Arxiv:1512.00090V1 [Astro-Ph.GA] 30 Nov 2015 Aaissc Si 8 Bsine L 01 N M51 and 2011) Al

Home , Embedded cluster

arXiv:1512.00090v1 [astro-ph.GA] 30 Nov 2015 aaissc si 8 Bsine l 01 n M51 and 2011) al. et (Bastian M83 in 2003; as Lada such & galaxies (Lada expulsion gas survive 2015a). clus- they Fujii open classical if into evolve ters, clusters embedded of Some 1 100 lses lblrcutr r l ( old are clusters Globular clusters. esaegnrlyrte on ( clus- young Open rather clusters. generally globular are and ters clusters open Galactic ie( sive trcutr nteMlyWy hs ye include types These Way. Milky open the among in variety wide clusters a star is there that indicate servations rpittpstuigL using typeset Preprint 2015 2, December version Draft AJFellow NAOJ on asv lsesaecmo nnab starburst nearby in common are clusters massive Young trcutr r lsial aeoie ntogroups, two in categorized classically are clusters Star − • • • & 10 PreisZate l 2010). al. et Zwart (Portegies y)adeteeydne( dense extremely and Myr) asv lses hc r loyug( born young un- also were are they which considered clusters, moment massive are 2011). the Zwart Portegies & from (Gieles which bound associations, 2002). al. (Figuerêdo et 0 tr navlm ihardu of radius a with several of volume composed a reside are in clusters and stars disk Embedded 100 Galactic the 2003). in cloud Lada gas natal & their (Lada in embedded still therefore and meddcutr,wihaevr young very are which clusters, embedded 10 rcsecletymthsteosrain.Acrigt u ca follows our associations rela to and mass-radius clusters According The open molecu for observations. from radius clusters. the path half-mass star matches evolutionary massive excellently distinct to tracks a or star indicate clou young associations molecular models massive and form densest Our galaxies, The Way. starburst in process. Milky typical relaxation are the but their Way to lost due but clumpy, whic scale more clusters time considerably embedded form initially were to tend associations Galaxy The Way Milky the for typical hs tla itiuin r otne ymaso direct c of molecular means the by of continued times simulat free-fall are Our initial distributions a an stellar are after stars these models. cloud The collapsing theoretical gravity. dy own the and its and in under formation simulations collapses that the of cloud molecular study combination We a associations. using and clusters, massive trclusters star eainfrcutr nteMlyWay. headings: Milky Subject the in clusters for relation 4 5 eetosrain aervae ait fyugsa cluste star young of variety a revealed have observations Recent tr,hratrw alte cascl open “classical” them call we hereafter stars, M ⊙ ,addne( dense and ), H OMTO N YAIA VLTO FYUGSA CLUSTER STAR YOUNG OF EVOLUTION DYNAMICAL AND FORMATION THE iiino hoeia srnm,Ntoa Astronomical National Astronomy, Theoretical of Division A 1. T E r tl mltajv 11/12/01 v. emulateapj style X ednOsraoy ednUiest,P o 53 30RA 2300 9513, Box PO University, Leiden Observatory, Leiden h introduction / c=0 = pc ehd:numerical methods: — aais trcutr:gnrl—(aay)oe lsesadass and clusters open (Galaxy:) — general clusters: star galaxies: & . 100 34( . M & t y)wt typically with Gyr) 1 age ⊙ 0Gr,mr mas- more Gyr), 10 pc / & Myr) − 3 .Rcn ob- Recent ). 10 2 / rf eso eebr2 2015 2, December version Draft 3 3 M ohted r ossetwt h bevdage-mass-radius observed the with consistent are trends Both . . .PreisZwart Portegies S. ∼ ⊙ [email protected] Myr 3 . pc pc 1 oy 181–8588 Tokyo ABSTRACT .S Fujii S. M. − 10 3 ) 1 assmlrt hs ftemsiecutr ( clusters massive the have of clusters those Leaky to similar clusters.” 10 mass “leaky a 14 clusters, Trumpler and star 2, and 2010). clus- al. 1 the et Westerlund includes Zwart category (Portegies 3603, latter are NGC This others the ters arms. and spiral Quintuplet, Galac- the and the in Arches Way. to i.e., Milky close the center, reside in tic clusters rare star are young they massive but Two 2011), al. et (Chandar tteeoehlsu odffrnit ewe h various environments. the stellar clustered between of differentiate families to and us classes evolution. helps in difference dif- therefore the the It and between conditions distinction initial here clear in more used ference a modeling make to numerical by us The clusters allows star for- simulations. young the of of paper, means evolution this dynamical in and discuss, mation we conglomerates stellar of system the system, Otherwise the of together. age bound unbound. dynamical is be the must exceed stars stars the the of age oito a emd ntertobtenteaeo the ( of system age the the of between time ratio as- dynamical the an and on and stars made cluster be open can an sociation between distinction the (2011) associations. clusters OB leaky as the (2009) classified Pfalzner in (2010) listed al. et Zwart Portegies 1 flnr(09 ugse nte yeo young of type another suggested (2009) Pfalzner na tep ocaiytevroscassadfamilies and classes various the clarify to attempt an In codn oteagmnaini ils&PreisZwart Portegies & Gieles in argumentation the to According 4 bevtr fJpn –11Oaa Mitaka, Osawa, 2–21–1 Japan, of Observatory M r h ⊙ N / 2 ,btwt uhlwrdniy( density lower much a with but ), c=2 = pc bd iuain.Temlclrclouds molecular The simulations. -body s nldn meddsses young systems, embedded including rs, smdt omi h ess regions densest the in form to ssumed lses hyide r aei the in rare are indeed They clusters. cltostetm vlto fthe of evolution time the lculations edn h Netherlands The Leiden, , infrbt ye fevolutionary of types both for tion aia vlto fteeclusters these of evolution namical . 7( s hc r beti h Milky the in absent are which ds, vlet eebeoe clusters. open resemble to evolve h od h yaia vlto of evolution dynamical The loud. reuaiyi bu dynamical a about in irregularity t age / ossatwt turbulent a with start ions pc) a lust pnclusters open to clouds lar 2 / 3 hra o massive for whereas , cain:general ociations: S t ∼ age 1–10 /t dyn M .I the If ). ⊙ pc − 3 ∼ ). 2 M. S. Fujii & S. Portegies Zwart

In previous papers, we performed direct N-body sim- 2.1. The Astronomical Multipurpose Software ulations using initial conditions constructed from the re- Environment sults of hydrodynamical simulations of turbulent molec- The hydrodynamical simulations and the data analy- ular clouds. There we found that young massive clus- ses in this study are performed using the AMUSE frame- ters form from turbulent molecular clouds, if the local work (Portegies Zwart et al. 2013; Pelupessy et al. 2013). star formation efficiency depends on the local gas den- AMUSE is not a single code, but a extensive library of sity (Fujii & Portegies Zwart 2015; Fujii 2015b). We also more than 50 high-performance simulation codes. The found that observed embedded clusters tend to evolve into AMUSE consortium is a spin-off from the MODEST com- classical open clusters (Fujii 2015b). Our simulations, munity, which upon three workshop in Lund, Amsterdam, however, did not provide a channel for forming associa- and Split culminated in a first implementation of, what tions (or leaky clusters, according to Pfalzner (2009)). at that time was called the Multi-User Software Environ- At this point it is still unclear how leaky clusters ment (or MUSE) (Portegies Zwart et al. 2009). Later the form. Pfalzner (2011) proposed that leaky clusters are package was extended from its primary objective of Noah’s born as embedded clusters, that their mass increases Arc (two codes per domain) to about a dozen codes per due to a prolonged phase of star formation, and that domain. the expansion is driven by the expulsion of the residual Apart from scientific production software, AMUSE also gas. This scenario was tested by means of simulations supports from generating initial conditions to data pro- in Pfalzner & Kaczmarek (2013), Parmentier & Pfalzner cessing. The fundamental package is written in the Python (2013), and Pfalzner et al. (2014), in which it was con- language and it is freely available via Github and via cluded that the known embedded clusters in the Galactic the project web page at http://amusecode.org. All the disk are the ancestors of leaky clusters. scripts used to run the simulations in this paper are avail- In our previous simulations we did not find leaky clus- able via this project web page. ters. This may have been a result of our selected initial conditions for the parental molecular cloud, for which 5 6 2.2. Hydrodynamical Simulations we chose rather massive (10 –10 M⊙) and dense (100– 1000 cm−3) structures. The molecular clouds observed in 2.2.1. Initial Conditions for Molecular Clouds the Milky Way tend to follow Larson’s relation (Larson All initial conditions are generated using the AMUSE 1981), which indicates a relation between cloud mass and framework. We adopt isothermal (30K) homogeneous density: According to this law massive clouds have a lower spheres as initial conditions of molecular clouds follow- density, if the clouds are close to be virialized. The ini- ing Bonnell et al. (2003). We give a divergence-free ran- tial conditions in our previous study would then biased dom Gaussian velocity field δv with a power spectrum towards too dense clouds compared to the typical massive δv 2 k−4 (Ostriker et al. 2001; Bonnell et al. 2003). clouds in the Milky Way. The| | spectral∝ index of 4 appears in the case of compressive In this paper, we expand on the initial parameter space, turbulence (Burgers turbulence),− and recent observations by also allowing massive clouds with a lower density. This of molecular clouds (Heyer & Brunt 2004) and numeri- expansion of the parameter space helps in the formation of cal simulations (Federrath et al. 2010; Roman-Duval et al. associations, as well as for making dense massive clusters. 2011; Federrath 2013a) also suggested values similar to 4. We support our numerical models with theoretical argu- Each model is run with a different random seed for a− re- ments in order to understand the the dynamical evolution alization of the initial conditions. of each type of star clusters (classical open, embedded, We adopt the virial ratio E / E = 1 (here E and E young massive, and leaky clusters or associations). k p k p are kinetic and potential energies)| | | and| three masses for the 4 5 6 molecular clouds of M = 10 , 4 10 , and 10 M⊙. The g × density of these molecular clouds are ρg = 17, 170, and −3 −3 1700 cm (which corresponds to 1, 10, and 100 M⊙pc 2. simulations assuming that the mean weight per particle is 2.33mH, We perform a series of N-body simulations based on the respectively). The initial conditions are summarized in results of hydrodynamical simulations of turbulent molec- Table 1. ular clouds. We first perform simulations of molecular Once we chose the cloud mass and density, the ra- clouds with a turbulent velocity field using an smoothed dius (Rg) and the velocity dispersion in three dimensions particle hydrodynamics (SPH) code. The resolution of the (σg) are determined. Some of our models (such as a hydrodynamical simulations is relatively low and therefore models m1M-d1-s15, m1M-d1-s16 and m1M-d1-s17, with 6 −3 the simulation cannot resolve the formation of individual Mg = 10 M⊙ and ρg = 17cm ) roughly follow Larson’s stars, but can resolve the clumpy structures of the gas. Af- relation (Larson 1981), ter around one free-fall time of the initial molecular clouds, L 0.5 we stop the hydrodynamical simulations and replace a part σ (kms−1), (1) of gas particles with stellar particles assuming a star for- ∼ 1pc mation efficiency depending on the local density. We then where σ is the velocity dispersion and L is the size of the remove all residual gas particles and perform direct N- cloud (Heyer & Brunt 2004; Mac Low & Klessen 2004). body simulations only with stellar particles. We describe In Figure 1 we present the distribution of mass and den- the details of the initial conditions and the simulations in sity for the simulations listed in Table 1. In order to de- the following (see also Fujii & Portegies Zwart 2015; Fujii termine the mass of a molecular cloud that is consistent 2015b). with Larson’s relation we adopt a velocity dispersion of The formation and evolution of young star clusters 3

103

m10k-d100m400k-d100 100 m1M-d100 ) 3 − m10k-d10 m400k-d10 pc

⊙ 10 M ( g

ρ Larson’s relation m1M-d1 1

0.1 100 103 104 105 106 107 108 Mg (M⊙)

Fig. 1.— Mass-density relation of our initial models (see Table 1). The dashed line indicates the mass-density relation from Larson’s relation 2 (Larson 1981) for virialized cloud (σg = GMg/Rg).

Table 1

Initial conditions for the hydrodynamical simulations. All models are super virial, with |Ek|/|Ep| = 1 Model Mass Radius Density Velocity dispersion Initial free-fall time −3 −1 Mg(M⊙) rg (pc) ρg(cm ) σg (kms ) tﬀ,i (Myr) m1M-d100-s7 1 106 13.4 1.7 103 19.6 0.81 m1M-d1-s15 1 × 106 62 17× 9.1 8.1 m1M-d1-s16 1 × 106 62 17 9.1 8.1 m1M-d1-s17 1 × 106 62 17 9.1 8.1 m400k-d100-s1 4 × 105 10 1.7 103 14.4 0.82 m400k-d100-s2 4 × 105 10 1.7 × 103 14.4 0.82 m400k-d100-s3 4 × 105 10 1.7 × 103 14.4 0.82 m400k-d10-s8 4 × 105 21 170× 9.9 2.5 m400k-d10-s9 4 × 105 21 170 9.9 2.5 m10k-d100-s4 1 × 104 2.87 1.7 103 4.2 0.81 m10k-d100-s5 1 × 104 2.87 1.7 × 103 4.2 0.81 m10k-d100-s6 1 × 104 2.87 1.7 × 103 4.2 0.81 m10k-d10-s11 1 × 104 6.2 170× 2.9 2.6 m10k-d10-s12 1 × 104 6.2 170 2.9 2.6 m10k-d10-s13 1 × 104 6.2 170 2.9 2.6 ×

’s’ indicates the random seeds for the turbulence. 4 M. S. Fujii & S. Portegies Zwart

σ GM /R . Some models have initially a higher ve- mass function is 1 M⊙, which corresponds to the mass of g ≃ g g locityp dispersion, which we motivate through cloud-cloud individual SPH particles. Mass in our simulations is there- collisions (Furukawa et al. 2009; Fukui et al. 2013, 2014) fore globally conserved, but not locally. or to simulate molecular clouds in starburst galaxies. We further motivate and discuss on our choice of the initial 2.4. N-body simulations conditions in 4. After the stellar particles are initialized (mass randomly § from the Salpeter mass function, and position and velocity 2.2.2. Smoothed Particle Hydrodynamics Simulations from the parent SPH particle), we remove the residual gas, We perform hydrodynamical simulations using the SPH leaving only the stellar particles in the simulations. The code Fi (Hernquist & Katz 1989; Gerritsen & Icke 1997; instantaneous removal of the gas has not a dramatic effect Pelupessy et al. 2004; Pelupessy 2005) in the AMUSE on the stellar distribution, because most stars are formed framework. Our calculations have relatively low mass in the densest regions where little low-density (residual) resolution of m = 1M⊙ per particle. The gravitational gas is present. The gas that is insufficiently dense to form softening length during the hydrodynamical simulations is stars tend to be enveloping the densest stellar conglomer- 0.1 pc, and the SPH softening length (h) is chosen such ates. 3 that ρgh = mNnb (Springel & Hernquist 2002). Here We now switch on the N-body code, for which we Nnb = 64 is the target number of neighbor particles. With adopted the direct sixth-order Hermite predictor-corrector the adopted isothermal gas temperature of 30K we can scheme (Nitadori & Makino 2008) without gravitational resolve the Jeans instability down to h 0.4 pc, which softening and with an accuracy parameter, η =0.1–0.25. is smaller than the typical size of known∼ embedded clus- The total energy error over the time span of the N-body − ters (1 pc) (Lada & Lada 2003) but somewhat larger than simulations remained below 10 3. ∼ the observed typical width of gas filaments ( 0.1 pc) The sizes of the stars we adopted from the zero-age (Arzoumanian et al. 2011). With these limitations,∼ we ob- main sequence radii for solar metallicity stars Hurley et al. viously cannot resolve the formation of individual stars, (2000). We allow stars to collide using the sticky sphere but we do resolve dense gas clumps. We think that the approach. New stellar radii are assumed to be the limited resolution of our hydrodynamical simulations does zero-age main sequence radii for the new mass. Stel- not pose a serious problem, because we are interested in lar mass-loss was incorporated only at the end of the the global dynamical structure of the molecular cloud after main sequence Hurley et al. (2000) (see Fujii et al. 2009; only about an initial free-fall time scale, tff,i (see Table 1 Fujii & Portegies Zwart 2013, for the details). for the free fall time scales for each of the initial models). We did not perform the N-body simulations for mod- 4 −3 In fact, after 0.9t we stop the hydrodynamical simu- els m10k-d10 (Mg = 10 M⊙ and ρg = 10M⊙pc = ff,i − lation to analyze the resulting gas distribution, initialize 170cm 3), because the hydrodynamical simulations re- stars, and continue the simulations using a gravitational sulted in less than 100 stars and we aim at & 100M⊙ star N-body code. clusters. In these simulations even the densest regions were −3 < 1000M⊙pc . 2.3. The star formation 3. results After stopping the hydrodynamical simulation (around 3.1. Formation of Embedded, Classical Open, and Young 0.9tff,i) we replace some of the SPH particles with stellar ∼particles. The selection of SPH particles is based, through Massive Clusters the local gas density ρ, on the local star formation effi- The N-body simulations are started at what we will call ciency (SFE) ǫloc: t = 0 Myr. The initial distribution of stars follows the dis- 0.5 tribution of the densest regions in the turbulent molecular ρ cloud. In Figure 2 we present a time series of snapshots of ǫloc = αsfe −3 . (2) 100 M⊙pc model m400k-d100-s3. The entire system continuously ex- Here αsfe is a free parameter in our simulations to control pands because not all stars are bound after gas expulsion. the SFE. the form of ǫloc (Eq. 2) is motivated by the obser- The distribution of stars is clumpy and it takes a few Myr vations of individual molecular clouds for which the star before the stars assemble in a more coherent aggregate. formation rate is argued to scales with local free-fall time We interrupt the simulations twice, at t = 2 and at scale (Krumholz et al. 2012; Federrath 2013b). t = 10 Myr, in order to analyze the stellar distribution, Here we adopt αsfe = 0.02, which reproduces the and detect clustered aggregates. Clumps are found in observed global SFE across an entire molecular cloud these snapshots by means of HOP (Eisenstein & Hut 1998) of several per cent, but also leads to a 10–30% SFE in AMUSE, using an outer cut-off density of ρout = −3 3 in dense regions (> 1000M⊙cm ) (Lada & Lada 2003; 4.5Ms/(4πRh) (three times the half-mass density of the 3 Higuchi et al. 2009; Federrath & Klessen 2013). In Table entire stellar system, ρh = Ms/(8πRh), where Ms is the 2 we present the global SFE (ǫ) and the SFE for the dense total stellar mass and Rh is the half-mass radius of the regions (ǫd) in our simulations. entire distribution of the stars), a saddle-point density Depending on the local SFE we replace individual gas threshold (ρsaddle =8ρout) and the peak density threshold particles to individual stellar particles conserving their po- (ρpeak = 10ρout) and the number of particles for neighbor sitions and velocities. For each selected particle we assign search (Ndense) as well as the number of particles to calcu- a mass from the Salpeter mass function (Salpeter 1955) late the local density (Nhop) are set to be 64. The number between 0.3 M⊙ and 100M⊙, irrespective of the mass of of neighbors is used to determine which two groups merge its parent SPH particle. The mean mass of the adopted Nmerge = 4. With these settings the detection limit of the The formation and evolution of young star clusters 5

Table 2 Models for N-body simulations Model Mass N of particles Virial ratio∗ SFE (Global) SFE (Dense) Ms(M⊙) Ns Ek / Ep ǫ ǫd m1M-d100-s7 1.1 105 109080| 0.9| | | 0.11 0.27 m1M-d1-s16 1.9 × 104 18760 0.50 0.019 0.63 m1M-d1-s16-t0.75 4.6 × 103 4566 19 0.0046 0.42 m1M-d1-s16-t0.65 3.9 × 103 3855 131 0.0039 0.083 m1M-d1-s15-t0.75 5.9 × 103 5902 1.6 0.0059 0.49 m1M-d1-s15-t0.65 4.0 × 103 3954 80 0.0040 0.12 m1M-d1-s17-t0.75 5.5 × 103 5506 6.4 0.0055 0.26 m1M-d1-s17-t0.65 4.3 × 103 4322 63 0.0043 0.088 m400k-d100-s1 3.2 × 104 31895 1.3 0.078 0.22 m400k-d100-s2 2.3 × 104 23273 4.2 0.057 0.16 m400k-d100-s3 4.3 × 104 42596 0.43 0.096 0.25 m400k-d10-s8 1.5 × 104 14978 1.4 0.037 0.38 m400k-d10-s9 2.8 × 104 27891 0.41 0.068 0.39 m10k-d100-s4 4.1 × 102 406 5.9 0.042 0.11 m10k-d100-s5 2.6 × 102 256 7.4 0.027 0.079 m10k-d100-s6 2.5 × 102 246 8.4 0.026 0.078 m10k-d10-s11 49× 49 - 0.0049 0.00 m10k-d10-s12 61 61 - 0.0061 0.00 m10k-d10-s13 65 65 - 0.0065 0.00

’s’ indicates the random seeds for the turbulence.

∗ |Ek| and |Ep| are the total kinetic and potential energies of the entire stellar system, respectively. For virialized systems the virial ratio equals 0.5. For models m10k-d10 we did not perform N-body simulations, and therefore their virial ratio is not calculated.

Fig. 2.— Snapshots of model m400k-d100-s3. The size of the dots indicate the masses of the stars: 8 < m/M⊙ < 16 for the small dots, 16 < m/M⊙ < 40 for middle sized and m> 40M⊙ for the largest dots. Stars with a mass m< 8M⊙ are plotted as small blue dots. 6 M. S. Fujii & S. Portegies Zwart clump mass is 100M⊙. Sometimes HOP identifies multi- by cloud-cloud collisions (Fukui et al. 2014). The veloc- ple clumps as one,∼ but by applying the method repeatedly ity dispersion of our dense model 20 kms−1, which we can separate those again. For this iterative procedure is comparable to the typical relative∼ velocity of molecu- we adopt ρout = ρh,c, where ρh,c is the half-mass density lar clouds associated with young massive clusters such as of a detected clump. We continue this procedure until NGC 3603 and Westerlund 2. For these clusters a collision ρh,c & 100ρh, after which the clumps are so dense com- between two molecular clouds was considered to trigger pared to the background that they do not separate any- their formation (Furukawa et al. 2009; Ohama et al. 2010; more in substructures (see Fujii 2015b, for the details). Fukui et al. 2013, 2014), which is consistent with our find- In Figure 3 we present the mass and half-mass ra- ings here. dius of the star clusters obtained from our simulations at For forming a star cluster in our simulations, the molec- t = 2 and at t = 10 Myr. For comparison, we added ular cloud must be compressive (a high velocity dispersion a number of observed open clusters (classical open, em- due to a high density), which is consistent with observa- bedded, young massive, and leaky clusters) to the same tions (Zinnecker & Yorke 2007). From various initial con- diagram. The majority of the identified clusters have ditions, we find that star clusters similar to open, known masses and radii consistent with those of classical open embedded, and young massive clusters form in these sim- 4 clusters (Piskunov et al. 2008) (see also Fujii 2015b) and ulations, but leaky clusters (M 10 M⊙ and rh 10 pc) of known embedded clusters (Lada & Lada 2003). The must form from different initial conditions.∼ We discuss∼ the densest initial molecular clouds (m1M-d100, m400k-d100, formation of leaky clusters in the following section ( 3.2). and m400k-d10) tend to to form massive compact clusters, § similar to young massive clusters. Such compact clusters 3.2. Formation of Leaky Clusters do not form in the less dense or less massive molecular clouds (such as m1M-d1 or m10k-d100). In the previous section, we show that known embedded, When observing the 10-Myr-old stellar conglomerates classical open, and young massive clusters form from tur- from a distance, they tend to blend in a single star forming bulent molecular clouds, but no leaky cluster is found in −3 region with an average density of 0.01M⊙pc , which our simulations. In this section we address the question: is comparable to the mean field density∼ in solar neighbor- how do leaky clusters form? Is the formation process dif- hood (Holmberg & Flynn 2000). Such conglomerates may ferent from the other clusters? remain unrecognizable as a cluster system. For those sim- Pfalzner (2011) proposed that observed embedded clus- ulations in which no clumps are detected down to a limit ters grow in mass and size due to star formation and of 100 M⊙, we adopt the median distance of the stars from become leaky clusters as a result of the expulsion of the cluster center. the residual gas. This scenario was later explored and The masses and half-mass radii of the clusters in our the evolutionary tracks of such a cluster on the mass- simulations mainly resemble the populations of observed radius diagram were suggested (Parmentier & Pfalzner embedded and classical open clusters. This result appears 2013; Pfalzner & Kaczmarek 2013; Pfalzner et al. 2014). to be independent of the initial molecular-cloud density. Portegies Zwart et al. (2010), however, classified the leaky Embedded and classical open clusters cluster around the clusters as OB associations. We here do not discuss if the point where the cluster age (tage) equals the dynamical leaky clusters are associations or clusters, but treat both time (tdyn) and the half-mass relaxation time (trh) Fujii leaky clusters and associations as less dense clustered sys- (see also 2015b). tems. Here the dynamical time and the half-mass relaxation We consider leaky clusters (and also OB associations) time are written as to form clumpy but that they lose this structure in the −1/2 3/2 early dynamical evolution, contrary to the arguments in 4 M rh tdyn 2 10 year (3) Pfalzner (2011). We support our argument with the simu- ∼ × 106M⊙ 1pc lation model m1M-d1-s16 (see the left panels in Figure 4). and This simulation started with a spherical molecular cloud M 1/2 r 3/2 that collapsed asymmetrically due to the turbulence ve- 8 h locity field. Stars formed mainly in the densest regions trh 2 10 6 year, (4) ∼ × 10 M⊙ 1pc which result in the stellar distribution being elongated and respectively (Portegies Zwart et al. 2010), where M is the clumpy. cluster mass, and rh is the half-mass radius. For clarity After the residual gas has been removed, the clusters we assumed that the virial radius of star clusters is com- tend to be super virial, and some stars escape right away parable to the half-mass radius and that the mean stellar (see the virial ratio given in Table 2). As a consequence, mass is 1 M⊙ (as is the case in our simulations). In Figure the entirely stellar distribution expands with time. At an 3 we present lines on which the relaxation (black full) and age of t = 10 Myr the density of the environment has de- dynamical (black dash-dotted) times are equal to the age creased substantially, and the spatial distribution of the of the clusters, respectively. Both lines move as well as all stars resembles leaky clusters and OB associations. In the symbols upward with time. Figure 5 we present the spatial distribution of O and B For the formation of young massive clusters, we find spectral-type stars in the association Scorpius OB2 (Sco that a dense massive molecular cloud is necessary. The OB2), which can be compared with our simulations in Fig- densities required to form such massive clusters exceed the ure 4. density expected by Larson’s relation; the velocity disper- Sco OB2 is composed of three subgroups; Upper sion necessary for the formation of young massive clusters Scorpius (USco), Upper Centaurus-Lupus (Upper Cen- is too high. Such an initial high density may be realized Lup), and Lower Centaurus-Crux (Lower Cen-Crux) The formation and evolution of young star clusters 7

100 100 m400k-d100 m1M-d100 t = 10Myr m400k-d10 m10k-d100 m1M-d1 ILac1UppCen-Crux OriIa LowCen-CruxUSco t = 2Myr OriIc IC1805 2 − = 2 10 = h 10 h OriIbNGC7380 ρ ρ NGC2244NGC6611 log CYgOB2loglog trh

(pc) = 3 (pc)

h h kiPer ONCQuintuplet hPer , r = 0 = 4 , r ρh ρh RSGC1 eﬀ log DBS2003 log eﬀ r Wd1 log r Wd2 trh 1 NGC3603 = 2 1 Tr14 Arches m400k-d100 m1M-d100 log trh m400k-d10 m10k-d100 = 1 m1M-d1 0.1 log trh = 0 0.1 10 100 103 104 105 106 10 100 103 104 105 106 M (M⊙) M (M⊙)

Fig. 3.— Mass-radius diagram of observed and simulated clusters at t = 2 and 10 Myr since the start of the N-body simulations. Colored dots are clusters obtained from simulations using a clump ﬁnding method. Crosses indicate the median radius and the total mass of the entire stellar system rather than the detected individual clusters. Red squares indicate observed clusters with an age of 1–5 Myr (left) and 5– 15 Myr (right). Data are from Piskunov et al. (2008); Winston et al. (2009); Luhman et al. (2003); Andersen et al. (2006); Fang et al. (2009); Levine et al. (2006); Flaherty & Muzerolle (2008); Bonatto & Bica (2011); Horner et al. (1997); Drew et al. (1997); Hodapp & Rayner (1991); Portegies Zwart et al. (2010). Observed clusters with names are the clusters listed in Pfalzner (2009) and Portegies Zwart et al. (2010). Black thick solid and dash-dotted lines indicate the line at which the relaxation time and the dynamical time are equal to the age of the stellar 4 −3 populations. Gray dashed lines indicate the half-mass density of 0.01, 1, 100, and 10 M⊙pc , and gray dotted lines indicate the half-mass relaxation time of 1000, 100, 10, and 1 Myr from top to bottom. We used the median radius for the observed leaky clusters (Wolﬀ et al. 2007; Pfalzner 2009).

(Wolff et al. 2007). These subgroups are listed in clumpiness of molecular clouds at a time when we stop Pfalzner (2009) as leaky clusters, and as associations in the hydrodynamical simulations (at 0.9tff,i). In ob- Portegies Zwart et al. (2010). They are all located at sim- served star forming regions, however, stars∼ appear to form ilar distances from the sun, at 145pc, 142, and 118pc, re- when the local density exceeds some threshold density for spectively (Wolff et al. 2007), and therefore they are con- self-gravitating clouds of 103cm−3 (McKee & Ostriker sidered to be a system. The distribution of massive (O and 2007), and feedback starts∼ to dominate the hydrodynam- B) stars in Sco OB2 is very similar to the distribution of ics as soon as the first massive star forms, which may hap- massive stars in model m1M-d1-s16 at an age of 10 Myr. pened well before a free-fall time scale. The free-fall time In figure 6 we present the result of our clump finding scale of model m1M-d1-s16 is 8 Myr, which is consid- analysis for model m1M-d1-s16 at 2Myr and at 10 Myr. erably longer than the formation∼ time for massive stars At t = 2 Myr we detected 20 clusters that are similar to ( 1Myr) (McKee & Ostriker 2007). In such a region, observed embedded star clusters.∼ At t = 10 Myr no clear where∼ the star forming time scale is considerably smaller massive clusters remain visible in the snapshot (see Figure than the free-fall time scale of the entire molecular cloud, 4), although we still detected several classic open cluster- stellar feedback is expected to terminate the star forma- like structures; in the epoch between 2 Myr to 10 Myr, the tion before the molecular cloud fully collapses. This would stellar distribution has dispersed. result in a less clumpy stellar distribution. When interpreting the entire system in each simulation Unfortunately in our simulations, we cannot take as a single association, the mass and radius are very sim- such gradual star formation and feedback processes ilar to those of observed leaky clusters and OB associa- into account, although they have been addressed with tions. In figure 6 we present these as crosses (to the top the AMUSE framework by Pelupessy & Portegies Zwart of the panels at 2Myr and 10Myr). Model m1M-d1-s16 (2012). In order to mimic the early star formation pro- has a similar appearance and dynamical structure as the cess, we experimented with stopping the hydrodynamical Sco OB2 system, rather than the individual sub-clusters simulations at an earlier epoch and replace the gas parti- USco, Upper Cen-Lups and Lower Cen-Crux. In this anal- cles with stellar particles. ysis we excluded single stars (those with a local density As in our previous simulations, we assumed that the −3 −3 ρ6 < 10 M⊙pc ) which is more than an order of mag- feedback terminates star formation which causes the resid- nitude lower than the mean density of the solar neighbor- ual gas to be ejected instantaneously. We stop the hy- hood (ρ6 here is the density measure within the 6 nearest drodynamical simulation for model m1M-d1-s16 at t = neighbors). 0.65tff,i and 0.75tff,i (5.3 and 6.2 Myr, respectively), and We still detected clumps consistent with open clusters replace gas particles to stellar particles using the same in model m1M-d1-s16. These clumps are the result of the way as for model m1M-d1-s16, i.e., assuming a local star 8 M. S. Fujii & S. Portegies Zwart formation efficiency given by equation (2) and the same 4. initial conditions of molecular clouds values for αsfe =0.02. The numbers of stars that form us- In the previous section, we showed that our dense mod- ing this procedure decrease considerably, and the resulting els tend to form young massive clusters and that less dense virial ratio of the stellar system increases. We also run the models lead to leaky clusters as well as known embedded same initial conditions but with different random seeds and classic open clusters. The types of the resulting star (m1M-d1-s15 and m1M-d1-s17). In Table 2 we present cluster is sensitive to the initial conditions of the parental some global parameters for these models. molecular clouds. In this section, we compare our initial Snapshots of these models (m1M-d1-s16-t0.75 and conditions with observed molecular clouds and discuss a m1M-d1-s16-t0.65) are shown in the middle and right pan- model for the formation of clusters in the Milky Way and els of Figure 4. The distribution of massive stars is less other nearby galaxies. clumpy compared with that of model m1M-d1-s16 (stan- In Figure 7, we present the mass and density of in- dard model, in which the hydrodynamical simulation is dividual molecular clouds observed in the Milky Way stopped at 0.9t ; see the left panels of Figure 4). We ff and those estimated for local disk and starburst galaxies also apply the clump finding algorithm to these mod- (Krumholz et al. 2012). We also show the initial condi- els, and the results of which are shown in Figure 6. At tions of our simulations. The dashed line in the Figure 7 t = 2Myr, several clumps are detected in both models, indicates the Larson’s relation. In order to estimate the but they are less dense compared with those detected in mass of molecular clouds following Larson’s law, we as- our standard model. In model m1M-d1-s16-t0.65 in par- sume that the molecular clouds are in virial equilibrium ticular, the density of the detected clumps is only slightly (i.e., they satisfy σ2 = GM /r , where σ , M , and r are elevated compared to the background density in the solar g g g g g g −3 the velocity dispersion, mass, and radius of the molecular neighborhood (0.01M⊙pc ) (Holmberg & Flynn 2000), clouds, respectively). Observed molecular clouds, how- and these clumps may therefore not be recognized as clus- ever, are not necessarily virialized. ters. In model m1M-d1-s16-t0.75 some clumps which re- Molecular clouds in the Milky Way tend to follow Lar- semble open clusters are still detected at t = 10 Myr, but son’s relation, but with a large scatter of the density. none in model m1M-d1-s16-t0.65. If we treat the entire On the other hand, not all of our initial conditions are system as a one cluster, the masses are similar to those of consistent with the mass and density of molecular clouds leaky clusters and associations, even though the size re- 4 observed in the Milky Way. Models m10k-d100 (10 M⊙ mains larger by about a factor of two. On Figure 6 we −3 −3 4 and 100M⊙pc 1700cm ) and m10k-d10 (10 M⊙ and present the resulting clusters with a mass and half-mass −3 ≃ −3 −3 −3 10M⊙pc 170cm ), for example, are initially indistin- radius in stars with ρ > 10 M⊙pc . 6 guishable from≃ typical molecular clouds in the Milky Way. Our assumption that star formation terminates instan- As we described in section 3, the number of stars formed taneously throughout the system after about one free- in model m10k-d10 was too small (fewer than 100 stars) fall time of the molecular cloud probably overestimates to be recognized as a cluster in our analysis. Model m10k- the effect of the feedback considerably. In observed star- d100 produces a sufficiently large number of stars but does forming regions the feedback from massive stars tend to not form a recognizable cluster after 2Myr. If we treat the limit star formation locally, but may not affect the en- entire region of this model as a cluster conglomerate, the tire ( 100 pc across) star forming region. In the simula- mass and radius is similar to that of an open cluster. From tions∼ of Pelupessy & Portegies Zwart (2012), the wind of this, we conclude that the molecular clouds typical in the one massive 30M⊙ star blows the residual gas from the ∼ Milky Way tend to form classical open clusters, but that clustered environment in a couple of Myr, which is much they are insufficiently massive and dense to form massive longer than adopted in our simulations. star clusters. If star formation proceeds as clumpy as simulated here, 6 −3 Model m1M-d1 (10 M⊙ and 1M⊙pc ) represents the the feedback is even more localized, which will result in most massive molecular cloud in the Milky Way (Murray a considerable age spread among subgroups. Our sim- 2011), and it follows Larson’s relation. This initial condi- ulations would then be representative for the formation tion results in several embedded cluster cores, that even- of cluster complexes such as USco, Upper Cen-Lups, and tually evolve to a conglomerate of associations. Lower Cen-Crux, or OB association such as Sco OB2. The initial conditions which tend to form young massive The ages of these three subgroups are slightly different clusters are considerably denser than the molecular clouds each other; 14–15, 11–12, and 5–6 Myr for Upper Cen- observed in the Milky Way (see Figure 7). To form a young Lup, Lower Cen-Crux, and USco, respectively (Wolff et al. massive clusters in our simulations a mass of at least sev- 2007). If we could assume local feedback processes, an 5 −3 −3 eral 10 M⊙ and a mean density of 10M⊙pc (170 cm ) association (or leaky clusters) similar to Sco OB2 might is required. Such initial conditions are common in local form from an initial condition such as models m1M-d1. starburst galaxies, but very rare in the Milky Way. Less dense clusters tend to have a wider age spreads In Figure 7 we present the estimated mass and molec- (Parmentier et al. 2014), which is also consistent with our ular clouds density typical for local starburst and disk simulations. We therefore argue that the ancestors of as- galaxies. This data is obtained from Krumholz et al. sociations are conglomerates of denser embedded clusters. (2012). We calculated the masses and densities for these We detect these as an environment with multiple low-mass molecular clouds from the free-fall time scale provided but rather dense clusters that disperse in time. The evap- by Krumholz et al. (2012) using the observed surface gas oration of these clusters is driven by relaxation and feed- densities (Σg). In Krumholz et al. (2012) they considered back, and this makes them resemble associations. two rather distinct regimes of molecular clouds; these are the molecular cloud regime and the Toomre regime. The The formation and evolution of young star clusters 9

Fig. 4.— Snapshots at t = 2 (top) and 10 (bottom) Myr for model d1-1M, but for diﬀerent timing of gas removal. t = 0.9, 0.75, and 0.65tﬀ,i (models m1M-d1-s16, m1M-d1-s16-t0.75, and m1M-d1-s16-t0.65) from left to right.

100

USco

0 (pc) y

Upper Cen-Lup −50

Lower Cen-Crux −100

−100 −50 0 50 100 x (pc)

Fig. 5.— Positions of B-type stars which belong to USco (magenta), Upper Cen-Lup (green), and Lower Cen-Crux (orange). Data is from Wolﬀ et al. (2007). We assume 140 pc as the distance (Wolﬀ et al. 2007). 10 M. S. Fujii & S. Portegies Zwart

100 100 0.65tff,i 0.9tff,i Sco OB2 0.75tff,i

ILac1UppCen-Crux OriIa LowCen-CruxUSco OriIcIC1805 2 log 2 log − t = 2 − t = 2 10 = rh h 10 = rh h h NGC7380 = 3 ρ h = 3 ρ ρ OriIbNGC2244NGC6611 ρ log CYgOB2 log log log (pc) (pc)

h h kiPer ONCQuintupletlog hPerlog , r = 0 trh = 4 , r = 0 trh = 4 ρh = 2 ρh ρh RSGC1= 2 ρh eﬀ log DBS2003 log eﬀ log log r Wd1 r Wd2 1 NGC3603 1 Tr14log log Archestrh trh = 1 = 1 t = 2Myr t = 10Myr 0.1 log trh = 0 0.1 log trh = 0 10 100 103 104 105 106 10 100 103 104 105 106 M (M⊙) M (M⊙)

Fig. 6.— Mass-radius diagram of detected clusters for models m1M-d1-s16 (green dots), m1M-d1-s16-t0.75 (blue triangles), and m1M-d1- s16-t0.65 (cyan diamonds) at t = 2 and 10 Myr. Red squares indicate observed young clusters; clusters with an age of 1–5 Myr and 5–15 Myr are plotted in top and bottom panels, respectively. The data for the observed clusters are the same as those in Figure 3. The mass and half-mass radii of simulations, interpreted as unresolved clusters, are shown as crosses; each cross represents a single simulation. molecular cloud regime is expected to be common in lo- also Krumholz & McKee 2005). From these equations, the cal disk galaxies. The molecular clouds are decoupled density of the molecular cloud becomes from their surrounding interstellar medium, and as a re- (β + 1)φ Ω2 ρ P . (8) sult self-gravitating (Krumholz et al. 2012). The Toomre g,T ≃ πGQ2 regime is common in starburst galaxies. In this case the Here we adopt Q 1 and β = 0 following Krumholz et al. interstellar medium is highly turbulent and therefore the (2012). If we assume∼ that the cloud is virialized — free-fall time scale of the molecular clouds should be esti- i.e., σ2 GM /r , where M and r are the mass g ∼ g,T g g,T g mated using the mid-plane pressure in the galactic disks and radius of the cloud— from equation (7) and ρg,T = (see Krumholz et al. 2012, for the details). 3 3Mg,T/(4πrg) we can estimate the cloud mass using: Following the description of Krumholz et al. (2012), 1 3 3/2 3 3 −1/2 we estimate the typical mass of molecular clouds for Mg,T G Σgtorbρg . (9) each galaxy listed in Krumholz et al. (2012). We take ∼ 32rπ the smaller free-fall time scale for the molecular cloud Because for each galaxy torb and Σg are given in and Toomre regimes (t and t , respectively) as Krumholz et al. (2012), we can estimate the cloud den- ff,GMC ff,T sity and mass from equation (8) and (9). Here we adopt the free-fall time scale (t ), which is consistent with ff φ 3, following Krumholz et al. (2012). Krumholz et al. (2012). We calculate the density through P ≃ the free-fall time scale using: For the molecular cloud regime (i.e., tff,GMC

103

m10k-d100m400k-d100 100 m1M-d100 ) 3 Starburst −

pc m10k-d10 m400k-d10

⊙ 10 M

( Milky Way g

ρ Larson’s relation m1M-d1 1 Disk

0.1 100 103 104 105 106 107 108 Mg (M⊙)

Fig. 7.— Mass-density relation of observed molecular clouds. Green diamonds indicate individual molecular clouds in the Milky Way galaxy. Blue circles and red triangles are for molecular clouds typical in individual local disk and starburst galaxies, respectively. Each point indicates one galaxy. The data is from Krumholz et al. (2012). Color squares indicate our initial conditions which are the same as those 2 shown in Figure 1. Dashed line indicates the mass-density relation following Larson’s law for virialized cloud (σg = GMg/Rg). this curve, and those have a relatively high concentration. creases. These processes proceed on the half-mass relax- As a consequence, their core densities exceed the local den- ation time: sity considerably, which helps to identify them as clusters 0.065σ3 t = . (12) in observational campaigns. rh G2 m ρ ln Λ The blue dash-dotted lines in Figure 8 indicate the mass- Here σ and ρ are the velocityh i dispersion and density of radius relation for which the dynamical time scale (see the cluster, respectively, and ln Λ is the Coulomb loga- equation (3)) is equal to the age of the cluster. Each rithm (Spitzer 1987). We rewrite equation (12) to include panel contains two lines, one for the minimum and one some common dimensions in equation (4). for the maximum age of the clusters shown in the pan- Gieles et al. (2011) modeled the post-collapse evolution els. The regions between these lines is shaded blue. of the half-mass radius and the density of star clusters Clusters between or below the blue lines will be recog- due to the energy flux from the core following the de- nizable as a bound systems unless the lines exceed the scription of Hénon (1965). We attempt to understand magenta dashed line. Portegies Zwart et al. (2010) and the dynamical evolution of young star clusters using their Gieles & Portegies Zwart (2011) argued that the ratio be- description. We ignore the pre-collapse phase and con- tween cluster age and the dynamical time provides a good sider only the evolution in the post-collapse (expansion) indicator for separating the bound from the unbound sys- phase, because the pre-collapse phase is much shorter than tems: They adopt as a criterion tage/tdyn & 3 to make post-collapse. The core-collapse time, which is the time this distinction. Using this criterion they categorized the for the pre-collapse phase, scales with the relaxation time leaky clusters in Pfalzner (2009) as associations. The blue (see equation (4) or (12)). This time scale depends on region in Figure 8 moves upward with time, together with the stellar mass function, and for clusters with a realistic the observed clusters. At t > 20 Myr (the right panel mass function the core collapse time is generally shorter in Figure 8) the blue lines are located above the magenta than one relaxation (Portegies Zwart & McMillan 2002; line, indicating that these clusters have a density too low Gürkan et al. 2004; Fujii & Portegies Zwart 2014). Since to be recognized as a cluster. most of young open clusters in our observed sample have a The evolution of dense star clusters is quite different relaxation time . 10 Myr (see the left panel of Figure 8), from those of open clusters or associations. Dense star they probably reach core collapse well within a few Myr. cluster evolution can roughly be divided in two phases; We also ignore the effect of the Galactic tidal field, because before core collapse and after core collapse. In the for- the time scale we treat here is short (< 100 Myr) com- mer phase the core radius of the star clusters shrinks, pared to the time scale for the tidal disruption ( 1Gyr) and as a consequence, its core density increases (Hénon (Gieles et al. 2007). ∼ 1965; Lynden-Bell & Wood 1968). From the moment the The time of the half-mass radius of clusters due to bi- first hard binaries form in the cluster core (Spitzer & Hart nary heating in the core is given by equation (B7) in 1971; Aarseth 1974), they act as an energy sources (Heggie Gieles et al. (2011): 1975; Hut 1983) causing the core to re-expand. From this 3G 1/3 moment on, the core- and half-mass radius of clusters in- r (125ζt)2/3. (13) h ≃ 4πN The formation and evolution of young star clusters 13

Here N is the initial number of stars in the cluster. In analytic models suggest two distinct populations of mas- 4 equation (B7) in Gieles et al. (2011), the cluster mass is sive ( 10 M⊙) clusters, which are called as starburst and assumed that M = m N, where m is the mean stel- leaky∼ clusters by Pfalzner (2009). We argue that these two lar mass. They adoptedh i a scaled massh i m = 0.5 and as populations naturally appear if we consider the formation a result their equation (B7) is slightly differenth i from our and the dynamical evolution process of star clusters. equation. If we assume that m = 0.5 M⊙, we can write this equation as h i 5.3. Time evolution of cluster radius: Leaky and starburst clusters −1/3 2/3 2/3 M tage 4 rh 2.0 ζ pc. (14) Young star clusters with M 10 M⊙ are divided ≃ M⊙ Myr into two groups, as can be seen∼ in Figure 8. Pfalzner Here we adopt t = tage because we ignore the pre-collapse (2009) named them starburst (young massive) clusters phase. The the expansion-rate coefficient, ζ, depends on and leaky clusters (following Portegies Zwart et al. (2010) the ratio of the maximum to the minimum mass in the we identify the latter category as associations). Pfalzner stellar mass function, µ mmax/mmin. In Gieles et al. (2009) and Pfalzner (2011) showed that both families (2011) we use ζ 0.2, which≡ corresponds to µ 10, and of clusters expand with time, but at a different rate: ≃ ≃ which is appropriate for globular clusters. Young clus- r/pc = 0.16(tage/Myr) for the starburst clusters and 2/3 ters however, should have a larger value of µ because of r/pc = 3.5(tage/Myr) for the associations. In this sec- the presence of massive stars. For some of these clus- tion, we discuss the origin of these different evolutionary 4 ters mmax/mmin 100 M⊙/0.01 M⊙ 10 . Following tracks. ≃ ≃ Gieles et al. (2011) and assuming ζ µ1/2, we obtain In Figure 9 we present the age and radius of observed 4 3 5 ζ 20 for µ 10 . Equation (14)∝ with ζ = 20 for in young star clusters with a mass of 10

100 100 100 1 < t < 5Myr 5 < t < 20Myr 20 < t < 100Myr

ILac1UppCen-Crux OriIa LowCen-CruxUSco OriIc IC1805 2 log 2 log 2 log − t = 2 − t = 2 − t = 2 10 = rh h 10 = rh h 10 = rh h h NGC7380 = 3 ρ h = 3 ρ h = 3 ρ ρ OriIbNGC2244NGC6611 ρ ρ log CYgOB2 log log RSGC3log log log (pc) (pc) (pc)

h h kiPer RSGC2 h log loghPer log

, r ONCQuintuplet , r , r = 0 trh = 4 = 0 trh = 4 = 0 trh = 4 ρh = 2 ρh ρh RSGC1= 2 ρh ρh = 2 ρh eff log DBS2003 eff log eff log r log r log r log Wd1 Wd2 1 NGC3603 1 1 Tr14log log log Archestrh trh trh = 1 = 1 = 1

log trh = 0 log trh = 0 log trh = 0 0.1 0.1 0.1 10 100 103 104 105 106 10 100 103 104 105 106 10 100 103 104 105 106 M (M⊙) M (M⊙) M (M⊙)

Fig. 8.— Mass-radius diagrams of observed young star clusters for tage =1–5, 5–20, and 20–100 Myr from left to right. The data is from Lada & Lada (2003) for embedded clusters (red circles), Piskunov et al. (2008); Winston et al. (2009); Luhman et al. (2003); Andersen et al. (2006); Fang et al. (2009); Levine et al. (2006); Flaherty & Muzerolle (2008); Bonatto & Bica (2011); Horner et al. (1997); Drew et al. (1997); Hodapp & Rayner (1991) for open clusters(red pluses), and Portegies Zwart et al. (2010) for young massive clusters and leaky clusters (red crosses). The data for the leaky clusters are overlapped with Pfalzner (2009). The clusters listed in Portegies Zwart et al. (2010) are shown with the names.

100

−3 UppCen-Crux h ILac1 ρ = 0.1M⊙pc OriIa USco LowCen-Crux IC1805 OriIc 10 OriIbNGC7380 NGC2244NGC6611 CYgOB2 RSGC3 (pc) h RSGC2

, r kiPer ONC Quintuplet hPer eﬀ RSGC1 DBS2003

r, r Wd1 1 Wd2 NGC3603 Tr14 Arches Our Model Pfalzner Observation 0.1 1 10 100 tage (Myr)

3 5 Fig. 9.— Cluster radius as a function of time for clusters with a mass of 10 –10 M⊙. Red plus-signs are from Portegies Zwart et al. (2010) and red pluses from the others (see the caption of Figure 8). Blue dashed lines are the relation given in Pfalzner (2011) and Pfalzner 2/3 (2009): r = 3.5tage and r = 0.16tage for top and bottom, respectively. Green lines are cluster radii as a function of time obtained from our 2/3 2/3 model: rh = 2.7tage and rh = 0.34tage for top and bottom, respectively. Note that for starburst clusters, we plot the half-mass radii given in Portegies Zwart et al. (2010) instead of “size” in Pfalzner (2009). The formation and evolution of young star clusters 15

4 for those observed in the Milky Way (10 M⊙ and 100–1000 retical models for cluster expansion due to the dynamical cm−3) lead to classical open clusters. More massive clouds evolution (Gieles et al. 2011). These models satisfactorily 5 6 (10 –10 M⊙) with the same density evolve into dense mas- explain the evolution in radius of simulated clusters as well sive clusters. These massive molecular clouds are common as the observed clusters. in starburst galaxies, but are very rare in local disk galax- We also found that the distribution of clusters on the ies such as the Milky Way. This result is consistent with mass-radius diagram is also limited by the density with observations that young massive clusters are common in which the dynamical time scale is equal to the cluster age. starburst galaxies, but only several has been found in the This implies that if the cluster age is much shorter than Milky Way. We argue that such massive clouds must be the dynamical time; such clusters cannot be recognized as able to form in the Milky Way Galaxy, even though they (bound) systems (Gieles & Portegies Zwart 2011). After are probably rare. 20 Myr the density of these associations drops below Dense massive clusters in our simulation form from the≃ background density and dissolve. 6 molecular clouds with a mass of 10 M⊙ and a density The gap of the radius distribution for associations and −3 −3 of 1000 cm (100M⊙pc ) leading to a velocity dis- young massive clusters suggested by Pfalzner (2009) is persion∼ of 20 kms−1. This is consistent with the rela- consistent with our simulation results. While young mas- tive velocity∼ of molecular clouds observed near young massive clusters evolve following the cluster expansion model, sive clusters in the Milky Way such as near NGC 3603 leaky clusters have tage tdyn. With our models, the (Fukui et al. 2014) and Westerlund 2 (Furukawa et al. evolution of radius for observed∼ leaky and young mas- 2/3 2009; Ohama et al. 2010). We argue that massive clusters sive clusters are described by rh/pc = 2.7(tage/pc) 2/3 in the Milky Way can therefore not form from individ- and rh/pc = 0.34(tage/Myr) , respectively. These are ual clouds, but their formation may have been initiated in also consistent with observations. Pfalzner et al. (2014) cloud-cloud collisions (Furukawa et al. 2009; Ohama et al. claimed that star formation continues in embedded clus- 2010; Fukui et al. 2014). ters and that after the gas expulsion they expand and 6 Molecular clouds with a mass of 10 M⊙ and a low become associations. Our models however indicate that −3 −3 ∼ density of 10cm ( 1M⊙pc ), which follow Larson’s clumpy star forming regions are observed as a conglomer- ∼ ∼ relation, tend to form associations (“leaky clusters” in the ate of embedded clusters, but at a later time these systems terminology of Pfalzner (2009)). These relatively low den- lose their clumpiness due to the expulsion of the resid- sity and massive molecular clouds form a number of small ual gas and two-body relaxation. Because our coverage of clumps. They might be detected as embedded or classical parameter space remains limited and much is still to be open clusters when they are young, but they evolve to less uncovered, we hope to explore a much wider range of ini- dense clusters due to the gas expulsion and relaxation. Af- tial conditions of molecular clouds (different masses, radii, ter several Myr, these systems lose their clumpiness and and density distributions) and other assumption for star become recognizable as associations. formation (different epochs for star formation and gradual In our simulations we assumed that stars form instan- gas removal rather than instantaneous gas expulsion). taneously upon the expulsion of the residual gas (after an Our results suggest that the difference in the parental initial free-fall time of the molecular cloud). Our prescrip- molecular clouds results in the formation of various types tion for star formation is simple compared to reality, in of star clusters if we assume the same star formation pro- which star formation triggers the expulsion of the residual cess and that the cluster formation process does not de- gas by means of feedback processes. Regardless the sim- pend on the condition of the galaxy, either normal disk or plicity of our approach, we are still able to make a distinc- starburst. tion between the formation of associations, open clusters, and massive star clusters. The young stellar system, Sco OB2, is an assembly of We thank the anonymous referee for useful comments. associations of slightly different ages, USco, Upper Cen- This work was supported by JSPS KAKENHI Grant Num- Lups, and Lower Cen-Crux. A stellar system similar ber 26800108 and NAOJ Fellowship, the Netherlands Re- to Sco OB2 naturally originates in our simulations of search Council NWO (grants #643.200.503, #639.073.803 relatively massive and low-density molecular clouds, al- and #614.061.608), and the Netherlands Research School though the age spread cannot be reproduced with our for Astronomy (NOVA). Numerical computations were method. The relation that less dense clusters have wider partially carried out on Cray XC30 at the Center for Com- age spreads of stars is observationally and theoretically putational Astrophysics (CfCA) of the National Astro- suggested (Parmentier et al. 2014). nomical Observatory of Japan and Little Green Machine In addition, we compared our simulations with theo- at Leiden Observatory.

REFERENCES

Aarseth, S. J. 1974, A&A, 35, 237 Drew, J. E., Busﬁeld, G., Hoare, M. G., et al. 1997, MNRAS, 286, Andersen, M., Meyer, M. R., Oppenheimer, B., Dougados, C., & 538 Carpenter, J. 2006, AJ, 132, 2296 Eisenstein, D. J., & Hut, P. 1998, ApJ, 498, 137 Arzoumanian, D., Andr´e, P., Didelon, P., et al. 2011, A&A, 529, L6 Fang, M., van Boekel, R., Wang, W., et al. 2009, A&A, 504, 461 Bastian, N., Adamo, A., Gieles, M., et al. 2011, MNRAS, 417, L6 Federrath, C. 2013a, MNRAS, 436, 1245 Bonatto, C., & Bica, E. 2011, MNRAS, 414, 3769 —. 2013b, MNRAS, 436, 3167 Bonnell, I. A., Bate, M. R., & Vine, S. G. 2003, MNRAS, 343, 413 Federrath, C., & Klessen, R. S. 2013, ApJ, 763, 51 Chandar, R., Whitmore, B. C., Calzetti, D., et al. 2011, ApJ, 727, Federrath, C., Roman-Duval, J., Klessen, R. S., Schmidt, W., & Mac 88 Low, M.-M. 2010, A&A, 512, A81 16 M. S. Fujii & S. Portegies Zwart

Figuerêdo, E., Blum, R. D., Damineli, A., & Conti, P. S. 2002, AJ, Luhman, K. L., Stauffer, J. R., Muench, A. A., et al. 2003, ApJ, 593, 124, 2739 1093 Flaherty, K. M., & Muzerolle, J. 2008, AJ, 135, 966 Lynden-Bell, D., & Wood, R. 1968, MNRAS, 138, 495 Fujii, M., Iwasawa, M., Funato, Y., & Makino, J. 2009, ApJ, 695, Mac Low, M.-M., & Klessen, R. S. 2004, Reviews of Modern Physics, 1421 76, 125 Fujii, M. S. 2015a, PASJ, 67, 59 McKee, C. F., & Ostriker, E. C. 2007, ARA&A, 45, 565 —. 2015b, PASJ, arXiv:1410.5540 Murray, N. 2011, ApJ, 729, 133 Fujii, M. S., & Portegies Zwart, S. 2013, MNRAS, 430, 1018 Nitadori, K., & Makino, J. 2008, New Astronomy, 13, 498 —. 2014, MNRAS, 439, 1003 Ohama, A., Dawson, J. R., Furukawa, N., et al. 2010, ApJ, 709, 975 —. 2015, MNRAS, 449, 726 Ostriker, E. C., Stone, J. M., & Gammie, C. F. 2001, ApJ, 546, 980 Fukui, Y., Ohama, A., Hanaoka, N., et al. 2013, ArXiv e-prints, Parmentier, G., & Pfalzner, S. 2013, A&A, 549, A132 arXiv:1306.2090 Parmentier, G., Pfalzner, S., & Grebel, E. K. 2014, ApJ, 791, 132 —. 2014, ApJ, 780, 36 Pelupessy, F. I. 2005, PhD thesis, Leiden Observatory, Leiden Furukawa, N., Dawson, J. R., Ohama, A., et al. 2009, ApJ, 696, L115 University, P.O. Box 9513, 2300 RA Leiden, The Netherlands Gerritsen, J. P. E., & Icke, V. 1997, A&A, 325, 972 Pelupessy, F. I., & Portegies Zwart, S. 2012, MNRAS, 420, 1503 Gieles, M., Athanassoula, E., & Portegies Zwart, S. F. 2007, MNRAS, Pelupessy, F. I., van der Werf, P. P., & Icke, V. 2004, A&A, 422, 55 376, 809 Pelupessy, F. I., van Elteren, A., de Vries, N., et al. 2013, A&A, 557, Gieles, M., Heggie, D. C., & Zhao, H. 2011, MNRAS, 413, 2509 A84 Gieles, M., & Portegies Zwart, S. F. 2011, MNRAS, 410, L6 Pfalzner, S. 2009, A&A, 498, L37 Girichidis, P., Federrath, C., Allison, R., Banerjee, R., & Klessen, —. 2011, A&A, 536, A90 R. S. 2012, MNRAS, 420, 3264 Pfalzner, S., & Kaczmarek, T. 2013, A&A, 559, A38 Girichidis, P., Federrath, C., Banerjee, R., & Klessen, R. S. 2011, Pfalzner, S., Parmentier, G., Steinhausen, M., Vincke, K., & Menten, MNRAS, 413, 2741 K. 2014, ApJ, 794, 147 Gürkan, M. A., Freitag, M., & Rasio, F. A. 2004, ApJ, 604, 632 Piskunov, A. E., Kharchenko, N. V., Schilbach, E., et al. 2008, A&A, Heggie, D. C. 1975, MNRAS, 173, 729 487, 557 Hénon, M. 1965, Annales d’Astrophysique, 28, 62 Portegies Zwart, S., McMillan, S. L. W., van Elteren, E., Pelupessy, Hernquist, L., & Katz, N. 1989, ApJS, 70, 419 I., & de Vries, N. 2013, Computer Physics Communications, 183, Heyer, M. H., & Brunt, C. M. 2004, ApJ, 615, L45 456 Higuchi, A. E., Kurono, Y., Saito, M., & Kawabe, R. 2009, ApJ, 705, Portegies Zwart, S., McMillan, S., Harfst, S., et al. 2009, New A, 14, 468 369 Hodapp, K.-W., & Rayner, J. 1991, AJ, 102, 1108 Portegies Zwart, S. F., & McMillan, S. L. W. 2002, ApJ, 576, 899 Holmberg, J., & Flynn, C. 2000, MNRAS, 313, 209 Portegies Zwart, S. F., McMillan, S. L. W., & Gieles, M. 2010, Horner, D. J., Lada, E. A., & Lada, C. J. 1997, AJ, 113, 1788 ARA&A, 48, 431 Hurley, J. R., Pols, O. R., & Tout, C. A. 2000, MNRAS, 315, 543 Roman-Duval, J., Federrath, C., Brunt, C., et al. 2011, ApJ, 740, Hut, P. 1983, ApJ, 272, L29 120 Inoue, T., & Fukui, Y. 2013, ApJ, 774, L31 Salpeter, E. E. 1955, ApJ, 121, 161 Kim, W.-T., & Ostriker, E. C. 2002, ApJ, 570, 132 Spitzer, L. 1987, Dynamical evolution of globular clusters King, I. R. 1966, AJ, 71, 64 Spitzer, L. J., & Hart, M. H. 1971, ApJ, 164, 399 Krumholz, M. R., Dekel, A., & McKee, C. F. 2012, ApJ, 745, 69 Springel, V., & Hernquist, L. 2002, MNRAS, 333, 649 Krumholz, M. R., & McKee, C. F. 2005, ApJ, 630, 250 Winston, E., Megeath, S. T., Wolk, S. J., et al. 2009, AJ, 137, 4777 Lada, C. J., & Lada, E. A. 2003, ARA&A, 41, 57 Wolff, S. C., Strom, S. E., Dror, D., & Venn, K. 2007, AJ, 133, 1092 Larson, R. B. 1981, MNRAS, 194, 809 Zinnecker, H., & Yorke, H. W. 2007, ARA&A, 45, 481 Levine, J. L., Steinhauer, A., Elston, R. J., & Lada, E. A. 2006, ApJ, 646, 1215