Multiwavelength Studies of Young OB Associations
aa r X i v : . [ a s t r o - ph . S R ] A p r Chapter 1
Multiwavelength Studies of Young OBAssociations
Eric D. Feigelson
Abstract
We discuss how contemporary multiwavelength observations ofyoung OB-dominated clusters address long-standing astrophysical questions:Do clusters form rapidly or slowly with an age spread? When do clustersexpand and disperse to constitute the field star population? Do rich clustersform by amalgamation of smaller subclusters? What is the pattern and du-ration of cluster formation in massive star forming regions (MSFRs)? Pastobservational difficulties in obtaining good stellar censuses of MSFRs havebeen alleviated in recent studies that combine X-ray and infrared surveysto obtain rich, though still incomplete, censuses of young stars in MSFRs.We describe here one of these efforts, the MYStIX project, that produced acatalog of 31,784 probable members of 20 MSFRs. We find that age spreadwithin clusters are real in the sense that the stars in the core formed afterthe cluster halo. Cluster expansion is seen in the ensemble of (sub)clusters,and older dispersing populations are found across MSFRs. Direct evidencefor subcluster merging is still unconvincing. Long-lived, asynchronous starformation is pervasive across MSFRs.
Galactic Plane star clusters, well-known to classical astronomers like 2 nd century Claudius Ptolemy and 10 th century Abd al-Rahman al-Sufi, werecatalogued in the 18-19 th centuries by Charles Messier and William and JohnHerschel. As astrophysical explanations for astronomical phenomena rose toprominence around the turn of the 20 th century, it was natural that theprocesses giving rise to clusters were investigated. We address here several Eric D. FeigelsonDepartment of Astronomy & Astrophysics, Pennsylvania State University, University ParkPA 16802, e-mail: [email protected] astrophysical themes of long-standing importance where, even today, theoryis not well-constrained by observation.The historically oldest issue is the argument that most stars are born inclusters that expand and disperse to comprise the field star population. Ina 1917 discussion of Kapteyn’s ‘systems of stars which travel together inparallel paths’, Charlier [9], director of Lund Observatory in Sweden, argues that the stars which now belong to such a system are only the insignificant remnantof a large cluster which at one time constituted a compact system in space.
Such questions could be investigated computationally, both by integratingdifficult differential equations and by Monte Carlo N-body calculations, inthe 1970s. In an important study, Tutukov [65] wrote:
It is generally believed that ... stars [form] in small groups which dissolve compar-atively quickly during very early stages of evolution, practically at the moment oftheir formation. ... It is natural to suppose that the gas not utilized for star forma-tion was blown away by hot stars, probably due to the ionizing radiation and stellarwind. If the mass of gas is higher than the mass of stars and the kinetic energy ofthe gas exceeds the binding energy of the cluster, then the disruption of a youngcluster seems inevitable.
The issue of stellar dispersal arose again when early-type stars were dis-covered far from their natal clouds away from the Galactic Plane. Greenstein& Sargent [24] noted:
The kinematical behavior of these stars is, however, quite strange ... The stars arenot kinematically relaxed; they are apparently observed soon after formation andejection. ... [This reveals] a fundamental problem that far too many, hot, high-velocity, apparently normal stars exist.
Some of these stars are clearly runaway stars ejected at high velocities fromhard binary interactions, but others some dispersed up to ∼
200 pc from thePlane could not easily be traced to rich clusters [11]. In a catalogue of stellarmembers in OB associations within 3 kpc, Garmany & Stencel [19] foundthat massive OB stars are commonly spread over large ( ∼
200 pc) regions;these did not appear to be high-velocity runaways.Another long-standing issue concerns the mechanism by which rich starclusters form. Aarseth & Hills [1] sought to evaluate two alternatives views:simultaneous formation of a monolithic rich cluster and its possible laterconstruction from pre-existing subclusters. They wrote:
The density distribution of stars in a stellar cluster usually gives every appearance ofbeing smoothing varying and non-clumpy. On the face of it, this is a bit surprisingsince elementary considerations from [Jeans gravitational collapse] star-formationtheory suggest that a cluster should initially be subdivided into a hierarchy of sub-clusters. ... The subdivision process terminates when the cloud becomes opaqueenough for the collapse time-scale to catch up with the cooling time-scale ... [so]that a cluster is initially composed of a hierarchy of subclusters.
Stellar subgroups were empirically found in a number of nearby rich OB as-sociations by Blaauw [6]. But it was unclear whether the primary process is
Multiwavelength Studies of Young OB Associations 3 fragmentation of an initially homogeneous cluster, or incomplete consolida-tion of smaller subclusters into a unified structure. The latter view came tothe fore when molecular clouds were discovered to be highly inhomogeneousdue to supersonic turbulence [41]. Maps obtained with the Herschel satellitefar-infrared imaging show that even the coldest and densest cloud structuresmostly have clumpy and filamentary structure [3].A third contentious issue is the duration of star formation in molecu-lar clouds. Various researchers argue, on both physical and observationalgrounds, that cluster formation is rapid, although a small number of starsmay form over an extended period before the principal starburst [12, 47, 28].Others suggest that regulation of star formation by magnetically induced tur-bulence in molecular clouds and feedback from nascent stars prevents large-scale free-fall gravitational collapse and rapid cluster formation [41, 5, 31, 32].The evidence outlined above for widely distributed early-type stars suggeststhat star formation in massive star forming regions are long-lived, so that ear-lier generation of massive stars have time to drift outward from still-activestar forming regions.
It is now clear that most stars form in rich clusters. The cluster luminosityfunction in the Milky Way Galaxy and nearby galaxies demonstrates thatthe majority of stars form in clusters with 10 − stars [40] and, dur-ing galactic starburst episodes, superclusters of 10 stars may dominate. Buteven the fundamental physical properties, processes and timescales of clusterformation and early evolution are observationally poorly established. Cogentarguments have been made that clusters form quickly [12] and slowly [61],that they form as a unified structure or are assembled from merging subclus-ters [43, 5], that they form in spherical cloud cores or in filamentary cloudstructures [56, 3]. Timescales for cluster formation and early dynamical evo-lution are poorly constrained by observation. Attempts to measure the agesof constituent stars of nearby clusters by fitting their location in Hertzprung-Russell diagrams (HRDs) to theoretical evolutionary tracks is beset withobservational difficulties, so that it is unclear whether the observed spreadsin HRDs represent true age spreads [54].The reasons for the failure to test competing astrophysical models of clus-ter formation can arguably be placed on practical observational difficulties indefining their member stars. Much progress has been made in studying theprogenitor molecular clouds through, for example, maps of coolant molecularlines with millimeter array telescopes and far-infrared imaging of continuumdust emission with the Herschel satellite. The environmental effects of the hotOB stars can also be traced across the Galactic Plane: ionized gas is easilymapped at radio wavelengths, and heated dust produces PAH band emission Eric D. Feigelson mapped with infrared space telescopes. But the actual stellar populations ofstar clusters beyond distances ∼ d ∼ b ∼ ◦ and longitudes in the inner quadrants, field stars have 10 −
100 timeshigher surface density than the cluster members over most of the cluster ex-tent at near-infrared magnitudes around the peak of the Initial Mass Func-tion. Interstellar absorption can reach A V ∼
30 mag along the line-of-sightto the cluster, and can vary by tens of magnitude within the star formingregion due to the local molecular cloud. Detection of faint infrared stars isdifficult amid the nebular H II region emission from heated dust.As a result of these problems, the census of young star cluster members hasoften been restricted to nearby lower-mass clusters or to special subpopula-tions of massive clusters: the inner cluster core where the surface density risesabove the field stars; OB stars that are brighter and bluer than ambient starsand easily confirmed with optical spectroscopy; and pre-main sequence starswith photometric infrared excesses (IRE) from dust protoplanetary disks.The IRE criterion is often used to define the population of ‘young stellarobjects’ (YSOs) but it is restricted to disk-bearing pre-main sequence stars(Class I-II). In many clusters, the bulk of the stars have lost their disks andare thus photometrically indistinguishable from contaminant field stars in theinfrared bands. Inferences regarding star formation histories may be flaweddue to the IRE sample bias towards younger systems with hot inner accretiondisks.However, a technique has emerged in recent years that overcomes, tosome degree, these observational difficulties and biases. Sensitive and high-resolution imaging of star forming regions with NASA’s Chandra X-ray Ob-servatory, sensitive in the 0 . − − . d ∼ − − A V ∼ Multiwavelength Studies of Young OB Associations 5 mag in some cases. Finally, X-ray selection is complementary to IRE selectionbecause it most efficiently captures disk-free (Class III) stars.The remainder of this chapter discusses a particular effort called MYS-tIX (Massive Young Stellar complexes study in Infrared and X-rays) thatcombines Chandra X-ray, UKIRT near-infrared, and Spitzer Space Telescopemid-infrared surveys of 20 OB-dominated star forming regions at distances0 . < d < ∼ § § § § The MYStIX effort seeks to construct an improved census of stars in richclusters and their environs in 20 MSFRs near the Sun. Populations that arenot dominated by an O or early-B star are omitted; thus MYStIX omitsnearby small star forming regions like the Taurus-Auriga, ρ Ophiuchi andChamaeleon complexes. Table 1 lists the MYStIX star forming regions withapproximate distance from the Sun and spectral type of the dominant star.The accompanying Figure 1 shows the location of the MYStIX regions on adiagram of the Milky Way Galaxy with the Sun at the middle. The MYStIXtargets do not constitute a complete sample in any way, but rather wereselected by practical considerations: they must have sufficiently deep coverageby the Chandra and Spitzer satellite imagers.
Eric D. Feigelson
Table 1
MYStIX Star Forming RegionsRegion D kpc * Region D kpc *Orion Neb 0.4 O7 NGC 6334 1.7 O8:Flame Neb 0.4 O8: NGC 6357 1.7 O3W 40 0.5 O: Eagle Neb 1.8 O9RCW 36 0.7 O8 M 17 2.0 O4NGC 2264 0.9 O7 W 3 2.0 O5Rosette Neb 1.3 O4 W 4 2.0 ...Lagoon Neb 1.3 O4 Carina Neb 2.3 O2NGC 2362 1.5 O9I Trifid Neb 2.7 O7DR 21 1.5 ... NGC 3576 2.8 O:RCW 38 1.7 O5 NGC 1893 3.6 O5
Figure 1. Galactic location of MYStIXstar forming regions (triangles)
Simply stated, the MPCM samples are the sum of probable complexmembers extracted from X-ray sources in the Chandra X-ray Observatoryimages, IRE sources from UKIRT near-infrared observations (often part ofthe UKIDSS Galactic Plane Survey) and the Spitzer Space Telescope mid-infrared observations, and published OB stars confirmed by published opticalspectroscopy. But the actual procedure for constructing the MPCM samplesis complicated by the need to reduce the often-overwhelming contaminationof Galactic field stars combined with spatially variable cloud absorption andnebular emission. Challenges overcome include:
X-ray source lists were obtained using the
ACIS Extract package and asso-ciated software developed for the Chandra ACIS instrument at Penn State[34, 64]. This allows detection of sources with as few as 3 − Near-infrared source lists were obtained with the UKIDSS pipeline soft-ware modified to accommodate very crowded Galactic plane fields withnebulosity [30].
Mid-infrared source lists were obtained with the Spitzer IRAC team soft-ware modified to accommodate crowding and nebulosity [35].
X-ray/infrared counterpart identifications were based on a probabilis-tic calculation of proximate sources that accounts for the magnitude dis-tribution expected for true complex members, in order to reduce falseassociations with fainter field stars [45].
Infrared excess stars were extracted based on a complicated decision treeof criteria designed to reduce the often-heavy contamination by field redgiants and false sources associated with nebular knots [52].
Multiwavelength Studies of Young OB Associations 7
The classified X-ray sources, IRE stars and published OB stars were thencombined into the MPCM catalog of 31,784 stars in the 20 regions of Table 1[8]. The MYStIX papers, and their electronic tables of intermediate and finalsamples, are collected at the Web site http://astro.psu.edu/mystix.The MPCM sample is far from a complete census. The X-ray samples aregenerally limited to stars with masses above ∼ . ⊙ , and thus miss the peakof the IMF of low-mass members. Various biases are present in the sample aswell (see Appendix B of Feigelson et al. 2013). Nonetheless, the MPCM sam-ples are the largest for most of the star forming regions under consideration.Tests of the sample reliability were made using the well-studied NGC 2264population; ∼
80% of previously identified H α and optically variable starswere recovered, and dozens of new members are proposed [17].Figures 1.2-1.3 illustrate the MPCM samples for four MYStIX star form-ing regions. The regions have complex structures though with some similarbehaviors. Lagoon Nebula (M 8)
In this MSFR, we see two major clusters: thepoorly characterized NGC 6523 cluster to the east with the famous mas-sive star Herschel 36; and the well characterized NGC 6530 cluster ina large cavity to the west. As one proceeds westward, the fraction ofIRE stars (red circles in Figure 1.2) decreases; it is not immediately clearwhether this is an age gradient or a selection effect due to the difficultyof finding IRE stars in the bright PAH nebulosity of the western region.A clump of stars is also seen to the far-southeast associated with a brightrimmed cloud; it includes the luminous embedded star M 8E.
NGC 6334
This is a large MSFR elongated along the Galactic Planewith both heavy absorption and complex bright nebular emission thatprecluded generation of a reliable stellar census in the past. The 1,667-member MPCM sample shows several distinct clusters, some dominatedby young IRE stars and others by older X-ray selected stars [15]. The mor-phology might represent a star formation wave from the southwest to thenortheast, but older clusters are sometimes superposed on younger clus-ters and a distributed young star component is also present. A selection oflikely protostars, based on MYStIX sources with ascending infrared spec-tral slopes or ultra-hard X-ray spectra, shows a distribution of very youngstars tracing the curved molecular filament to the northeast [59].
NGC 6357
This region has 2,235 MPCMs, very few of which had previ-ously been identified by optical or infrared surveys even though this isa very active star forming region in the Carina spiral arm. Three veryrich clusters are seen; Pismis 24 to the northwest has several ∼
100 M ⊙ O3 stars. In each cluster, we can see spatial displacements between theinfrared and X-ray selected subsamples. The IRE selection method is inef-fective around the brightest nebular emission of the northwest H II region.Two dozen new absorbed (4 < A V <
24 mag) candidate OB stars areidentified in the MYStIX catalog in this region [53].
Eric D. Feigelson
05 05 05 04 04 18:04 04 04 03 0312141618-24:20222426283032
Lagoon Nebula
NGC 6334
Fig. 1.2
Spatial location of MYStIX Probable Complex Members (MPCMs) for the La-goon Nebula and NGC 6334 fields [8]. Infrared excess stars are noted by red circles, X-rayselected stars by yellow dots, and published OB stars by cyan circles. The stars are super-posed on Spitzer IRAC 8.0 µ m maps. Each Chandra field subtends 17 ′ × ′ . Multiwavelength Studies of Young OB Associations 9 Fig. 1.3
The MPCMs in the NGC 6357 and Eagle Nebula complexes superposed onSpitzer IRAC 8 µ m maps [8]. Yellow dots are X-ray selected members, red circles areinfrared-excess members, and cyan symbols are published OB stars. Fig. 1.4
Statistically defined subclusters in the NGC 6357 and Eagle Nebula complex areshown as black ellipses superposed on smoothed maps of the MPCM stellar distribution.
Eagle Nebula (M 16)
Here the southwestern rich cluster is dominatedby disk-free X-ray selected members, while the sparser subclusters to thenorth and west are dominated by disk-bearing IRE members. As in mostMYStIX regions, the X-ray selected stars outnumber the IRE stars, imply-ing that the star formation has endured for many millions of years beyondthe typical longevity of infrared-emitting disks.
To reveal the spatio-temporal history of star formation in MYStIX regions,it would be very desirable to obtain reliable ages of different (sub)clusters ofMPCM stars. Two pre-main sequence chronometers are traditionally used:a star’s location in the HRD compared to theoretical evolutionary tracks;and the presence of a star’s infrared-emitting circumstellar disk [27, 57]. Butneither are very effective for MSFRs. HRD locations are not available be-cause the stars are often too reddened to readily obtain optical spectra, andin any case several extraneous problems render HRD-derived ages uncer-tain [54]. Disk fractions or classification (Class 0-I-II-III) derived from in-frared photometry are inaccurate and difficult to calibrate. For example, IREpopulations are reduced by local H II region contamination, differences ininfrared-to-X-ray sensitivities can systematically bias disk fraction compar-isons between MYStIX regions, and individual disk dissipation timescalesrange over 0 . − L x , produced by magnetic re-connection flares, and stellar mass M in pre-main sequence stars. This L x − M relation is best calibrated in the Taurus-Auriga population [62]. The astro-physical cause of this correlation is poorly understood (presumably relatedto magnetic dynamos in fully convective stellar interiors), but it accountsfor much of the 10 range of L x in young stellar populations. MYStIX L x measurements, after correction for soft X-ray absorption from interveninginterstellar gas, thus give mass estimates for each star. MYStIX also givesmeasures photospheric luminosities L bol ; Getman et al. use dereddened J band magnitudes M J as a proxy for L bol . M values inferred from L x andmeasured M J values combined with standard theoretical evolutionary tracksgive stellar age estimates for each star, nicknamed Age JX . Each Age JX valueis quite inaccurate, but obtaining the median Age JX for a spatially definedsubsample of young stars appears to be effective for elucidating histories ofstar formation within and between clusters. The MYStIX fields are mostly centered on rich OB associations with opti-cally bright H II region, often with names like ‘Rosette Nebula’ and ‘LagoonNebula’ that date to the 19 th century. But examination of the MPCM spatial Multiwavelength Studies of Young OB Associations 11 distributions show considerable diversity in clustering behavior − a simpledichotomy between rich clusters and distributed star formation is clearlyinadequate. Global statistics of spatial point processes, such as Ripley’s K function and the related two-point correlation function [29], are not directlyuseful as they are strongly affected by the richest clusters and do not reflectthe diversity of patterns within a single field. Defining stellar ‘clusters’ or‘groups’ by surface density enhancements [16] also has the disadvantage ofrequiring an arbitrary threshold.We therefore proceeded to locate ‘clusters’ using a parametric statisti-cal regression approach known as ‘mixture models’ [42]. Here we require thatcluster structure have a specific mathematical form corresponding an isother-mal sphere or ellipsoid [36]. A likelihood function giving the probability thatthe observed celestial locations of MPCM stars corresponds to a specifiedmixture of isothermal ellipsoids. When a flat ‘distributed’ stellar populationis added, a model with k clusters has 6 k + 1 parameters. The best fit modelis obtained by maximum likelihood estimation for a range of k , and the op-timal number of clusters is obtained by maximizing the Akaike InformationCriterion, a well-accepted penalized likelihood measure for model selection.Note that the method permits hierarchical structures with one ellipsoid ly-ing within or overlapping another ellipsoid. Model fits are generally excellentwith no strong features in the residual spatial maps. The resulting spatial de-compositions for the NGC 6357 and Eagle Nebula MYStIX fields are shownin Figure 1.4.The result of this analysis is the assignment of each of the ∼ J − H color index andthe sample median of the individual median energies of the X-rays from theconstituent stars. For example, the Eagle Nebula has 12 statistically signif-icant subclusters (Fig. 1.4) with sample populations ranging from 7 to 451MPCM stars, core radii from 0.07 pc to 1.0 pc, ellipticities from 7% to 64%,and absorptions from A V ∼ Age JX valuesof the constituent stars (Sec. 1.4). Ages for the Eagle (sub)clusters rangefrom 0.8 to 2.4 Myr. Second, the total stellar population can be inferred byscaling the sample X-ray luminosity function (truncated at different limitingX-ray sensitivities) to the full-sampled X-ray luminosity function of the OrionNebula Cluster [37]. The total populations inferred from X-ray luminosityfunctions agree well with a parallel analysis based on dereddened J bandmagnitudes scaled to a standard Initial Mass Function. Combining the estimated total population with the (sub)cluster structuralparameters like core radius, unbiased estimates can be made of importantquantities such as total stellar mass (in M ⊙ ), central surface densities (instars/pc ), central volume densities (in stars/pc ), characteristic crossing andrelaxation times (in Myr) [38]. Comparisons of MPCM stellar spatial distributions in maps like Figs. 1.3-1.4 can be misleading due to inhomogeneity in sensitivity. This particularlyaffects the X-ray measurements. First, within each Chandra ACIS field thesensitivity is highest at the field center and degrades by a factor of ∼ L x -Mass relationship) and mass function aretruncated at different levels.However, as outlined in Sec. 1.5, these problems can be overcome [37].We first ‘flatten’ the intra-ACIS sensitivity variation by omitting the faintsources near the field center. The stellar surface densities are then normalizedto the full IMF assuming all regions have the same intrinsic X-ray luminosityfunction. Although the lower mass stars missed by Chandra cannot be indi-vidually identified, the surface densities can be scaled upward to compensatefor the different truncation levels. Note it is more difficult to corrected themaps for variations in the surface densities of IRE sources, which are deficientin the brightest H II nebular regions.The result is Fig. 1.5 a remarkable new view of the stellar distributions inmassive star forming clouds [37]. The densities correspond to the full intrin-sic stellar populations down to the M ∼ .
08 M ⊙ limit shown on a uniformphysical scale (see the 5 pc scale bar) and a uniform color scale in stars/pc (see color calibration bar). We find, for example, that the both the embeddedclusters and the revealed massive cluster of the Rosette Nebula region havelow surface densities of 10 stars/pc . But the RCW 38, Orion Nebula Clus-ter, and M 17 clusters have extremely high central surface densities around10 stars/pc .Diversity, rather than consistency, is the premier result from these surfacedensity maps. The main Rosette Nebula cluster NGC 2244 must be in acompletely different dynamical state than the RCW 38 or W 40 clusters; andindeed this may be related to the complete absence of mass segregation inNGC 2244 [66]. Until these maps were compared, it was not realized that Multiwavelength Studies of Young OB Associations 13
Fig. 1.5
Montage of maps of surface density of X-ray selected stars in MYStIX regions,shown to the same physical scale in parsecs [37]. These intrinsic surface densities have beencorrected for variations of X-ray sensitivities within and between fields, with density valuesscaled to the full IMF.
RCW 38 (which is badly contaminated in the IR bands due to nebulosity)has the densest collection of stars of any cluster in the nearby Galaxy. It thusprovides an excellent laboratory to study dynamical effects of close stellarencounters [51, 50].The MYStIX maps showing of a wide range of central surface densities, < to ∼ × stars/pc (Fig. 1.5), stands in conflict with the findings ofBressert and colleagues who report that young stellar clusters exhibit a char-acteristic central surface density distribution with mean around 20 stars/pc [7]. Their study is limited to nearby molecular clouds where clusters are gen-erally small and, most importantly, their sample is limited to IRE stars andthus miss the disk-free X-ray selected stars that dominate many star form- ing regions. The MYStIX findings on stellar surface densities, although stillsubject to limitations and biases, are probably more reliable than the moreconstrained IRE-only results. We now discuss how the MYStIX project – specifically the MPCM sample of31,747 young stars in 142 (sub)clusters associated with 20 MSFRs – addressesthe astrophysical questions outlined in § The MYStIX dataset shows many cases of the expected range of clusterstructures: compact clusters embedded in their molecular cores, larger clus-ters following molecular gas ejection, and older stars dispersing into the fieldpopulation.Direct evidence for cluster expansion is shown in Fig. 1.6 [36, 38]. The firstpanel shows that MYStIX (sub)cluster core radii systematically increase asclusters range from heavily absorbed to lightly absorbed. The X-ray medianenergy range is roughly equivalent to 0 < A V <
40; the same result is seenusing J − H as an absorption measure. The other panels show show therelationships between core radii or central density and median Age JX valuesfor the subclusters. Here we see roughly a factor of 10 increase in radius,and a factor of 1000 decrease in central core density, as (sub)clusters agefrom ∼ . ∼ −
20 Myr.
Multiwavelength Studies of Young OB Associations 15
Fig. 1.6
Expansion of young clusters.
Left:
Subcluster core radius vs.
X-ray median en-ergy, a measure of interstellar absorption with nonparametric regression curve [36].
Centerand right:
Bivariate scatter diagrams showing subcluster radius and central density vs.Age JX with nonparametric regression curve [38]. Pre-MYStIX studies had reported that X-ray selected stars, includingearly-type OB stars, were often dispersed from the molecular cores that activeform stars today [15]. In the Carina complex, half of the X-ray stars lie out-side the regions dominated by the Trumpler 14-15-16 clusters and the SouthPillars clouds [16]. This pattern is seen in most MYStIX regions. Dispersedstellar surface densities range from near-zero to tens of stars/pc in the dif-ferent regions [36, 38]. Age JX analysis shows that, in nearly all cases, the dis-persed stars are older (typically 3 to > The MYStIX (sub)cluster sample gives ample opportunity to reveal mergingof smaller subclusters as an important process of building up large equi-librated clusters as predicted in models of cluster formation in turbulentmolecular clouds [5]. Yet the evidence is unclear.First, consider the geometric properties of MYStIX (sub)clusters withoutinclusion of physical quantities such as age and mass [36]. As exemplifiedin NGC 6357 and the Eagle Nebula decompositions in Fig. 1.4, some richclusters are consistent with simple smooth ellipsoidal stellar distributions,while others are clumpy and require several ellipsoids to be adequately mod- eled. Fig. 1.7 is a diagram of the ellipsoidal structures in 15 MYStIX regionsplaced into a heuristic classification of simple, linear chain, core-halo, andcomplex clumpy classes [36]. As in § Fig. 1.7
Heuristic classification of the star formation complex morphology of 15 MYStIXregions based on the ellisoidal subcluster spatial decomposition [36]. Regions are shown onthe same physical scale (see the 5 pc calibration bar), and line thickness is scaled to thesubcluster stellar surface density of each subcluster.
It is tempting to interpret the morphological classes as an evolutionarysequence where star formation begins as linear chains in filamentary clouds,passes through a clumpy stage as subclusters merge, and ends with core-halo and simple structures that may be in dynamical equilibrium. However,when
Age JX values are examined for these morphological classes, no evidencefor an evolutionary sequence is found [38]. Perhaps linear morphologies (likeDR 21 and NGC 2264) disperse rather than merge into simpler spherical mor-phologies (like W 40 and the three clusters of NGC 6357). However, it seems Multiwavelength Studies of Young OB Associations 17 physically reasonable to suggest that the dense but clumpy configuration ofM 17 will equilibrate into a unified rich cluster.A second failure to detect (sub)cluster merging is from a scatter plot oftotal stellar population vs.
Age JX for MYStIX (sub)clusters. No indicationof cluster population growth is seen [38]. It is possible that the statisticaldecomposition of stellar clustering into 142 isothermal ellipsoids masqueradesa growth effect.A third test, however, gives a hint of cluster growth. A strong anti-correlation between (sub)cluster central star densities and core radii naturallyappears in ensembles of young clusters. A relationship ρ ∝ r − c is expectedfrom a collection of clusters of uniform and constant mass seen at differ-ent phases of expansion. The MYStIX sample shows ρ ∝ r − . ± . c over therange 0 . ≤ r c ≤ . ≤ log ρ ≤ [38]. This relationshipappears shallower than a − The MYStIX and related studies give unequivocal evidence that long-livedstar formation is pervasive, both across MSFRs and within rich clusters.The acquisition of
Age JX estimates for dozens of spatially well-defined(sub)clusters allows us to study the history of star formation across MYStIXstar forming regions. Getman and colleagues find a clear and consistent pat-tern: more heavily absorbed clusters have younger ages than lightly absorbedclusters [21]. This is shown for two MYStIX regions in Fig. 1.8, RCW 36with a ‘simple’ structure and Rosette Nebula with a ‘complex’ structure. InRCW 36 the ages range from 0.9 to 1.9 Myr, while in Rosette they range from1 to 4 Myr. Ages are also available for stars that are not assigned to clusters;these distributed stars always show older ages than absorbed clusters. These results confirm with widespread belief that clusters are formed in-side dusty molecular cores (high J − H color environments) and later expeltheir molecular material (low J − H environments). But there were few quan-titative measures of this expectation prior to the MYStIX analysis. Previousdemonstrations of age gradients were based on spatial correlations betweenClass I-II-III (disk-bearing to disk-free) populations and absorption in theW 40 and Rosette Nebula regions [33, 69]. Both of these quantities are notcalibrated to age in Myr, and the situation is often not so simple; in the OrionL1541 cloud, for example, two clusters dominated by older disk-free stars arelightly obscured while one is heavily obscured [48]. Fig. 1.8
Age differences for selected MYStIX subclusters as a function of J − H colorindex, a measure of cloud absorption, in the RCW 36 and Rosette Nebula fields [21]. Eachpoint refers to subclusters identified in [36]. The ‘U’ designation refers to unclustered stars. A more surprising result is the age spread, and spatial age gradient, foundby Getman and colleagues within two nearby rich clusters, in addition tothe gradients found earlier between (sub)clusters [22]. The cluster cores are
Multiwavelength Studies of Young OB Associations 19
Fig. 1.9
Age differences within the Flame Nebula and Orion Nebula Clusters showingthat the cores are younger than the halos [22]. much younger than the cluster outer regions (Fig. 1.9). In the Flame Nebulacluster, stars within 0.2 pc of the center are 0.2 Myr old while stars 1 pcfrom the center are 1.6 Myr old. In the Orion Nebula cluster, the age rangesfrom 1.2 Myr to 2.0 Myr. This measurement is based entirely on analysis ofsolar-type stars, and thus does not conflate age and mass segregation.The result is startling because naive models for cluster formation (basedon Jeans gravitational collapse in an isothermal cloud core) expect that starfill form first in the dense center, and thus would later appear to have theoldest, not the youngest stars. Other models tend to homogenize the youngerand older stars during a subcluster merging process [5]. More complex clusterformation scenarios might explain the observed phenomenon; for example, theolder stars may have kinematically dispersed from the core, and/or the coremay have been supplied with infalling molecular gas to allow star formationafter the gas was depleted in the halo [22].But the MYStIX intracluster age gradient also resolves a long-standingcontroversy concerning apparent stellar age spreads in HRDs [54]. The agespread appears to be real, at least in part, because it represents a spatialsegregation of older and younger stars. Thus models based on rapid clusterformation in a single collapse time [12] are not consistent with the findings,at least for the rich clusters in the Orion cloud complex.
We emerge with some optimism that a frustrating period is ending whenmodels for clustered star formation were largely unconstrained by empiri-cal results concerning the outcomes of star formation processes (properties of the young stellar populations) to complement empirical results on theinputs to star formation processes (molecular cloud properties). A multi-wavelength approach provides the key: X-ray surveys to isolate the pre-mainsequence population from the contaminating field star population and toavoid strong nebular emission; near-infrared imaging replacing optical obser-vations to penetrate regions of high absorption; and mid-infrared photometryto discriminate the important subpopulation of disk-bearing young stars fromoften-overwhelming Galactic field star contamination.The diversity of clustering patterns found in MYStIX regions (Fig. 1.7)points to the importance of studying star formation in multiple environments.The observational strategy of MYStIX can easily be extended to more starforming regions in the nearby (roughly distances < ∼
20 regionswith distances ≤ . − . . < M < . ⊙ ), and to the richest star formingregions of the Galaxy lying ∼ −
12 kpc from the Sun. Million-second Chan-dra exposures are needed to acquire sufficient X-ray sensitivity, and infraredfollowup requires both high resolution and high sensitivity. Fortunately, theChandra satellite is in good health since launch in 1999 and is likely to lastfor a considerable time into the future. Infrared technologies are continu-ously improving: the VISTA Via Lactea project gives wide-field, multi-epochphotometry of large portions of the the Galactic Plane [44]; the KMOS andMOSFIRE multi-object spectrographs offer efficient spectroscopic capabili-ties on 8-meter class telescopes; and the James Webb Space Telescope willgreatly advance infrared imaging and spectroscopy in a few years. Theseobservational capabilities give confidence that fruitful interactions betweentheory and observations can become the norm in the study of clustered starformation.
Acknowledgements
This review rests on the labor and talents of the MYStIX team,particularly Patrick Broos, Konstantin Getman, Michael Kuhn, Tim Naylor, MatthewPovich, and Leisa Townsley. Many of the astrophysical results appear in the dissertationof Michael Kuhn and work led by Kostantin Getman; the author is especially grateful fortheir collaborative energy and thoughtful analysis. The MYStIX Project was principallysupported at Penn State by NASA grant NNX09AC74G, NSF grant AST-0908038, andSAO/CXC grant AR7-18002X and ACIS Team contract SV-74018.
References
1. Aarseth, S. J., & Hills, J. G. 1972,
Astro. Astrophys , 21, 2552. Adams, F. C., Hollenbach, D., Laughlin, G., & Gorti, U. 2004,
Astrophs. J. , 611, 360 Multiwavelength Studies of Young OB Associations 213. Andr´e, P., Men’shchikov, A., Bontemps, S., et al. 2010,
Astron. Astrophys. , 518, LL1024. Banerjee, S., & Kroupa, P. 2013,
Astrphys. J. , 764, 295. Bate, M. R. 2009,
Mon. Not. Royal Astro. Soc. , 392, 5906. Blaauw, A. 1964,
Ann. Rev. Astro. Astrophys. , 2, 2137. Bressert, E., Bastian, N., Gutermuth, R., et al. 2010,
Mon. Not. Royal Astro. Soc. ,409, L548. Broos, P. S., Getman, K. V., Povich, M. S., et al. 2013,
Astrophys. J. Suppl. , 209, 329. Charlier, C. V. L. 1917, The Observatory, 40, 38710. Damiani, F., Micela, G., Sciortino, S. 2016,
Astron. Astrophys. , 596,
Astro. Astrophys. , 437, 24712. Elmegreen, B. G. 2000,
Astrophys. J. , 530, 27713. Feigelson, E. D., & Montmerle, T. 1999,
Ann. Rev. Astron. Rev. , 37, 36314. Feigelson, E. D., & Townsley, L. K. 2008,
Astrophys. J. , 673, 35415. Feigelson, E. D., Martin, A. L., McNeill, C. J., et al. 2009,
Astron. J. , 138, 22716. Feigelson, E. D., Getman, K. V., Townsley, L. K., et al. 2011,
Astrophys. J. Suppl. ,194, 917. Feigelson, E. D., Townsley, L. K., et al. 2013,
Astrophys. J. Suppl. , 209, 2618. Figer, D. F. 2008, IAU Symposium, 250, 24719. Garmany, C. D., & Stencel, R. E. 1992,
Astron. Astrophys. Suppl. , 94, 21120. Getman, K. V., Flaccomio, E., Broos, P. S., et al. 2005,
Astrophys. J. Suppl. , 160, 31921. Getman, K. V., Feigelson, E. D., Kuhn, M. A., et al. 2014,
Astrophys. J. , 787, 10822. Getman, K. V., Feigelson, E. D., & Kuhn, M. A. 2014,
Astrophys. J.
Astrophys. J. Suppl.
Astrophys. J. Suppl. , 28, 15725. Guarcello, M. G., Prisinzano, L., Micela, G., et al. 2007,
Astro. Astrophys. , 462, 24526. Guarcello, M. G., Micela, G., Peres, G., et al. 2010,
Astron. Astrophys. , 521, AA6127. Haisch, K. E., Jr., Lada, E. A., & Lada, C. J. 2001,
Astrophys. J. Lett. , 553, L15328. Hartmann, L., Ballesteros-Paredes, J., & Heitsch, F. 2012,
Mon. Not. Royal Astro.Soc. , 420, 145729. Illian, J., Pentinnen,A., Stoyan, H. & Stoyan. D. 2008,
Statistical Analysis and Mod-eling of Spatial Point Patterns , Wiley30. King, R. R., Naylor, T., Broos, P. S., et al. 2013,
Astrophys. J. Suppl. , 209, 2831. Krumholz, M. R., & Tan, J. C. 2007,
Astrophys. J. , 654, 30432. Krumholz, M. R., Klein, R. I., & McKee, C. F. 2012,
Astrophys. J. , 754, 7133. Kuhn, M. A., Getman, K. V., Feigelson, E. D., et al. 2010,
Astrophys. J. , 725, 248534. Kuhn, M. A., Getman, K. V., Broos, P. S., et al. 2013,
Astrophys. J. Suppl. , 209, 2735. Kuhn, M. A., Povich, M. S., Luhman, K. L., et al. 2013,
Astrophys. J. Suppl. , 209, 2936. Kuhn, M. A., Feigelson, E. D., Getman, K. V., et al. 2014,
Astrophys. J. , 787, 10737. Kuhn, M. A., Feigelson, E. D. & Getman, K. V. 2014,
Astrophys. J. , 802,
Astrophys. J. ,812,
Astrophys. J. , submitted40. Lada, C. J., & Lada, E. A. 2003,
Ann. Rev. Astron. Astrophys. , 41, 5741. Mac Low, M.-M., & Klessen, R. S. 2004,
Rev. Mod. Phys. , 76, 12542. McLachlan, G. & Peel, D. 2000,
Finite Mixture Models , Wiley43. McMillan, S. L. W., Vesperini, E., et al. 2007,
Astrophys. J. Lett. , 655, L4544. Minniti, D., Lucas, P. W., Emerson, J. P., et al. 2010,
New Astron. , 15, 43345. Naylor, T., Broos, P. S., & Feigelson, E. D. 2013,
Astrophys. J. Suppl. , 209, 3046. O’dell, C. R. 1998,
Astron. J. , 115, 26347. Palla, F., & Stahler, S. W. 2000,
Astrophys. J. , 540, 25548. Pillitteri, I., Wolk, S. J., Megeath, S. T., et al. 2013,
Astrophys. J. , 768, 9949. Pfalzner, S. 2009,
Astron. Astrophys. , 498, L3750. Pfalzner, S., & Kaczmarek, T. 2013,
Astron. Astrophys. , 559, A3851. Pflamm-Altenburg, J., & Kroupa, P. 2006,
Mon. Not. Royal Astro. Soc. , 373, 29552. Povich, M. S., Kuhn, M. A., Getman, K. V., et al. 2013,
Astrophys. J. Suppl. , 209, 312 Eric D. Feigelson53. Povich, M. S., Busk, H. A., Feigelson, E. D., et al. 2017,
Astrophys. J. , 838, 6154. Preibisch, T. 2012,
Res. Astron. Astrophys. , 12, 155. Prisinzano, L., Sanz-Forcada, J., Micela, G., et al. 2011,
Astron. Astrophys. , 527, AA7756. Rathborne, J. M., Jackson, J. M., & Simon, R. 2006,
Astrophys. J. , 641, 38957. Richert, A. J., Getman, K. V., Feigelson. E. D., et al.
Astrophys. J. , submitted58. Rivera-G´alvez, S., Rom´an-Z´u˜niga, C. G., Jim´enez-Bai´on, E., et al. 2015,
Astron. J. ,150, 19159. Romine, G., Feigelson, E. D., Getman, K. V., et al. 2016,
Astrophys. J. , 833, 19360. Sharma, S., Pandey, A. K., Ojha, D. K., et al. 2017,
Mon. Not. Royal Astr. Soc. , 467,294361. Tan, J. C., Krumholz, M. R., & McKee, C. F. 2006,
Astrophys. J. Lett. , 641, L12162. Telleschi, A., G¨udel, M., Briggs, K. R., et al. 2007,
Astron. Astrophys. , 468, 42563. Townsley, L. K., Broos, P S., Corcoran, M. F., et al. 2011,
Astrophys. J. Suppl. , 194,
Astrophys. J. Suppl. , 213, 165. Tutukov, A. V. 1978,
Astron. Astrophys. , 70, 5766. Wang, J., Townsley, L. K., Feigelson, E. D., et al. 2008,
Astrophs. J. , 675, 46467. Wright, N. J., Parker, et al. 2014,
Mon. Not. Royal Astro. Soc. , 438, 63968. Wright, N. J., Drake, J. J., Guarcello, M. G., et al. arXiv:1408.657969. Ybarra, J. E., Lada, E. A., Rom´an-Z´u˜niga, C. G., et al. 2013,
Astrophys. J. , 769, 14070. Zwintz, K., Fossati, L., Ryabchikova, T., et al. 2014,