Triggering of merger-induced starbursts by the tidal field of galaxy groups and clusters
aa r X i v : . [ a s t r o - ph ] D ec Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 12 November 2018 (MN L A TEX style file v2.2)
Triggering of merger-induced starbursts by the tidal fieldof galaxy groups and clusters
M. Martig ⋆ and F. Bournaud Laboratoire AIM, CEA/DSM - CNRS - Universit´e Paris Diderot. Dapnia/SAp, 91191 Gif-sur Yvette, France
Accepted xxx. Received xxx; in original form xxx
ABSTRACT
Star formation in galaxies is for a part driven by galaxy mergers. At low redshift,star formation activity is low in high-density environments like groups and clusters,and the star formation activity of galaxies increases with their isolation. This starformation – density relation is observed to be reversed at z ∼
1, which is not explainedby theoretical models so far. We study the influence of the tidal field of a galaxygroup or cluster on the star formation activity of merging galaxies, using N-bodysimulations including gas dynamics and star formation. We find that the merger-driven star formation is significantly more active in the vicinity of such cosmologicalstructures compared to mergers in the field. The large-scale tidal field can thus enhancethe activity of galaxies in dense cosmic structures, and should be particularly efficientat high redshift before quenching processes take effect in densest regions.
Key words: galaxies: evolution – galaxies : interactions – galaxies: starburst – galax-ies: clusters: general
In the Local Universe ( z ≃ z ≃ ⋆ E-mail: [email protected] regions of the field. In clusters, star formation is even lessactive, which is explained by a variety of phenomena includ-ing the ram-pressure stripping (Quilis et al. 2000), galaxyharassment (Moore et al. 1996), and galaxy strangulation(Kawata & Mulchaey 2007).Cosmological ΛCDM models (Millenium, Springel et al.2005) explain the Local star formation activity – environ-mental density relation, but predict that it gets reversed onlyat very high redshift z >
2. Yet, it has been recently discov-ered that the star formation – density relation is already re-versed at z ≃
1, where the star formation activity of galaxiesincreases with the local density of the surrounding galaxies,except in the very densest regions (Elbaz et al. 2007, Cooperet al. 2007). This reversal of the star formation-density re-lation at z ∼ c (cid:13) M. Martig and F. Bournaud tidal tails developed in galaxy mergers, but did not studythe star formation in this context. Here we show that galaxymergers are statistically more efficient in triggering strongstarbursts if they take place in the vicinity of a larger struc-ture. This should contribute to the triggering of star forma-tion in dense environments at z ∼
1. Indeed the triggeringof star formation by the large-scale tidal field should be ef-ficient in particular in young groups and near forming clus-ters, where the quenching factors did not have time to actyet. The numerical simulations are described in Section 2,and the results are presented in Section 3. Our conclusionsare discussed and summarized in Sections 4 and 5.
Galaxies are modelled as stars, gas and dark matter parti-cles. The gravitational potential is computed with an FFT-based particle-mesh technique, with a spatial resolution andsoftening of 190 pc, as described in Bournaud & Combes(2002, 2003). Gas dynamics is modelled with a sticky-particles scheme with parameters β r =0.7 and β t =0.5. Starformation is computed using a local Schmidt law : the starformation rate is proportional to the local gas density to theexponent 1.5 (Kennicutt 1998).In order to directly compare the SFR of an interactinggalaxy and of the same galaxy isolated, at fixed gas mass, weimplement a gas disc of 10 particles in this galaxy, togetherwith 7 × stellar particles and 5 × dark matter parti-cles. The particle mass resolution is 1.5 × M ⊙ for gas,1.8 × M ⊙ for stars, 10.8 × M ⊙ for dark matter. Themerging companion is modelled with 7 × stellar particlesand 5 × dark matter particles. Because we study star for-mation mainly in a context of z >
0, we model galaxies witha moderate visible mass of 1 . × M ⊙ . At z = 0 this cor-responds to somewhat small spirals (similar to M 33). Thebulge:disc mass ratio is 0.24, the gas mass fraction in thedisc is 15%. The bulge has a scale-length of 600 pc, the dischas a Toomre profile with a radial scale-length of 1.6 kpcfor stars and 4.6 kpc for gas, truncated at 5.6 kpc. A darkhalo of mass 5 . × M ⊙ is implemented with a Plummerprofile of scale-length 6 kpc truncated at 20 kpc, giving acircular velocity V circ ≃
100 km s − .Galaxies are evolved as isolated systems for 500 Myrbefore simulations are started. This way, galaxies alreadyhave acquired a realistic (barred) spiral structure when theyinteract and merge, without having time to undergo a ma-jor secular evolution of their bulge mass or disc size. Starformation is shut down during this initial period, so that in-teractions really start with the gas fraction indicated above.The evolution of the SFR is thus related to the interac-tion/merger, without any bias introduced by the transitionfrom the initial axisymmetric model to a realistic spiral disc. We performed simulations of binary equal-mass mergers oftwo spiral galaxies corresponding to the model above. Theorbital parameters of the merging pair are as follows: • the inclination i of the orbital plane with respect to each galactic disc was fixed to 33 degrees. This is the averagevalue R i p ( i ) di , the probability of an inclination i being p ( i ) ∝ i in spherical geometry. This way we model typicalorbits that are not coplanar nor polar. • the initial velocity V was varied to 0.2, 0.4, 0.6 and 0.8,in units of the circular velocity V circ . • the impact parameter b (defined as the initial separationin the direction perpendicular to the initial velocity) wasvaried to 3, 4, 5, and 6 times the gas disc radius. • the orientation was varied to prograde and retrograde.We model groups and clusters gravitational potentialusing a Plummer profile. This choice is discussed in Sec-tion 3.3, and should be representative of most groups andclusters tidal field at least in the peripheral regions studiedhere. The modelled cluster has a mass of 10 M ⊙ , and aradial scale-length of 400 kpc (this choice would be reason-ably representative for instance of the Virgo cluster (e.g.,Fouqu´e et al. 2001) and the group a mass of 5 × M ⊙ and a scale-length of 150 kpc, which could be representativeof the Local Group depending on its dark:visible ratio.The galaxy pair was initially placed at 400 kpc from thecentre of the cluster (resp. 150 kpc from the group centre).Galaxies are not placed specifically in central regions, but inthe periphery. This is a more general choice, and in the caseof clusters it ensures that galaxies there can still contain gasreservoirs and form stars.We chose four possible configurations for the relativeposition of the galaxy pair and the group or cluster: • configuration 1: the group/cluster centre is in the or-bital plane, along the axis supporting the initial relative ve-locity of the galaxy pair. • configuration 2: the group/cluster centre is in the or-bital plane, along the axis perpendicular to the initial rela-tive velocity of the galaxy pair. • configuration 3: the group/group centre is in the orbitalplane, along the bisector of the two previous directions. • configuration 4: the group/cluster centre is at 45 de-grees from the orbital plane, with a projected position inthe orbital plane similar to configuration 3.Each orbital parameter for the merging galaxies hasbeen simulated without any external field, and with thegroup and the cluster in each configuration; the total num-ber of cases is then as large as 320. We restrict ourselves tocases leading to a merger, otherwise the parameter space toexplore would be too large. We make the choice of the galaxypair having no initial velocity w.r.t. the group/cluster (butfree to move within in) as justified in Sect. 3.2. In the following, ‘relative SFR’ refers to the SFR of a galaxywith a merging companion and/or a group or cluster tidalfield, divided by the SFR of the same galaxy isolated. Thestarbursts are described with the maximum value of the rel-ative SFR, i.e. SFR peak /SFR isol on Fig. 2.
We show in Fig. 1 the evolution of the relative SFR for asingle galaxy in the group and cluster tidal fields (without c (cid:13) , 000–000 tar formation triggering near groups and clusters Time (Myr) R e l a t i v e S F R single galaxy with cluster (1)single galaxy with group (2)isolated galaxy Figure 1.
SFR of a single galaxy with a cluster in configuration1 or a group in configuration 2, relative to the SFR of the samegalaxy without the external field. any interacting companion) for two of the simulated con-figurations. The average relative SFR is in both cases onlyslightly larger than 1. Thus, the tidal field of the cluster orthe group can weakly trigger the star formation in a singlegalaxy, but without driving significant starbursts as mergerscould do.
The effect of the group or cluster tidal field on star formationcan be larger when we replace the single galaxy by a mergingpair. Before the systematical statistical study of Sect. 3.3,we show here an example of galaxy pair in the cluster field.Fig. 2 shows the relative SFR of a galaxy merging with anequal mass companion on a direct orbit with V = 0 . V circ and b = 4 R disc , with and without the cluster tidal field. Thegalaxy merger induces a starburst, the intensity of which issignificantly affected by the cluster field: the peak intensityof the relative SFR is increased by the presence of the clus-ter, at various levels depending on the configuration. Thetriggering can be quite significant, with for example on Fig.2 a peak SFR increased by a factor three when the clusteris in configuration 1, compared to the merging pair withoutthe cluster field.We tested in such simulations the influence of the ini-tial velocity of the galaxy pair w.r.t the group/cluster, start-ing with radial and tangential velocities of 100 km.s − and300 km.s − . The change in the SFR is found to be withinfive per cent. This weak influence justifies our choice to notvary systematically the initial velocity of the merging galaxypair w.r.t the group/cluster. To perform a statistical analysis on our whole simulationsample, we take into account the fact that the various config-urations leading to mergers that we simulated have differentlikelihoods, and must then be weighted accordingly. Withinthe simple assumption of a random distribution of compan-ions, the collision rate varies as the velocity and the cross-section πb for an impact parameter b . Thus, each simulationshould be attributed a probability ∝ b V f ( V ) (e.g., Mihos
100 300 500 700
Time (Myr) R e l a t i v e S F R isolated galaxymerger without clustermerger with cluster (1)merger with cluster (2)merger with cluster (3) SFR peak
SFR isol
Figure 2.
Comparison of the SFR (relative to the isolated case)for mergers without and with a cluster, in different configurations.Parameters are described in text (Sect. 3.2). The thin blue curvesshow SFR evolutions in the target galaxy when the perturbinggalaxy also contains gas; the red curves show test simulationswith a mass resolution of dark matter particles enhanced by afactor 3. f ( V ) is the velocity distribution of galaxies.The distribution f is generally unknown. The real distribu-tion f ( V ) should increase with V (in particular in clusters,and for the moderate velocities leading to mergers that westudy). In the following we present the results for f ( V ) = 1and f ( V ) ∝ V , assuming that the real distribution is likelyin between. The variations caused by different assumptionson f ( V ) are anyway small (see values below), and we mainlyfocus on the f ( V ) = 1 assumption that is found to give aconservative limit to the final result, the conclusions beingquantitatively stronger (in minor proportions) if we assumean increasing form for f .We show on Fig. 3 the statistical distribution of themaximum relative SFR (SFR peak /SFR isol on Fig. 2) formerging galaxy pairs with/without the external gravita-tional potential. We notice that the major mergers are sig-nificantly more efficient to trigger star formation if they takeplace in the gravitational field of a cluster or a group. Indeed,for a merging pair in the field, the fraction of ‘significant’bursts (SFR multiplied by a factor 2 at least compared tothe isolated reference disc) is 45% for f ( V ) = 1 (respectively35% for f ( V ) = V ). This fraction becomes 81% (resp. 85%)for the model group potential and 90% (resp. 92%) in the pe-riphery of the model cluster. Similarly for ‘strong’ starbursts(say, SFR multiplied by at least a factor 5), only 6% (3%) ofthe mergers without any external potential reach this level,while this fraction is increased to 11% (21%) in a group tidalfield as well as in a cluster tidal field. Thus, starbursts atall levels are triggered by the presence of the group/clusterpotential: over our sample of mergers, the maximal instan-taneous SFR is on average doubled, depending on the shapeof f ( V ), compared to mergers in the field far from groupsand clusters. The enhancement is even larger in some cases,in particular for the highest initial velocities. This trigger-ing effect is also important if we consider the integrated starformation over the burst duration (i.e. the total mass ofstars formed) that is on average multiplied by 1.3–1.4 in thevicinity of both the model group and cluster.We thus find a noticeable enhancement of the star for- c (cid:13)000
Comparison of the SFR (relative to the isolated case)for mergers without and with a cluster, in different configurations.Parameters are described in text (Sect. 3.2). The thin blue curvesshow SFR evolutions in the target galaxy when the perturbinggalaxy also contains gas; the red curves show test simulationswith a mass resolution of dark matter particles enhanced by afactor 3. f ( V ) is the velocity distribution of galaxies.The distribution f is generally unknown. The real distribu-tion f ( V ) should increase with V (in particular in clusters,and for the moderate velocities leading to mergers that westudy). In the following we present the results for f ( V ) = 1and f ( V ) ∝ V , assuming that the real distribution is likelyin between. The variations caused by different assumptionson f ( V ) are anyway small (see values below), and we mainlyfocus on the f ( V ) = 1 assumption that is found to give aconservative limit to the final result, the conclusions beingquantitatively stronger (in minor proportions) if we assumean increasing form for f .We show on Fig. 3 the statistical distribution of themaximum relative SFR (SFR peak /SFR isol on Fig. 2) formerging galaxy pairs with/without the external gravita-tional potential. We notice that the major mergers are sig-nificantly more efficient to trigger star formation if they takeplace in the gravitational field of a cluster or a group. Indeed,for a merging pair in the field, the fraction of ‘significant’bursts (SFR multiplied by a factor 2 at least compared tothe isolated reference disc) is 45% for f ( V ) = 1 (respectively35% for f ( V ) = V ). This fraction becomes 81% (resp. 85%)for the model group potential and 90% (resp. 92%) in the pe-riphery of the model cluster. Similarly for ‘strong’ starbursts(say, SFR multiplied by at least a factor 5), only 6% (3%) ofthe mergers without any external potential reach this level,while this fraction is increased to 11% (21%) in a group tidalfield as well as in a cluster tidal field. Thus, starbursts atall levels are triggered by the presence of the group/clusterpotential: over our sample of mergers, the maximal instan-taneous SFR is on average doubled, depending on the shapeof f ( V ), compared to mergers in the field far from groupsand clusters. The enhancement is even larger in some cases,in particular for the highest initial velocities. This trigger-ing effect is also important if we consider the integrated starformation over the burst duration (i.e. the total mass ofstars formed) that is on average multiplied by 1.3–1.4 in thevicinity of both the model group and cluster.We thus find a noticeable enhancement of the star for- c (cid:13)000 , 000–000 M. Martig and F. Bournaud
Maximum relative SFR P r obab ili t y mergers of field galaxiesmergers near a clustermergers near a group Figure 3.
Probability to have a maximum relative SFR largerthan the value specified on the x-axis for mergers of field galaxiesvs galaxies in a group or cluster tidal field. The results shown arefor the conservative assumption f ( V ) = 1. mation activity resulting from the addition of the cluster orgroup potential. This effect is comparable in amplitude tothe primary effect of star formation triggering by a majormerger compared to an isolated galaxy, hence a significantone. Nevertheless the quantitative results are exact only forthe adopted group or cluster mass, radius, and distance ofthe merging pair to the group/cluster centre. Other struc-tures would be more or less efficient in triggering star for-mation depending on the intensity of their tidal field on themerging galaxies. We have assumed a Plummer profile for the potential of themodelled group and cluster. In the case of clusters, real po-tentials are more likely cuspy (Pointecouteau et al. 2005).A cuspy Hernquist profile with the same half-mass radiusand total mass as our Plummer profile would actually exerta larger tidal field in its central regions near the cusp, anda comparable one (but always slightly larger) in the outerregions. The Plummer profile is thus a conservative choice,which could only lead to somewhat underestimating the ef-fect. Changing the mass or scale-length of the structure bya factor two would lead to similar variations in its tidal fieldas changing its profile. Our choice was also motivated bythe fact that forming groups and clusters at z ∼ Our results show that a given galaxy pair that merges inthe vicinity of a group or cluster has an SFR enhanced bythis external tidal field. In the statistical analysis, we havecompared results assuming a similar velocity distributionfor galaxies in the field and those near groups/clusters (ei-ther f ( V ) = 1 or f ( V ) = V in all cases). This should beclose to reality in groups, but one could expect higher ve-locities to dominate in the vicinity of clusters. Taking thisinto account would not however result in major changes; theaverage SFR enhancement is affected by 10% if we assume f ( V ) = 1 in the field and f ( V ) = V near the cluster. Onealso expects an increase in the proportion of fly-bys not fol-lowed by mergers in the vicinity of clusters (not necessarilynear groups). Di Matteo et al. (2007) showed that the SFRsare of the same order of magnitude for mergers and fly-bys,so that our results should be at first order extendable tohigher-velocity fly-bys. When weighting our results with thecross-section πb , the impact parameter b was simply es-timated at the beginning of our simulations ( t = 0). Theexternal tidal field would actually modify the orbits evenbefore t = 0 and possibly bias this parameter b comparedto the real impact parameter b ∞ at infinite distance. How-ever, the difference between b and b ∞ is minimized whenthe group/cluster is in configuration 1 (along the directionof the initial velocity), and a significant enhancement of theSFR is still found when we restrict ourselves to this config-uration (even 20% higher than in the other configurations).The star formation triggering that we found cannot then bean artefact resulting from the assimilation of b to b ∞ .The comparable effect found in our model group andmodel cluster indicates that the main requirement is thepresence of a tidal field, while the total size and mass of thestructure play only a secondary role in determining the ex-act level of triggering. One can however wonder if this reallyimplies an observable triggering of the SF activity, because c (cid:13) , 000–000 tar formation triggering near groups and clusters the merging pairs in dense and low-density regions are notnecessarily similar. Galaxies in groups can still contain im-portant gas reservoirs, but galaxies in the central regions ofrelaxed clusters at z ∼ In this paper, we have shown that a major galaxy merger ismore efficient to trigger an intense burst of star formation ifit takes place in the tidal field of a galaxy group or cluster.While the group/cluster fields do not trigger much the starformation in a single galaxy, the effect on merging pairs isimportant. The starbursts in our simulations are amplifiedby a factor 2 on average (sometimes much more) for typi-cal groups and clusters. More massive structures could evenhave larger quantitative effects depending on the intensityof the tidal field. A pair of M33-like spiral galaxies mergingin the vicinity of a Local-like group or a Virgo-like clusterwould see the intensity of its starburst amplified by the ex-ternal field by typically a factor ∼
2, but possibly more onsome orbits or regions of high tidal field. In a forthcomingpaper, we will study the dynamical response of gas duringmergers within such an external tidal field and the connec-tion with the enhancement of starbursts.Dense cosmological structures trigger the merger-induced star formation by the action of their tidal field. Thisholds at least from groups to clusters, the level dependingon their mass and size. Because gas dynamics and star for-mation must be spatially well resolved for SFRs to be accu-rately computed in galaxy mergers , large-volume cosmolog-ical simulations may miss or underestimate this effect. Thistriggering of merger-induced starbursts by the tidal field ofdense cosmological structures should be particularly efficientat high redshift ( z ≃ z ≃
0, quenching mechanismshave acted in the highest density regions resulting in a lessactive star formation there.Our results unveil a new star-formation triggeringmechanism in groups and at the periphery of clusters, whichcan act in particular at high redshift before star formationis quenched in dense regions. This can contribute to explainwhy LIRGs are often found in proto-cluster environments(Laag 2006) and the high frequency of blue star forminggalaxies in young high-redshift clusters (Butcher & Oemler1984). More generally, this can trigger the star formation ac-tivity in group mergers, and contribute to the reversal of thestar formation – density relation with increasing redshift.
ACKNOWLEDGMENTS
Simulations were performed on NEC-SX8 vector computersat CEA/CCRT and CNRS/IDRIS, as part of the Horizonproject. Discussions with David Elbaz, Pierre-Alain Duc,Giovanna Temporin and Emanuele Daddi motivated thisstudy and improved the manuscript. We are grateful to YvesRevaz for providing his visualisation software pNbody, PaolaDi Matteo, Shardha Jogee, Chanda Jog, Fran¸coise Combesand Romain Teyssier for useful discussions, and to an any-mous referee for constructive remarks.
REFERENCES
Bell, E.F., Papovich, C., Wolf, C., et al. 2005, ApJ, 625, 23Bournaud, F. & Combes, F. 2002, A&A, 392, 83Bournaud , F. & Combes, F. 2003, A&A, 401, 817Bridge, C. R., Appleton, P. N., Conselice, C. J., et al. 2007,ApJ, 659, 931Butcher, H. & Oemler, A. Jr. 1984, ApJ, 285, 426Chung, A., van Gorkom, J. H., Kenney, J. D. P., &Vollmer, B. 2007, ApJL, 659, L115Conselice, C. J., Chapman, S. C., & Windhorst, R. A. 2003,ApJL, 596, L5Cooper, M. C., Newman, J. A., Weiner, B. J., et al. 2007,astro-ph/0706.4089Cox, T. J., Jonsson, P., Somerville, R. S., Primack, J. R.,& Dekel, A. 2007, MNRAS in prep., astro-ph:0709.3511Cox, T. J., Jonsson, P., Primack, J. R., & Somerville, R. S.2006, MNRAS, 373, 1013Daddi, E., Dickinson, M., Morrison, G., et al. 2007, ApJ670, 156Di Matteo, P., Combes, F., Melchior, A.-L., & Semelin, B.2007a, A&A, 468, 61Duc, P.-A., Mirabel, I. F., & Maza, J. 1997, A&AS, 124,533Egami, E., Misselt, K. A., Rieke, G. H., et al. 2006, ApJ,647, 922Elbaz, D., Daddi, E., Le Borgne, D., et al. 2007, A&A, 468,33Fouqu´e, P., Solanes, J. M., Sanchis, T., & Balkowski, C.2001, A&A, 375, 770Jogee, S. et al. 2007, ApJ, in prep.Kauffmann, G., White, S. D. M., Heckman, T. M., et al.2004, MNRAS, 353, 713Kawata, D. & Mulchaey, J. S. 2007, astro-ph/0707.3814Kennicutt, R. C., Jr. 1998, ApJ, 498, 541Laag, E. A. 2006, in AAS Meeting Abstracts, Vol. 209Lewis, I., Balogh, M., De Propris, R., et al. 2002, MNRAS,334, 673Mihos, J. C. 2004, Clusters of Galaxies: Probes of Cosmo-logical Structure and Galaxy Evolution, 277Mihos, J. C. & Hernquist, L. 1996, ApJ, 464, 641Moore, B., Katz, N., Lake, G., Dressler, A., & Oemler, A.1996, Nature, 379, 613Naab, T., Burkert, A., & Hernquist, L. 1999, ApJ, 523,L133Pointecouteau, E., Arnaud, M., & Pratt, G. W. 2005, A&A,435, 1Quilis, V., Moore, B., & Bower, R. 2000, Science, 288, 1617Springel, V., White, S. D. M., Jenkins, A., et al. 2005,Nature, 435, 629 c (cid:13)000