotropy than do the locations of

42: blue, low--mass satellites with high star formation rates.  There are two

43: contributors to this result.  First, the blue satellites

44: have only recently arrived within their hosts' halos, while the red

45: satellites arrived in the far distant past.  Second, the sample

46: of blue satellites is heavily contaminated by interlopers, which suppresses

47: the measured anisotropy compared to the intrinsic anisotropy.

48: \end{abstract}

49: 

50: \keywords{

51: dark matter --- galaxies: dwarf --- galaxies: fundamental

52: parameters --- galaxies: halos --- galaxies: structure}

53: 

54: \section{Introduction}

55: The locations of satellite galaxies, measured with respect to the symmetry

56: axes of their hosts, may hold important clues to the formation

57: of large galaxies.  This is especially true for Cold Dark Matter 

58: (CDM) models in which the dark matter halos of galaxies are

59: mildly flattened, and galaxy formation and mass accretion occur within

60: filaments.  Some early studies of the locations of satellite galaxies

61: suggested that satellites had a preference for being located near the

62: minor axes of their hosts (e.g., Holmberg 1969; Zaritsky et al.\ 1997), an observation

63: that is sometimes known as the ``Holmberg effect''.    

64: Valtonen et al.\ (1978) found exactly the opposite effect, and concluded that

65: compact satellites tended to be aligned with the major axes of their hosts.

66: Other early studies suggested

67: that any tendency for satellite galaxies to be found in preferred locations

68: was at best rather weak, and perhaps non--existent (e.g., Hawley \& Peebles 1975;

69: Sharp et al.\ 1979; MacGillivray et al.\ 1982).  

70: All of these early studies were based on relatively small samples of between

71: $\sim 10$ and $\sim 200$ satellite galaxies and as

72: modern, extensive redshift surveys have

73: become available, the observed number of host--satellite systems has

74: increased enormously.  Based upon these large surveys it

75: now appears that, when averaged over all host--satellite pairs, the 

76: satellites of relatively isolated host galaxies

77: have a tendency to be found near the

78: major axes of their hosts (see, e.g., Brainerd 2005).

79: There is, however, increasing evidence that the locations of the satellites  

80: depend upon host type (e.g., red vs.\ blue), as well as satellite type.

81: 

82: In an analysis of the locations of the satellites of relatively isolated

83: host galaxies in the Two Degree Field Galaxy Redshift Survey (2dFGRS; Colless

84: et al.\ 2001, 2003), Sales \& Lambas (2004) found a tendency for the

85: satellites of early--type hosts to be located near the major axes

86: of the hosts, while the satellites of late--type hosts were consistent

87: with being distributed isotropically (see the erratum by Sales \& Lambas

88: 2009). In addition, they found a tendency for

89: the locations of

90: satellites with low star formation rates to show a greater degree of 

91: anisotropy than satellites with high star formation rates. 

92: Azzaro et al.\ (2007, hereafter APPZ) concluded that, as a whole, 

93: the satellites of relatively isolated host galaxies in the Sloan

94: Digital Sky Survey (SDSS; Fukugita et al.\ 1996; Hogg et al.\ 2001; Smith

95: et al.\ 2002; Strauss et al.\ 2002; York et al.\ 2000) were found preferentially

96: near the major axes of their hosts.  Further, APPZ found that the degree of

97: anisotropy was greatest for the red satellites of red host galaxies, while

98: the locations of the satellites of blue host galaxies were consistent with

99: an isotropic distribution.  Similar results were found by Siverd et al.\ (2009)

100: in a more recent analysis of the SDSS, where they showed that the satellites

101: of red, centrally--concentrated hosts are found preferentially close to the

102: major axes of the hosts, and the effect is strongest for red, centrally--concentrated

103: satellites.  In a study of extremely isolated SDSS host galaxies, Bailin

104: et al.\ (2008) found that the satellites of spheroidal host galaxies were

105: located preferentially close to the major axes of the hosts, while the satellites

106: of blue disk hosts were distributed isotropically.

107: 

108: The dependence of

109: satellite location on the color of the host has also been observed

110: within group environments by Yang et al.\ 

111: (2006, hereafter Yang06),

112: who found that the satellites of red central galaxies in

113: the SDSS had

114: a strong tendency to be aligned with the major axes of the central galaxies, while

115: the satellites of blue central galaxies were distributed isotropically about the

116: central galaxies.  Further, Yang06 found that the red satellites of

117: red central galaxies were distributed much 

118: more anisotropically than were the blue

119: satellites of red central galaxies, and the degree of anisotropy in the 

120: satellite locations increased only weakly

121: with the mass of the surrounding halo.

122: 

123: Here we further investigate the anisotropic distribution

124: of satellite galaxies around relatively isolated hosts,  focussing

125: on the dependence of the anisotropy on various physical parameters 

126: of the hosts and the satellites (e.g., rest--frame color, specific star formation

127: rate and stellar mass).   We also investigate the effects of ``interlopers''

128: (i.e., false satellites) on the locations of the satellites, as well

129: as the dependence

130: of satellite location on projected distance from the host.

131: The locations of satellites 

132: in the observed Universe are computed using SDSS galaxies, and these are

133: compared to the locations of satellites in the

134: $\Lambda$CDM Millennium Run simulation.  

135: Our work here is similar in spirit to that of Kang et al.\ (2007; hereafter Kang07), who

136: used a simulation that combined N-body calculations with

137: semi-analytic galaxy formation

138: to compare the locations of satellite

139: galaxies in a $\Lambda$CDM universe to the results obtained by Yang06

140: for SDSS satellites.  Our work differs from that of Kang07

141: in a number of ways, however.  First, we focus on the satellites of

142: relatively isolated host galaxies whereas

143: Yang06 and Kang07 focus 

144: primarily on group systems.

145: Second, in our work we use the stellar masses

146: of the host and satellite galaxies when exploring the dependence of the 

147: satellite locations on mass.  In contrast, Yang06 and Kang07

148: use a group luminosity function to assign masses to the dark matter halos

149: that surround their groups.

150: Third, we divide our theoretical

151: galaxies into two broad classes, elliptical and non--elliptical, and we

152: use different prescriptions to assign shape parameters to the luminous

153: portions of these galaxies.  Kang07, however, did not divide their

154: theoretical galaxies into different classes and they 

155: used identical prescriptions

156: to assign shape parameters to the luminous portions of all of their galaxies.

157: 

158: We note that Sales et al.\ (2007) have also investigated the locations of

159: satellite galaxies of relatively isolated host galaxies in the Millennium

160: Run.  Their approach, however, was rather different than our own.  Sales

161: et al.\ (2007) use the full information of the simulation (in particular, 

162: 3D distances) to select their hosts and satellites, while we focus on samples

163: that are selected using the same selection criteria that 

164: are used to select hosts and satellites from large redshift surveys.  Having

165: full 3D information, Sales et al.\ (2007) selected all satellites 

166: with $M_r < -17$ that were 

167: found within the virial radii of their hosts and computed the locations of 

168: the satellites.  The result was preference for the satellites to populate

169: a plane that is perpendicular to the angular momentum axis of the host's

170: halo (i.e., the reverse of the Holmberg effect).

171:  

172: The outline of the paper is as follows. In \S2 we describe the

173: SDSS data, the Millennium Run simulation, and the way in which we

174: define images for the luminous host galaxies in the Millennium

175: Run.  In \S3 we discuss the

176: selection criteria for finding hosts and satellites, and we highlight some

177: of the properties of the host and satellite galaxies in the Millennium Run.

178: In \S4 we compute the locations of the satellite galaxies and we compare

179: the results obtained with SDSS galaxies to those obtained with the Millennium

180: Run galaxies.  We summarize our results and compare them

181: to previous, similar studies in

182: \S5, and we present our conclusions in \S6.

183: Throughout we adopt cosmological parameters 

184: $H_0 = 73$~km~sec$^{-1}$~Mpc$^{-1}$, $\Omega_{m0} = 0.25$, and

185: $\Omega_{\Lambda 0} =

186: 0.75$.

187: 

188: 

189: \section{Observational and Theoretical Data Sets}

190: 

191: Our goal in this paper is to compute the locations of the

192: satellites of relatively isolated host galaxies for: (i) observed

193: galaxies in our Universe and (ii) theoretical galaxies in a $\Lambda$CDM

194: universe.  Below we outline the details of the observational and theoretical

195: data sets that are used in our analysis.

196: 

197: \subsection{Observed Galaxies: SDSS}

198: 

199: The SDSS is a large imaging and spectroscopic survey that has

200: mapped roughly one quarter of the sky. The spectroscopic portion of the

201: SDSS is complete to a reddening--corrected 

202: Petrosian magnitude of $r=17.77$ (see, e.g., Strauss et al. 2002).  Our primary

203: observational data set consists of the

204: seventh data release of the SDSS (DR7; Abazajian et al.\ 2009), including

205: all of the photometric and spectroscopic information for objects with high

206: quality redshifts (zconf $>$ 0.9) that have galaxy--type spectra

207: (specClass = 2), $r\le 17.77$, and redshifts in the range $0.01 \le z \le 0.15$.

208: 

209: We use the de--reddened Petrosian

210: $ugriz$ magnitudes (e.g., petroMag\_r-extinction\_r), and we 

211: select the

212: position angles, semi--minor axes, and semi--major

213: axes of our galaxies from the Petrosian $r$--band data.  In addition,

214: the IDL code by Blanton et al. (2003; v4\_1\_4) was used

215: to K-correct the SDSS galaxy colors to the present epoch (i.e., $z=0$).

216: Further, in some of the analyzes below we will supplement the data provided

217: directly by the SDSS with stellar mass estimates and

218: star formation rates.

219: Stellar masses are available for the 

220: vast majority of the galaxies in the DR7, but at the moment star formation

221: rates are only available for galaxies in the fourth SDSS data release (DR4;

222: Adelman-McCarthy et al.\ 2006).  Therefore, our galaxy sample will necessarily

223: be restricted when we look at the dependence of satellite location on

224: star formation rate. The 

225: stellar masses and star formation rates for

226: the SDSS galaxies are publicly available at {\tt http://www.mpa-garching.mpg.de/SDSS/}.

227: Stellar masses in these catalogs were computed using the philosophy of

228: Kauffmann et

229: al.\ (2003) and Salim et al.\ (2007). Star formation rates

230: were computed using various emission

231: lines in the SDSS spectra as described in Brinchmann et al.\

232: (2004). Throughout our analysis we use the specific star formation

233: rate (SSFR) of the SDSS galaxies, which is

234: defined to be the ratio of the star formation rate (in $M_\odot$~yr$^{-1}$) to

235: the stellar mass (in solar units), and we use the average values of

236: the likelihood distributions of the total SSFR obtained by

237: Brinchmann et al.\ (2004).   

238: 

239: 

240: \subsection{Theoretical Galaxies: Millennium Run Simulation}

241: 

242: The Millennium Run simulation\footnote{http://www.mpa-garching.mpg.de/millennium}

243: (MRS) follows the

244: growth of cosmic structure in a $\Lambda$CDM ``concordance'' cosmology

245: ($H_0 = 73$~km~sec$^{-1}$~Mpc$^{-1}$,$\Omega_{m0} + \Omega_{b0} = 0.25$,

246: $\Omega_{b0} = 0.04$, $\Omega_{\Lambda 0} = 0.75$, $n=1$,

247: $\sigma_8 = 0.9$).  The simulation was

248: completed by the Virgo Consortium in summer 2004 using the Max

249: Planck Society's supercomputer center in Garching, Germany, and

250: is described in Springel et al.\ (2005).

251: The simulation follows the evolution of the dark matter

252: distribution from $z = 127$ to $z = 0$ using $N = 2160^3 \simeq

253: 10^{10}$ particles of mass $m_p = 8.6 \times 10^8 h^{-1} M_{\sun}$.

254: The simulation volume is a cubical box with periodic 

255: boundary conditions and a comoving side length of $L = 500 h^{-1}$~Mpc.

256: A TreePM method is used to evaluate the 

257: gravitational force law, and a softening length of $5

258: h^{-1}$~kpc is used.  The simulation thus achieves a truly impressive dynamic

259: range of $10^5$ in length. 

260: Since one of our goals is to construct an

261: accurate catalog of simulated host galaxies and their satellites, 

262: it is important for us to use a high--resolution

263: simulation that follows the fate of satellite galaxies accurately

264: as they orbit within the halo of the 

265: central host galaxy. The combination of high spatial and mass resolution

266: therefore makes the MRS

267: ideal for our purposes.

268: 

269: The stored output of the MRS allows semi--analytic models

270: of galaxy formation to be implemented by collecting

271: the detailed assembly histories of all resolved halos and

272: subhalos, then simulating the formation and

273: evolution of galaxies within these structures for a variety of

274: assumptions about the physics that is involved. The data on the

275: halo, subhalo, and galaxy populations which have been produced by such

276: efforts can be used to address a wide range of questions

277: about galaxy and structure evolution

278: (e.g., Croton et al.\ 2006). As part of the activities of the

279: German Astrophysical Virtual Observatory, detailed

280: information about the halos, subhalos, and galaxies have been

281: publicly released for two independent models of galaxy

282: formation (Lemson et al.\ 2006). 

283: 

284: In order to compare to the SDSS,

285: we need to analyze the MRS in the same way 

286: in which one would analyze a combined imaging and redshift survey of the

287: observed Universe.   To do this, we make use of the MRS all--sky mock galaxy redshift

288: catalog\footnote{http://www.g-vo.org/Millennium/Help?page=databases/mpamocks/blaizot2006\_\_allsky}

289: that was constructed by Blaizot et al.\ (2005) using the Mock

290: Map Facility (MoMaF).   The MRS mock redshift survey is intended to 

291: mimic the SDSS, having a nearly identical redshift distribution and very similar

292: color distributions for the galaxies.   The mock redshift survey incorporates

293: the semi--analytic galaxy formation model of De Lucia \& Blaizot (2007) for the

294: MRS galaxies. Therefore, galaxy fluxes in all of the SDSS bandpasses, as well as star formation

295: rates, stellar masses,

296: and $B$--band bulge--to--disk ratios, are available for the MRS galaxies.

297: 

298: In order to make the most direct comparison to

299: the SDSS, we need to include the galaxy images 

300: that one would have in a real observational survey.  That is, our goal is

301: to determine the locations of satellite galaxies, measured with respect to

302: the major axes of the images of their luminous host galaxies.

303: There are, however, no actual {\it images} of the simulated galaxies, and

304: we must therefore {\it define} images for the MRS host galaxies.  

305: As an aid to defining the image shapes, the bulge--to--disk

306: ratios from the semi--analytic galaxy formation model may be used

307: to assign rough intrinsic morphologies to the MRS hosts.

308: Following De Lucia et al. (2006) we therefore use the $B$--band

309: bulge--to--disk ratios to classify MRS host galaxies with $\Delta M(B)

310: < 0.4$ as ellipticals, where $\Delta M(B) =

311: M(B)_{\rm bulge}-M(B)_{\rm total}$.  Similarly, we classify MRS host galaxies with 

312: $\Delta M(B) \geq 0.4$ as ``non--ellipticals''.

313: We also note that visual inspection

314: of the images of the SDSS host galaxies has revealed these objects to 

315: be ``regular'' systems (i.e.,

316: ellipticals, lenticulars, or spirals).  Therefore, it is reasonable to assume

317: that the non--elliptical MRS hosts

318: are disk systems with

319: significant net angular momentum, and we will treat all non--elliptical MRS

320: hosts as though they were disk galaxies below.

321: 

322: Following Heavens et al. (2000) we assume that 

323: elliptical MRS host galaxies share the shapes of their dark matter halos.  During a

324: collaborative visit to the Max Planck Institute for Astrophysics (MPA), we were 

325: fortunate to be granted access to the particle data files that resulted from

326: the MRS.  The enormous size of the particle files precludes them from being made 

327: publicly-available; thus, at present,

328: it is only possible to work with the files on site at MPA.   During the 

329: visit to MPA the particles within the virial radii ($r_{200}$)  of the

330: elliptical MRS host galaxies were identified, and these particles 

331: were then used to compute 

332: equivalent ellipsoids of inertia for the elliptical hosts.  A total of 98\%

333: of the elliptical MRS hosts contain more than 1000 particles within their

334: virial radii, so the equivalent ellipsoids of inertia are well--determined.

335: The major axes of projections of these equivalent ellipsoids of

336: inertia onto the sky then

337: define the orientations of the major axes of the elliptical MRS host galaxies.  

338: 

339: In the case of the non--elliptical MRS hosts, it is natural to assume that the

340: net angular momentum of the disk will be perpendicular to the disk.

341: In addition, recent numerical simulations have indicated 

342: that the angular momenta of disk galaxies

343: and their dark matter halos are reasonably well--aligned (e.g., Libeskind

344: et al.\ 2007). Furthermore, the disk angular momentum vectors show a

345: tendency to be aligned with the minor axis of the surrounding mass

346: with a mean misalignment of $\sim 25^{\circ}$ (Bailin \&

347: Steinmetz 2005).  We, therefore,

348: computed the angular momentum vectors of the halos of the non--elliptical

349: MRS hosts using all particles contained within the virial radii.  These were

350: then used to place thin disks within the halos, oriented such

351: that the disks are perpendicular to the net angular momenta of the halos.

352: The major axes of the projections of these thin disks onto the sky then

353: define the orientations of the major axes of the non--elliptical MRS hosts.  

354: We note that

355: the angular momentum vectors of the host halos are well--determined, and 62\% of the

356: hosts contain more than 1000 particles that were used to compute the angular

357: momentum.  

358: 

359: \section{Host--Satellite Catalogs}

360: 

361: Although the MRS contains full 6--dimensional phase space information (i.e., positions

362: and velocities) for all of the galaxies, such is of course not the case for the observed 

363: Universe.  That is, since there is no direct distance information for the vast majority

364: of the galaxies in the SDSS, we are forced to select host

365: galaxies and their satellites using proximity criteria in redshift space, rather

366: than real space.  Again, in order to compare the simulation results as directly

367: as possible to the results from the SDSS, we select host and satellite galaxies

368: in the MRS in the same way that they are selected in the SDSS.  Below we discuss

369: our selection criteria and the resulting catalogs.

370: 

371: \subsection{Host--Satellite Selection Criteria}

372: 

373: Hosts and satellites are selected by requiring the hosts to be

374: relatively isolated.  In addition, hosts and satellites must be nearby 

375: one another in terms of projected separation on the sky, $r_p$,

376: and radial velocity difference, $|dv|$. Throughout we use the Sample 1 selection

377: criteria from Brainerd (2005). Specifically, hosts must be 2.5

378: times more luminous than any other galaxy that falls within $r_p

379: \leq 700$~kpc and $|dv| \leq 1000$~km~sec$^{-1}$. Satellites must

380: be at least 6.25 times less luminous than their host, and they must be

381: located within $r_p \leq 500$~kpc and $|dv| \leq 500$~kpc.

382: In order to eliminate a small number of systems that

383: pass the above tests but which are, in reality, more likely to be

384: representative of cluster environments instead of relatively isolated

385: host--satellite systems, we impose two further restrictions: (1)

386: the sum total of the luminosities of the satellites of a given

387: host must be less than the luminosity of the host, and (2) the

388: observed total number of satellites of a given host must not

389: exceed 9. Our selection criteria yield relatively isolated hosts and their

390: satellites, and it is worth noting that both the Milky Way and M31

391: would be rejected as host galaxies under our selection criteria.

392: We also note that, although we have adopted one particular

393: host--satellite selection algorithm, the results are not particularly

394: sensitive to the details of the selection algorithm (see, e.g., Brainerd 2005;

395: Agustsson \& Brainerd 2006, hereafter AB06).

396: 

397: We know from the MRS that the hosts will span a wide 

398: range of virial masses and, hence, a wide range of virial radii.  Therefore,

399: very different parts of the halos are probed by applying a fixed search

400: aperture of 500~kpc for the satellites.  The selection technique that we

401: have used is, however, fairly standard in the literature, has the advantage

402: that it is simple to implement, and does not depend on any specific {\it a priori}

403: assumption that the luminosity of a galaxy is correlated with its mass.

404: There are some indications from previous studies (e.g., Yang06) that the

405: satellite anisotropy may be a function of radius and we will explore this

406: in the following section.

407: 

408: In addition, the simple host--satellite selection criteria that we 

409: adopt allow, at least in principle, for ``multi--homed'' satellites.  That 

410: is, in principle a given satellite could be paired with more than one host.

411: In practice, we find that this occurs extremely rarely 

412: and the results we present below are completely unaffected by the presence

413: of mutli--homed satellites.  Also, it is true that the selection criteria

414: allow for the presence of galaxies with luminosities

415: $L_{\rm host}/6.25 \le L \le L_{\rm host}/2.5$ nearby to the host, and these

416: galaxies are not used in our analyzes.  In practice $\sim 48$\%

417: of the hosts have ``non--selected'' satellites nearby to them.  Of

418: the hosts that have non--selected satellites, the vast majority (77\%) 

419: have only one (54\%) or two (46\%) non--selected satellites.   Because

420: of this, we refer to our host galaxies as being merely ``relatively''

421: isolated.

422: 

423: 

424: \subsection{SDSS Host--Satellite Catalog}

425: 

426: In addition to selection criteria above, we require that the images of the

427: SDSS galaxies are not associated with obvious aberrations in the imaging 

428: (for which we performed a visual check). We also require that

429: the host galaxies are not

430: located close to a survey edge (i.e., the host must be

431: surrounded by spectroscopic targets from the SDSS, within the area

432: of interest).  We limit our study to the redshift range $0.01 \le

433: z \leq 0.15$, where the lower limit helps ensure that

434: the peculiar velocities do not dominate over the Hubble flow,

435: and the upper limit simply reflects the fact that very few hosts can be

436: found beyond this redshift.  After imposing all of our selection criteria,

437: our primary SDSS catalog consists of 

438: 4,487 hosts and 7,399 satellites.  Note, however, that the size of the 

439: SDSS catalog is reduced when, below,

440: we further restrict our analyzes to SDSS

441: galaxies with measured stellar masses and specific star formation rates

442: (see Table~1).

443: 

444: \subsection{MRS Host--Satellite Catalog}

445: 

446: We select host and satellite galaxies from the mock redshift survey of

447: the MRS using the same 

448: redshift space proximity criteria that we

449: used for the SDSS.  Because of the shear size of the simulation, this 

450: results in a very large sample consisting of 70,882 hosts (of which

451: 30\% are classified as elliptical) and 140,712 satellites.

452: In addition we note that the semi-analytic model of De Lucia \&

453: Blaizot (2007) distinguishes each MRS

454: galaxy according to three distinct types: type 0, type 1, and type 2.

455: Type 0 galaxies are the central galaxies of their 

456: friends-of-friends (FOF) halos.  These objects

457: are fed by radiative cooling from the surrounding halo.

458: Type 1 galaxies are the central galaxies of

459: ``subhalos'', and they have their own self--bound dark matter subhalo.  Type 2

460: galaxies have been stripped of their dark matter and they lack distinct

461: substructure.  In our catalog of MRS

462: host--satellite pairs,

463: 94\% of the hosts are the central galaxies of

464: their own FOF halo (i.e., they are type 0 objects).  This 

465: assures us that our prescription for finding host

466: galaxies is working well.  In contrast to the MRS hosts,

467: the MRS satellites are primarily type 1 objects (41\% of the sample)

468: or type 2 objects (39\% of the sample).  That is, the vast majority of the MRS 

469: satellites that are selected by proximity to the host in redshift space

470: are, indeed, contained within a larger halo.

471: However, 20\%  of the

472: MRS satellites are central galaxies of their

473: own FOF halo (i.e., they are type 0 objects).  These latter objects are

474: examples of ``interlopers'' --

475: objects which pass the redshift space proximity tests but which

476: are not necessarily

477: nearby to a host galaxy.  Without actual distance information for the

478: galaxies, a certain amount of interloper contamination of the satellite

479: population cannot be avoided.  However, since

480: the SDSS and MRS

481: host--satellite catalogs were selected in the same way, we expect

482: that the contamination of the SDSS satellite sample by interlopers will

483: be similar to that for the MRS sample.  We will investigate the effects

484: of interlopers on the observed locations of satellite galaxies in 

485: \S4.2 below.

486: 

487: A summary of the basic properties of the hosts and satellites in the SDSS

488: (left panels) and the MRS (right panels) is shown in Fig.~1.  From top

489: to bottom, the panels of Fig.~1 show probability distributions for the

490: number of satellites per host (panels a and b), 

491: the redshift distributions of the hosts (panels c and d), the

492: distribution of apparent magnitudes for the hosts

493: and satellites (panels e and f), the

494: distribution of absolute magnitudes for the hosts

495: and satellites (panels g and h), and the distribution of stellar

496: masses for the hosts and satellites (panels i and j).  Although

497: the distributions are not identical for the SDSS and MRS, they

498: are sufficiently similar that a direct comparison of the locations of the

499: satellites in the SDSS and the MRS should be meaningful.

500: 

501: One of the great luxuries of simulations (as opposed to observations

502: of the real Universe) is that all the information about the simulated 

503: galaxies is known.  In the remainder of this section we highlight some

504: of the information about the MRS hosts and satellites that, for

505: the most part, is not known for the SDSS hosts and satellites. 

506: Fig.~2 shows the relationship between the halo virial mass and the

507: stellar mass for the MRS hosts (left panel), the dependence of the halo

508: virial mass on absolute $r$--band magnitude for the MRS hosts

509: (middle panel), and the variation of stellar mass with 

510: $(g-r)$ for the MRS hosts (right panel).  

511: From Fig.~2, then, it is clear that the stellar mass of

512: the MRS hosts correlates well with the virial mass of the halo and, therefore,

513: the absolute magnitude.  In addition, it is clear that the reddest MRS host

514: galaxies are also the most massive hosts in the simulation. 

515: 

516: Fig.~3 highlights information that is known about the MRS satellites.

517: To construct this figure, we use only those objects

518: which we consider to be genuine satellites in the host--satellite catalog.

519: We make this restriction for Fig.~3 because here we

520: are interested in the properties of the genuine satellites, not the

521: properties of the interlopers.  Here we accept as genuine satellites those

522: objects that are located within a physical distance, $r_{3D} \le 500$~kpc,

523: of a host galaxy.  This is a rather non--restrictive definition of a 

524: genuine satellite and is based simply upon a match to the search radius

525: (i.e., $r_p \le 500$~kpc) that is used in our host--satellite selection

526: criteria (see \S3.1).  In addition we define the redshift at which the satellite

527: first enters its host's halo to be the redshift at which the satellite first

528: becomes a member of the FOF group of particles to which the host belongs.

529: The top panels of Fig.~3 show that the stellar masses

530: of the MRS satellites correlate well with the absolute magnitude (panel a),

531: $(g-r)$ color (panel b), and the redshift at which the satellites first entered

532: the halos of their hosts (panel c).  That is, the more massive is a satellite,

533: the more luminous is the satellite, the redder it is at the present day, and

534: the earlier it first entered the halo of its host.

535: This agrees well with the results of Kang07 from

536: their analysis of the redshifts at which satellite galaxies with various masses

537: and colors first entered the halos surrounding central galaxies in

538: group systems.   Fig.~3d) shows that there

539: is a strong correlation of the present--day color of a satellite and the

540: redshift at which it first entered its host's halo; the very reddest satellites

541: entered the halo more than 10~Gyr in the past, and the very bluest satellites

542: entered the halo within the past 1.5~Gyr.  Fig.~3f) shows that the projected

543: distance at which a satellite is found at the present day is also a strong function

544: of the redshift at which the satellite first entered the halo; on average,

545: satellites

546: at $r_p < 50$~kpc entered their hosts' halos $\sim 3.5$~Gyr earlier than satellites

547: at $r_p \sim 400$~kpc.  Fig.~3e) shows the ratio of the satellite to host stellar

548: mass as a function of the redshift at which the satellites first entered their

549: hosts' halos.  The majority of host--satellite pairs (84\%) have

550: mass ratios $\le 0.15$, and in the case of these pairs there is a monotonic

551: trend of mass ratio with $z_{\rm entry}$: the smaller is the mass ratio, the

552: more recently the satellite entered its hosts' halo.  In the case of the small

553: percentage of host--satellite pairs with mass ratios $> 0.15$, the trend

554: is reversed: the larger is the mass ratio, the more recently the satellite

555: entered its hosts' halo.

556: 

557: \section{Satellite Galaxy Locations: Analysis and Results}

558: 

559: The location of a satellite galaxy with respect to its host is

560: computed by measuring the angle, $\phi$, between the major axis of

561: the host and the direction vector on the sky that connects the

562: centroid of the satellite to the centroid of its host. 

563: Throughout we will refer to the angle $\phi$ as the ``location''

564: of the satellite.  Because we

565: are simply interested in investigating any preferential alignment

566: of the satellite locations with the semi--major axes of the hosts,

567: $\phi$ is restricted to the range [0$^\circ$, 90$^\circ$].

568: By definition, a value of $\phi = 0^{\circ}$ indicates alignment

569: with the host major axis, while a value of $\phi = 90^{\circ}$

570: indicates alignment with the host minor axis. 

571: 

572: Fig.~4 shows the probability distribution for the locations of the

573: satellite

574: galaxies in the SDSS (left panels) and the MRS (right panels) that were

575: selected using the redshift space proximity criteria from \S3.1.  In

576: this figure we have computed $\phi$ for all satellites and we have made

577: no subdivisions of the data based on host properties, satellite 

578: properties, or the projected distances at which the satellites are 

579: found.  The top

580: panels of Fig.~4 show the differential probability distributions, $P(\phi)$,

581: where the error bars have been computed from 1000 bootstrap resamplings

582: of the data.  Also shown in the top panels of Fig.~4 is the mean

583: satellite location, $\left< \phi \right> $, along with

584: the confidence levels at which

585: the $\chi^2$ test rejects uniform distributions for $P(\phi)$.

586: The bottom panels of Fig.~4

587: show the cumulative probability distributions for the satellite locations, 

588: $P(\phi \le \phi_{\rm max})$, along with the confidence levels at which the

589: Kolmogorov--Smirnov (KS) test rejects uniform distributions for 

590: $P(\phi \le \phi_{\rm max})$.  It is clear from Fig.~4 that the satellites

591: in both the SDSS and the MRS are located preferentially near the major

592: axes of their hosts, and the effect is detected with very high significance.

593: However, the tendency for satellites to be found near the major axes

594: of their hosts is stronger in the MRS than it is in the SDSS.

595: It is likely that

596: this discrepancy is due to the rather idealized way in which the

597: MRS host galaxies have been placed within their halos, and may point

598: to a modest misalignment between mass and light in the host galaxies

599: (e.g., AB06, Kang07, Bailin et al.\ 2005).

600: 

601: \subsection{Dependence of the Anisotropy on Host \& Satellite Properties}

602: 

603: In this subsection we explore ways in which the locations

604: of satellite galaxies may depend upon various physical properties of the

605: hosts and satellites. 

606: Fig.~5 shows results for the dependence of satellite location on 

607: various properties of the hosts.  Results for the SDSS satellites are

608: shown in the left panels of Fig.~5

609: and results for the MRS satellites are shown in the right panels.

610: The top panels of Fig.~5 show

611: the mean satellite location, $\left< \phi \right>$, as a function

612: of the host's $(g-r)$ color, computed at $z=0$.  In the case of the SDSS satellites,

613: $\left< \phi \right>$ is a strong function of host color, with the satellites

614: of the reddest MRS hosts exhibiting a large degree of anisotropy, while the satellites of

615: the bluest SDSS hosts are 

616: consistent with being distributed isotropically around their hosts.

617: In the case of the MRS satellites,

618: the satellites of red hosts are also

619: distributed much more anisotropically than are

620: the satellites of blue hosts. However, there is

621: also a clear anisotropy present in the locations of the satellites of the bluest

622: MRS hosts that is not seen for the satellites of the bluest SDSS hosts.

623: 

624: The middle panels of Fig.~5 show the dependence of $\left< \phi \right>$

625: on the specific star formation rate (SSFR) of the host.  

626: Here is it clear that in both the SDSS

627: and the MRS, the mean satellite location is a strong function of the

628: SSFR; the lower the SSFR, the more anisotropically distributed are the 

629: satellites. 

630: The bottom panels of Fig.~5 show the dependence of $\left< \phi \right>$

631: on the stellar mass of the host.  

632: From these panels, then, the mean locations of the satellites in 

633: both the SDSS and the MRS are functions of the stellar mass of the host;

634: the greater is the mass of the host, the more anisotropic are the locations

635: of the satellites. Overall,

636: the dependence of the mean satellite location, $\left< \phi

637: \right>$, on host color, SSFR, and stellar mass agrees fairly well between the

638: SDSS and MRS satellites.  While the precise values of $\left< \phi \right> $

639: are not identical in the two samples, a general trend is clear in both cases.

640: The satellites of hosts that are red, massive, and have low SSFR are distributed

641: much more anisotropically than are the satellites of hosts that are blue,

642: low mass, and have high SSFR.

643: 

644: Our results in Fig.~5e and Fig.~5f are somewhat at odds with the results of

645: Yang06 and Kang07 (i.e., we find that the

646: locations of the satellites are a function of the stellar mass of the host).  Yang06

647: found a weak tendency for the anisotropy in the locations of the

648: satellites of primary galaxies in SDSS group systems to

649: increase with the mass of the halos.  In particular, Yang06 found

650: that the mean location of the satellites of primaries with halo masses in the

651: range $1.4\times 10^{12}~ M_\odot \le M \le 1.4\times 10^{13}~ M_\odot$

652: was $\left< \phi \right> = 43.1^\circ\pm 0.4^\circ$ while the mean location

653: of the satellites of primaries with halo masses in the range

654: $1.4\times 10^{14}~ M_\odot \le M \le 1.4\times 10^{15}~ M_\odot$

655: was $\left< \phi \right> = 40.7^\circ\pm 0.5^\circ$.  That is, an increase in

656: the masses of the halos by a factor of $\sim 100$ resulted in a

657: decrease in $\left< \phi \right>$ of $2.4^\circ \pm 0.6^\circ$.  We 

658: find $\left< \phi \right> = 44.4^\circ \pm 0.6^\circ$ for the satellites of

659: relatively isolated SDSS hosts with $M_{\rm stellar} \sim 3\times 10^{10}~ M_\odot$

660: and $\left< \phi \right> = 41.3^\circ \pm 0.6^\circ$ for the satellites of

661: relatively isolated SDSS hosts with $M_{\rm stellar} \sim 3\times 10^{11}~ M_\odot$;

662: i.e., we see a decrease in $\left< \phi \right>$ of $3.1^\circ \pm 0.8^\circ$.

663: We do not know the masses of the halos of our SDSS hosts, but from Fig.~3a 

664: (i.e., the correlation of $M_{\rm stellar}$ with $M_{\rm virial}$ for 

665: the MRS hosts) we expect that this stellar mass range for our SDSS 

666: hosts corresponds to a 

667: factor of $\sim 30$ in halo mass.   Therefore, we see a similar decrease

668: in the value of $\left< \phi \right>$ in only $\sim 1$ order of magnitude in

669: mass for our sample as Yang06 saw in $\sim 2$ orders of magnitude in mass

670: for their sample.

671: Based on a simple extrapolation of our

672: results for the satellites of relatively isolated

673: SDSS hosts, we might therefore have expected the

674: satellites in the study of Yang06 to show a greater difference (by factor

675: of $\sim 3$ to 4) in the dependence of their locations on halo mass. 

676: 

677: In their simulation, Kang07 found no dependence of the satellite locations

678: on the masses of the halos that surrounded the primaries, and they explain that

679: this is due to the fact that the greater flattening of the higher mass halos

680: is counterbalanced by the satellites of lower mass halos having locations that

681: are somewhat flatter than the mass of the surrounding halo.

682: In an attempt to understand

683: the discrepancy between our results and those of

684: Kang07, we expand upon our result

685: for the dependence of the satellite locations on host mass in Fig.~6, where

686: we investigate the effects of the host image assignment prescription on

687: $\left< \phi \right>$ for the MRS galaxies.  The left panels of Fig.~6 show

688: results for MRS hosts that are classified as elliptical.  These are objects

689: for which the luminous galaxy is assumed to share the shape of the surrounding

690: dark matter halo.  The right panels of Fig.~6 show results for MRS hosts

691: that are classified as non--elliptical.  These are objects for which the

692: luminous galaxy is assumed to be a thin disk, oriented such that the angular

693: momentum of the disk aligns with the net angular momentum of the surrounding

694: halo.  The top panels of Fig.~6 show $\left< \phi \right>$ as a function of

695: host color.  From these panels, it is clear that the satellite anisotropy is

696: stronger for the very reddest elliptical MRS hosts than it is for the bluest

697: elliptical MRS hosts, however there is essentially no dependence on host color

698: for the locations of the satellites of non--elliptical MRS hosts.  It is also

699: clear that the satellites of the elliptical MRS hosts show a much greater

700: degree of anisotropy in their locations compared to the satellites of

701: non--elliptical MRS hosts.  This is due to the fact that strict alignment of

702: mass and light in the numerical galaxies, as was done for the elliptical 

703: MRS hosts, maximizes the anisotropy of the

704: satellite locations (see AB06 and Kang07).

705: 

706: The bottom panels of Fig.~6 show the dependence of the satellite locations

707: on the stellar masses of the MRS hosts.  From these panels

708: it is clear that, at fixed host mass, the satellites of the elliptical

709: MRS hosts show a greater degree of anisotropy in their locations than do the

710: satellites of non--elliptical MRS hosts.  In addition,

711: {\it within a given

712: class of MRS host galaxy} there is no clear trend of $\left< \phi \right>$ with

713: the stellar mass of the host.  That is, the trend with host stellar mass that

714: we see in panel f) of Fig.~5 is due to the fact that the lowest mass MRS hosts

715: are non--ellipticals (whose satellites show a relatively small degree of

716: anisotropy in their locations) while the highest mass MRS hosts are ellipticals

717: (whose satellites show a much greater anisotropy in their locations).  The fact

718: that, within a particular image assignment prescription for the MRS hosts, 

719: we see no dependence of

720: $\left< \phi \right>$ on host mass probably explains why Kang07 did not see

721: a strong dependence of the satellite anisotropy on the masses of the central

722: galaxies in their simulation.  Kang07 did not assign galaxy types to

723: their central galaxies, and they used the same prescription to assign image shapes

724: to all the luminous galaxies in their simulation.  

725: 

726: In Fig.~7 we demonstrate the effect on $\left< \phi \right>$

727: if we use the same image assignment 

728: scheme for all of the MRS hosts.  That is, Fig.~7 shows how $\left< \phi \right>$ is

729: affected if we do not adjust our image assignment scheme according to whether or

730: not the MRS host galaxy is an ``elliptical'' or a ``non--elliptical''.  Open triangles 

731: in Fig.~7 show the dependence of $\left< \phi \right>$ on host color (left panel),

732: SSFR (middle panel), and stellar mass (right panel) under the assumption that all

733: MRS host galaxies share the shapes of their dark matter halos.  That is, the open triangles

734: in this figure show the resulting values of $\left< \phi \right>$ if we simply apply

735: the ``elliptical'' image assignment scheme to all MRS hosts.  Open circles in

736: Fig.~7 show the result of simply applying the ``non--elliptical'' image assignment

737: scheme to all MRS hosts.  That is, the open circles show the result that occurs if

738: all MRS hosts are assumed to be thin disks, oriented such that the angular momentum

739: of the disk is perfectly aligned with the net angular momentum of the halo.  For

740: comparison, solid squares show the results from Fig.~5 for the SDSS hosts and

741: satellites.  From Fig.~7, then,

742: if we adopt the same image assignment scheme for all MRS hosts, independent

743: of their bulge--to--disk ratios, we cannot 

744: reproduce the observed dependence of $\left< \phi \right>$ on host color, SSFR,

745: and stellar mass that we find for the SDSS galaxies.

746: If we use a single image assignment scheme for all MRS hosts, 

747: $\left< \phi \right>$ for the MRS satellites

748: generally has a much weaker dependence on host color, SSFR, and stellar mass than we see

749: in the SDSS, and sometimes the dependence of $\left< \phi \right>$ on host property

750: is actually opposite to what we see in the SDSS.  Fig.~7 then argues rather strongly

751: for the need for two distinct image assignment schemes as we have adopted for the

752: elliptical and non--elliptical MRS hosts.  It also suggests that luminous elliptical 

753: galaxies and luminous spiral galaxies in the observed Universe are oriented within

754: their dark matter halos in rather different ways.

755: 

756: In Figs.~8 through 10 we expand upon our results in Fig.~5 for the

757: dependence of the satellite locations on host color, 

758: and we do this by splitting

759: our sample into ``red'' hosts and ``blue'' hosts.

760: To define ``red'' and ``blue'', we fit

761: the distributions of $(g-r)$ host colors in the top panels of Fig.~5

762: by the sum of two Gaussians (e.g., Strateva et al.\ 2001; Weinmann et al.\

763: 2006).   We find that the division between

764: the two Gaussians lies at $(g-r) = 0.7$ for the SDSS galaxies and

765: at $(g-r) = 0.75$ for the

766: MRS galaxies.  We therefore define SDSS

767: hosts with $(g-r) < 0.7$ to be ``blue'' and SDSS hosts with $(g-r) \ge

768: 0.7$ to be ``red''.  Similarly, we define MRS hosts with $(g-r) < 0.75$ to

769: be ``blue'' and MRS hosts with $(g-r) \ge 0.75$ to be ``red''.

770: Figs.~8 and 9 then show $P(\phi)$ and $P(\phi \le

771: \phi_{\rm max})$ for satellites of the red and blue hosts, respectively.

772: 

773: It is clear from Figs.~8 and 9 that the satellites of red hosts

774: have a much stronger preference for being located near the major axes

775: of their hosts than do the satellites of blue hosts.  This is true for 

776: both the SDSS and MRS satellites.  In addition,

777: the MRS satellites show a stronger

778: preference for being located near the major axes of their hosts than do the

779: SDSS satellites.  The satellites of blue SDSS hosts

780: are consistent with having an isotropic distribution around their hosts, while

781: the satellites of red SDSS hosts have a strong preference for being located

782: near the major axes of their hosts.

783: Such a disparity in the locations of the satellites of red and blue host

784: galaxies was also found by APPZ, Kang07, Yang06, Bailin et al.\ (2008), 

785: and Siverd et al.\ (2009),

786: with the satellites of blue hosts showing little to no preference

787: for a particular location relative to their hosts.

788: 

789: In the case of APPZ, small number statistics (i.e., a relatively 

790: small number of host--satellite pairs in these studies) 

791: prevented them from placing a strong

792: constraint on whether or not the locations of the satellites of blue hosts

793: were, in fact,  truly different from the locations of

794: the satellites of the red hosts.   The cause of this is

795: two--fold.  First, 

796: the majority of SDSS hosts are red (see Table~1).  

797: Second, the blue hosts

798: tend to have fewer satellites

799: than do their red counterparts.  This results in a paucity of host-satellite pairs in 

800: which the host is blue.  Here, however,

801: our sample of SDSS hosts

802: and satellites is sufficiently large that we can make a definitive 

803: statement about the locations of the satellites of blue hosts versus

804: the locations of the satellites of red hosts.  

805: To do this, we computed a two--sample KS test using the 

806: cumulative probability distributions from the bottom left panels of Figs.~8

807: and 9.  The result is that, at the 99.9\% confidence level,

808: the KS test rejects the null hypothesis that

809: the locations of the satellites of red SDSS hosts are drawn from the same

810: distribution as the locations of the satellites of blue SDSS hosts.  That

811: is, with high significance, the locations of the satellites of red and blue

812: SDSS hosts are truly different.

813: 

814: Fig.~10 illustrates

815: the underlying cause of the ``lack'' of anisotropy in the locations of

816: the satellites of the blue SDSS hosts.  Here we plot the

817: mean satellite location, $\left< \phi \right>$, as a function of 

818: projected distance.  The left panels of Fig.~10 show the results for the

819: satellites of red hosts, while the right panels show the results for the

820: satellites of blue hosts.  In the case of the satellites of red hosts, $\left<

821: \phi \right>$ is largely independent of $r_p$.  Hence, when we average the

822: satellite locations over all projected distances, $r_p \le 500$~kpc (i.e.,

823: as in Figs.~5, 8 and 9), the result is that the satellites of red hosts exhibit

824: a strong degree of anisotropy.  In the case of the satellites of blue hosts,

825: however, $\left< \phi \right>$ is a function of $r_p$.  Satellites of

826: blue hosts that are located at small projected distances have a tendency

827: to be found close to the

828: major axes of their hosts, while satellites of blue hosts with larger projected 

829: distances exhibit a different degree of anisotropy.  In particular,

830: satellites of blue SDSS hosts that have large values of $r_p$  have a tendency

831: to be found close to the {\it minor} axes of their hosts, and when

832: the locations of all satellites of the blue SDSS hosts are averaged

833: over all projected distances, $r_p \le 500$~kpc, the result is consistent with

834: an isotropic distribution (i.e., top left panel of Fig.~9).  The satellites

835: of blue MRS hosts show a preference for being located close to the major axes

836: of their hosts for projected distances $r_p < 300$~kpc, but at larger 

837: projected distances the satellite locations become consistent with a random

838: distribution.  Therefore,

839: the net anisotropy of the MRS satellites of blue hosts

840: is substantially reduced when averaged over all

841: values of $r_p \le 500$~kpc (i.e., top right panel of Fig.~9).

842: 

843: Fig.~11 shows the dependence of the mean satellite location as

844: a function of various properties of the satellites.

845: Panels a) and b) show the dependence of $\left< \phi \right>$ on

846: $(g-r)$, panels c) and d) show the dependence of $\left< \phi 

847: \right>$ on specific star formation rate, panels e) and f) show the

848: dependence of $\left< \phi \right>$ on the stellar mass, and

849: panels g) and h) show the dependence of $\left< \phi \right>$ on the 

850: projected distances at which the satellites are found.  

851: As in Fig.~5, there is generally

852: good agreement between the results for SDSS satellites (left panels) and

853: MRS satellites (right panels), 

854: with the greatest degree of anisotropy being shown

855: by the reddest, most massive, and lowest--SSFR satellites.  The 

856: locations of the bluest, least massive, and highest--SSFR satellites show

857: little to no anisotropy.  This is in part attributable to the fact that

858: these objects are likely to have been accreted in the very

859: recent past (see, e.g.,

860: Fig.~3); however, as we will see in the next section this is also partially

861: attributable to the fact that our blue satellite population is heavily

862: contaminated with interlopers whose effect is to strongly suppress the anisotropy.

863: 

864: Finally, we note

865: that the locations of the satellites are weakly--dependent upon the

866: projected distances at which they are found (panels g and h of Fig.~11), with

867: the satellites found at $r_p \sim 450$~kpc showing less anisotropy than

868: satellites found at smaller projected distances.  This is, of course,

869: unsurprising since the objects that are found

870: at large $r_p$ are most likely to be either genuine satellites that have been

871: accreted very recently (see, e.g., panel f of Fig.~3) or 

872: interlopers.  In addition, we note that, contrary to the claims

873: of Bailin et al.\ (2008) very few of our SDSS satellites are found at

874: projected distances $r_p < 50$~kpc (see the histogram in 

875: Fig.~5g).  The lack of SDSS satellites

876: at small projected distances is caused primarily by the fact

877: that fiber collisions prevent the simultaneous measurement of the redshifts of

878: two galaxies that are very close to each other on the sky.  So, it is only in regions

879: of the sky that were observed multiple times that satellites with small values

880: of $r_p$ may be found.  Also, because we have performed a visual check

881: of each and every host galaxy, we know for certain that the satellites that

882: we do identify at $r_p < 50$~kpc are, indeed, separate from their host.  That is,

883: the satellites at these projected distances are not, say, H-II regions or

884: bright blue knots within the host galaxy

885: that have been misidentified as objects that are distinct from the host galaxy.

886: 

887: \subsection{Effects

888: of Interlopers and $z_{\rm entry}$}

889: 

890: When discussing the satellites, it is important to remember that 

891: at least some fraction

892: of the satellites that are found using the selection criteria in \S3.1 are not

893: genuine satellites at all.  Rather, they are interlopers that are not

894: necessarily nearby to a host galaxy, but they happen to pass all of

895: the proximity and magnitude criteria in order 

896: to be included as satellites in the

897: catalog.  In the case of the SDSS satellites, we have no way of knowing which

898: of the satellites in our catalog are real and which are interlopers.  In the

899: case of the MRS satellites, however, we have full phase--space information

900: and we know the physical distances of each of the satellites in the 

901: catalog from their respective hosts.

902: Until now, all of our

903: calculations of the locations of satellite galaxies in the MRS have included

904: both the satellites that are physically close to host galaxies, as well

905: as the interlopers.  This was done in order to

906: better compare the MRS to the SDSS via identical procedures for the 

907: identification of hosts and satellites. In this section 

908: we will examine the effects of the interlopers on the observed

909: anisotropic distribution of the satellites, as well as the effect of

910: the redshift at which the satellites first entered their hosts' halos.

911: 

912: Here we adopt the same

913: rather non--restrictive definition of a genuine satellite as in \S3.3 and we

914: accept as genuine

915: satellites those objects that are located within a physical distance

916: $r_{3D} \le 500$~kpc of a host galaxy.  The mean location of all MRS satellites

917: that are found within $r_{3D} \le 500$~kpc of a host galaxy is

918: $\left< \phi \right> = 39.12^\circ \pm 0.08^\circ$, while the mean location

919: of the interlopers is $\left< \phi \right> = 43.6^\circ \pm 0.1^\circ$.  Clearly,

920: then, the presence of the interlopers in the full data set reduces the

921: measured anisotropy in the satellite locations compared to what one would

922: measure in the absence of the interlopers.  Interestingly, the interlopers are

923: not randomly--distributed around the hosts.  Instead, on average the interlopers

924: show a weak preference for being located near the major axes of the hosts.

925: This is due to the fact that relatively

926: few interlopers are located at extremely large distances from the host galaxies.

927: The median distance of the interlopers from the hosts is only 630~kpc, indicating

928: that by and large they are within the local vicinity of the hosts.  

929: 

930: Shown in Fig.~12 are the results for the differential probability 

931: distribution, $P(\phi)$, for MRS satellites, with and without the 

932: contribution of interlopers.  

933: The open points in Fig.~12 show

934: $P(\phi)$, computed using all satellites in the MRS catalog, including

935: the interlopers.  

936: The filled points show $P(\phi)$, computed using only

937: the satellites in the MRS catalog that are located within

938: a physical distance $r_{3D} \le 500$~kpc of their host.  

939: Included in 

940: each of the panels of Fig.~12 is the value of the mean satellite location,

941: with and without the contribution of interlopers, along with the

942: fraction of satellites in the MRS catalog that are interlopers 

943: (i.e., objects which have $r_{3D} > 500$~kpc).

944: As above, the net effect of interlopers is to reduce

945: the value of $\left< \phi \right>$.

946: The top panels of Fig.~12 show $\left< \phi \right>$ for the satellites

947: of red MRS hosts (left panel) and the satellites of blue MRS hosts

948: (right panel).  The fraction of interlopers 

949: is nearly identical;

950: interlopers account for 32\% of the satellites of red MRS hosts

951: and 35\% of the satellites of blue MRS hosts.

952: The presence of the interlopers reduces $\left< \phi \right>$ by similar

953: amounts for the satellites of both the red and blue MRS hosts.  

954: 

955: We note

956: that the presence of interlopers is not the cause of the reduced anisotropy 

957: for the satellites of the blue hosts compared to the satellites of the

958: red hosts.  That is, the removal of the interlopers from the MRS sample

959: does not result in the locations of the satellites of blue MRS hosts being

960: the same as those of red MRS hosts.  Formally, when the interlopers

961: are removed, the mean location of the

962: MRS satellites surrounding blue hosts differs from the mean location of the

963: MRS satellites surrounding red hosts by more than $20\sigma$.  This 

964: differs from the conclusions

965: of Kang07 who found that removing the interlopers from their sample resulted

966: in the locations of the satellites of blue central galaxies being the

967: same as the locations of the satellites of red central galaxies.

968: However, as with the dependence of satellite anisotropy on

969: host mass, this difference may be simply attributable to the two

970: different prescriptions that we

971: have used to assign images to the luminous MRS host galaxies.  That is,

972: on average, the red MRS hosts are ellipticals and the blue MRS hosts

973: are non--ellipticals.  From Fig.~6, then, we would automatically expect 

974: the satellites of red MRS hosts to show a greater degree of anisotropy

975: in their locations than the satellites of blue MRS hosts because of the

976: strong correlation of the satellite anisotropy with the 

977: host image assignment scheme (i.e., our ``elliptical'' image assignment

978: scheme maximizes the satellite anisotropy).

979: 

980: The bottom panels of Fig.~12 show $\left< \phi \right>$ for red

981: MRS satellites (left panel) and blue MRS satellites (right panel).

982: Here the interloper fraction is strikingly different; only 19\% 

983: of the red MRS satellites are interlopers,

984: while 57\% of the blue MRS satellites are interlopers.

985: Therefore, the presence of a large number of interlopers in the sample of

986: blue satellites is a major factor in the reduced anisotropy of blue 

987: satellites compared to red satellites (e.g., panels a) and b) of Fig.~11).

988: 

989: As noted by Kang07, the redshift at which a genuine satellite

990: first enters the halo of its host is a strong function of the mass of

991: the satellite and the present--day color of the satellite.  From panels c)

992: and d) of Fig.~3, the more massive

993: the satellite and the redder

994: is its present--day $(g-r)$ color, the earlier the satellite

995: made its first entry into

996: the halo of its host (see also Kang07).  

997: One would naturally expect that it would take a few crossing

998: times for satellites to have their trajectories affected to the point

999: where the locations of the satellites would provide a good

1000: proxy for the distribution of the mass with the host's halo.  For a CDM

1001: halo with a mass of $\sim 10^{12} M_\odot$ and virial 

1002: radius $\sim 180h^{-1}$~kpc,

1003: the crossing time will be of order $\tau_{\rm cross} \simeq

1004: R/v \simeq 1.7$~Gyr for $v \sim 150$~km~sec$^{-1}$.  Therefore, unless 

1005: the infall of satellites is highly non--spherical, we would expect

1006: satellites that arrived within their host's halo within the

1007: past billion years should show markedly less anisotropy than satellites

1008: that arrived within their host's halo in the much more distant past.

1009: 

1010: Solid squares in the top panel of Fig.~13 show the mean satellite location, $\left<

1011: \phi \right>$,  as a function of the redshift at which the genuine MRS 

1012: satellites first entered their hosts' halos.  From this figure, 

1013: satellites that first entered their host's halo within the past $\sim 1.25$~Gyr

1014: (i.e., $z_{\rm entry} \sim 0.1$) show

1015: considerably less anisotropy than do those which first entered their host's

1016: halo at earlier times. 

1017: Referring to the bottom left panel of Fig.~3, the bluest

1018: MRS satellites are those which

1019: first entered their host's halo at redshifts $z_{\rm entry} \sim 0.1$, while the

1020: reddest MRS satellites are those which first entered their

1021: host's halo at redshifts $z_{\rm entry} > 2$.  Therefore,

1022: it is unsurprising that, after the removal of interlopers with 

1023: $r_{3D} > 500$~kpc,

1024: the degree of anisotropy exhibited by the 

1025: blue MRS satellites (bottom right panel of Fig.~12, $\left< \phi \right> =

1026: 41.9^\circ \pm 0.2^\circ$) is considerably

1027: less than the degree of anisotropy exhibited by the red genuine MRS

1028: satellites (bottom left panel of Fig.~12, $\left< \phi \right> = 

1029: 38.2^\circ \pm 0.1^\circ$).  

1030: Also shown in the top panel

1031: of Fig.~13 is the mean satellite location, $\left< \phi \right>$,

1032: as a function of $z_{\rm entry}$ 

1033: for the genuine satellites of red MRS hosts (open triangles) and 

1034: the the genuine satellites of blue MRS hosts (open circles).  From this

1035: figure, then, it is clear that satellites began arriving within the halos

1036: of the red MRS hosts much earlier than did the satellites of blue MRS hosts.

1037: The bottom panel of Fig.~13 shows the probability of the entry redshift,

1038: $P(z_{\rm entry})$, for the type 1 and type 2 MRS satellites.

1039: The type 2 satellites are the objects

1040: that have been stripped of their dark matter and, as expected, Fig.~13

1041: shows that $z_{\rm entry}$ is, on average, considerably earlier for the type~2

1042: satellites than it is for the type 1 satellites (which still retain

1043: their dark matter).

1044: 

1045: \section{Summary and Comparison to Previous Results}

1046: 

1047: Here we summarize the major results of our study and compare them to 

1048: results of previous, similar investigations.  The major results

1049: that we have obtained by computing the mean satellite location,

1050: $\left< \phi \right>$, using all satellites (including interlopers) are: 

1051: 

1052: \begin{itemize}

1053: \item[1.] $\left< \phi \right>$ is a function of the host color,

1054: specific star formation rate, and stellar mass.  Satellites of red, massive

1055: hosts with low SSFR show considerably more anisotropy than do satellites of

1056: blue, low mass hosts with high SSFR

1057: (Fig.~5).

1058: 

1059: \item[2.] In order to reproduce the observed trends for the dependence of

1060: $\left< \phi \right>$ on host color, SSFR, and stellar mass,  we require two

1061: distinct image assignment prescriptions for the simulated galaxies: ellipticals

1062: share the shapes of their dark matter halos and non--ellipticals have their

1063: angular momentum vectors aligned with the net angular momentum of the halo.

1064: (Fig.~7)

1065: 

1066: \item[3.] $\left< \phi \right>$ is a function of the satellite color,

1067: specific star formation rate, and stellar mass.  Red, massive satellites with

1068: low SSFR show considerably more anisotropy than do blue, low mass satellites

1069: with high SSFR (Fig.~11).

1070: 

1071: \item[4.] Averaged over all satellites at all

1072: projected distances, the locations of the satellites of blue SDSS host galaxies

1073: are consistent with

1074: an isotropic distribution, while the satellites of red

1075: SDSS host galaxies have a strong preference for being found near the major

1076: axes of their hosts.  At the 99.9\% confidence level, the two 

1077: distributions are inconsistent with having

1078: been drawn from the same parent distribution (Figs.~8 and 9).

1079: 

1080: \item[5.] Satellites of blue MRS host galaxies are found preferentially

1081: close to the major axes of their hosts, however the degree of anisotropy

1082: is considerably less than that shown by the satellites of red MRS

1083: host galaxies (Figs.~8 and 9).

1084: 

1085: \item[6.] $\left< \phi \right>$ for the satellites

1086: of red host galaxies is approximately independent of $r_p$, while

1087: $\left< \phi \right>$ for the satellites of blue host galaxies is an

1088: increasing function of $r_p$  (Fig.~10).

1089: 

1090: 

1091: \end{itemize}

1092: 

1093: \noindent

1094: The major results that we have obtained with regards to interlopers

1095: are:

1096: 

1097: \begin{itemize}

1098: \item[7.]

1099: The interloper contamination is similar (32\% and 35\%, respectively) for the satellites

1100: of red MRS hosts and blue MRS hosts (Fig.~12, top panels).

1101: 

1102: \item[8.] Interlopers are not the cause of the different amount of anisotropy

1103: shown by the locations of the satellites of blue MRS hosts versus 

1104: the satellites of red MRS hosts.  

1105: The genuine satellites

1106: of red MRS hosts show considerably more anisotropy than do the genuine

1107: satellites of blue MRS hosts, and the significance is 

1108: greater than $20\sigma$ (Fig.~12, top panels). 

1109: 

1110: \item[9.] Our host--satellite selection criteria result in

1111: 57\% of the blue satellites in the MRS catalog being interlopers and

1112: 19\% of the red satellites being interlopers

1113: (Fig.~12, bottom panels).

1114: 

1115: \item[10.] At the 16$\sigma$ level, the red genuine MRS satellites show

1116: considerably more anisotropy in their locations than do the blue genuine

1117: MRS satellites  (Fig.~12, bottom panels).  This is due to the fact that

1118: the blue satellites have only recently arrived within their hosts' halos, 

1119: while the red satellites arrived in the far distant past.

1120: 

1121: \end{itemize}

1122: 

1123: As mentioned above, the general trend for the satellites of red hosts

1124: to show considerably more anisotropy than those of blue hosts has

1125: been observed by others (e.g., APPZ; Yang06; Kang07; Bailin et al.\ 2008;

1126: Siverd et al.\ 2009),

1127: and our results agree well with these previous results.  Further, 

1128: we have demonstrated conclusively that in the case of relatively isolated 

1129: host--satellite systems, the satellites of blue host galaxies are distributed

1130: differently around their hosts than are the satellites of red host galaxies.

1131: 

1132: Also as mentioned above, although our results for the satellites of  SDSS 

1133: host galaxies show trends that are very similar to our results for the satellites

1134: of MRS host galaxies,  the satellites of MRS host galaxies exhibit a greater

1135: degree of anisotropy in their locations.  This is probably attributable

1136: to the simple prescriptions that we have used to define the images 

1137: of the MRS host galaxies, and may indicate that a certain degree of

1138: misalignment of the galaxy images from our idealized prescriptions

1139: is necessary

1140: (see also AB06; Kang07; Okumura et al.\ 2009;

1141: Faltenbacher et al.\ 2009; Okumura \& Jing 2009). 

1142: To estimate the degree of misalignment that is necessary

1143: for the anisotropy of the locations of the

1144: satellites of the MRS galaxies to match those of the SDSS galaxies, we add

1145: Gaussian--random errors to the orientations of the MRS host galaxy images (as

1146: viewed in projection on the sky).  When we do this, we find that a mean

1147: misalignment of $|\delta \theta| \sim 20^\circ$ (measured relative to the

1148: ``idealized'' MRS host image)  reduces the anisotropy in

1149: the locations of the satellites of the MRS hosts to the point that,

1150: when averaged over $r_p \le 500$~kpc, the result agrees with the result for

1151: the satellites of SDSS hosts.   We note that, although we have phrased

1152: this in terms of a misalignment of the host galaxy image from the idealized

1153: prescription, this should not be strictly interpreted as the mass and light of the

1154: SDSS galaxies being misaligned by an average of

1155: $\sim 20^\circ$.  While there may be some

1156: degree of true misalignment, it is always important to keep in mind that 

1157: there are observational errors associated with the measurement of the 

1158: position angles of observed galaxies, and these can be particularly large

1159: in the case of very round galaxies, or galaxies with well--resolved spiral

1160: arms.  Such errors in the determination of the position

1161: angles of the SDSS galaxies will, therefore, contribute some amount

1162: to a need for 

1163: misalignment of the host images in the MRS in order to match

1164: the observations.  Unfortunately, errors for the position angles of the

1165: SDSS galaxies are not yet available in the data base, so we are unable to

1166: estimate the contribution of position angle errors to the value of

1167: $|\delta \theta |$ above.

1168: 

1169: Although our work is very similar in spirit to that of Kang07, we arrive

1170: at some different conclusions.  First, we find that the degree

1171: of anisotropy in the satellite locations depends upon the

1172: stellar mass of the host galaxy, while Kang07 found no dependence of

1173: the satellite locations on the mass of the surrounding halo.

1174: The discrepancy between our theoretical results and the

1175: theoretical results of Kang07 is probably due to the fact that

1176: we have chosen to use two different image assignment schemes for the MRS

1177: hosts (ellipticals vs.\ non--ellipticals), while Kang07 use the same

1178: image assignment scheme for all of their central galaxies.  We find that

1179: {\it within a given image assignment scheme} there is no dependence of 

1180: $\left< \phi \right>$ on host mass; however, there is considerably more

1181: anisotropy shown by the satellites of elliptical MRS hosts than non--elliptical

1182: MRS hosts.  This, combined with the fact that the least massive MRS hosts

1183: are non--ellipticals and the most massive MRS hosts are ellipticals leads

1184: to the trend of satellite anisotropy with host mass that we see in the simulation.

1185: 

1186: In their study of the locations of satellites in SDSS group systems, Yang06 found

1187: a rather weak dependence of satellite location on the mass of the

1188: surrounding halo; over two orders of magnitude in halo mass, the value of

1189: $\left< \phi \right>$ decreased by only $2.4^\circ \pm 0.6^\circ$.  

1190: By contrast, we appear to find a somewhat stronger trend of 

1191: satellite location with host mass.  Over $\sim 1$~order of

1192: magnitude in host mass we find a decrease in the value of $\left< \phi \right>$

1193: that is similar to the value found by Yang06: $3.1^\circ \pm 0.8^\circ$.

1194: A simple extrapolation of our results to much higher masses would suggest that

1195: over the mass range of their sample, Yang06 should have found a greater

1196: change in $\left< \phi \right>$.

1197: The resolution of this discrepancy is unclear, but it could 

1198: have to do with the fact

1199: that we are investigating somewhat different systems (i.e., relatively isolated

1200: hosts vs.\ group environments, where perhaps the central galaxy is not located

1201: precisely at the dynamical center).   In addition, we use stellar masses to define the

1202: masses of our host galaxies while Yang06 derive masses for the halos of their

1203: groups using a conditional luminosity function.  This discrepancy

1204: certainly warrants

1205: further investigation in the future, particularly since $\Lambda$CDM 

1206: predicts that the flattening of the dark matter halos of galaxies should

1207: increase with halo virial mass (e.g.,  Warren et al.\ 1992; Jing \& Suto 2002;

1208: Bailin \& Steinmetz 2005; Kasun \& Evrard 2005; Allgood et al.\ 2006). 

1209: 

1210: Additionally, in their simulation

1211: Kang07 find that the reason the satellites of blue

1212: central galaxies show less anisotropy than the satellites of red central

1213: galaxies is that the presence of a large number of interlopers around the

1214: blue central galaxies 

1215: suppresses the anisotropy.  This is because Kang07 find

1216: that there is a considerably larger number of interlopers in the sample 

1217: of satellites around blue central galaxies ($\sim 35$\%) than there are

1218: in the sample of satellites around red central galaxies ($\sim 15$\%).  

1219: When Kang07 remove the interlopers, they find that the degree of

1220: anisotropy shown by the genuine satellites of red and blue centrals is 

1221: identical.  In our work we find a nearly identical interloper fraction 

1222: for the satellites of red and blue host galaxies (32\% for red hosts 

1223: and 35\% for blue hosts).   However, it is important to note that

1224: we have used a simple non--iterative

1225: technique to identify host and satellite galaxies,

1226: while Kang07 use a sophisticated,  iterative technique which is supposed

1227: to reduce the number of interlopers on average.  So, it is unsurprising

1228: that our relative number of interlopers would differ.

1229: 

1230: When we remove the interlopers from the MRS host and satellite catalog, we find

1231: that the satellites of blue hosts still show much less anisotropy

1232: than do the satellites of red hosts.  In our analysis, there appear

1233: to be two causes of the differences between the locations of the 

1234: satellites of red and blue hosts.  First, $\left< \phi \right>$ is

1235: largely independent of $r_p$ for the satellites of red hosts.  Therefore,

1236: when $\left< \phi \right>$ is averaged over all projected distances,

1237: $r_p \le 500$~kpc, the satellites of the red hosts show a great deal

1238: of anisotropy.  In contrast, $\left< \phi \right>$ for the satellites

1239: of blue hosts is a function of $r_p$, with satellites located at small

1240: $r_p$ being found near the major axes of their hosts and satellites

1241: located at larger distances having different locations (nearly isotropic

1242: in the case of the MRS satellites, and near the minor axes of the hosts

1243: in the case of the SDSS satellites).  Therefore, when $\left< \phi \right>$

1244: is averaged over all projected distances, $r_p \le 500$~kpc, the satellites

1245: of blue hosts show a markedly reduced anisotropy.  In addition, we know that

1246: the blue MRS hosts are by and

1247: large disk systems (``non--ellipticals'') and the satellites of the

1248: non--elliptical MRS hosts are distributed much less anisotropically than are the

1249: satellites of the elliptical MRS hosts due to our

1250: image assignment schemes.  Thus, as with the discrepancy

1251: regarding the trend of satellite anisotropy with host mass, the discrepancy

1252: between our results and those of Kang07 for the origin of the different

1253: amount of anisotropy shown by satellites of red and blue hosts may be due

1254: in large part to the two different assignment schemes that we have used to

1255: define the images of the MRS host galaxies.

1256: 

1257: Now, it is, of course, extremely important not to put too much significance on one

1258: data point, especially in the case of a figure in which the data points

1259: are inherently correlated.  Nevertheless, the value of $\left< \phi \right>$

1260: for the satellites of blue SDSS hosts that are located at $r_p \sim 400$~kpc is

1261: intriguing because it suggests a ``reversal'' of the anisotropy signal at

1262: large distances (right panel of Fig.~10).  

1263: In their sample of extremely isolated SDSS host galaxies

1264: (much more isolated than our sample), Bailin

1265: et al.\ (2008) found no statistically--significant

1266: dependence of $\left< \phi \right>$ on $r_p$; however, their sample size is

1267: much smaller than we have used here (337 hosts and 388 satellites).

1268: A weak tendency for the satellites of isolated disk galaxies

1269: to be aligned with the minor axes of the hosts was seen 

1270: by Zaritsky et al.\ (1997) when the satellite locations were averaged out

1271: to large projected distances ($r_p \sim 500$~kpc).  More recently,

1272: Siverd et al.\ (2009) found a weak tendency for extremely faint satellites

1273: of highly--inclined blue SDSS galaxies to have a minor axis preference when

1274: the locations of the satellites were averaged out to similarly large 

1275: projected distances.  This is tantalizing in light of the results

1276: of Zhang et al.\ (2009) who found that the spin axes of dark matter halos

1277: with mass $\lesssim 10^{13} M_\odot$ tend to be aligned along the filament

1278: in which the halo resides.   In addition, Bailin et al.\ (2008) found

1279: that satellites that are most likely to have been accreted recently

1280: have a tendency to be found along the same axis as the large--scale

1281: structure that surrounds the host galaxy.

1282: Thus, a ``reversal'' of the anisotropy for

1283: the locations of the satellites of disk host galaxies at large projected 

1284: distances could indicate preferential infall of satellites along filaments.

1285: Establishing the existence of

1286: such a reversal of the anisotropy at large projected

1287: distance will, of course, take a great deal more effort (see, e.g., Siverd

1288: et al. 2009 who conclude that the discrepancies between previous investigations

1289: are largely attributable to sample selection).

1290: 

1291: We have shown that satellites that are very blue, have low masses

1292: and high SSFR tend to show little to no anisotropy in their locations while

1293: satellites that are very red, have high masses and low SSFR show a great

1294: deal of anisotropy in their locations.  Similar results have been seen Yang06,

1295: Kang07, and Siverd et al.\ (2009).  Using their simulation, Kang07 interpret

1296: this effect to be due to the fact that the reddest, most massive

1297: satellites are those which entered their hosts' halos in the far distant past,

1298: while the bluest, least massive satellites have only recently arrived within

1299: the halo.  Our work with the MRS hosts and satellites

1300: directly supports this conclusion, however there is an additional component to

1301: the effect in our case.  The redshift space selection criteria that we have

1302: adopted result in the majority of blue satellites (57\%) being interlopers,

1303: the presence of which

1304: reduces the anisotropy exhibited by the genuine blue satellites by a substantial

1305: amount (a $7\sigma$ effect; see the bottom right panel of Fig.~12). 

1306: 

1307: It is, of course, a tremendous simplification to use the global

1308: dark matter halo properties to obtain properties of the luminous central galaxy

1309: as we have done here.  This is due to the fact that the scale size of

1310: the luminous galaxy is far smaller than that of the halo in which it resides.

1311: Therefore, it is not necessarily the case that the net halo shape or net halo

1312: momentum will be reflected in the shape or angular momentum of the central

1313: galaxy.  Given these caveats, it is really quite

1314: remarkable that such naive prescriptions as we have adopted here give 

1315: rise to a fair agreement between theory and observation.  If nothing

1316: else, our results lend credence to the idea that large luminous galaxies

1317: have some knowledge of the halo in which they reside, despite the fact that

1318: the luminous galaxy may be an order of magnitude smaller in extent than its

1319: dark matter halo.  While mass may not directly trace light within galaxies,

1320: it would not be possible to have such similar results for the locations of

1321: satellite galaxies in the observed Universe and $\Lambda$CDM if mass and 

1322: light were not strongly coupled within the host galaxies.

1323: 

1324: \section{Conclusions}

1325: 

1326: 

1327: Here we have shown that the locations of the

1328: satellites of relatively isolated host galaxies in the

1329: SDSS and the Millennium Run simulation (MRS) show very similar trends,

1330: provided that we adopt two distinct image assignment prescriptions for

1331: the MRS hosts: elliptical hosts share the shapes of their dark matter

1332: halos while non--elliptical hosts have their angular momentum vectors

1333: aligned with the net angular momentum of their halos.  If we use only

1334: a single image assignment prescription for all MRS hosts, it is not

1335: possible to reproduce the dependencies of the mean satellite location

1336: on host properties that we see in the SDSS.

1337: Averaged over all projected distances, $r_p$, the degree to

1338: which satellites are found preferentially close to the major axes of

1339: their hosts is a function of the host's stellar

1340: mass, SSFR, and $(g-r)$ color.  The satellites of red, massive hosts

1341: with low SSFR show a strong tendency for being located near the

1342: major axes of their hosts, while the satellites of blue, low--mass hosts

1343: with high SSFR show little to no anisotropy in their locations.  

1344: Red, massive satellites with

1345: low SSFR show a strong tendency for being located near the 

1346: major axes of their hosts, while blue, low--mass satellites with high

1347: SSFR show little to no anisotropy in their locations.  

1348: This last trend can be understood in part by the

1349: different times at which satellites entered their hosts' halos.  That is,

1350: redder, more massive satellites entered their hosts' halos in the far distant

1351: past while bluer, less massive satellites have only recently entered

1352: their hosts' halos. Therefore, the blue satellites have had their kinematics affected 

1353: less by their hosts than have the red satellites.  In the case of the 

1354: blue satellites, however, there is an additional factor that reduces

1355: the observed anisotropy.  From our analysis

1356: of the MRS, we expect that the majority of the blue satellites are interlopers,

1357: not genuine satellites, and the presence of these

1358: objects greatly suppress the value of the measured anisotropy in

1359: comparison to the intrinsic anisotropy.

1360: 

1361: Overall, the presence of interlopers in the satellite catalogs suppresses

1362: the degree to which the satellites exhibit an anisotropy in their

1363: locations.  However, even after the removal of the interlopers from

1364: the catalog of MRS satellites, the satellites of blue MRS host

1365: galaxies show substantially less anisotropy in their locations than

1366: do the satellites of red MRS host galaxies.    There are two causes

1367: for the reduction of the anisotropy for the satellites of blue hosts

1368: versus the satellites of red hosts.  First, there is a marked difference

1369: of the dependence of the mean satellite location on projected distance for 

1370: the satellites of red hosts compared to the satellites of blue hosts.

1371: In the case of the red SDSS and MRS hosts, the locations of the satellites are largely

1372: independent of the projected distances at which they are found.

1373: In the case of the satellites of blue SDSS hosts, we find that at large

1374: projected distances ($r_p \sim 400$~kpc), there is a tendency for the satellites

1375: to be found close to the {\it minor}  axes of their hosts, while at

1376: smaller projected distances ($r_p \sim 100$~kpc) the satellites have a tendency

1377: to be found close to the major axes of their hosts.  The satellites

1378: of the blue MRS hosts that are found at small projected distances are located 

1379: preferentially close to the major axes of the hosts, while at large projected

1380: distances the locations of the satellites are essentially isotropic.  Therefore,

1381: when the locations of the satellites of blue host galaxies are averaged over

1382: all projected distances ($r_p \le 500$~kpc) there is a substantial reduction 

1383: in the signal compared to when the locations of the satellites of red host

1384: galaxies are averaged over all projected distances.

1385: 

1386: In addition, we find that the prescriptions we use to assign images

1387: to the MRS host galaxies give rise to different degrees of anisotropy in the

1388: satellite locations. 

1389: The satellites of elliptical MRS hosts are distributed

1390: much more anisotropically than are the satellites of non--elliptical MRS 

1391: hosts.  Further, the red MRS hosts are by and large ellipticals, while the

1392: blue MRS hosts are by and large non--ellipticals.  Therefore, at fixed

1393: host mass, we find a substantial

1394: reduction in the anisotropy of the satellites of blue MRS hosts compared to 

1395: red MRS hosts due to the different methods by which the luminous host galaxies

1396: have been embedded within their halos.

1397: 

1398: The locations of satellite galaxies with respect to the symmetry

1399: axes of their hosts may, at first glance, seem to be a mere curiosity. 

1400: However, the

1401: current investigations are beginning to show that the locations of 

1402: satellite galaxies can be used as direct probes of the large--scale

1403: potentials of dark matter halos, and even provide clues to the

1404: orientations of the host galaxies within their halos.    Out of necessity,

1405: the resulting constraints are statistical in nature (since each

1406: host galaxy generally has only 1 or 2 satellites), but this

1407: makes the use of satellite galaxies as halo probes very complementary

1408: to weak gravitational lensing techniques.  Considerably

1409: larger samples of hosts and satellites than those used here may reveal a 

1410: wealth of information about the sizes and shapes of the dark matter halos

1411: of the hosts, the orientation of the hosts within their halos, and the

1412: history of mass accretion by large, bright galaxies.

1413: 

1414: 

1415: 

1416: \section*{Acknowledgments}

1417: 

1418: It is a great pleasure to thank Simon White and the Max Planck Institute

1419: for Astrophysics for hospitality and financial

1420: support of a collaborative visit that allowed us to work directly

1421: with the MRS particle files.   We are also very pleased to thank the

1422: referee for thoughtful, constructive remarks that truly improved the manuscript.

1423: Support under NSF contracts AST-0406844 and AST-0708468 is

1424: gratefully acknowledged.  Funding for the SDSS has been provided by

1425: the Alfred P. Sloan Foundation, the Participating Institutions, NASA,

1426: the NSF, the US Department of Energy, the Japanese Monbukagakusho, and

1427: the Max Planck Society.  The SDSS is managed by the Astrophysical

1428: Research Consortium for the Participating Institutions (the University

1429: of Chicago, Fermilab, the Institute for Advanced Study, the Japan 

1430: Participation Group, The Johns Hopkins University, Los Alamos National

1431: Laboratory, the Max Planck Institute for Astronomy, the Max Planck

1432: Institute for Astrophysics, New Mexico State University, the

1433: University of Pittsburgh, Princeton University, the US Naval Observatory,

1434: and the University of Washington.  The SDSS Web site is 

1435: \url{http://www.sdss.org}.

1436: 

1437: 

1438: \begin{thebibliography}{99}

1439: 

1440: \bibitem{} Abazajian, K. Et al.\ 2009, ApJS, 182, 543

1441: 

1442: \bibitem{} Adelman-McCarthy, J. K. et al.\ 2006, ApJS, 162, 38

1443: 

1444: \bibitem{} Agustsson, I. \& Brainerd, T. G. 2006, ApJ, 650, 500 (AB06)

1445: 

1446: \bibitem{} Allgood, B., Flores, R. A., Primack, J. R., Kravtsov, A. V.,

1447: Wechsler, R. H., Faltenbacher, A. \& Bullock, J. S. 2006, MNRAS, 367, 1781

1448: 

1449: \bibitem{} Azzaro, M., Patiri, S. G., Prada, F., \& Zentner, A.

1450: R. 2007, MNRAS, 376, L43

1451: 

1452: \bibitem{} Bailin, J., \& Steinmetz, M. 2005, ApJ, 627, 647

1453: 

1454: \bibitem{} Bailin, J., Kawata, D., Gibson, B. K., Steinmetz, M., 

1455: Navarro, J. F., Brook, C. B., Gill, S. P. D., Ibata, R. A.,

1456: Knebe, A., Lewis, G. F. \& Okamoto, T. 2005, ApJ, 627, L17

1457: 

1458: \bibitem{} Bailin, J., Power, C., Norberg, P., Zaritsky, D. \&

1459: Gibson, B. K. 2008, MNRAS, 390, 1133

1460: 

1461: \bibitem{} Blanton, M. R., Lin H., Lupton, R. H., Maley, F. M.,

1462: Young, N., Zehavi, I. \& Loveday, J. 2003, AJ, 125, 2276

1463: 

1464: \bibitem{} Blaizot, J., Wadadekar, Y., Guiderdoni, B., Colombi, S. T.,

1465: Bertin, E., Bouchet, F. R., Devriendt, J. E. G., \& Hatton, S. 2005,

1466: MNRAS, 360, 159

1467: 

1468: \bibitem{} Brainerd, T. G. 2005, 628, L101 

1469: 

1470: \bibitem{} Brinchmann J., Charlot S., White S. D. M., Tremonti

1471: C., Kauffmann G., Heckman T., Brinkmann J., 2004, MNRAS, 351, 1151

1472: 

1473: \bibitem{} Colless, M., et al. 2001, MNRAS, 328, 1039

1474: 

1475: \bibitem{} Colless, M., et al., 2003, web publication (astro-ph/0306581)

1476: 

1477: \bibitem{} Croton, D. J., Springel, V., White, S. D. M., De

1478: Lucia, G., Frenk, C. S., Gao, L., Jenkins, A., Kauffmann, G.,

1479: Navarro, J. F., \& Yoshida, N. 2006, MNRAS, 365, 11

1480: 

1481: \bibitem{} De Lucia, G., \& Blaizot, J. 2007

1482: MNRAS, 375, 2

1483: 

1484: \bibitem{} De Lucia, G., Springel, V., White, S. D. M., Croton,

1485: D., \& Kauffmann, G. 2006, MNRAS, 366, 499

1486: 

1487: \bibitem{} Faltenbacher, A., Li, C., White, S. D. M., Jing, Y. P., Mao, S. \&

1488: Wang, J. 2009, RAA, 9, 41

1489: 

1490: \bibitem{} Fukugita, M., Ichikawa, T., Gunn, J. E., Doi, M., Shimasaku, K.,

1491: \& Schneider, D. P. 1996, AJ, 111, 1748

1492: 

1493: \bibitem{} Hawley, D. L. \& Peebles, P. J. E. 1975, AJ, 80, 477

1494: 

1495: \bibitem{} Heavens, A., Refregier, A., \& Heymans, C. 2000,

1496: MNRAS, 319, 649

1497: 

1498: \bibitem{} Hogg, D. W., Finkbeiner, D. P., Schelgel, D. J., \& Gunn, J. E.

1499: 2001, AJ, 122, 2129

1500: 

1501: \bibitem{} Holmberg, E. 1969 Ark. Astron., 5, 305

1502: 

1503: \bibitem{} Jing, Y. P. \& Suto, Y. 2002, ApJ, 574, 538

1504: 

1505: \bibitem{} Kang, X., van den Bosch, F. C., Yang, X., Mao, S., Mo, H. J.,

1506: Li, C. \& Jing, Y. P. 2007, MNRAS, 378, 1531 (Kang07)

1507: 

1508: \bibitem{} Kasun, S. F. \& Evrard, A. E. 2005, ApJ, 629, 781

1509: 

1510: \bibitem{} Kauffmann G.,  Heckman, T. M., White, S. D. M., Charlot, S.,

1511: Tremonti, C., Peng, E. W., Seibert, M., Brinkmann, J.,

1512: Nichol, R. C., SubbaRao, M., \& York, D. 2003, MNRAS, 341, 33

1513: 

1514: \bibitem{} Lemson, G., \& the Virgo Consortium 2006, web publication

1515: (astro-ph/0608019)

1516: 

1517: \bibitem{} Libeskind N. I., Cole S., Frenk C. S., Okamoto, T., \&

1518: Jenkins A., 2007, MNRAS, 374, 16L

1519: 

1520: \bibitem{} MacGillivray, H. T., Dodd, R. J., McNally, B. V.

1521: \& Corwin, H. G., Jr. 1982, 198, 605

1522: 

1523: \bibitem{} Okumura, T., Jing, Y. P. \& Li, C. 2009, ApJ, 694, 214

1524: 

1525: \bibitem{} Okumura, T. \& Jing, Y. P. 2009, ApJ, 694, L83

1526: 

1527: \bibitem{} Sales, L. \& Lambas, D. G. 2004, MNRAS, 348, 1236 

1528: 

1529: \bibitem{} Sales, L. \& Lambas, D. G. 2009, MNRAS, 395, 1184

1530: 

1531: \bibitem{} Sales, L., Navarro, J. F., Lambas, D. G., White, S. D. M. \&

1532: Croton, D. J. 2007, MNRAS, 382, 1901

1533: 

1534: \bibitem{} Salim, S. et al.\ 2007, ApJS, 173, 267

1535: 

1536: \bibitem{} Sharp, N. A., Lin, D. N. C. \& White, S. D. M. 1979,

1537: MNRAS, 187, 287

1538: 

1539: \bibitem{} Siverd, R., Ryden, B. \& Gaudi, B. S. 2009, astro-ph/0903.2264

1540: 

1541: \bibitem{} Smith, J. A. et al.\ 2002, AJ, 123, 2121

1542: 

1543: \bibitem{} Springel, V., et al. 2005, Nature, 435, 629

1544: 

1545: \bibitem{} Strateva I., et al., 2001, ApJ, 122, 1861

1546: 

1547: \bibitem{} Strauss, M.A., Weinberg, D.H., et al. 2002, AJ, 124,

1548: 1810

1549: 

1550: \bibitem{} Valtonen, M., Teerikorpi, P. \& Argue, A. 1978, AJ, 83, 135

1551: 

1552: \bibitem{} Warren, M. S., Quinn, P. J., Salmon, J. K. \& Zurek, W. H.

1553: 1992, ApJ, 399,405

1554: 

1555: \bibitem{} Weinmann, S. M., van den Bosch, F. C., Yang, X., \&

1556: Mo, H. J., 2006, MNRAS, 366, 2

1557: 

1558: \bibitem{} Yang, X., van den Bosch, F. C.,Mo, H. J., Mao, S., Kang,

1559: X., Weinmann, S. M. \& Jing, Y. P. 2006, MNRAS, 369, 1293 (Yang06)

1560: 

1561: \bibitem{} York D. G., et al.\ 2000, AJ, 120, 1579

1562: 

1563: \bibitem{} Zaritsky, D., Smith, R., Frenk, C. \& White, S. D.

1564: M., 1997, ApJ, 478, L53

1565: 

1566: \bibitem{} Zhang, Y., Yang, X., Faltenbacher, A., Springel, V., Lin, W. \&

1567: Wang, H. 2009, ApJ in press (astro-ph/0906.1654)

1568: 

1569: \end{thebibliography}

1570: 

1571: \clearpage

1572: 

1573: \centerline{Table 1: Numbers of Hosts and Satellites} 

1574: \bigskip

1575: \centerline{

1576: \begin{tabular}{lcccc}

1577: \hline\hline

1578: \multicolumn{1}{c}{ } &

1579: \multicolumn{2}{c}{SDSS} &

1580: \multicolumn{2}{c}{MRS} \\

1581:   & hosts & satellites & hosts & satellites \\ \hline

1582:   primary sample (all galaxies) & 4,487  & 7,399  & 70,882 & 140,712 \\

1583: galaxies with known $M_{\rm stellar}$ & 4,412 & 7,296 & 70,882 & 140,712 \\

1584: galaxies with known SSFR & 2,421 & 4,004 & 47,157 & 79,812 \\

1585: red galaxies & 2,926 & 2,334 & 37,022 & 86,178 \\

1586: blue galaxies & 1,561 & 5,065 & 33,860 & 54,534 \\

1587: \hline

1588: \end{tabular}

1589: }

1590: 

1591: \clearpage

1592: \begin{figure}

1593: \centerline{\scalebox{0.80}{\includegraphics{f1.eps} } }%

1594: \vskip -0.0cm%

1595: \caption{Summary of basic properties of the host--satellite

1596: pairs in the SDSS (left panels) and the MRS (right

1597: panels).  From top to bottom, the panels show probability distributions

1598: for the number of satellites per host, the redshift distribution of the

1599: hosts, the $r$--band apparent magnitude distributions of the hosts and satellites,

1600: the $r$--band absolute magnitude distributions of the hosts and satellites,

1601: and the distribution of stellar masses for the hosts and satellites.  In 

1602: panels e) through j) dotted lines indicate results for the satellites and

1603: solid lines indicate results for the hosts.}

1604: \label{fig1}

1605: \end{figure}

1606: 

1607: \begin{figure}

1608: \centerline{\scalebox{0.90}{\includegraphics{f2.eps} } }%

1609: \vskip -0.0cm%

1610: \caption{Properties of MRS host galaxies.  a) Mean host halo virial mass as a

1611: function of stellar mass.  b) Mean host halo virial mass

1612: as a function of absolute $r$--band magnitude.  c) Mean host stellar mass as a 

1613: function of $(g-r)$, computed at $z=0$.  In all panels the data have been

1614: binned such that there are an equal number of objects per bin.

1615: In all cases the standard deviations in the mean values are comparable

1616: to or smaller than the data points.

1617: }

1618: \label{fig2}

1619: \end{figure}

1620: 

1621: \begin{figure}

1622: \centerline{\scalebox{0.90}{\includegraphics{f3.eps} } }%

1623: \vskip -0.0cm%

1624: \caption{Properties of 

1625: satellite galaxies in the MRS that are located within a physical distance

1626: $r_{3D} \le 500$~kpc of a host galaxy. {\it Top:} Mean satellite stellar

1627: mass as a function of absolute $r$--band magnitude (panel a), 

1628: $(g-r)$ at $z=0$ (panel b), and

1629: redshift at which the

1630: satellite first entered its host's halo (panel c). 

1631: {\it Bottom:} Mean redshift at

1632: which a satellite first entered the halo of its host as a function of

1633: $(g-r)$ at $z=0$ (panel d), ratio of satellite to host stellar mass

1634: (panel e), and projected distance at which the satellite is found (panel f).

1635: In each panel the data have been binned such that there are an equal number of objects

1636: per bin.

1637: In all cases the standard deviations in the mean values are comparable

1638: to or smaller than the data points.}

1639: \label{fig3}

1640: \end{figure}

1641: 

1642: \begin{figure}

1643: \centerline{\scalebox{1.00}{\includegraphics{f4.eps} } }%

1644: \vskip 0.0cm%

1645: \caption{ {\it Top:} Differential probability distribution,

1646: $P(\phi)$, for the locations of all satellites,  measured with

1647: respect to the major axes of the hosts.

1648: Dotted line shows the expectation for a uniform

1649: (i.e., circularly--symmetric) distribution of satellites. 

1650: The mean satellite location, $\left< \phi \right>$, and 

1651: the confidence level at which the $\chi^2$ test rejects a uniform distribution

1652: distribution are shown in the panels.

1653: Error bars are omitted when they are

1654: comparable to or smaller than the data point.

1655: {\it Bottom:}

1656: Cumulative probability distribution, $P(\phi \le \phi_{\rm max})$,

1657: for the locations

1658: of the satellites with respect to the major

1659: axes of the hosts (solid line).  Also shown is

1660: $P(\phi \le \phi_{\rm max})$ for a uniform distribution (dotted line).

1661: The median satellite location, $\phi_{\rm med}$, and 

1662: the confidence level at which the KS test rejects a uniform distribution 

1663: are shown in the panels.

1664: {\it Left:} Satellites in the SDSS.  {\it Right:} Satellites in the

1665: MRS.

1666: All satellites with $r_p \le 500$~kpc have been used in the calculations.}

1667: \label{fig4}

1668: \end{figure}

1669: 

1670: \begin{figure}

1671: \centerline{\scalebox{1.00}{\includegraphics{f5.eps} } }%

1672: \vskip 0.0cm%

1673: \caption{Data points with error bars show the

1674: mean satellite location, $\left< \phi \right>$,

1675: for SDSS satellites

1676: (left panels) and MRS satellites (right panels), as a function of

1677: various properties of the hosts.  

1678: Histograms show the distribution

1679: of the host property in each panel.

1680: {\it Top:} $\left< \phi \right>$ as a function of the host's

1681: $(g-r)$ color, computed at $z=0$.  {\it Middle:} $\left< \phi \right>$ as a function

1682: of host

1683: specific star formation rate, SSFR.  {\it Bottom:} $\left< \phi \right>$

1684: as function of host stellar mass. 

1685: All satellites with $r_p \le 500$~kpc have been used in the calculations.

1686: In each panel

1687: the data have been binned such that there are an equal number of 

1688: objects per bin in the calculation of $\left< \phi \right>$.  

1689: Error bars are omitted when the standard

1690: deviation in the mean value of $\phi$ is smaller than

1691: the data point.}

1692: \label{fig5}

1693: \end{figure}

1694: 

1695: \begin{figure}

1696: \centerline{ \scalebox{1.00}{\includegraphics{f6.eps} } }%

1697: \vskip 0.0cm%

1698: \caption{Data points show the mean satellite location, 

1699: $\left< \phi \right>$, for MRS satellites

1700: as a function of host properties for elliptical MRS hosts (left

1701: panels) and non--elliptical MRS hosts (right panels).  

1702: Histograms show the distribution

1703: of the host property in each panel.  {\it Top:} Mean satellite location 

1704: as a function of host $(g-r)$ color.  {\it Bottom:} Mean 

1705: satellite location as a function of host stellar mass.

1706: All satellites with $r_p \le 500$~kpc have been used in the calculations.

1707: In all panels the data have been binned such that there are an equal

1708: number of objects per data point.

1709: In all cases the standard deviation in the mean value of $\phi$ is

1710: comparable to or smaller than the data points.

1711: }

1712: \label{fig6}

1713: \end{figure}

1714: 

1715: \begin{figure}

1716: \centerline{\scalebox{1.00}{\includegraphics{f7.eps} } }%

1717: \vskip 0.0cm%

1718: \caption{Same as Fig.~5, except here single image 

1719: assignment prescriptions are used to define the major axes of the MRS hosts.

1720: Open circles: Major axes of all MRS hosts are obtained from

1721: projections of circular disks onto the sky, where the angular

1722: momenta of the disks are aligned with the angular momenta of the

1723: halos.  

1724: Open triangles: Major axes of all MRS hosts are 

1725: obtained from projections of the halo equivalent ellipsoids onto

1726: the sky.  

1727: Solid squares: SDSS results from Fig.~5. 

1728: Error bars are omitted when the standard deviation in the

1729: mean value of $\phi$ is comparable to or smaller than the data point.

1730: }

1731: \label{fig7}

1732: 

1733: \end{figure}

1734: \begin{figure}

1735: \centerline{\scalebox{1.00}{\includegraphics{f8.eps} } }%

1736: \vskip 0.0cm%

1737: \caption{Same as Fig.~4, but for the satellites of red

1738: hosts. 

1739: All satellites with projected distances $r_p \le 500$~kpc

1740: have been used in the calculations.

1741: }

1742: \label{fig8}

1743: \end{figure}

1744: 

1745: \begin{figure}

1746: \centerline{\scalebox{1.00}{\includegraphics{f9.eps} } }%

1747: \vskip 0.0cm%

1748: \caption{ Same as Fig.~4, but for the satellites of

1749: blue hosts.  

1750: All satellites with projected distances $r_p \le

1751: 500$~kpc have been used in the calculation.

1752: }

1753: \label{fig9}

1754: \end{figure}

1755: 

1756: \begin{figure}

1757: \centerline{\scalebox{0.90}{\includegraphics{f10.eps} } }%

1758: \vskip 0.0cm%

1759: \caption{Mean satellite location as a function of projected

1760: distance, $r_p$, for the satellites of SDSS hosts (solid squares) and

1761: MRS hosts (crosses).  {\it Left: } Satellites of red hosts.

1762: {\it Right:} Satellites of blue hosts.  Error bars are omitted when

1763: the standard deviation in the mean value of $\phi$ is comparable

1764: to or smaller than

1765: the data point.

1766: }

1767: \label{fig10}

1768: \end{figure}

1769: 

1770: \begin{figure}

1771: \centerline{ \scalebox{1.00}{\includegraphics{f11.eps} } }%

1772: \vskip -0.5cm%

1773: \caption{Data points with error bars show the

1774:  mean satellite location, $\left< \phi \right>$, for SDSS satellites

1775: (left panels) and MRS satellites (right panels), as a function of

1776: various properties of the satellites.  

1777: Histograms show the distribution of the satellite property in each 

1778: panel.  From top to bottom the panels show

1779:  $\left< \phi \right>$ as a function of $(g-r)$,

1780: $\left< \phi \right>$ as a function of satellite

1781: specific star formation rate (SSFR), $\left< \phi \right>$ as

1782: as a function of satellite stellar mass, and 

1783: $\left< \phi \right>$ as a function of the projected

1784: distance at which the satellites are found. 

1785: In each panel the data have been binned such that there are an equal 

1786: number of objects per bin in the calculation of $\left< \phi \right>$.  

1787: Error bars are omitted when the standard

1788: deviation in the mean value of $\phi$ is comparable to or

1789: smaller than the data point.}

1790: \label{fig11}

1791: \end{figure}

1792: 

1793: \begin{figure}

1794: \centerline{ \scalebox{0.90}{\includegraphics{f12.eps} } }%

1795: \caption{

1796: Effects of interlopers on the satellite locations in the MRS.   Open

1797: points show $P(\phi)$ using all objects that were identified as satellites

1798: according to the selection criteria in \S3.1.  In all cases the

1799: error in $P(\phi)$ is smaller than the data points. Solid points show 

1800: $P(\phi)$ after all interlopers have been removed from the 

1801: satellite sample (see text).  {\it Top panels:} 

1802: $P(\phi)$ for red (left) and blue (right) MRS hosts.  {\it Bottom panels:}

1803: $P(\phi)$ for red (left) and blue(right) MRS satellites.

1804: }

1805: \label{fig12}

1806: \end{figure}

1807: 

1808: \begin{figure}

1809: \centerline{ \scalebox{1.00}{\includegraphics{f13.eps} } }%

1810: \vskip 0.0cm%

1811: \caption{{\it Top:} Mean satellite location at $z=0$ for genuine MRS satellites as a 

1812: function of the redshift at which they first entered their host's halo.  

1813: Here all satellites are located within a physical distance

1814: of $r_{3D} \le 500$~kpc of the host at the

1815: present day. The data have been binned such that there are an 

1816: equal number of objects per bin, and error bars are omitted when the

1817: standard deviation in the mean value of $\phi$ is comparable to or smaller than

1818: the data point. Solid squares: satellites of all MRS hosts.  Open circles:

1819: satellites of blue MRS hosts.  Open triangles: satellites of red MRS hosts.

1820: {\it Bottom:} Probability distribution for the redshift

1821: at which the genuine MRS satellites first entered their host's halo. 

1822: }

1823: \label{fig13}

1824: \end{figure}

1825: 

1826: 

1827: \end{document}

1828: