The luminosities of backsplash galaxies in constrained simulations of the Local Group
Alexander Knebe, Noam I Libeskind, Steffen R. Knollmann, Luis A. Martinez-Vaquero, Gustavo Yepes, Stefan Gottloeber, Yehuda Hoffman
aa r X i v : . [ a s t r o - ph . C O ] O c t Mon. Not. R. Astron. Soc. , 1–9 (2008) Printed 15 March 2018 (MN L A TEX style file v2.2)
The luminosities of backsplash galaxies in constrained simulationsof the Local Group
Alexander Knebe , Noam I Libeskind , Ste ff en R. Knollmann , Luis A. Martinez-Vaquero ,Gustavo Yepes , Stefan Gottl ¨ober , Yehuda Ho ff man Grupo de Astrof´ısica, Departamento de Fisica Teorica, Modulo C-15, Universidad Aut´onoma de Madrid, Cantoblanco E-28049, Spain Astrophysikalisches Institut Potsdam, An der Sternwarte 16, D-14482 Potsdam, Germany Racah Institute of Physics, The Hebrew University of Jerusalem, Jerusalem 91904, Israel
First draft
ABSTRACT
We study the di ff erences and similarities in the luminosities of bound, infalling and the so-called backsplash (Gill et al. 2005) galaxies of the Milky Way and M31 using a hydrody-namical simulation performed within the Constrained Local UniversE Simulation (CLUES)project. The simulation models the formation of the Local Group within a self-consistent cos-mological framework. We find that even though backsplash galaxies passed through the virialradius of their host halo and hence may have lost a (significant) fraction of their mass, theirstellar populations are hardly a ff ected. This leaves us with comparable luminosity functionsfor infalling and backsplash galaxies and hence little hope to decipher their past (and di ff er-ent) formation and evolutionary histories by luminosity measurements alone. Nevertheless,due to the tidal stripping of dark matter we find that the mass-to-light ratios have changedwhen comparing the various populations against each other: they are highest for the infallinggalaxies and lowest for the bound satellites with the backsplash galaxies in-between. Key words: methods: n-body simulations – methods: numerical – galaxies: formation –galaxies: haloes
Ever since Klypin et al. (1999) and Moore et al. (1999) pointed outthat dark matter simulations of cosmic structure formation lead toan excess of subhaloes as compared to the number of observed(luminous) satellite galaxies visibly surrounding the Milky Way(MW) and M31, the industry for simulating and studying substruc-ture in cosmological (dark matter) haloes has boomed. The tensionhas been marginally loosened with the discovery of a substantialnumber of new ultra-faint satellites galaxies in the Local Groupthanks to the SDSS data (Adelman-McCarthy 2007): within thepast couple of years the number of known MW and M31 satel-lites has nearly doubled. And taking into account the detectionlimits and the sky coverage of the SDSS survey we will most cer-tainly stumble across even more galactic satellites in the near futurewhen, for instance, upcoming data from GAIA and panSTARRShave been analysed.As noted by several groups before (Moore et al. 2004;Gill et al. 2005), there exists a prominent population of galaxiesthat are found outside the virial region of their host at the presentday, but whose orbits took them inside the virial radius at earliertimes. While their studies were based upon cosmological simula-tions of galaxy clusters, the existence of this “backsplash popu-lation” has also been found for MW-type objects (Warnick et al.2008; Ludlow et al. 2009). This raises the question whether or not (and how) one can distinguish infalling and backsplash galaxiesfrom each other. Gill et al. (2005) suggested to use the line-of-sightvelocity distribution: as shown in their Fig.8 the distribution of line-of-sight velocities of subhaloes relative to the host (and convolvedwith the 2dF velocity uncertainty of 100 km / sec) is di ff erent forthe infalling and the backsplash population. However, there maybe a simpler way that does not involve spectroscopy. Since back-splash satellites had, at one point in their orbit, a closer approachto the central galaxy than infalling satellites, the tidal influence ofthe host must have been stronger for the backsplash populationthen for infalling satellites. Could this di ff erence in tidal forces ef-fect the initial stellar population (if existent), and can it be used todiscriminate between the two populations? It has been shown byGill et al. (2005) that backsplash galaxies loose on average 40%of their initial mass when grazing their host. But what about thestellar content? As baryons are able to cluster more strongly in thecentre of the potential well the stars are also more centrally concen-trated. Therefore, will the cold baryonic component be safe fromtidal stripping when the backsplash galaxy (briefly) flies throughits host? This question is the major motivation for this work. Weaddress the issue of separating the three types of galaxies (boundsatellites, backsplash and infalling) by means of luminosity (andpossibly mass) measurements only. c (cid:13) Knebe et al.
In this Section we describe the simulations used throughout thisstudy and the methodology employed to identify host haloes andtheir substructure.
We use the same set of simulations already presented inLibeskind et al. (2010) and Knebe et al. (2010) and refer the readerto these papers for a more exhaustive discussion and presentationof these constrained simulations of the Local Group that form partof the CLUES project. However, we briefly summarize their mainproperties here for clarity.We choose to run our simulations using standard Λ CDM ini-tial conditions, that assume a WMAP3 cosmology (Spergel et al.2007), i.e. Ω m = . Ω b = . Ω Λ = .
76. We use a nor-malization of σ = .
73 and a n = .
95 slope of the powerspectrum. We used the PMTree-SPH MPI code
GADGET2 (Springel2005) to simulate the evolution of a cosmological box with sidelength of L box = h − Mpc. Within this box we identified (ina lower-resolution run utilizing 1024 particles) the position of amodel local group that closely resembles the real Local Group (cf.Libeskind et al. 2010). This Local Group has then been re-sampledwith 64 times higher mass resolution in a region of 2 h − Mpc aboutits centre giving a nominal resolution equivalent to 4096 particlesgiving a mass resolution of m DM = . × h − M ⊙ for the darkmatter and m gas = . × h − M ⊙ for the gas particles. For moredetails we refer to the reader to Gottl¨ober et al. (2010).For this particular study we focus on the gas dynamical SPHsimulation, in which we follow the feedback and star forma-tion rules of Springel & Hernquist (2003): the interstellar medium(ISM) is modeled as a two phase medium composed of hot am-bient gas and cold gas clouds in pressure equilibrium. The ther-modynamic properties of the gas are computed in the presence ofa uniform but evolving ultra-violet cosmic background generatedfrom QSOs and AGNs and switched on at z = K is also ignored. Cold gas cloud for-mation by thermal instability, star formation, the evaporation of gasclouds, and the heating of ambient gas by supernova driven windsare assumed to all occur simultaneously.
In order to identify halos and subhaloes in our simulation we haverun the MPI + OpenMP hybrid halo finder
AHF described in detailin Knollmann & Knebe (2009). AHF is an improvement of the
MHF halo finder (Gill et al. 2004), which locates local over-densities inan adaptively smoothed density field as prospective halo centres.The local potential minima are computed for each of these densitypeaks and the gravitationally bound particles are determined. Onlypeaks with at least 20 bound particles are considered as haloes andretained for further analysis (even though we place a tighter con-straint on the number of particles for the present analysis, cf. be-low). We like to stress that our halo finding algorithm automatically AMIGA halo finder, to be downloaded freely from identifies haloes, sub-haloes, sub-subhaloes, etc. For more detailson the mode of operation and actual functionality we though referthe reader to the code description paper by Knollmann & Knebe(2009).Subhaloes are defined as haloes which lie within the virialregion of a more massive halo, the so-called host halo. We buildmerger trees by cross-correlating haloes in consecutive simulationoutputs. For this purpose, we use a tool that comes with the
AHF package called
MergerTree , that follows each halo (either host orsubhalo) identified at redshift z = and is closest to it in mass. Again,for more elaborate details we point to the reader to Libeskind et al.(2010). The stellar population synthesis model STARDUST (seeDevriendt et al. 1999, and references therein for a detaileddescription) has been used to derive luminosities from the starsformed in our simulation. This model computes the spectral energydistribution from the far-UV to the radio, for an instantaneous star-burst of a given mass, age and metalicity. The stellar contributionto the total flux is calculated assuming a Kennicutt initial massfunction (Kennicutt 1998).
The prime target of this study is to find possible di ff erences inthe properties of backsplash, bound and infalling galaxies with re-spects to luminosity. We explicitly use the term “galaxies” as wefocus solely on subhaloes with a luminous stellar component; allother objects will be neglected for this particular investigation. Wefurther only consider satellites of the (simulated) MW and An-dromeda (M31) galaxy; the subhaloes of both these host haloeswill be stacked in the subsequent plots presented here. In additionto the requirement for subhaloes to contain stars we also apply alower mass cut of M > × h − M ⊙ roughly corresponding to100 particles in total (note that particles have di ff erent masses asthey represent dark matter, gas and stars). Before examining the properties of backsplash galaxies we wishto first confirm their existence. To this extent we plot in Fig.1 theclosest approach (normalized to the virial radius of the satellite’shost at the time of minimum distance) vs. its present-day distance(normalized to its host’s virial radius). The number of objects inthe respective population are given in the legend. Note that we onlyplot those subhaloes that contain a stellar component. This figurecontains three distinct parts defining the three di ff erent populations.First, those subhaloes whose minimum distance equals its present-day distance are the infalling population: they are continuouslyfalling towards their host. Second, galaxies that entered the virialradius of their host and remained inside ever since. Even thoughthe host radius is increasing in size since the time a subhalo en-tered, we nevertheless find that there are no subhaloes above the1:1 line; we therefore conclude that the increase in host radius asmeasured by R t now host / R t min host is smaller than the ratio D now / D min . Thiscomes as no surprise as we do not expect satellites to orbit on cir-cular orbits ( D now = D min ); subhaloes may have (highly) eccentric c (cid:13) , 1–9 acksplash Galaxies in the Local Group Figure 2.
The orbits of all considered subhaloes. The left panel shows the backsplash galaxies, the middle panel the bound and the right panel the infallingpopulation.
Figure 1.
Minimum distance D min as a function of present-day distance D now both normalized to the virial radius of the host at the respective time. orbits taking them close to the centre of their host (cf. Figs.7 and 8in Gill et al. 2004). Third and last, there are galaxies that once wereinside their host’s virial radius but are presently found outside, i.e.the backsplash population.Fig.1 indicates that we might expect to find of order 40% to bebacksplash galaxies in the vicinity of the Milky Way and / or M31– a percentage in agreement with previous studies of this class ofobjects(cf. Gill et al. 2005; Warnick et al. 2008). The question nowis whether or not we will be able to distinguish these populationsand find the backsplash galaxies, respectively, by quantifying theirluminosities.Before proceeding we would like to add a cautionary remarkclarifying our terminology: we call subhaloes that are inside theirhost’s virial at z =
0, “bound”. Those subhaloes that were onceinside their hosts virial radius but are found at z = z = . z =
0. The orbitshave been normalized to the virial radius of the respective host (atredshift z ) of each galaxy and hence the solid line D sat / R host = all populations werestill infalling: this can be seen at a redshift z > .
5. We will returnto this redshift later as we expect the properties of galaxies to bedrawn from the same statistical distribution at that time: no galaxyhas yet entered the host (or left again) which may (or may not) havecaused changes in the internal properties and – in particular – theluminosities.
In this section we look at the luminosity of the stellar compo-nents of galaxies identified as bound, backsplash and infalling.We start by comparing, in Fig.3, the Johnson V-Band luminos-ity of the bound satellites to the backsplash and infalling popu-lation of galaxies as well as the observational data as taken fromKoposov et al. (2008) and Macci`o et al. (2010), respectively (thinsolid line, referred to as “Maccio’ sample” from now on): thesedata are a combination of the volume corrected MW satellite lu-minosity function (Koposov et al. 2008) augmented with informa-tion from Mateo (1998) and Macci`o et al. (2010) kindly provided tous by Andrea Maccio (personal communication). And even thoughour bound luminosity function agrees with the Maccio data ratherwell, we stress that we included the observational data merely asa reference to guide the eye. It is not our prime objective to repro- c (cid:13) , 1–9 Knebe et al.
Figure 3.
The luminosity function of subhaloes in the Johnson V-Band. The“Maccio” observational data (thin solid line) is a combination of the volumecorrected MW luminosity function Koposov et al. (2008) augmented withinformation from Mateo (1998) and Macci`o et al. (2010) under the assump-tion of an NFW-like radial distributions of satellites. Note that the compar-ison to the observational data is not the prime target of this study and onlyserves as a reference, respectively.comparison pz = z = . Table 1.
Kolmogorov-Smirno ff (KS) probabilities p for various compar-isons of the luminosity functions presented in Fig.3. duce the MW and / or M31 luminosity function of satellite galaxieswith our simulation. However, a close match (as seen in Fig.3) re-assures us that our simulation is not too far fetched and that ourmethod for lighting up subhaloes (cf. Section 2.3) yields credibleresults. The central theme of this paper is the comparison betweenthe (numerically obtained) infalling and backsplash population andpossibilities to decipher them photometrically.To better quantify the di ff erences and similarities betweenthe respective simulated luminosity functions we applied theKolmogorov-Smirno ff (KS) test that provides us with the signifi-cance level p that the null hypothesis that two data sets are drawnfrom the same parent distribution; small values of p ∈ [0 ,
1] showthat the two cumulative distribution functions (i.e. in our case twoluminosity functions) are significantly di ff erent. We find that theKS probability that the backsplash and infalling distributions havebeen drawn from the same parent function is 67%. The significance We utilized the routine kstwo() as described in Press et al. (1992).
Figure 4.
The luminosity function of subhaloes in the Johnson V-Band atredshift z = . level is just 11% when comparing the backsplash with the boundpopulation and 19% when comparing the infalling with the boundsatellites. These numbers and probabilities, respectively, have beensummarized in Table 1.In addition to calculating the KS probability p that these distri-butions stem from the same parent distribution we also performedthe experiment of randomly drawing N back galaxies from the in-falling and bound sample where N back is the number of backsplashgalaxies. Comparing the resulting down-sampled luminosity func-tions again using a KS test we find a probability p of p ≈ . p ≈ .
12 when comparing the backsplash to the infalling or boundsatellites population. All this hints at similarities between back-splash and infalling satellites whereas the bound population haslikely evolved di ff erently.Since all bound and backsplash galaxies themselves were, atsome stage, infalling satellites the di ff erences between the boundand backsplash / infalling luminosity function at redshift z = z = . z = . / or di ff erences.Even though our primary motivation is to find a way to dis-tinguish backsplash from infalling satellites that only utilizes pho-tometry, we nevertheless present another (observable) correlation:the luminosity vs. the velocity dispersion σ v , in Fig.5. As alreadyalluded to above when discussing the luminosity function, we also c (cid:13) , 1–9 acksplash Galaxies in the Local Group Figure 5.
The relation between Johnson V-Band luminosity and subhalovelocity dispersion at redshift z =
0. The observational data is taken formWalker et al. (2009). added observational data (taken from Walker et al. (2009, their Ta-ble 1)) simply to guide the eye. While we also recover the ob-served trend in our numerical data, the focus should lie with theinfalling and backsplash galaxies. To better quantify the correla-tions between σ v and M V we calculated the Spearman rank coef-ficients R S : for the observational data it amounts to R S = . R S = . ff erent magnitude limits of both the observa-tional and numerical data, i.e. the two data sets do only cover thesame magnitudes in the range M V ∈ [ − , − ffi cients R S for the backsplashand infalling populations are R S = .
373 and R S = .
573 respec-tively. While there are di ff erences in the strengths of the correla-tion we find it di ffi cult to utilize this interdependence to separatebacksplash from infalling satellites: while the Spearman rank sig-nificances S S are very close to zero for the bound and observationaldata (indicating a reliable determination of the respective R S value)they are of order 0.2 for the backsplash and infalling population,probably due to the small statistical sample.Above, we showed that while all three populations do followthe same trend for the M V − σ v relation (coinciding with the trendfound in observational data), there are nevertheless subtle di ff er-ences in the strength of this correlation, especially for the back-splash and infalling population (cf. the di ff erent Spearman rank co-e ffi cients R S ). However, the most prominent and well pronounceddi ff erence can be found when studying the mass-to-light ratios M / L V presented in Fig.6 as a function of V-band magnitude M V .We stress that the mass M used in this plot is actually the masswithin the visible radius of the subhalo; we found the distance of the The Spearman rank coe ffi cient R S is a non-parametric measure of cor-relation: it assesses how well an arbitrary monotonic function describesthe relationship between two variables, without making any other assump-tions about the particular nature of the relationship between the variables(Kendall & Gibbons 1990). Its significance S S is a value between 0 and 1and a small value indicates a significant correlation. We use the IDL routine R CORRELATE() to calculate both these numbers.
Figure 6.
Mass-to-light ratios (in terms of solar values) as a function ofV-band luminosity M V . The thin solid line represents the observational re-lation as found by Mateo (1998). The other lines are the best-fit curves(with the amplitude as a free parameter) to the bound (dashed), backsplash(dotted), and infalling (dot-dashed) population, respectively, with the leg-end listing the respective value of the amplitude, too. Note that we used the“mass M inside the visible radius” as described in the text for this plot. farthest stellar particle and used the total mass interior to this radiusas M . Wadepuhl & Springel (2010) already noted that a (substan-tial) shift (i.e. A ≈ .
2) of the observationally determined analyticalrelation M / L ( M / L ) ⊙ = A . + L / L ⊙ ! (1)is required (Mateo 1998, A = A as a free parameter and using only the bound,backsplash, and infalling satellites we find A = . ± . A = . ± . A = . ± . Thesedi ff erent amplitudes are naturally explained by the di ff ering histo-ries and (strengths of) interactions with the host. We will see be-low in Section 3.3 that bound galaxies lost the largest amount oftheir dark matter when compared with the other two populations;infalling satellites are in fact still gaining mass through accretion.Therefore, taken together with the fact that their luminosities arenevertheless still similar, we may infer that the mass-to-light ratiosshould be significantly di ff erent. This notion opens up the possi-bility to use the relation presented in Fig.6 to separate the threepopulations from each other. In practice this requires not only pho-tometric measurements but also a (proxy for the) mass estimation.However, using the mass inside the stellar radius also may ex-plain the di ff erences found in Fig.6: stars in real satellites may bemore compact relative to the dark matter than in our simulation,and might therefore be less susceptible to tidal stripping (togetherwith the dark matter inside the “visible” radius). This would alsosuggest that the di ff erences between our three di ff erent populationsmight be smaller if the luminous parts of the satellites were morecompact. The fitting to Eq.(1), i.e. a function M / L ( L ), has been done using IDL’s CURVEFIT routine using equal weights for the data points M / L vs. M V ; thereported standard deviations had been returned by CURVEFIT , too.c (cid:13) , 1–9
Knebe et al.
Figure 7. Di ff erence between Johnson V-Band luminosity at present dayand redshift z = . M . We further like to mention that we not only used the mass in-side the stellar radius as a measure for the mass entering the mass-to-light ratio. We also applied various other definitions, e.g. the to-tal mass inside the virial radius as well as the mass as determinedfrom the velocity dispersion under the assumption of virialisationand an NFW density profile (both at the virial radius and at 15%of the virial radius). While the amplitudes A are certainly di ff er-ent when using di ff erent mass estimates, the general trend remainsunaltered: the M / L ratios for the infalling satellites are shifted up-wards with respects to the backsplash population which itself hashigher ratios than the bound subhaloes.However, we also need to bear in mind one of the subtleties ofhalo finding, especially subhalo finding: the definition of mass andthe edge of a subhalo, respectively. While it is straight forward todefine an outer edge for an (isolated) field halo (usually defined asthe radius at which the mean interior density drops below 200 timesthe critical density), the situation is more tricky for subhaloes: theyhave to be truncated at the point where their density profile startsto rise again due to the host’s background density. Therefore, thesame subhalo placed inside and outside of a host will have di ff er-ent masses due to the nature of (our) halo finding technique. Thisexplains at least in part the o ff set in the mass-to-light ratios forbound / backsplash and infalling galaxies: the infalling ones havein general higher masses. And part of the gap between bound andbacksplash may be explained by the same phenomenon, though notall of it; there certainly is no uncertainty that bound galaxies havelost more mass than backsplash subhaloes.The di ff erences between the luminosities of the populationsat redshift z = z = .
5) calls for a closer look at the evolution of satellite galaxyluminosity. To this extent we plot in Fig.7 the di ff erence betweenthe Johnson V-Band luminosity M V at redshift z = z = . M for eachgalaxy considered in this paper, using di ff erent symbols for the dif-ferent populations (stars for backsplash, plus-signs for bound, anddiamonds for infalling galaxies).Fig.7 now shows several things. For a substantial number ofsatellites (especially the backsplash and infalling population) we Figure 8.
Ratio of stellar to gas mass at redshift z = . M . only observe a “constant” decrease in luminosity of approx. 1.5magnitudes. However, the luminosity of the bound galaxies dropssignificantly – especially on the low-mass end – while some of thehigher mass ones gain luminosity. As luminosity is directly linkedto stellar content we are left with the question of how these di ff er-ences relate to changes in the stellar population and / or removal (orgain) of star particles from a subhalo. We study these issues in thefollowing subsection.Studying luminosities is also closely related to colours, i.e.ratios of luminosities in di ff erent wave-bands. It therefore appearsnatural to ask the question - and use the data available to us - tohave a closer look at di ff erences in colours for our three popula-tions. When plotting the B − V colour as a function of halo mass M (not presented here) we observe that there are practically no di ff er-ences at all amongst the various subhaloes and populations, respec-tively. Neither is there are correlation with mass. Colour appearsuna ff ected when categorizing galaxies as bound, infalling or back-splash. Before investigating the stellar component directly we would liketo start with a few words on the subhalo gas content. We find thathardly any of the subhaloes under consideration contain a signif-icant gas content at redshift z =
0. When expressed in terms ofthe cosmic baryon fraction the amount of mass in gas is of order < − for more than 90% of the subhaloes. However, their stel-lar mass fractions (again in terms of the cosmic baryon fraction) is > − for all of them (which is a direct outcome of restricting our-selves to a sample of subhaloes that contain a stellar component).The situation though is rather di ff erent at redshift z = . x -axis and the ratio ofstellar-to-gas mass at redshift z = . y -axis. The fact thatnone of the gas is left at redshift z = / removed through interactions with the c (cid:13) , 1–9 acksplash Galaxies in the Local Group Figure 9.
Ratio of stellar mass at present day and redshift z = . M . host halo (e.g. ram pressure stripping) or other influences prior toinfall. If the former is true we would expect to observe an increasein stellar mass since redshift z = .
5, unless there is a conspiracyat work: the existing stars may be stripped at the same rate as starformation may convert gas into new stars, leaving the number ofstars unchanged. However, this scenario is rather unlikely. We havealso checked for the influence of the cosmological UV background:recall that in our simulations the thermodynamic properties of thegas are computed in the presence of a uniform but evolving UVcosmic background generated from quasi-stellar objects and activegalactic nuclei and switched on at z = M c ( z ) as given by Eq. (6) in Hoeft et al.(2006). When plotting the mass accretion histories of all our sub-haloes under investigation here and comparing it to the aforemen-tioned formula in Hoeft et al. (2006), we find that all the backsplashand infalling galaxies are in fact below the evaporation limit. Forthe bound objects we find that 1 / z = .
5, abovethat mass limit. However, they also drop below it by z = z =
0, due to photo-evaporation by the UV background.The question now is, whether or not we find an evolution ofthe stellar component between redshifts z = . z =
0. Wetherefore plot the ratio of the stellar content at these redshifts as afunction of (present-day) subhalo mass in Fig.9. We observe thatthe backsplash (as well as the infalling) subhaloes hardly lost anystars since z = .
5. Note that our simulations do not model stel-lar mass loss and hence the stellar mass remains constant when nostar particles are stripped or newly created. However, this is still inagreement with the evolution of the luminosity as found in Fig.7:subhaloes with a constant number of stellar particles merely evolvepassively from z = . z = ffi ciently stripped, the stellar component remainedmore or less una ff ected – at least for the infalling and backsplashpopulation which are of prime interest in the present study. Figure 10.
Ratio of total bound mass at present day and at redshift z = . M . As we expect the stellar component to be concentrated at thecentre of a subhalo, the previous finding on stellar mass loss forbound subhaloes immediately leads to questions regarding the na-ture of mass loss in general. To this extent we show in Fig.10the ratio of total bound masses M (again as a function of today’smass) at redshifts z = z = .
5. We note that outside theinfluence of a host halo, subhaloes behave like field haloes andgrow in mass through accretion processes; this is clearly confirmedfor the infalling population (albeit with the exception of two ob-jects). We also observe mass loss via tidal stripping, especially forthe bound subhaloes. And while backsplash galaxies may at timesloose as much as 40% of their original mass (Gill et al. 2005) wealso find the odd backsplash galaxy in our particular sample thatgained mass. Nevertheless, the picture is more or less clear: eventhough backsplash galaxies loose mass, their stellar component re-mains una ff ected. This is not the situation for bound galaxies thatloose both dark matter and stars due to the tidal interactions withthe host, as expected. The picture drawn here therefore naturallyexplains the di ff erences in the (amplitude of the) M / L V ratios eventhough for that particular study only the “mass inside the visibleradius” has been considered: when using the total bound mass (notpresented here) we recover the same relations amongst the di ff erentsubhalo populations with the ratios in amplitude unchanged; how-ever, the absolute value of the amplitudes is more than a factor twohigher. In this study we set out to examine the di ff erences of the lumi-nosities of backsplash, bound, and infalling satellite galaxies in aconstrained cosmological hydrodynamical simulation of the LocalGroup. Our prime question is: Is it possible to distinguish thesedi ff erent population by mere photometry? While we find marginaldi ff erences in the bound vs. backsplash / infalling galaxies, the twopopulations residing in the outskirts of the host halo appear to havestrikingly similar properties in terms of luminosity. The time back-splash subhaloes spent under the influence of the host is thereforenot long enough to a ff ect the stellar component: they loose mass,but primarily dark matter and / or gas particles are stripped - the star c (cid:13)000
5. We note that outside theinfluence of a host halo, subhaloes behave like field haloes andgrow in mass through accretion processes; this is clearly confirmedfor the infalling population (albeit with the exception of two ob-jects). We also observe mass loss via tidal stripping, especially forthe bound subhaloes. And while backsplash galaxies may at timesloose as much as 40% of their original mass (Gill et al. 2005) wealso find the odd backsplash galaxy in our particular sample thatgained mass. Nevertheless, the picture is more or less clear: eventhough backsplash galaxies loose mass, their stellar component re-mains una ff ected. This is not the situation for bound galaxies thatloose both dark matter and stars due to the tidal interactions withthe host, as expected. The picture drawn here therefore naturallyexplains the di ff erences in the (amplitude of the) M / L V ratios eventhough for that particular study only the “mass inside the visibleradius” has been considered: when using the total bound mass (notpresented here) we recover the same relations amongst the di ff erentsubhalo populations with the ratios in amplitude unchanged; how-ever, the absolute value of the amplitudes is more than a factor twohigher. In this study we set out to examine the di ff erences of the lumi-nosities of backsplash, bound, and infalling satellite galaxies in aconstrained cosmological hydrodynamical simulation of the LocalGroup. Our prime question is: Is it possible to distinguish thesedi ff erent population by mere photometry? While we find marginaldi ff erences in the bound vs. backsplash / infalling galaxies, the twopopulations residing in the outskirts of the host halo appear to havestrikingly similar properties in terms of luminosity. The time back-splash subhaloes spent under the influence of the host is thereforenot long enough to a ff ect the stellar component: they loose mass,but primarily dark matter and / or gas particles are stripped - the star c (cid:13)000 , 1–9 Knebe et al. particles remain more or less una ff ected by the host’s tidal field.Therefore, their luminosity function and luminosities in general re-main akin to the infalling population.Nevertheless, when allowing for not only photometric infor-mation but also adding “mass” to our analysis, we found that themass-to-light ratios (as a function of magnitude) are significantlyhigher in infalling than in backsplash galaxies, which are in turnboth higher than for bound satellites. Fitting the observationallydetermined relation presented in Mateo (1998) for M / L V vs . M V byleaving the amplitude as a free parameter we find di ff erences in theamplitude of a factor of 1.5 and 2.5 between backsplash and boundand infalling and backsplash galaxies, respectively. We note, how-ever, that part of this shift can be explained by (our method of) halofinding and certain endemic limitations when comparing field andsubhaloes. The radial extent of a subhalo has to be truncated due tothe embedding within the host’s background and hence has a lowermass than in the case when the same subhalo is found in isolation(i.e. exterior to a host halo) even though we explicitly only con-sidered the “mass inside the visible radius” for this particular partof the investigation. We also need to acknowledge that the originalrelation had to be shifted by a factor of 7.2 to bring it into agree-ment with our numerical data which nevertheless is not the primetarget of the present paper and its explanation left to a future study,respectively.Even though there still remains a lot to be quantified, we be-lieve that this di ff erence may provide a new window on distinguish-ing between infalling and backsplash galaxies that could be appliedto observational data. Its origin is readily explained by the fact thatwhile backsplash and bound galaxies both lose mass the mass lossis greater for bound than for backsplash galaxies; therefore, if thestellar population is una ff ected (as found in our simulations) weobserve an enhanced decrease in the mass-to-light ratios for boundgalaxies and – more importantly for our purposes – a decrease whencomparing infalling against backsplash.However, the apparent discrepancy between the simulationpresented and used here and the observational data yet remainsunexplained. It could be possible that stars in real satellites aremore compact relative to the dark matter than in our simulation,and might therefore be less susceptible to tidal stripping (togetherwith the dark matter inside the “visible” radius). But this wouldalso suggest that the di ff erences between our three di ff erent popu-lations might be smaller if the luminous parts of the satellites weremore compact closing the aforementioned “window” again. In thatregards, we remind the reader that we not only used the mass in-side the stellar radius as a measure for the mass entering the mass-to-light ratio. We also applied various other possibilities, e.g. thetotal mass inside the virial radius as well as the mass as determinedfrom the velocity dispersion under the assumption of virialisationand an NFW density profile (both at the virial radius and at 15% ofthe virial radius). While the ratios of the M / L curves are certainlydi ff erent when using di ff erent mass estimates, the forcited trend re-mained unaltered.A closer inspection of Fig.2 reveals that most of the back-splash galaxies fell into their host at approximately the same time.When studying the distribution of infall times (not shown herethough) there appears to be a continuous infall of bound galax-ies whereas the backsplash objects all cluster at about redshift z ≈ .
55. As pointed out by several other authors recently, subhaloes Note that Wadepuhl & Springel (2010) also required a shift by a factor of5.2 for their simulation data. may have the tendency to fall into (Milky Way like) hosts in groups(cf. Klimentowski et al. 2010; Li & Helmi 2009; D’Onghia & Lake2008; Li & Helmi 2008). Hence could it be that all our backsplashgalaxies are part of a larger group? We explicitly checked for thisconjecture by studying their 3D orbits and cannot confirm it: ourbacksplash galaxies come from various directions yet fall in at asimilar redshift. However, we also need to acknowledge that thesedirections are not random but rather correlated – however, this hasbeen studied in detail in a companion paper Libeskind et al. (2010).
ACKNOWLEDGEMENTS
AK is supported by the Ministerio de Ciencia e Innovacion(MICINN) in Spain through the Ramon y Cajal programme and fur-ther acknowledges support by the Ministerio de Education (MEC)grant AYA 2009-13875-C03-02. SRK acknowledges support by theMICINN too under the Consolider-Ingenio, SyeC project CSD-2007 -00050. We thank DEISA for granting us supercomputingtime on MareNostrum at BSC and in SGI- Altix 4700 at LRZ, torun these simulations under the DECI- SIMU-LU and SIMUGAL-LU projects. We acknowledge support of MICINN through theConsolider-Ingenio 2010 Programme under grant MULTIDARKCSD2009-00064. We also thank ASTROSIM for giving us di ff er-ent travel grants to visit our respective institutions. GY acknowl-edges financial support from MEC (Spain) under project AYA2009-13875-C03-02 and the ASTROMADRID project financed byComunidad de Madrid. We thank Andrea Maccio for kindly provid-ing us with the data of the observed luminosity function (averageof MW and M31). REFERENCES
Adelman-McCarthy J. K. e. a., 2007, ApJS, 172, 634Devriendt J. E. G., Guiderdoni B., Sadat R., 1999, A&A, 350, 381D’Onghia E., Lake G., 2008, ApJ, 686, L61Gill S. P. D., Knebe A., Gibson B. K., 2004, MNRAS, 351, 399Gill S. P. D., Knebe A., Gibson B. K., 2005, MNRAS, 356, 1327Gill S. P. D., Knebe A., Gibson B. K., Dopita M. A., 2004, MN-RAS, 351, 410Gottl¨ober S., Ho ff man Y., Yepes G., 2010, in S. Wagner, M. Stein-metz, A. Bode, M.M. M¨uller ed., High Performance Computingin Science and Engineering Springer, Constrained Local Uni-versE Simulations (CLUES). pp 309–323Haardt F., Madau P., 1996, ApJ, 461, 20Hoeft M., Yepes G., Gottl¨ober S., Springel V., 2006, MNRAS,371, 401Kendall M., Gibbons J. D., 1990, Rank Correlation Methods, 5edn. A Charles Gri ffi n TitleKennicutt Jr. R. C., 1998, in G. Gilmore & D. Howell ed., TheStellar Initial Mass Function (38th Herstmonceux Conference)Vol. 142 of Astronomical Society of the Pacific Conference Se-ries, Overview: The Initial Mass Function in Galaxies. pp 1– + Klimentowski J., Łokas E. L., Knebe A., Gottl¨ober S., Martinez-Vaquero L. A., Yepes G., Ho ff man Y., 2010, MNRAS, 402, 1899Klypin A., Kravtsov A. V., Valenzuela O., Prada F., 1999, ApJ,522, 82Knebe A., Libeskind N. I., Knollmann S. R., Yepes G., Gottlø”berS., Ho ff man Y., 2010, ArXiv e-printsKnollmann S. R., Knebe A., 2009, ApJS, 182, 608 c (cid:13) , 1–9 acksplash Galaxies in the Local Group Koposov S., Belokurov V., Evans N. W., Hewett P. C., Irwin M. J.,Gilmore G., Zucker D. B., Rix H., Fellhauer M., Bell E. F.,Glushkova E. V., 2008, ApJ, 686, 279Li Y., Helmi A., 2008, MNRAS, 385, 1365Li Y., Helmi A., 2009, in J. Andersen, J. Bland-Hawthorn, &B. Nordstr¨om ed., IAU Symposium Vol. 254 of IAU Symposium,Group infall of substructures on to a Milky Way-like dark halo.pp 263–268Libeskind N. I., Knebe A., Ho ff man Y., Gottloeber S., Yepes G.,Steinmetz M., 2010, ArXiv e-printsLibeskind N. I., Yepes G., Knebe A., Gottl¨ober S., Ho ff man Y.,Knollmann S. R., 2010, MNRAS, 401, 1889Ludlow A. D., Navarro J. F., Springel V., Jenkins A., Frenk C. S.,Helmi A., 2009, ApJ, 692, 931Macci`o A. V., Kang X., Fontanot F., Somerville R. S., KoposovS., Monaco P., 2010, MNRAS, 402, 1995Mateo M. L., 1998, ARA&A, 36, 435Moore B., Diemand J., Stadel J., 2004, in Diaferio A., ed., IAUColloq. 195: Outskirts of Galaxy Clusters: Intense Life in theSuburbs On the age-radius relation and orbital history of clustergalaxies. pp 513–518Moore B., Ghigna S., Governato F., Lake G., Quinn T., Stadel J.,Tozzi P., 1999, ApJ, 524, L19Press W. H., Teukolsky S. A., Vetterling W. T., Flannery B. P.,1992, Numerical recipes in C. The art of scientific computing.Cambridge: University Press, —c1992, 2nd ed.Spergel et al. D. N., 2007, ApJS, 170, 377Springel V., 2005, MNRAS, 364, 1105Springel V., Hernquist L., 2003, MNRAS, 339, 289Wadepuhl M., Springel V., 2010, ArXiv e-printsWalker M. G., Mateo M., Olszewski E. W., Pe˜narrubia J., WynEvans N., Gilmore G., 2009, ApJ, 704, 1274Warnick K., Knebe A., Power C., 2008, MNRAS, 385, 1859This paper has been typeset from a TEX / L A TEX file prepared by theauthor. c (cid:13)000