1: %\documentclass[useAMS,usenatbib]{mn2e}
2: \documentclass[usenatbib]{mn2e}
3: \bibliographystyle{mn2e}
4: %\def\baselinestretch{1}
5: %\usepackage[dvips]{graphicx}
6: \usepackage{graphicx}
7: \usepackage{times}
8: \usepackage{amssymb}
9: \usepackage{natbib}
10: %%%%%%%%%% user-defined commands %%%%%%%%%%%%%%%%%
11: \def\gsim { \lower .75ex \hbox{$\sim$} \llap{\raise .27ex \hbox{$>$}} }
12: \def\lsim { \lower .75ex \hbox{$\sim$} \llap{\raise .27ex \hbox{$<$}} }
13: %%%%%%%%%%%%%%%%%%%%%%%%%%%
14:
15: \newcommand{\apj}{ApJ}
16: \newcommand{\apjl}{ApJL}
17: \newcommand{\apjs}{ApJS}
18: \newcommand{\aj}{AJ}
19: \newcommand{\mnras}{MNRAS}
20: \newcommand{\nat}{Nature}
21: \newcommand{\pasj}{PASJ}
22: \newcommand{\araa}{ARA\&A}
23: \newcommand{\aapr}{A\&ARv}
24: \newcommand{\aap}{A\&A}
25: \newcommand{\prd}{PhRvD}
26: \newcommand{\physrep}{PhysRep}
27:
28: %%%%%%%%%% end user-defined commands %%%%%%%%%%%%%%%%%
29:
30: \begin{document}
31:
32: \title[Satellites of Simulated Galaxies]
33: {Satellites of Simulated Galaxies: survival, merging, and their relation to the dark and
34: stellar halos}
35:
36: \author[Sales, Navarro, Abadi \& Steinmetz]{Laura V. Sales$^{1,2}$, Julio F. Navarro,$^{3,4}$\thanks{Fellow of the Canadian Institute for Advanced Research.} Mario G. Abadi $^{1,2,3}$
37: and Matthias Steinmetz$^{5}$
38: \\
39: $^{1}$ Observatorio Astron\'omico, Universidad Nacional de C\'ordoba, Laprida
40: 854, 5000 C\'ordoba, Argentina.
41: \\
42: $^{2}$ Instituto de Astronom\'{\i}a Te\'orica y Experimental, Conicet, Argentina.
43: \\
44: $^{3}$Department of Physics and Astronomy, University of Victoria, Victoria, BC V8P 5C2,
45: Canada\\
46: $^{4}$Max-Planck Institut f\"ur Astrophysik, Karl-Schwarzschild Strasse 1,
47: Garching, D-85741, Germany\\
48: %
49: $^{5}$Astrophysikalisches Institut Potsdam, An der Sternwarte 16, Potsdam 14482, Germany\\
50: %
51: %
52: }
53:
54: \maketitle
55:
56: \begin{abstract}
57: We study the population of satellite galaxies formed in a suite of
58: N-body/gasdynamical simulations of galaxy formation in a $\Lambda$CDM
59: universe. The simulations resolve the $\sim 10$ most luminous
60: satellites around each host, and probe systems up to six or seven
61: magnitudes fainter than the primary. We find little spatial or
62: kinematic bias between the dark matter and the satellite
63: population. The radius containing half of all satellites is comparable
64: to the half-mass radius of the dark matter component, and the velocity
65: dispersion of the satellites is a good indicator of the virial
66: velocity of the halo; $\sigma_{\rm sat}/V_{\rm vir} \sim 0.9 \pm 0.2$.
67: Applied to the Local Group, this result suggests that the virial
68: velocity of the Milky Way and M31 might be substantially lower than
69: the rotation speed of their disk components; we find $V_{\rm vir}^{\rm
70: MW} \sim 109\pm 22$ km/s and $V_{\rm vir}^{\rm M31} \sim 138 \pm 35$
71: km/s, respectively, compared with $V_{\rm rot}^{\rm MW} \sim 220$ km/s
72: and $V_{\rm rot}^{\rm M31} \sim 260$ km/s. Although the uncertainties
73: are large, it is intriguing that both estimates are significantly
74: lower than expected from some semianalytic models, which predict
75: a smaller difference between $V_{\rm vir}$ and $V_{\rm rot}$. The
76: detailed kinematics of simulated satellites and dark matter are also
77: in good agreement: both components show a steadily decreasing velocity
78: dispersion profile and a mild radial anisotropy in their velocity
79: distribution. By contrast, the stellar halo of the simulated galaxies,
80: which consists predominantly of stellar debris from {\it disrupted}
81: satellites, is kinematically and spatially distinct from the
82: population of {\it surviving} satellites. This is because the survival
83: of a satellite as a self-bound entity depends sensitively on mass and
84: on time of accretion; surviving satellites are significantly biased
85: toward low-mass systems that have been recently accreted by the
86: galaxy. Our results support recent proposals for the origin of the
87: systematic differences between stars in the Galactic halo and in
88: Galactic satellites: the elusive ``building blocks'' of the Milky Way
89: stellar halo were on average more massive, and were accreted (and
90: disrupted) earlier than the population of dwarfs that has survived
91: self-bound until the present.
92: \end{abstract}
93:
94: \begin{keywords}
95: galaxies: haloes - galaxies: formation -
96: galaxies: evolution.
97: \end{keywords}
98:
99: \section{Introduction}
100: \label{sec:intro}
101:
102: The satellite companions of bright galaxies are exceptionally useful
103: probes of the process of galaxy formation. Studies of the dynamics of
104: satellites around bright galaxies, for example, have provided
105: incontrovertible evidence for the ubiquitous presence of massive dark
106: halos surrounding luminous galaxies, a cornerstone of the present
107: galaxy formation paradigm. Following the pioneering work of
108: \cite{holmberg69,zaritsky93,zaritsky97a} compiled perhaps the first
109: statistically-sound sample of satellite-primary systems with accurate
110: kinematics, and were able to provide persuasive evidence that the dark
111: matter halos hinted at by the flat rotation curves of spiral galaxies
112: (\citealt{sofue01} and references therein) truly dwarf the mass of
113: the luminous component and extend well beyond the luminous radius of
114: the central galaxy.
115:
116: Satellite dynamical studies have entered a new realm since the advent
117: of large redshift surveys, such as the 2dfGRS \citep{colless01} and
118: the SDSS \citep{york00,strauss02}, which have increased
119: many-fold the number of primary-satellite systems known. Recent work
120: based on these datasets have corroborated and extended the results of
121: Zaritsky et al, and their conclusions now appear secure. The dynamics
122: of satellites confirm (i) that dark matter halos extend to large
123: radii, (ii) that more massive halos surround brighter galaxies, and
124: (iii) that early-type galaxies are surrounded by halos about twice as
125: massive as late-type systems of similar luminosity (\citealt{mckay02,
126: prada03,brainerd04a, vandenbosch05}; see
127: \citealt{brainerd04b} for a recent review).
128:
129: Satellites may also be thought of as probes of the faint end of the
130: luminosity function. After all, satellite galaxies are, by definition,
131: dwarf systems, thought to be themselves surrounded by their own
132: low-mass dark matter halos. These low-mass halos are expected to be
133: the sites where the astrophysical processes that regulate galaxy
134: formation (i.e., feedback) operate at maximum efficiency. Thus, the
135: internal structure, star formation history, and chemical enrichment of
136: satellites provide important constraints on the process of galaxy
137: formation in systems where theoretical models predict a highly
138: non-trivial relation between dark mass and luminosity (see, e.g.,
139: \citealt{whiteandrees78, kauffmann93, cole94}; see as well
140: \citealt{cole00} and \citealt{benson02} for a more
141: detailed list of references).
142:
143: The anticipated highly non-linear mapping between dark matter and
144: light at the faint-end of the luminosity function is perhaps best
145: appreciated in the satellite population of the Local Group, where the
146: relatively few known satellites stand in contrast with the {\it
147: hundreds} of ``substructure'' cold dark matter (CDM) halos of
148: comparable mass found in cosmological N-body simulations
149: (\citealt{klypin99b,moore99}). Possible resolutions of this ``satellite
150: crisis'' have been discussed by a number of authors, and there is
151: reasonably broad consensus that it originates from inefficiencies in
152: star formation caused by the combined effects of energetic feedback
153: from evolving stars and by the diminished supply of cold gas due to
154: reionization \citep[see, e.g.][]{kauffmann93, bullock00,
155: somerville01, benson02}. These effects combine to reduce
156: dramatically the star formation activity in substructure halos, and
157: can reconcile, under plausible assumptions, the substructure halo mass
158: function with the faint end of the satellite luminosity function
159: \citep{stoehr02, kazantzidis04, penarrubia07}.
160:
161: The price paid for reconciling cold dark matter substructure with the
162: Local Group satellite population is one of simplicity, as the
163: ``feedback'' processes invoked involve complex astrophysics that is
164: not yet well understood nor constrained. It is not yet clear, for
165: example, whether the brighter satellites inhabit the more massive
166: substructures, or whether, in fact, there is even a monotonic relation
167: between light and mass amongst satellites. This issue is further
168: complicated by the possibility that a substantial fraction of a
169: satellite's mass may have been lost to tides. Tidal stripping is
170: expected to affect stars and dark matter differently, complicating
171: further the detailed relation between light and mass in substructure
172: halos (\citealt{hayashi03, kravtsov04}, Strigari et al. 2007a,b).
173: \nocite{strigari07a,strigari07b}
174:
175: These uncertainties hinder as well the interpretation of satellites as
176: relics of the hierarchical galaxy assembly process, and consensus has
177: yet to emerge regarding the severity of the biases that the various
178: effects mentioned above may engender. Do the spatial distribution of
179: satellites follow the dark matter? Is the kinematics of the satellite
180: population substantially biased relative to the dark matter's? Have
181: satellites lost a substantial fraction of their stars/dark matter to
182: stripping? Are surviving satellites fair tracers of the population of
183: accreted dwarfs?
184:
185: Of particular interest is whether satellites may be considered relics
186: of the ``building blocks'' that coalesced to form the early
187: Galaxy. Indeed, the stellar halo of the Milky Way is regarded, in
188: hierarchical models, to consist of the overlap of the debris of many
189: accreted satellites which have now merged and mixed to form a
190: kinematically hot, monolithic stellar spheroid \citep{searleandzinn78,
191: bullockandjohnston05, abadi06, moore06} . A
192: challenge to this view comes from detailed observation of stellar
193: abundance patterns in satellite galaxies in the vicinity of the Milky
194: Way. At given metallicity, the stellar halo (at least as sampled by
195: stars in the solar neighbourhood) is systematically more enriched in
196: $\alpha$-elements than stars in Galactic satellites \citep{fuhrmann98,
197: shetrone01, shetrone03, venn04}, a result that remains
198: true even when attempting to match stars of various ages or
199: metallicities \citep{unavane96, gilmoreandwyse98, pritzl05}.
200: Can hierarchical models explain why satellites identified today
201: around the Milky Way differ from the ones that fused to form the
202: Galactic halo?
203:
204: Preliminary clues to these questions have been provided by the
205: semianalytic approach of Bullock, Johnston and collaborators
206: \citep{bullockandjohnston05, font06a, font06b}, who argue that hierarchical
207: models predict naturally well-defined distinctions between the halo
208: and satellite stellar populations. Detailed answers, however, depend
209: critically on which and when substructure halos are ``lit up'' and how
210: they evolve within ``live'' dark matter halos. These are perhaps best
211: addressed with direct numerical simulation that incorporates the
212: proper cosmological context of accretion as well as the gasdynamical
213: effects of cooling and star formation in an evolving population of
214: dark matter halos. The study we present here aims to address these
215: issues by analyzing the properties of the satellite population of
216: $L_*$ galaxies simulated in the $\Lambda$CDM scenario. We introduce
217: briefly the simulations in \S~\ref{sec:numexp}, analyze and discuss
218: them in \S~\ref{sec:analysis} and we conclude with a summary in
219: \S~\ref{sec:conc}.
220:
221: %%%%%%%%%%%%%%%%%%%%%%%%%%
222: \begin{figure*}
223: \begin{center}
224: \includegraphics[width=\linewidth,clip]{figs/fig1_lr.ps}
225: \end{center}
226: \caption{Spatial distribution of the stellar component of four of our
227: simulated galaxies at z=0. Each panel corresponds to a different
228: simulation, projected so that the inner galaxy is seen approximately
229: ``edge-on''. The virial radius of the system is marked by the outer
230: green circle in each panel. The inner circle has a radius of $20$
231: kpc, where most the stars in each galaxy are found. Stars that have
232: formed in satellites that survive as self-bound entities until $z=0$
233: are shown in yellow. ``In situ'' stars, i.e., those formed in the
234: most massive progenitor of the galaxy, are shown in cyan, whereas
235: those formed in satellites that have been accreted and disrupted by
236: the main galaxy are shown in red. Note that the diffuse outer
237: stellar halo reaches almost out to the virial radius, and consists
238: almost exclusively of accreted stars. The inner galaxy, on the other
239: hand, is dominated by stars formed ``in situ''.
240: \label{fig:xypanel}}
241: \end{figure*}
242: %%%%%%%%%%%%%%%%%%%%%%%%%%
243:
244: \section{The Numerical Simulations}
245: \label{sec:numexp}
246:
247: Our suite of eight simulations of the formation of $L_*$ galaxies in
248: the $\Lambda$CDM scenario is the same discussed recently by
249: Abadi, Navarro and Steinmetz 2006. The simulations are similar to the one
250: originally presented by \citet{steinmetzandnavarro02}, and have been
251: analyzed in detail in several recent papers, which the interested
252: reader may wish to consult for details on the numerical setup
253: \citep{abadi03a, abadi03b, meza03, meza05, navarro04}.
254:
255: Briefly, each simulation follows the evolution of a small region of
256: the universe chosen so as to encompass the mass of an $L_{*}$ galaxy
257: system. This region is chosen from a large periodic box and
258: resimulated at higher resolution preserving the tidal fields from the
259: whole box. The simulation includes the gravitational effects of dark
260: matter, gas and stars, and follows the hydrodynamical evolution of the
261: gaseous component using the Smooth Particle Hydrodynamics (SPH)
262: technique \citep{steinmetz96}. We adopt the following cosmological
263: parameters for the $\Lambda$CDM scenario: $H_0=65$ km/s/Mpc,
264: $\sigma_8=0.9$, $\Omega_{\Lambda}=0.7$, $\Omega_{\rm CDM}=0.255$,
265: $\Omega_{\rm bar}=0.045$, with no tilt in the primordial power
266: spectrum.
267:
268: All simulations start at redshift $z_{\rm init}=50$, have force
269: resolution of order $1$ kpc, and the mass resolution is chosen so that
270: each galaxy is represented on average, at $z=0$, with $\sim 50,000$
271: gas/dark matter particles and $\sim 125,000$ star particles. Each
272: re-simulation follows a single $\sim L_*$ galaxy in detail, and
273: resolves a number of smaller, self-bound systems we shall call
274: generically ``satellites''. We shall hereafter refer to the main
275: galaxy indistinctly as ``primary'' or ``host''.
276:
277: Gas is allowed to turn into stars at rates consistent with the
278: empirical Schmidt-like law of \citet{kennicutt98} in collapsed regions at
279: the center of dark matter halos. Because star formation proceeds
280: efficiently only in high-density regions, the stellar components of
281: primary and satellites are strongly segregated spatially from the dark
282: matter. We include the energetic feedback of evolving stars, although
283: its implementation mainly as a heating term on the (dense) gas
284: surrounding regions of active star formation implies that most of this
285: energy is lost to radiation and that feedback is ineffective at
286: curtailing star formation. The transformation of gas into stars thus
287: tracks closely the rate at which gas cools and condenses at the center
288: of dark matter halos. This results in an early onset of star-forming
289: activity in the many progenitors of the galaxy that collapse at high
290: redshift, as well as in many of the satellite systems we analyze here.
291:
292: Another consequence of our inefficient feedback algorithm is that gas
293: cooling and, therefore, star formation, proceed with similar
294: efficiency in all well-resolved dark matter halos, irrespective of
295: their mass. As a result, the total stellar mass of a satellite
296: correlates quite well with the ``original'' mass of its progenitor
297: dark halo; i.e., the total mass of the satellite before its accretion
298: into the virial radius of its host. We define the {\it virial} radius,
299: $r_{\rm vir}$, of a system as the radius of a sphere of mean density
300: $\Delta_{\rm vir}(z)$ times the critical density for closure. This
301: definition defines implicitly the virial mass, $M_{\rm vir}$, as that
302: enclosed within $r_{\rm vir}$, and the virial velocity, $V_{\rm vir}$,
303: as the circular velocity measured at $r_{\rm vir}$. Quantities
304: characterizing a system will be measured within $r_{\rm vir}$, unless
305: otherwise specified. The virial density contrast, $\Delta_{\rm
306: vir}(z)$ is given by $\Delta_{\rm vir}(z)=18\pi^2+82f(z)-39f(z)^2$,
307: where $f(z)=[\Omega_0(1+z)^3/(\Omega_0(1+z)^3+\Omega_\Lambda))]-1$ and
308: $\Omega_0=\Omega_{\rm CDM}+\Omega_{\rm
309: bar}$\citep{bryanandnorman98}. $\Delta_{\rm vir}\sim 100$ at $z=0$.
310:
311: It is likely that improvements to our feedback algorithms may lead to
312: revisions in the efficiency and timing of star formation in these
313: galaxies, and especially in the satellites, but we think our results
314: will nonetheless apply provided that these revisions do not
315: compromise the relatively simple relation between stellar mass and
316: halo mass that underpins many of our results. For example, we expect
317: that modifications of the star formation algorithm will affect
318: principally the number, age, and chemical composition of stars, rather
319: than the dynamical properties of the satellites. This is because the
320: latter depend mainly on the mass, orbit, and timing of the merging
321: progenitors, which are largely dictated by the assumed cosmological
322: model. These properties are less sensitive to the complex astrophysics
323: of star formation and feedback, and therefore our analysis focuses on
324: the kinematics and dynamical evolution of the satellite population
325: around the eight galaxies in our simulation suite.
326:
327:
328: \begin{center}
329: \begin{figure}
330: \includegraphics[width=84mm]{figs/fig2.ps}
331: \caption{Cumulative luminosity distribution of simulated satellites
332: (filled circles), averaged over our eight simulations, and compared
333: with the Milky Way (blue dashed line) and M31 (red dotted curve)
334: satellite systems. Satellite luminosities are scaled to the luminosity
335: of the host. Error bars in the simulated data indicate Poisson
336: uncertainties in the computation of the average. The flattening of the
337: simulated satellite distribution below $0.1\%$ of the primary
338: luminosity is due to numerical limitations. The Local Group data are
339: taken from \citet{vandenbergh99}. For the MW and M31 systems we include
340: only satellites at distances closer than $300$ kpc from the central
341: galaxy.}
342: \label{fig:nlv}
343: \end{figure}
344: \end{center}
345:
346: \begin{center}
347: \begin{figure}
348: \includegraphics[width=84mm]{figs/fig3.ps}
349: \caption{Number density profile of simulated satellites, after scaling
350: their positions to the virial radius of each host and stacking all
351: eight simulations (solid circles; error bars denote Poisson
352: uncertainties associated with the total number of satellites in each
353: radial bin). The dotted line corresponds to the average dark matter
354: density profile, and the dashed line to the stars in the outer stellar
355: halo. The vertical normalization for the satellite and stellar halo
356: profiles has been chosen so that all profiles approximately coincide
357: at $r\sim 0.15 \, r_{\rm vir}$. Note that the spatial distribution of
358: satellites is similar to the dark matter, and that stars in the
359: stellar halo are significantly more centrally concentrated. Arrows
360: mark the radius containing half the objects in each component. See
361: text for further discussion.}
362: \label{fig:ndprof}
363: \end{figure}
364: \end{center}
365:
366:
367: \section{Results and Discussion}
368: \label{sec:analysis}
369:
370: \subsection{Characterization of the satellite population}
371: \label{ssec:char}
372:
373: Figure~\ref{fig:xypanel} shows the spatial distribution of all star
374: particles in four of our simulated galaxies. Stars are assigned to one
375: of three components and colored accordingly. Particles in cyan are
376: ``in-situ'' stars; i.e., stars that formed in the main progenitor of
377: the primary galaxy. Stars in red formed in satellites that have since
378: been accreted and fully disrupted by the tidal field of the
379: galaxy. Stars in yellow formed in systems that survive as recognizable
380: self-bound satellites until $z=0$. As discussed in detail by Abadi et
381: al (2006), the tidal debris of fully disrupted satellites makes up the
382: majority of the smooth outer stellar halo component. ``In-situ''
383: stars, on the other hand, dominate the inner galaxy, whereas surviving
384: satellites are easily identifiable as overdense, tightly bound clumps
385: of stars.
386:
387: In practice, we identify satellite systems using a friends-of-friends
388: algorithm to construct a list of potential stellar groupings, each of
389: which is checked to make sure that (i) they are self-bound, and that
390: (ii) they contain at least $35$ star particles. This minimum number of
391: stars (which corresponds roughly to $\sim 0.03\%$ of the stellar mass
392: of the primary at $z=0$) is enough to ensure the reliable
393: identification of the satellite at various times and the robust
394: measurement of their orbital properties, but is insufficient to study
395: the internal structure of the satellite. The satellite identification
396: procedure is run for all snapshots stored for our simulations,
397: allowing us to track the evolution of individual satellites.
398:
399: With these constraints, our simulations resolve, at $z=0$, an average
400: of about $10$ satellites within the virial radius of each simulated
401: galaxy. The cumulative luminosity distribution of these satellites
402: (computed in the $V$ band{\footnote{Luminosity estimates in various
403: bands are made by convolving the masses and ages of star particles
404: with standard spectrophotometric models, see, e.g., Abadi et al 2006
405: for details.}} for ease of comparison with data available for the Local
406: Group satellites) is shown in Figure~\ref{fig:nlv}. The brightest
407: satellite is, on average, about $\sim 12\%$ as bright as the primary,
408: in reasonable agreement with the most luminous satellite around the
409: Milky Way and M31: the LMC and M33 are, respectively, $11\%$ and
410: $8\%$ as bright as the MW and M31 (van den Bergh 1999).
411:
412: At brightnesses below $0.2\%$ of $L_{\rm host}$ the number of
413: simulated satellites levels off as a result of numerical
414: limitations. Independent tests (Abadi et al, in preparation) show that
415: this brightness limit corresponds to where satellite identification in
416: the simulations becomes severely incomplete. We note that this
417: limitation precludes us from addressing the ``satellite crisis''
418: alluded to in \S1: our simulations lack the numerical resolution
419: needed to resolve the hundreds of low-mass substructure halos found in
420: higher-resolution CDM simulations. On average, the $10$th brightest
421: satellite in our simulations is $\sim 5.6$ mag fainter than the
422: primary; for comparison, the MW and M31 have only $2$ and $5$
423: satellites as bright as that.
424:
425: Given the small number of systems involved and the considerable
426: scatter from simulation to simulation (the number of bright satellites
427: ranges from $4$ to $21$ in our eight simulations) we conclude that
428: there is no dramatic discrepancy between observations and simulations
429: at the bright end of the satellite luminosity function. Applying our
430: results to the full Local Group satellite population, including, in
431: particular, the extremely faint dwarfs being discovered by panoramic
432: surveys of M31 and by the SDSS (\citealt{zucker04,zucker06,willman05b,
433: martin06,belokurov06,belokurov07,irwin07,majewski07}, Ibata et al. 2007
434: submitted), involves a
435: fairly large extrapolation, and should be undertaken with
436: caution (see, e.g., Pe\~narrubia, McConnachie \& Navarro 2007
437: for a recent discussion).
438:
439:
440: \begin{center}
441: \begin{figure}
442: \includegraphics[width=84mm]{figs/fig4.ps}
443: \caption{Spherical components of satellite velocities at $z=0$ as a
444: function of their distance to the center of the host galaxy. Each
445: system has been rotated so that the angular momentum of the inner
446: galaxy is aligned with the direction of the $z$-coordinate
447: axis. Positions and velocities have been scaled to the virial radius
448: and velocity of each host halo. Panels on the right show the velocity
449: distributions of the satellite population within $r_{\rm vir}$ (solid
450: lines) and compare it to the dark matter particles (dotted lines).
451: The velocity dispersions are given in each panel. Note the slight
452: asymmetry in the satellites' $V_{\phi}$ velocity distribution, which
453: results from the net co-rotation of satellites around the primary.}
454: \label{fig:vrtp}
455: \end{figure}
456: \end{center}
457:
458:
459: \begin{center}
460: \begin{figure}
461: \includegraphics[width=84mm]{figs/fig5.ps}
462: \caption{{\it Top panel:} Radial and tangential velocity dispersion
463: profiles of satellites, dark matter, and stellar halo, computed after
464: scaling to virial values and stacking all simulations in our
465: series. {\it Bottom panel:} Anisotropy parameter as a function of
466: radius for the satellite population, compared with dark matter
467: particles and with the stellar halo. Note that satellites are only
468: slightly more radially anisotropic than the dark matter and
469: kinematically distinct from the stellar halo.}
470: \label{fig:rdisp}
471: \end{figure}
472: \end{center}
473:
474: \subsection{Spatial distribution}
475: \label{ssec:satvr}
476:
477: Figure~\ref{fig:xypanel} shows that satellites are found throughout
478: the virial radius of the host and that, unlike stars in the smooth
479: stellar halo, satellites show little obvious preference for clustering
480: in the vicinity of the central galaxy. This is confirmed in
481: Figure~\ref{fig:ndprof}, where the solid circles show the number
482: density profile of satellites, after rescaling their positions to the
483: virial radius of each host and stacking all eight simulations. The
484: dashed and dotted lines in this figure correspond, respectively, to
485: the density profile of the stellar and dark matter halos, scaled and
486: stacked in a similar way. The vertical normalization of the satellite
487: and stellar halo profiles is arbitrary, and has been chosen so that
488: all profiles approximately match at $r\sim 0.15 \, r_{\rm vir}$.
489:
490: There is little difference in the shape of the dark matter and
491: satellite profiles: half of the satellites are contained within $\sim
492: 0.37 \, r_{\rm vir}$, a radius similar to the half-mass radius of the
493: dark matter, $\sim 0.3 \, r_{\rm vir}$. We conclude that, within the
494: uncertainties, the satellites follow the dark matter. The stellar
495: halo, on the other hand, is much more centrally concentrated than the
496: dark matter and satellites; its half-mass radius is only $\sim 0.05 \,
497: r_{\rm vir}$, as shown by the arrows in Figure~\ref{fig:ndprof}.
498:
499: This result implies that the spatial distribution of simulated
500: satellites is distinct from that of CDM substructure halos, whose
501: density profile is known to be significantly shallower than the dark
502: matter's \citep{ghigna98, ghigna00, gao04a, diemand04}. This suggests
503: that the ``mapping'' between dark and luminous substructure is highly
504: non-trivial, as argued by \citet{springel01} and \citet{delucia04}.
505: Our results, which are based on direct numerical simulation, validate
506: these arguments and illustrate the complex relation between galaxies
507: and the subhalos in which they may reside \citep[see
508: also][]{kravtsov04, nagaiandkravtsov05, gnedin06, weinberg06, libeskind07}.
509: Luminous satellites are resilient to disruption by tides, and they can
510: survive as self-bound entities closer to the primary, where
511: substructures in dark matter-only simulations may quickly disrupt, as
512: first pointed out by White \& Rees (1978).
513:
514: We conclude that using dark matter substructures to trace directly the
515: properties of luminous satellites is likely to incur substantial and
516: subtle biases which may be difficult to avoid. Models that attempt to
517: follow the evolution of dark matter substructures and their luminous
518: components are likely to fare better \citep[see, e.g.][]{croton06,
519: bower06}. At the low mass end, the inclusion of some treatment of the
520: substructure mass loss and tidal shocks is needed to put in better
521: agreement semianalytic models with the results from numerical simulations
522: \citep{taylorandbabul01,benson02a}.
523: Definitive conclusions will probably need to
524: wait until realistic simulations with enhanced numerical resolution
525: and improved treatment of star formation become available.
526:
527: \subsection{Kinematics}
528:
529: The likeness in the spatial distribution of satellites and dark matter
530: anticipates a similar result for their kinematics. This is illustrated
531: in Figure ~\ref{fig:vrtp}, where the panels on the left show the
532: spherical components of the satellites' velocities (in the rest frame
533: of the host and scaled to its virial velocity) versus galactocentric
534: distance (in units of the virial radius of the host). Velocity
535: components are computed after rotating each system so that the
536: $z$-axis (the origin of the polar angle $\theta$) coincides with the
537: rotation axis of the inner galaxy. The corresponding velocity
538: distributions are shown by the thick solid lines in the panels on the
539: right, and compared with those corresponding to the dark matter
540: particles (dotted lines).
541:
542: The velocity distribution of each component is reasonably symmetric
543: and may be well approximated by a Gaussian, except perhaps for the
544: satellites' $V_{\phi}$-component, which is clearly asymmetric. This is
545: a result of net rotation around the $z$ axis: the satellite population
546: has a tendency to co-rotate with the galaxy's inner body which is more
547: pronounced than the dark matter's. Indeed, we find that on average the
548: specific angular momentum of satellites is $\sim 50\%$ higher than the
549: dark matter, and a factor of $\sim 10$ higher than the stellar
550: halo. This result likely arises as a consequence of the accretion and
551: survival biases discussed below; surviving satellites accrete late and
552: from large turnaround radii, making them especially susceptible to the
553: tidal torques responsible for spinning up the galaxy. The overall
554: effect, however, is quite small, and rotation provides a negligible
555: amount of centrifugal support to the satellite population.
556:
557: The velocity dispersion of both satellites and dark matter particles
558: drops steadily from the center outwards, as shown in
559: Figure~\ref{fig:rdisp}. The top panel shows that the drop is similar in
560: all components, and that the velocity dispersion decreases from its
561: central value by a factor of $\sim 2$ at the virial radius. This
562: figure also shows that the velocity distribution is radially
563: anisotropic, and that the anisotropy becomes more pronounced in the
564: outer regions. The trends are again similar for satellites and dark
565: matter, rising slowly with radius and reaching $\beta\sim 0.4$ at the
566: virial radius. (The anisotropy parameter, $\beta$, is given by $\beta
567: = 1-({\sigma_t}^2/2{\sigma_r}^2)$, where $\sigma_r$ is the radial
568: velocity dispersion and
569: $\sigma_t=\sqrt{(\sigma_\phi^2+\sigma_\theta^2)/2}$ is the tangential
570: velocity dispersion.)
571:
572: The stellar halo, on the other hand, is kinematically distinct from
573: the satellites and from the dark matter. Overall, its velocity
574: dispersion is lower, and its anisotropy is more pronounced, rising
575: steeply from the center outwards and becoming extremely anisotropic
576: ($\beta \sim 0.8$) in the outer regions. As discussed in detail by
577: Abadi et al (2006), this reflects the origin of the stellar halo as
578: debris from satellite disruption, which occur at small radii, where
579: tidal forces are maximal. Stars lost during disruption (merging)
580: events and that now populate the outer halo must therefore be on
581: rather eccentric orbits, as witnessed by the prevalence of radial
582: motions in Figure~\ref{fig:rdisp}. The kinematical distinction between
583: satellites and stellar halo thus suggests that few halo stars have been
584: contributed by stripping of satellites that have survived self-bound
585: until the present. We shall return to this issue below.
586:
587: \begin{center}
588: \begin{figure}
589: \includegraphics[width=84mm]{figs/fig6.ps}
590: \caption{{\it Top panel:} Orbital evolution of two satellites, chosen
591: to illustrate the case of a system that merges quickly with the
592: primary and of another that survives as a self-bound entity until
593: $z=0$. Curves show the distance from the primary to the self-bound
594: stellar core of the satellite as a function of time. The dotted line
595: shows the evolution of the virial radius of the primary galaxy, and
596: the arrow indicates the time, $t_{\rm acc}$, when the satellites are
597: first accreted into the primary's halo. Although both satellites are
598: accreted more or less at the same time, they are not a physical pair
599: and evolve independently. {\it Bottom panel:} The evolution of the
600: satellites' bound mass of stars and dark matter, normalized to the
601: values computed at the time of accretion. Note that the stellar
602: component is much more resilient to the effect of tides.}
603: \label{fig:orbmass}
604: \end{figure}
605: \end{center}
606:
607:
608: \subsection{Application to the Local Group}
609:
610: The lack of strong kinematical bias between satellites and dark matter
611: may be used to estimate the virial velocity of the Milky Way and M31.
612: For example, assuming that the radial velocity dispersion of the
613: satellites is related to the virial velocity by $\sigma_r \sim 0.9 \,
614: (\pm 0.2) V_{\rm vir}$ (see Figure~\ref{fig:vrtp}; the uncertainty is
615: just the rms scatter from our eight simulations), we obtain $V_{\rm
616: vir}^{\rm MW}\sim 109\pm 22$ km/s from the $\sim 99$ km/s
617: Galactocentric radial velocity dispersion of the eleven brightest
618: satellites (see, e.g., the compilation of van den Bergh 1999).
619:
620: The same procedure may be applied to M31 satellites. Taking into
621: account projection effects, we find that the line-of-sight satellite
622: velocity dispersion is $\sigma_{\rm los} \sim 0.8 (\pm 0.2) V_{\rm
623: vir}$. Taking the $16$ brightest satellites within $300$ kpc from the
624: center of M31, we find $\sigma_{\rm los} \sim 111$ km/s, implying
625: $V_{\rm vir}^{\rm M31} \sim 138 \pm 35$ km/s. We use here the
626: compilation of \citet{mcconnachieandirwin06}, complemented with data
627: for And XIV from \citet{majewski07}, and for And XII from (Chapman et
628: al 2007, submitted).
629:
630: These results imply that the virial radius of the Milky Way is $r_{\rm
631: vir}^{\rm MW} \sim 240$ kpc. Our simulations predict that half of the
632: brightest satellites should be enclosed within $\sim 90$ kpc, which
633: compares favourably with observations: half of the eleven brightest
634: satellites are within $\sim 90.1$ kpc from the center of the Milky
635: Way. Contrary to the arguments of \citet{taylor05}, no
636: substantial bias between satellites and dark matter is required to
637: explain the MW satellite spatial distribution, provided that one
638: accepts a virial radius as small as $\sim 240$ kpc.
639:
640: The same argument, applied to M31, suggests that half of the 16
641: satellites within its virial radius ($r_{\rm vir}^{\rm M31} \sim 300$
642: kpc) must be within $\sim 111$ kpc, compared with the observational
643: value of $\sim 165 $ kpc. Note that these radii are actual distances to M31,
644: rather than projections.
645:
646: Despite the sizable statistical uncertainty inherent to the small
647: number of satellites in these samples, it is interesting that both of
648: the virial velocity estimates mentioned above are significantly lower
649: than the rotation speed measured for these galaxies in the inner
650: regions; $V_{\rm rot}^{\rm MW}\sim 220$ km/s and $V_{\rm rot}^{\rm
651: M31} \sim 260 $ km/s. These low virial velocity estimates are in line
652: with recent work that advocates relatively low masses for the giant
653: spirals in the Local Group \citep{klypin02, seigar06,abadi06, smith06}.
654:
655: If confirmed, this would imply that the circular velocity should drop
656: steadily with radius in the outer regions of these galaxies. As
657: discussed by Abadi et al (2006), this may be the result of ``adiabatic
658: contraction'' of the dark matter halo following the assembly of the
659: luminous galaxy. However, such result may be difficult to reconcile
660: with semianalytic models of galaxy formation, which favor a better
661: match between $V_{\rm rot}$ and $V_{\rm vir}$. Croton et al (2006)
662: argue that $V_{\rm rot}$ should be similar to the maximum circular
663: velocity of the dark matter halo, which is only about $\sim 20\%$
664: larger than $V_{\rm vir}$ for typical concentrations. It is possible
665: that taking into account the effects of the adiabatic contraction and
666: including the self-gravity of the baryon material might induce a large
667: scatter and allow rotation speeds as high as $V_{\rm rot}\sim 1.5-2$
668: times $V_{\rm vir}$ (A.Benson, private communication). Final word on
669: this issue needs further data to place better constraints on the mass
670: of the halo of the Local Group spirals at large distances, as well as
671: improved semianalytic modeling that re-examines critically the
672: response of the dark halo to the formation of the luminous galaxy. At
673: least from the observational point of view, the steady pace of
674: discovery of new satellites of M31 and MW facilitated by digital sky
675: surveys implies that it should be possible to revisit this issue in
676: the near future with much improved statistics.
677:
678: \begin{center}
679: \begin{figure}
680: \includegraphics[width=84mm]{figs/fig7.ps}
681: \caption{{\it Top panel:} Orbital decay timescale of satellites,
682: $\tau$, shown as a function of satellite mass. Decay timescales are
683: computed by fitting an exponential law to the evolution of the
684: apocentric radius of a satellite, and is shown in units of the
685: (radial) orbital period measured at accretion time. Satellite masses
686: (dark+baryons) are scaled to the total mass of the host at $t_{\rm
687: acc}$. Filled and open circles correspond to satellites that have,
688: respectively, survived or merged with the primary by $z=0$. Filled
689: squares show the median decay timescale after splitting the sample
690: into equal-number mass bins. More massive satellites spiral in faster
691: due to the effects of dynamical friction. {\it
692: Bottom panel:} Histogram of surviving and merged satellites as a
693: function of satellite mass. Note the strong mass bias of surviving
694: satellites relative to merged ones. }
695: \label{fig:orbdec}
696: \end{figure}
697: \end{center}
698:
699:
700: \begin{center}
701: \begin{figure}
702: \includegraphics[width=84mm]{figs/fig8.ps}
703: \caption{{\it Top panel:} Orbital pericenter-to-apocenter ratio
704: measured at two different times during the evolution of a
705: satellite. Values on the horizontal axis correspond to the time of
706: accretion whereas values on the vertical axis are computed once
707: dynamical friction has eroded the apocentric distance to $\sim e^{-1}$
708: of its turnaround value. Most satellites lie above the $1$:$1$ dotted
709: line, indicating significant orbital circularization by dynamical
710: friction. Open and filled circles correspond, respectively, to merged
711: or surviving satellites at $z=0$. Open and filled squares mark the
712: median of each of those populations, respectively. {\it Bottom panel:}
713: Histogram of pericenter-to-apocenter ratio at the time of accretion
714: for surviving and merged satellites. Note that satellites originally
715: on more eccentric orbits tend to merge faster.}
716: \label{fig:circ}
717: \end{figure}
718: \end{center}
719:
720:
721: \subsection{Satellite evolution}
722:
723: \subsubsection{Merging and survival}
724:
725: Satellites are affected strongly by the tidal field of the primary,
726: and evolve steadily after being accreted into the halo of the host
727: galaxy. This is illustrated in Figure~\ref{fig:orbmass}, where the
728: upper panel shows the evolution of the galactocentric distance for two
729: satellites in one of our simulations. These two satellites follow
730: independent accretion paths into the halo of the primary galaxy; after
731: initially drifting away from the galaxy due to the universal
732: expansion, they reach a turnaround radius of a few hundred kpc and are
733: then accreted into the virial radius of the primary at similar times,
734: $\sim 4.5$ Gyr after the Big Bang ($z \sim 1.5$). The accretion is
735: indicated by the intersection between the trajectory of each satellite
736: in the upper panel of Figure~\ref{fig:orbmass} and the dotted line,
737: which shows the evolution of the virial radius of the main progenitor
738: of the primary.
739:
740: We define the time that the satellite first enters the virial radius
741: of the primary as the {\it accretion time}, $t_{\rm
742: acc}$, or $z_{\rm acc}$, if it is expressed as a redshift. Because
743: masses, radii, and other characteristic properties of a satellite are
744: modified strongly by the tides that operate inside the halo of the
745: primary, it is useful to define the satellite's properties at the time
746: of accretion, and to refer the evolution to the values measured at
747: that time.
748:
749: One example of the effect of tides is provided by the self-bound mass
750: of the satellite, whose evolution is shown in the bottom panel of
751: Figure~\ref{fig:orbmass}. The dark matter that remains bound to the
752: satellite (relative to that measured at accretion time) is shown by
753: open symbols; the bound mass in stars is shown by solid triangles. One
754: of the satellites (dashed lines) sees its orbit eroded quickly by
755: dynamical friction, and merges with the primary less than $4$ Gyr
756: after accretion, at which point the self-bound mass of the dark matter
757: and stellar components drops to zero. The orbital period decreases rapidly as
758: the satellite sinks in; we are able to trace almost 5 complete orbits
759: before disruption although, altogether, the satellite takes only $2.5$
760: Gyr to merge after the first pericentric passage, a time comparable to
761: just half the orbital period at accretion time.
762:
763: As the satellite is dragged inwards by dynamical friction dark matter
764: is lost much more readily than stars; after the first pericentric
765: passage only about $40\%$ of the original dark mass remains attached
766: to the satellite, compared with $85\%$ of the stars. This is a result
767: of the strong spatial segregation between stars and dark matter which
768: results from gas cooling and condensing at the center of dark halos
769: before turning into stars. Stars are only lost in large numbers at the
770: time of merger, when the satellite is fully disrupted by the tides.
771:
772: The second satellite (solid lines in Figure~\ref{fig:orbmass})
773: survives as a self-bound entity until the end of the simulation. Its
774: orbit is affected by dynamical friction, but not as drastically as the
775: merged satellite: after completing 3 orbits, its apocentric distance
776: has only dropped from $\sim 250$ kpc at turnaround ($t_{\rm ta}\sim 3$
777: Gyr) to $\sim 180$ kpc at $z=0$. The stars in the satellite survive
778: almost unscathed; more than $85\%$ of stars remain bound to the
779: satellite at the end of the simulation, although only $\sim 45\%$ of the
780: dark matter is still attached to the satellite then.
781:
782: As expected from simple dynamical friction arguments, the final fate
783: of a satellite regarding merging or survival depends mainly on its
784: mass and on the eccentricity of its orbit. The ``merged'' satellite in
785: Figure~\ref{fig:orbmass} is $\sim 6$ times more massive than the
786: ``surviving'' one and is on a much more eccentric orbit: its first
787: pericentric radius is just $\sim 20$ kpc, compared with $45$ kpc for
788: the surviving satellite. More massive satellites on eccentric orbits
789: spiral in faster than low-mass ones, making themselves more vulnerable
790: to tides and full disruption.
791:
792: This is confirmed in Figure~\ref{fig:orbdec}, where we show the
793: orbital decay timescale of all satellites identified in our
794: simulations as a function of their mass. Satellite masses are shown in
795: units of the mass of the primary galaxy at the time of accretion, and
796: decay timescales, $\tau$, are normalized to the orbital period of the
797: satellite, measured at the same time. (The timescale $\tau$ is
798: computed by fitting the evolution of the apocentric distance of the
799: satellite, a good proxy for the orbital energy, to an exponential
800: law.)
801:
802: Surviving satellites are shown as filled circles in
803: Figure~\ref{fig:orbdec}, whereas open circles denote merged
804: satellites. More massive satellites clearly spiral in faster: $\tau$
805: is typically less than an orbital period for a satellite whose mass
806: exceeds $\sim 20\%$ of the primary. On the other hand, decay
807: timescales are often larger than $\sim 10$ orbital periods for
808: satellites with masses below $1\%$ of the primary. The dotted line
809: shows the $\tau \propto m^{-1}$ relation expected from simple
810: dynamical friction arguments \citep{binneyandtremaine87}. Most
811: satellites follow this trend, except perhaps for the most massive
812: systems, but this may just reflect difficulties estimating $\tau$ for
813: systems on very rapidly decaying orbits, because of poor time
814: sampling. The main result of these trends is a severe
815: underrepresentation of surviving satellites amongst massive
816: satellites, as shown by the distribution of satellite masses in the
817: bottom panel of Figure~\ref{fig:orbdec}.
818:
819: \begin{center}
820: \begin{figure}
821: \includegraphics[width=84mm]{figs/fig9.ps}
822: \caption{ Distribution of satellite stellar masses measured at the
823: time of accretion into the host halo, and normalized to the stellar
824: mass of the primary at $z=0$ (bottom panel). The shaded histogram
825: corresponds to satellites that remain self-bound at $z=0$; the other
826: histogram corresponds to satellites that merge with the primary before
827: $z=0$. The curves in the top panel indicate the cumulative fraction of
828: all {\it accreted stars} contributed by each of these two
829: populations. Note that the ``building blocks'' of the stellar halo are
830: significantly more massive than the average surviving satellite. On
831: average, accretion events bring about $25\%$ of the total number of
832: stars into the primary, $40\%$ of which remains attached to satellites
833: until $z=0$. The remainder belongs to ``merged'' satellites, the
834: majority of which make up the stellar halo. The total number of stars
835: contributed by disrupted satellites exceed those locked in surviving
836: satellites by $\sim 50\%$.}
837: \label{fig:minf}
838: \end{figure}
839: \end{center}
840:
841: \begin{center}
842: \begin{figure}
843: \includegraphics[width=84mm]{figs/fig10.ps}
844: \caption{Accretion redshift distribution of surviving (bottom panel),
845: merged (middle) and all (top) satellites in our simulations. All
846: histograms are scaled to the total number of satellites for ease of
847: comparison between panels. Dashed vertical lines indicate the
848: (average) redshift where the primary galaxy has accreted $25\%$,
849: $50\%$ and $75\%$ of its {\it total} mass at $z=0$. In each panel the
850: arrow shows the median satellite accretion redshift. The dotted curves
851: trace the cumulative distribution of satellites (by number) as a
852: function of $z_{\rm acc}$ (scale on right). Solid lines are like dotted
853: ones, but by mass. }
854: \label{fig:tinf}
855: \end{figure}
856: \end{center}
857:
858: \subsubsection{Orbital circularization}
859:
860: As they are dragged inwards by dynamical friction, the orbital energy
861: of the satellites is affected more than its angular momentum and, as a
862: result, the satellites' orbits become gradually more circular. This is
863: shown in Figure~\ref{fig:circ}, where we plot the ratio between
864: apocentric and pericentric distance, $r_{\rm per}/r_{\rm apo}$, at the
865: time of accretion versus the same quantity, but measured after
866: dynamical friction has eroded $r_{\rm apo}$ to $e^{-1}$ of its value
867: at accretion.
868:
869: As in Figure~\ref{fig:orbdec}, open and filled circles indicate
870: ``merged'' and ``surviving'' satellites at $z=0$. The vast majority of
871: the points lie above the $1$:$1$ line, indicating that the orbits have
872: become significantly less eccentric with time. Some points lie below
873: the dotted line, indicating the opposite effect; however, most of
874: these cases correspond to complex accretion where the satellite comes
875: as a member of a pair of satellites and is subject to three-body
876: interactions during accretion. (See Sales et al 2007 for further
877: details.)
878:
879: The large open and filled squares indicate the median $r_{\rm
880: per}/r_{\rm apo}$ for merged and surviving satellites,
881: respectively. Clearly, the eccentricity of the orbit is important for
882: the chances of survival of a satellite: most satellites originally on
883: very eccentric orbits have merged with the primary by $z=0$, and the
884: reverse is true for surviving satellites (see bottom panel in
885: Figure~\ref{fig:circ}).
886:
887: Satellites that merge with the primary by $z=0$ experience on average
888: a more substantial circularization of their orbits; the median $r_{\rm
889: per}/r_{\rm apo}$ evolves from $0.06$ to roughly $0.15$ in the time
890: it takes their orbital energies to decrease by $e^{-1}$. Further
891: circularization may be expected by the time that the satellite merges
892: with the primary and, under the right circumstances, a satellite may
893: even reach a nearly circular orbit before merging (see, e.g., Abadi et
894: al 2003b, Meza et al 2005).
895:
896: Orbital circularization has been proposed as an important factor to
897: consider when interpreting the effects of satellite accretion events
898: (although see Colpi et al 1999 \nocite{colpi99} for a different
899: viewpoint). Abadi et al
900: (2003b) argue, for example, that a satellite on a circularized orbit
901: might have contributed a significant fraction of the thick-disk stars
902: (and perhaps even some old thin-disk stars) of the Milky Way. A
903: further example is provided by the ``ring'' of stars discovered by the
904: SDSS in the anti-galactic center direction \citep{newberg02, yanny03,
905: helmi03}, which has been successfully modeled as
906: debris from the recent disruption of a satellite on a nearly circular
907: orbit in the outskirts of the Galactic disk \citep{penarrubia06}.
908: Since it is unlikely that the satellite formed on such orbit
909: (otherwise it would have been disrupted much earlier) its orbit has
910: probably evolved to become more bound and less eccentric as dynamical
911: friction brought the satellite nearer the Galactic disk, in agreement
912: with the trend shown in Figure~\ref{fig:circ}.
913:
914: \begin{center}
915: \begin{figure}
916: \includegraphics[width=84mm]{figs/fig11.ps}
917: \caption{ Mass fraction attached to surviving satellites at $z=0$,
918: shown as a function of radius, normalized to the virial radius of the
919: host. The open circles are the results of the dark matter-only
920: simulations of \citet{gao04b}, which are in very good agreement with
921: ours. This figure shows that, although surviving satellites have lost
922: a significant fraction of their dark mass to tides, their stellar
923: components have survived almost unscathed. Overall, satellites inside
924: the virial radius have conserved about $40\%$ of their original dark
925: mass, and $\sim 75\%$ of their stars. This suggests that stars
926: stripped off surviving satellites are in general an unimportant
927: contributor to the stellar halo, and highlights the need for
928: simulations that include gas cooling and star formation to estimate
929: the importance of tidal stripping in the satellite population. }
930: \label{fig:mrfin}
931: \end{figure}
932: \end{center}
933:
934: \subsection{Satellites and stellar halo: similarities and differences}
935:
936: The main result of the trends discussed in the preceding section is
937: the obvious mass bias present in the population of surviving
938: satellites: massive satellites merge too quickly to be fairly
939: represented amongst satellites present at any given time. This is
940: shown in the bottom panel of Figure~\ref{fig:orbdec}; although the
941: accretion of satellites with masses exceeding $10\%$ of the host (at
942: the time of accretion) is not unusual, few have survived self-bound
943: until $z=0$.
944:
945: This is also true when expressed in terms of the total stellar mass
946: that these accretion events have contributed to the simulated
947: galaxy. As shown in Figure~\ref{fig:minf}, merged satellites dominate
948: the high-mass end of the distribution of accreted satellites, and make
949: up on average $\sim 60\%$ of all accreted stars. Half of this
950: contribution comes in just a few massive satellites exceeding $10\%$
951: of the final mass in stars of the host (see upper panel in
952: Figure~\ref{fig:minf}). On the other hand, surviving satellites
953: contribute on average $\sim 40\%$ of all accreted stars and have a
954: combined stellar mass of about $12\%$ of the host at $z=0$. Half of
955: them are contributed by satellites less than $\sim 3\%$ as massive as
956: the host at $z=0$.
957:
958: Because of the strong orbital decay dependence on mass, surviving
959: satellites are also biased relative to the overall population of
960: accreted material in terms of accretion time. This is shown
961: quantitatively in Figure~\ref{fig:tinf}, which shows the $z_{\rm acc}$
962: distribution for all satellites accreted since $z=4$ (top panel). The
963: bottom and middle panels, respectively, split this sample between
964: satellites that have either survived or merged with the host by
965: $z=0$. The vertical lines in this figure illustrate the average mass
966: accretion history of the hosts in our simulation series: from left to
967: right, the vertical lines indicate the average redshift when the last
968: $25\%$, $50\%$, and $75\%$ of the mass were assembled into the virial
969: radius of the host.
970:
971: The accreted satellites, as a whole, trace very well this accretion
972: history, as may be seen from the histogram in the top panel, or by the
973: dotted line, which indicates the cumulative accretion history (scale
974: on right). Just like the total mass, half of all satellites were
975: accreted before $z\sim 1.8$ (see arrow labeled ``$50\%$''). The
976: results are quite different for ``merged satellites''; half of them
977: were actually accreted before $z=2.4$, which corresponds to a lookback
978: time of $\sim 2.7$ Gyr. Essentially no satellite accreted after $z=0.5$
979: has merged with the primary. Surviving satellites, on the other hand,
980: are substantially biased towards late accretion. Half of them were
981: only accreted after $z=1.4$, and the last $25\%$ since $z \sim 1$.
982:
983: Since stars brought into the galaxy by merged satellites contribute
984: predominantly to the stellar halo (see, e.g., Abadi et al 2006), this
985: result shows convincingly that substantial differences must be
986: expected between the stellar halo and surviving satellite population
987: in a galaxy built hierarchically. {\it The ``building blocks'' of the
988: stellar halo were on average more massive and were accreted and
989: disrupted much earlier than the population of satellites that survive
990: until the present.}
991:
992: Our results provide strong support for the semianalytic modeling
993: results of \citet{bullockandjohnston05}. Despite the differences in
994: modeling techniques (these authors use theoretical merger trees to
995: simulate Monte Carlo accretion histories and a semianalytic approach
996: to dinstinguish stars and dark matter within accreted satellites), our
997: results agree well. For example, they find that $\sim 80\%$ of the
998: stellar halo is contributed by the $\sim 15$ most massive disrupted
999: satellites; we find, on average, $70\%$. The median accretion time for
1000: disrupted satellites is $\sim 9$ Gyr ago; we find $\sim 10.5$ Gyr. Lastly,
1001: they find that the median accretion time of surviving satellites was
1002: as recently as $\sim 5$ Gyr in the past; we find $\sim 8.5$ Gyr.
1003:
1004: As discussed by \citet{font06a,font06b}, these results may help to
1005: explain the differences between the abundance patterns of halo stars
1006: in the solar neighbourhood and in Galactic dwarfs \citep{fuhrmann98,
1007: shetrone01, shetrone03, venn04}. Although stars in both
1008: the halo and satellites are metal-poor, the stellar halo is, at fixed
1009: [Fe/H], more enhanced in $\alpha$ elements than stars in the dwarfs,
1010: suggesting that its star formation and enrichment proceeded more
1011: quickly and thoroughly than in Galactic satellites. This is
1012: qualitatively consistent with the biases in the surviving satellite
1013: population mentioned above. Because of the limited numerical
1014: resolution of our simulations and our inefficient feedback recipe, we
1015: are unable to follow accurately the metal enrichment of stars in our
1016: simulations. Although this precludes a more detailed quantitative
1017: comparison between simulations and observations, we regard the
1018: distinction between satellite and stellar halo reported here as
1019: certainly encouraging.
1020:
1021: One final issue to consider is that, in principle, stars may also end
1022: up in the stellar halo as a result of partial stripping of surviving
1023: satellites. If substantial, this process might make stars in the
1024: stellar halo difficult to differentiate from those attached to
1025: satellites, despite the biases in mass and accretion time discussed
1026: above. As it turns out, stripping of surviving satellites adds an
1027: insignificant fraction of stars to the halo in our simulations; stars
1028: stripped from surviving satellites make up a small fraction ($\sim 6\%$)
1029: of all halo stars.
1030:
1031: This is shown in Figure~\ref{fig:mrfin}, where we plot the fraction of
1032: stars and dark matter that remains attached to surviving satellites as
1033: a function of the distance to the center of the galaxy. As shown by
1034: the filled triangles, more than $75\%$ of the stars brought into the
1035: system by surviving satellites remain attached to them at $z=0$. We
1036: conclude that the bulk of the halo population is not affected by stars
1037: stripped from existing satellites, and that the substantial difference
1038: between the stellar population of Galactic dwarfs and of the stellar
1039: halo predicted above is robust.
1040:
1041: \section{SUMMARY}
1042: \label{sec:conc}
1043:
1044: We have analyzed the properties of satellite galaxies formed in a
1045: suite of eight N-body/gasdynamical simulations of galaxy formation in
1046: a $\Lambda$CDM universe. Our simulations are able to resolve, at
1047: $z=0$, the $\sim 10$ most luminous satellites orbiting around $\sim
1048: L_*$ galaxies. We also track satellites that have merged with, or been
1049: disrupted fully by, the primary galaxy at earlier times, giving us a
1050: full picture of the contribution of accreted stars to the various
1051: dynamical components of the galaxy.
1052:
1053: As discussed in an earlier paper of our group (Abadi et al 2006), the
1054: stellar halo consists of stars stripped from satellites that have been
1055: fully disrupted by the tidal field of the primary. Our analysis
1056: here focuses on the spatial distribution, kinematics, and merging history
1057: of the population of surviving and merged satellites, and on their
1058: significance for the formation of the stellar halo. Our main results
1059: may be summarized as follows.
1060:
1061: \begin{itemize}
1062:
1063: \item The spatial distribution of satellites at $z=0$ is consistent
1064: with that of the dark matter in the primary galaxy's halo, and
1065: is significantly more extended than the stellar halo. On average, half of
1066: the $\sim 10$ brightest satellites are found within $0.37\, r_{\rm
1067: vir}$, comparable to the half-mass radius of the dark matter
1068: component. The half-mass radius of the stellar halo is, on the other
1069: hand, only $0.05\, r_{\rm vir}$.
1070:
1071: \item The kinematics of the satellite population is also similar to
1072: the dark matter's. Satellite velocities are mildly anisotropic in the
1073: radial direction, with $\beta_{\rm sat}\sim 0.3$-$0.4$, but not as
1074: extreme as stars in the halo, which are found to have $\beta_{\rm
1075: halo} \sim 0.6$-$0.8$ in the outskirts of the system. Satellite
1076: velocity dispersions drop from the center outwards, and decrease by
1077: about a factor of two at the virial radius from their central
1078: value. Overall, the velocity dispersion of the satellite population is
1079: found to provide a reasonable estimate of the halo's virial velocity:
1080: $\sigma_{\rm sat}/V_{\rm vir} \sim 0.9 \pm 0.2$, where the
1081: uncertainty is the rms of the eight simulations.
1082:
1083: \item The orbits of satellites evolve strongly after accretion as a
1084: result of dynamical friction with the host halo and of mass stripping
1085: by tides. More massive satellites spiral in faster than less massive
1086: systems and are disrupted quickly as they merge with the primary,
1087: adding their stars mainly to the stellar halo. The orbits of
1088: satellites with masses exceeding $10\%$ of the host mass decay on
1089: exponential timescales shorter than an orbital period, and merge
1090: shortly after accretion. Merged satellites typically make up $\sim
1091: 63\%$ of all accreted stars in a galaxy, a substantial fraction of
1092: which ($57\%$) was contributed by these few most massive satellites.
1093:
1094: \item Surviving satellites are a substantially biased tracer of the
1095: whole population of stars accreted by a galaxy. In contrast with the
1096: ``merged'' satellites that build up the halo, surviving satellites are
1097: predominantly low-mass systems that have been accreted recently. Half
1098: of the stars in the stellar halo were accreted before $z\sim 2.2$, and
1099: were in satellites more massive than $\sim 6\%$ of the host at the
1100: time of accretion. In contrast, half of the stars in surviving
1101: satellites were brought into the system as recently as $z\sim 1.6$,
1102: and formed in systems with masses less than $3\%$ of the host.
1103:
1104: \item Satellite orbits are continuously circularized by dynamical
1105: friction as they orbit within the primary's halo. The
1106: pericenter-to-apocenter ratio typically doubles once the orbital
1107: binding energy of the satellite has increased by a factor of $e$.
1108:
1109: \item Stars stripped from satellites that remain self-bound until the
1110: present make up an insignificant fraction of all stars accreted by a
1111: galaxy, showing that, once started, the disruption process of the
1112: stellar component of a satellite progresses on a very short
1113: timescale. Surviving satellites conserve at $z=0$ about $75\%$ of the
1114: stars they had at accretion time. Their surrounding dark halos, on the
1115: other hand, have been stripped of more than $\sim 40\%$ of their mass.
1116:
1117: \end{itemize}
1118:
1119: Our results offer a framework for interpreting observations of the
1120: satellite population around luminous galaxies and for extracting
1121: information regarding their dark matter halos. They also show that
1122: hierarchical galaxy formation models may explain naturally the
1123: differences in the properties of stars in the stellar halo and in
1124: Galactic satellites highlighted by recent observational work. Although
1125: our modeling of star formation is too simplistic (and our numerical
1126: resolution too poor) to allow for a closer, quantitative assessment of
1127: this issue, it is encouraging to see that, despite their differences,
1128: stellar halos and satellites may actually be both the result of the
1129: many accretion events that characterize galaxy formation in a
1130: hierarchically clustering universe.
1131:
1132: \vskip 0.5cm
1133:
1134: \section*{Acknowledgements}
1135: \label{acknowledgements}
1136:
1137: LVS and MGA are grateful for the hospitality of the Max-Planck
1138: Institute for Astrophysics in Garching, Germany, where much of the
1139: work reported here was carried out. LVS thanks financial support from
1140: the Exchange of Astronomers Programme of the IAU and to the ALFA-LENAC
1141: network. JFN acknowledges support from
1142: Canada's NSERC, from the Leverhulme Trust, and from the Alexander von
1143: Humboldt Foundation, as well as useful discussions with Simon White,
1144: Alan McConnachie, and Jorge Pe\~narrubia. MS acknowledges support by
1145: the German Science foundation (DFG) under Grant STE 710/4-1. We thank
1146: Scott Chapman and collaborators for sharing their results on
1147: Andromeda XII in advance of publication. We also thanks to the referee Andrew
1148: Benson for useful suggestions and comments on this paper.
1149:
1150: \bibliography{references}
1151:
1152:
1153: \end{document}
1154: