Surveying the Giant H ii Regions of the Milky Way with SOFIA. I. W51A

IRS 1—The e region of G49.5–0.4 encompasses the entire 15 arc-shaped IRS 1 infrared region and its surroundings east of the d complex. The IRS 1 arc as seen in the MIR is bisected by dark lanes (Figure 2), first discussed by Goldader & Wynn-Williams (1994) in their work with 2 μm images of W51A. These dark lanes are centered at the locations shown with star symbols in Figure 2. Goldader & Wynn-Williams (1994) pointed out several lines of evidence including the fact that the radio continuum maps show no gaps at these infrared-dark locations to conclude that the dark lanes are cold and dense dust filaments seen in absorption against the bright emission of the e arc. The SOFIA data show that these features are suppressed in their infrared emission even out to 37 μm. If this suppression is due to extinction from a dense, cold dust filament, it would be expected that, at long-enough wavelengths, one would see the continuum emission from the cold dust concentrated in these dark lanes. Interestingly, there is no indication of concentrated emission from these dark lanes in the Herschel 70 and 160 μm data. In fact, the northernmost dark lane is clearly suppressed in emission out to 160 μm. Perhaps more importantly, there are no indications of the two northern dark lanes having any enhanced emission in the 1.3 mm ALMA continuum maps of Ginsburg et al. (2017). This may indicate that the gaps are not actually due to dense cold dust filaments, but may simply be areas with less dust, contrary to previous assessments.

Zoom In Zoom Out Reset image size

The brightest radio continuum peak in the e arc is coincident with a peak seen at both 20 and 37 μm, as well in all Spitzer–IRAC bands (except 8 μm, which is saturated), which we label IRS 1/#9. Interestingly, there does not seem to be a peak at this location in the Herschel data.

The W51 e1/e2 cluster—There is a heavily studied massive star protocluster in the area ∼30'' interior to (east of) the arc, with radio sources designated e1, e2, e3, e4, e8n, e8s, e9, and e10 (Figure 3). This area is rich in maser emission and is often given the moniker W51 MAIN (or just W51 M) in maser studies of the region.

Zoom In Zoom Out Reset image size

Scott (1978) first found the two UCH ii regions in this area and named them e1 and e2, due to their proximity to the main e feature. Later, Gaume et al. (1993) discovered two more HCH ii regions near e1 and e2 at 3.6 cm, which were named e3 and e4. Zhang & Ho (1997) discovered an additional source at 1.3 cm that lies between e4 and e1, which they named e8. This was later split into two sources, e8n and e8s, which were found to also be separate HCH ii regions (Ginsburg et al. 2016). Recently, Ginsburg et al. (2016) discovered two more HCH ii regions designated e9 and e10. Furthermore, there are two hot molecular cores in this area, first seen as ammonia clumps by Ho et al. (1983), one very close to e2 and the other coincident with e8 (Figure 3). These hot cores have a rich line chemistry (Ginsburg et al. 2017) and are surrounded by multiple species of masers, which are the signposts of early massive star formation.

De Buizer et al. (2005) observed the W51 e region from the ground at arcsecond resolution at both 11.7 and 20.8 μm with the IRTF. At both wavelengths, only a single point source was detected in the region near e1, but not coincident with it. The new observations made here with SOFIA at 20 μm with better astrometric accuracy confirm that this MIR emission is not coming from e1 (Figure 3). Instead, it appears that the MIR point source is coincident with a newly detected radio continuum source, e9 (Ginsburg et al. 2016), seen at 6.5 cm, which is characterized as being a HCH ii region.

Our image of the e9 source looks much different at 37 μm. The morphology looks more like an arc, starting at the location of the 20 μm point source and stretching for ∼10'' to the east, wrapping around, but avoiding the radio sources e1 and e3. To see this emission in a little more detail, we deconvolved the 37 μm data, which yielded an image with ∼2'' resolution (Figure 3). We see from this image a "peanut" of emission with two peaks at 37 μm, a fainter one coincident with the e9 source and a brighter one peaking just to the east of the e4/e8 sources. There is also a finger of fainter emission extending north of the brightest 37 μm peak, which reaches the location of e2.

Although it seems clear from the deconvolved image that the peak at e9 seen at all wavelengths is clearly coming from the HCH ii region at this location, there are a couple of possible interpretations of why we see the brightest peak at 37 μm just east of the e4/e8 area. First is that the combined emission from the multiple UCH ii and HCH ii regions is simply escaping from an area of lower extinction, which is located east of the e4/e8 region. Evidence for this comes from the fact that the NH₃ (3, 3) peak (Ho et al. 1983), which is a dense gas tracer, peaks around the same location of the e4/e8 area, and the millimeter continuum emission appears to lie in a linear structure running more or less north–south just west of the ammonia peak and coincident with the "waist" of the peanut in 37 μm emission (Figure 3(c)). This all points to a possible gradient in density in this area, with the density falling off to the east from e4/e8.

A second possible scenario is that the brightest peak at 37 μm, and the finger of emission that connects it to the e2 area, are due to a cavity carved out of the surrounding medium by the CO outflow from e2 (Shi et al. 2010; Ginsburg et al. 2017). MIR emission is often seen coming from the outflow cavities carved out by the blueshifted side of the outflows in heavily obscured MYSO regions (De Buizer 2006; De Buizer et al. 2017). The CO outflow from the e2 area does indeed have a blueshifted outflow lobe pointing to the southeast of e2 at a position angle of 145° (see blue arrow in Figure 3(c)).

Interestingly, Barbosa et al. (2016) claimed that the Spitzer IRAC–GLIMPSE data detected infrared emission from both the e1 and e2 sources. Further scrutiny of the data shows that the infrared peak, misidentified as coming from e1, is actually the same peak seen at other MIR wavelengths presented here and coming from e9. The second source seen in the IRAC data is actually equidistant between e2 and e4, and not coming from e2 (see Figure 3; blue stars). It does not correspond to any known point source seen at any other infrared wavelength, but does appear to come from within the confines of the extended 37 μm emission. The source also does not appear in Two Micron All Sky Survey (2MASS) J, H, or K data of this region, meaning it is likely not a foreground source. This source is apparently below the detection limits of the MIR facilities that previously observed this region, but has a steep-enough spectral energy SED that we are beginning to pick it up at 37 μm with SOFIA.

Other detections in the IRS 1/e region—Within the extended emission of the northern stretch of the e arc, there is an infrared point source that was detected at 20 and 37 μm, which was first identified by Barbosa et al. (2016) as IRS 1/#1 (Figure 4). Several other compact radio sources (e4, e5, and e11–e23) have been identified in other areas within and around the e arc (see Figure 2). We detect compact or point-like sources in the MIR at the locations of e7, e15, and e5 (also called IRS 1/#2 by Barbosa et al. 2016; see Figure 4) in the SOFIA data. Although we do not resolve a point source at the location of the radio point source e11, there is an unlabeled, resolved, circular (r ∼ 3'') radio continuum source situated ∼4'' to the northeast of e11, where we do detect a diffuse infrared emission of about the same extent at 20 μm; however, it appears as an arc-shaped structure at 37 μm. We name this source IRS 1/#8 (Figure 2). We do not resolve MIR point sources at the locations of the centimeter radio continuum point sources labeled e12–e14 and e17–23, though there is extended MIR emission throughout the areas where they are situated (Figure 2). Faint infrared emission is also detected with SOFIA at the location of e6 only at 37 μm, although it can be seen in the Spitzer 8 μm data (Barbosa et al. 2016). Radio point source e16 was also detected in our MIR data, but only at 37 μm (Figures 2 and 5).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We detect several sources not seen in radio continuum emission. There is a bright resolved source at both 20 and 37 μm that was first detected at 2 μm by Goldader & Wynn-Williams (1994) and labeled IRS 3 (Figure 2). It is also seen at 11.7 μm in the IRTF data from De Buizer et al. (2005), at 8 μm in the Spitzer IRAC data, and at 2 μm in the 2MASS data of the area. There are three point sources detected for the first time and only seen at 37 μm in the vicinity of e15 (Figure 2). Following the nomenclature of Barbosa et al. (2016), we dub these sources IRS 1/#3 , IRS 1/#4, and IRS 1/#5. There are also two resolved regions of MIR emission in the northern part of the e region where there are no significant radio continuum emission peaks; both of these are seen in the NIR with 2MASS and in the MIR with SOFIA and Spitzer, which we will call IRS 1/#6 (see Figure 4) and IRS 1/#7 (see Figure 2), respectively.

The radio point source e18 of Ginsburg et al. (2016) is actually a double separated by only ∼05 and is coincident with the peak emission from a bar-shaped 37 μm source on the bottom of the e arc, just west of the dark lane (Figure 5). At 20 μm, the peak in emission is shifted to the peak of a ∼5'' in diameter radio continuum clump located ∼35 southeast of the e18 binary named e18d, which was identified by Ginsburg et al. (2016) as a H ii region (Figure 5(b)). At 37 μm, the source is very bright, and it appears to get brighter with increasing wavelength; at 70 μm, the emission peaks at the same location as the 37 μm peak, and this source appears as the fifth brightest source in all of W51 A (Figure 5(b)). It is the fourth brightest source in all of W51A at 160 μm after IRS 2, IRS 1 (peaked at IRS 1/#1), and the e1/e2 cluster region. We will call this infrared region IRS 4, in keeping with the major IR-emitting source nomenclature. IRS 4 is the most steeply rising subcomponent from 20 to 37 μm in this study, which, along with the high FIR intensities, indicates the source is highly embedded and/or young. As we will see in a later section, the best-fit SED model for this source yields a bolometric luminosity of 6.48 × 10⁵ L_⊙, which is the single star equivalent spectral type of O4.5, but the SED can be fit with MYSO models with masses in the range of 24–96 M_☉. However, due to the presence of multiple centimeter radio continuum sources (e16, the e18 binary, e18d), this location is likely to be an embedded core or clump that is in the process of forming a young massive protocluster.

In both the radio continuum and infrared imaging data, IRS 2 breaks up into several subcomponents surrounded by a ∼15'' × 15'' cloud of emission at high spatial resolution. This extended emission was first found to be peanut-shaped in the 2 μm images of Goldader & Wynn-Williams (1994), and they named the two peaks IRS 2E and IRS 2W. They also argued that the IRS 2 region is a small cluster of ongoing star formation, identifying at least a dozen NIR sources. This area is also rich in masers (which are typically signposts of massive star formation), and maser studies typically refer to this region as W51NORTH (Schneps et al. 1981).

High spatial resolution MIR imaging by Okamoto et al. (2001) and Barbosa et al. (2016) showed that the IRS 2W component is an extended region of emission with no discernible point sources. This source is coincident with the brightest centimeter radio continuum feature, a cometary UCH ii region, with similar appearance in the radio (i.e., Wood & Churchwell 1989; Gaume et al. 1993) and MIR. On the other hand, the IRS 2E component is found to contain a cluster of four point-source components with ∼1'' separations. Our SOFIA 20 μm image of the area is shown in Figure 4. We deconvolved the image to try to resolve out as many components in the IRS 2 region as possible, though at the limits of our deconvolution we still cannot resolve the individual components within the IRS 2E cluster.

In the outskirts of the extended IRS 2 region, there are several point sources nested within the diffuse halo of infrared emission. Several of these sources have radio counterparts (as seen by Ginsburg et al. 2016)—some have been identified before in infrared observations, and others we will identify here for the first time. Radio sources d4e&w, d6, and d7 all have infrared counterparts seen with SOFIA (Figure 4). Barbosa et al. (2016) previously have identified the infrared emission from d7 (which they call IRS 2/#3), and this was also seen in the MIR images of Kraemer et al. (2001) and named KJD 9. It can be seen in the Spitzer IRAC 8 μm GLIMPSE image as well. Sources d4e&w and d6 do not seem to have counterparts in the Spitzer 8 μm image; however, d6 was detected in the MIR by Kraemer et al. (2001) and named KJD 11.

Ginsburg et al. (2016) identified a diffuse radio source that they label d3; however, this is the previously identified radio source b2 (Mehringer 1994). The b2 source does have an MIR counterpart, but we will discuss it in a later section.

In addition to IRS  2/#3 (KJD 9), Barbosa et al. (2016) also identified three more point-like infrared sources, which they label IRS 2/#1, IRS 2/#2, and IRS 2/#4 on the eastern outskirts of IRS 2. These were also seen by Kraemer et al. (2001) and labeled KJD 7, KJD 8, and KJD 10, respectively. We see all four of these sources in the SOFIA 20 and 37 μm images (see Figures 2 and 4). IRS 2/#4 can also be seen as a point source in the radio continuum images of Ginsburg et al. (2016), though it was not labeled.

We also detect five more infrared sources as of yet not identified in the IRS 2/d region. Continuing the nomenclature of Barbosa et al. (2016), we will call these IRS 2/#6–#10. IRS 2/#7 appears to not be a point source, with a slight extension from SE to NW (Figure 2). IRS 2/#6 appears in the Spitzer 8 μm image, is weakly detected in the SOFIA 20 μm image, and is not present in the 37 μm SOFIA image (Figure 2).

We do not detect a source in the SOFIA data at the location of KDJ 6. Though Kraemer et al. (2001) claimed a source is present at this location, there is no information on the flux density, or, more importantly, the significance of the detection in their paper. The source is not present in the shorter Spitzer–IRAC bands, and the 5.8 and 8.0 μm data are not helpful because the presumed source location resides in a region of the image that is saturated.

The extended source b appears as a cometary H ii region or arc in the centimeter radio continuum images of Mehringer (1994), who also found that there is a velocity gradient from the SW to the NE as seen in H92α. Apart from this, little else is known about this region. In the infrared, the source is bisected by a dark lane that is clearly visible in the Spitzer IRAC data and all the way out to 37 μm (Figure 6). The dark lane appears to be almost perpendicular to the velocity gradient seen in the radio line emission. There is a submillimeter core here, as seen in the Herschel 160 μm data and in the 450 μm data of Hill et al. (2006), with a peak close to the location of the dark lane. This may be the case of an outflowing source (or sources) buried within the dark lane; however, the morphology of the radio continuum does not resemble a (partially) ionized jet or wind. The MIR appearance is knotty (Figure 6). However, our source-finding algorithm found peaks at slightly different locations for sources in the 20 and 37 μm data. This indicates that these are not likely to be individual centrally heated sources. These sources are likely externally heated knots of dust or optically thin holes in the dust clump surrounding the central protostar(s) (as traced by the submillimeter and radio peak) where MIR emission is escaping. As we will discuss in Section 4, this region appears to be the least evolved (i.e., youngest) region in all of W51A, and therefore it may yet be too embedded for us to detect the YSOs within it even at wavelengths as long as 37 μm.

Zoom In Zoom Out Reset image size

The j radio region appears as an elliptical shell in radio continuum maps (e.g., Mehringer 1994). In the infrared, the dust emission is fully contained within this shell tracing the ring-like structure (Figure 7). The 8 μm Spitzer image also shows a bright point source at the center of this shell. It is faint but detected in the SOFIA 20 μm images (but not at 37 μm) and is very prominent at shorter wavelengths like the NIR.

Zoom In Zoom Out Reset image size

Okumura et al. (2000) was the first to suggest the ring structure to be a wind-blown bubble driven by the star seen at its center in the NIR, claiming that it is a "P Cygni-type supergiant." This is a class of luminous blue variable (LBV) star, which is thought to be a short-lived (10⁴–10⁵ yr) stage of massive stellar evolution between the main-sequence O phase and the Wolf-Rayet phase (Morris et al. 1996). This short-lived phase is a time of great instability, leading to high mass loss, resulting in the shedding of material that eventually forms circumstellar shells which can be seen readily in the infrared (Wachter et al. 2010). Given the observations of Clark et al. (2009) and the latest derivation of the distance to W51 of 5.4 kpc from Sato et al. (2010), it can be concluded that this LBV candidate, dubbed [OMN2000] LS1 (hereafter LS1), has a luminosity of ∼5 × 10⁵ L_⊙.

Given their fleeting nature, LBVs are rare and only a couple dozen verified LBVs have been found in our Galaxy, with about another 50 candidates awaiting confirmation (Agliozzo et al. 2017). Though naively one might think that such an evolved star should not be found in a region of active star formation, this is one of several known LBVs coincident with massive star-forming complexes (e.g., Orion, G305, W43, Westerlund 1, and the Galactic Center), which represents a significant portion of the known population of LBVs. This means that these presently active star-forming regions have had a long history of sustained star formation, and Clark et al. (2009) claimed that the data on LS1 point to an age of 3–6 Myr for the oldest observed epoch of star formation in W51A.

There is very little study of the remaining G49.5–0.4 regions, which we will discuss all together in this section.

a—This region is an extended, round region of radio continuum emission with a diameter of about 20'', with a brighter area of emission on the northern side (Mehringer 1994). In the infrared, this source takes on a different morphology at almost every wavelength (Figure 8). At 20 μm, it appears to be a "hamburger" with a brighter top than bottom, with a darker lane bisecting it through the middle. At 37 μm it looks similar, but with more extended structures than at 20 μm. As with the radio centimeter continuum images, both SOFIA wavelengths show no embedded point sources or peaks that resemble point-like sources. At 8 μm, the dust emission is more clumpy and wispy. Comparing the 8 μm emission to the 6 cm radio continuum emission shows the peaks to be anticorrelated (see color image in Figure 8), and therefore the radio maybe tracing the more extinguished regions and the 8 μm may be clumpy and wispy in appearance because it is escaping through holes that are less optically thick. Both the Spitzer 8 μm and SOFIA 37 μm images show an arc or bubble to the south. This arc is also seen in the Spitzer–MIPS 24 μm image, but not in our SOFIA 20 μm image, so is likely fainter than our detection limit at that wavelength.

Zoom In Zoom Out Reset image size

b1 and b3—Radio source b1 appears as a large, circularly symmetric source in the low-resolution radio images of the region (Mehringer 1994). In the Spitzer 8 μm image (Figure 8), it consists of a subcomponent surrounded to the north and west by a narrow arc structure (∼20'' in diameter). In the SOFIA data, the 20 μm image shows a slightly extended source with a peak at the location of the 8 μm compact source peak. There is very little emission at 20 μm from the arc. At 37 μm, emission tracing the arc seen at 8 μm is detected, with emission also filling in the arc interior, looking more like a cometary UCH ii region perhaps caused by a bow shock, with a broad peak near the subcomponent location.

Source b3 looks like a slightly extended emission region on the northeast border of the b1 arc at 37 μm and has a similar appearance in the Spitzer 8 μm, and SOFIA 20 and 37 μm images (Figure 8). The peak at all three wavelengths is coincident with the radio peak. Like b1, this source has a bow-shock appearance.

Interestingly, the brightest 70 μm emission is located in between b1 and b3 (see the color image for this source in Figure 8).

b2—Radio source b2 is a symmetric and compact source in SOFIA 20 and 37 μm images (Figure 2), but has a peak offset to the west in the Spitzer 8 μm image.

We also detect one more subcomponent in the SOFIA images near b2, which does not have a radio continuum component. G49.5–0.4 b2/#1 (see Figure 2) is located ∼20'' southwest of b2, which appears as an unresolved point source at 20 μm, but is resolved and slightly extended at 37 μm (and in the Spitzer 8 μm image).

c1 and c—The naming convention for the radio emission in W51A has been to name the large regions of emission with letters, while individual peaks and subcomponents within or near these regions are indexed with numbers. It is puzzling that there does not appear to be an extended radio region labeled c, but only the individual source peak c1 has been identified. Though there is a large and diffuse radio continuum region surrounding c1 and extending east toward the b region, it has never been labeled in radio studies and is simply referred to as "the arc-like area between regions b and c1" by Mehringer (1994). The peak of c1 lies in an arc-shaped structure in the southeastern edge of a larger (r ∼ 40''), diffuse region of extended MIR and radio continuum emission (Figure 2). This region appears to be separated from c1 and b by gaps in radio continuum emission; however, the 20 and 37 μm maps look very different, with diffuse infrared emission from this region forming a continuous region of dust emission all the way east to c1. In keeping with previous nomenclature, we will refer to this entire extended radio and infrared continuum region as region c.

Figuerêdo et al. (2008) identified two revealed O9 stars (sources #62 and #64 in their list) near the peak of radio source c1. The radio continuum source identified as e15 by Ginsburg et al. (2016) and the three newly discovered MIR sources (IRS 1/#3 , IRS 1/#4, and IRS 1/#5) all lie in the northern edge of the extended c region (Figure 2).

f and g—Observations in the NIR by Okumura et al. (2000) find five revealed O stars and 23 early B stars in the combined f and g regions. Koo (1997) used H i absorption studies to determine that f and g are located either near the front or northern edge of the molecular cloud containing W51A, while components a, b, and e are likely to be embedded in or behind it.

With SOFIA, we see the same morphology and extent as what is seen in the low spatial resolution radio continuum images of this region (see the color image for this source in Figure 8). Spitzer images at 3–8 μm show extended emission from the e region continuing north and surrounding the f and g radio regions to the south, east, and west. Although there seems to be some emission and NIR point sources near the peak of radio source g, NIR emission is conspicuously absent from the areas of most of the extended radio and MIR continuum emission of the f and g regions. This suggests that this region is being carved out by the O stars present here, heating and ionizing the areas we see in the SOFIA MIR and radio continuum images, consistent with the hypothesis by Koo (1997) that the f and g regions are likely in front of the W51A molecular cloud.

Hill et al. (2005) find a 1.2 mm dust clump coincident with the peak of the g source, and estimate it has 180 M_⊙ of dust. We detect a subcomponent ∼1' southwest of the center of radio source f in the SOFIA data at both 20 and 37 μm (Figure 8), which we label as f/#1. There are some peaks at 20 μm not present at 37 μm, and vice versa, within the extended MIR emission of f and g, but no discernible embedded or point-like sources.

h—This region was found to contain class II methanol masers (Szymczak et al. 2000; Liu et al. 2010), which are a tracer of the earliest stages of massive star formation. In the NIR, Okumura et al. (2000) found dozens of revealed B stars around the h radio region, with an evolved O5 star near its center. This star can be seen in the Spitzer images of the region (Figure 7). It is bordered to the southeast by two concentric arcs, the nearest bright at both 20 and 37 μm, but the outer arc is only bright at wavelengths 37 μm and longer. Encircling the O5 star and the two arcs is an outer bright-rimmed bubble that can be most easily seen in the 70 μm Herschel image, which is filled in by radio continuum emission (Figure 7(b)). The radio continuum peak is close to the 20 and 37 μm peak, indicating that the whole h region may be ionized and heated by the O5 star located near there. This is unlike region j, which abuts the rim of h to the west, which is devoid of emission inside its wind-blown shell at infrared and centimeter radio wavelengths. Okumura et al. (2000) stated that the h and j regions have the lowest extinction in the whole of G49.5–0.4, which is likely due to the evolved state of these two regions.

i—One O9 star and one B1 star is seen in the NIR in this region by Okumura et al. (2000). The region appears to be a multipeaked, extended region with a radius of ∼14'' in the Spitzer 8 μm image (Figure 8). Interestingly, the 20 μm SOFIA image shows a much less extended emission with a peak coincident with the southwestern peak seen at 8 μm. The color image for this source in Figure 8 shows that the combined emission across all MIR wavelengths is fan-shaped, with the 20 μm emission being most compact, the 37 μm emission extending out to the north and west beyond that, with the 8 μm emission extending yet farther beyond both the 20 and 37 μm emission to the north and west. Given the morphology as a function of wavelength in the infrared could be a "blister"-type H ii where the source lies on the edge of a dense region and the emission is breaking out on one side (where the density is lowest).

In addition to the infrared-dark lanes discussed above in the previous sections, the Herschel 160 μm image show that the infrared-dark area south of b2, west of d and e, north of c, and east of b and a is "filled in" by 160 μm dust emission. This signifies that this area is infrared-dark due to the presence of widespread cold dust (Figure 9).

Zoom In Zoom Out Reset image size

The 160 μm emission is strongest around the d and e1/e2 regions and mimics the shape seen by SOFIA of those regions to first order. However, the brightest 160 μm emission actually wraps around and "avoids" the hot infrared emission seen by SOFIA of the b and c sources. Farther to the north, the outskirts of the 160 μm emission also look like they wrap around and avoid sources g and f. This appears to indicate that much of the appearance of G49.5–0.4 in the MIR is dominated by us only seeing emission on the surfaces of the subcloud structure and/or leaking out through less dense areas devoid of large dust grains carved out by ionization fronts and outflows within this region of the W51A molecular cloud.

Harvey et al. (1986) resolved this region into two components in the far-infrared with the Kuiper Airborne Observatory. The brightest peak in the far-infrared is near the centimeter radio continuum peak b. But there is a secondary peak ∼1' to the northeast in the far-infrared which they named b-east (Figure 10). This peak is seen in the 20 cm images of Mehringer (1994), but was not labeled. At all wavelengths from the NIR to the radio, there is a dark gap or decrease in emission running NW to SE and separating the southwestern part of source b from b-east and is therefore likely due to a decrease or absence of gas and dust at that location.

Zoom In Zoom Out Reset image size

Source b has a peak in the centimeter radio continuum that is close to, but not exactly coincident with, the far-infrared peak seen at 70 μm (∼5'' offset). Both components appear to reside in an infrared-dark area (as seen in Spitzer IRAC and SOFIA data) that bisects the b source and runs NE to SW (Figure 10). The 160 μm Herschel peak seems to be exactly centered and the same shape as the "lane" in the near- and MIR emission. Given the fact that this infrared-dark area has radio continuum emission and water maser emission (Cesaroni et al. 1988), and is surrounded by YSOs (Saral et al. 2017), it is likely the site of very embedded massive star formation that is infrared-dark at wavelengths <40 μm due to very high extinction. Most of the peaks within this region shift as a function of wavelength in the MIR, indicating that they are externally heated knots or holes in the otherwise optically thick emission in the region where MIR light is escaping. However, we find two sources where the peaks do not change with wavelength and are therefore likely to be MYSO candidates, which we label b/#3 and b/#4 (Figure 10).

Source b-east is a much more diffuse area of MIR emission, but it does appear to have one embedded point source, which we name b/#1 (Figure 10). It is likely an MYSO, because it is apparent in the Spitzer IRAC data, is seen at both 20 and 37 μm, and is coincident with a radio continuum peak at 1.5 cm (Ginsburg et al. 2016).

We have found in the SOFIA data a resolved but compact source detected 10'' east of the extended b-east region that is also seen in Spitzer 8 μm, which we have named b/#2 (Figure 10). This source is also a very bright object in the Herschel 70 and 160 μm images of this region but has no associated centimeter radio continuum emission. Given its high flux in the mid- and far-infrared and lack of radio continuum emission, we believe that it is likely an MYSO in a very young evolutionary state prior to the onset of a UCH ii region (as we will see in a later section, this source does indeed appear to be a MYSO from SED model fitting). We also see another isolated source ∼15'' southeast of b/#2, which also has a radio component that we label b/#5.

These four sources encircle the main intensity peak near source b, and all have cometary or shell-like structure in the radio and in the infrared (Figures 1 and 11). It appears from the morphologies of the sources in the Spitzer NIR data alone that these sources (and indeed most of this region's structure) are due to wind-blown bubbles and/or are bright-rimmed clouds. The outer layer of the shells is generally demarcated by the dust emission as seen in the Spitzer IRAC and SOFIA images, and generally the interiors of the shells/arcs are filled with centimeter radio continuum emission.

a—This source has an interesting double-arc structure, with the 8 μm emission, 37 μm emission, and centimeter radio continuum tracing both arcs. Interestingly, the SOFIA 20 μm emission dominantly traces the interior arc. The peak fluxes of the inner and outer arcs differ only by ∼20% in 8 and 37 μm, while at 20 μm there is almost an order of magnitude differences in flux at the same positions. As we will discuss later in Section 4.1, we can use a color–color diagram to determine if a source has flux in the IRAC bands that is dominated by PAH emission. Using that method, we have found that this source falls well within the definition of being PAH-dominated. Therefore, a plausible explanation for the behavior of the flux of this source as a function of wavelength is that the continuum emission of the outer arc may be low at wavelengths ≤20 μm, and that the IRAC 8 μm flux is high because of strong PAH emission.

It appears that there is a cluster of YSOs identified by Saral et al. (2017) located interior to (or east of) this double arc (Figure 11), which we assume is most likely responsible for the shaping, heating, and ionizing of source a. The four massive YSO candidates from Saral et al. (2017) are shown in Figure 11, although we only detect sources in the SOFIA data at the locations of the sources labeled SHA17 3 and SHA17 4. (We will show in the section on SED model fitting that these sources are unlikely to be MYSOs.)

Zoom In Zoom Out Reset image size

c—This source also has a double-rimmed structure, with the eastern arc traced by Spitzer 8 μm emission, and the western arc traced by the SOFIA 20 and 37 μm emission (Figure 11). The centimeter radio continuum emission fills in the area interior to the eastern arc, with a peak near the inner, western arc. Interestingly, Saral et al. (2017) found a cluster of ∼15 YSOs to the east and south of source c. Feedback from these YSOs may be responsible for shaping the arc-shaped dust structure seen here in the MIR; however, there is no evidence of any truly energetic YSOs in the cluster given that none of the cluster members display radio continuum emission, and the region appears to be devoid of continuum sources from the NIR out to 160 μm (i.e., no indication of massive and/or young and active cluster members).

d—This source has a ring shape with a radius of ∼1', which is brightest to the southeast and faintest to the northwest. This southeastern rim appears as a bright arc in the Spitzer NIR data and Herschel far-IR data (see the large arc of 70 μm emission to the west of source e in Figure 1), but we barely detect it at 37 μm and do not see any evidence of it at 20 μm.

e—This source is a very tiny bright-rimmed source located on the eastern rim of source d (Figure 10). This rim can be seen in the Spitzer NIR and SOFIA 37 μm data, and the center is filled by unresolved radio continuum emission at centimeter wavelengths. Only the brighter eastern part of the shell is detected at 20 μm. There is what appears to be a point source in the MIR, just to the southwest of e. We name this source e/#1 (Figure 10). There is no centimeter radio emission detected from this source.

f —This source is a smaller ring-shaped region (r ∼ 25''), with the outer rim radiating brightly in the Spitzer NIR images as well as the SOFIA 37 μm image (Figure 11). Interior to this is a ring seen in radio continuum emission as well as 20 μm, and so is likely an ionized bubble (i.e., Strömgren sphere). At all wavelengths, the ring is brightest to the west, giving it a cometary UCH ii appearance.

The L/M is considered as a good tracer of stellar cluster formation and molecular clump evolution, where L is the bolometric luminosity of young stellar clusters (or molecular clump) and M is the mass of the cluster. The theoretical study of Krumholz & Tan (2007) showed that the L/M of a massive stellar cluster (1000 M_☉) had a positive correlation with the age of the cluster, i.e., L/M increases with the evolutionary stage of the stellar cluster. Ma et al. (2013) studied 303 massive molecular clumps defined by HCO⁺ (1–0) emission (Barnes et al. 2011) in order to constrain physical properties along the complete span of protocluster evolution. They analyzed MIR to submillimeter dust continuum images to derive the bolometric luminosity and cold component dust temperature (T_c), and adopted a mass estimate from HCO⁺ (1–0). They found that the L/M of all molecular clumps were in the range from ∼0.1 to ∼1000 L_⊙/M_⊙. The L/M of the molecular clumps showed a positive correlation with T_c as well as MIR surface brightness (as measured in Spitzer–IRAC data). These results imply that L/M traces star cluster formation and evolution.

We estimated the mass of each extended source by producing a mass surface density (Σ) map and using the estimated distance of W51A (5.4 kpc; Sato et al. 2010). The pixel-by-pixel Σ values were derived via the method investigated in Lim et al. (2016). In this method, the optically thin assumption of dust continuum emission is adopted to perform the graybody fit (i.e., modified blackbody fit). We used Herschel–PACS 160 μm and –SPIRE 250, 350, and 500 μm images for the fitting, while the PACS 70 μm data were excluded because that wavelength could still be optically thick under the conditions encountered in these regions (Lim & Tan 2014; Ginsburg et al. 2015). The convolution of 160, 250, and 350 μm data to match the angular resolution of the SPIRE 500 μm images (∼36'') was performed with the methods introduced by Gordon et al. (2008). We then estimated the diffuse Galactic background emission via the Galactic Gaussian method (Battersby et al. 2011; Lim et al. 2016) that assumes the Galactic background follows Gaussian profiles along the latitude. Each pixel of the background-subtracted flux density maps (160–500 μm) was treated to derive Σ using the standard graybody equation,

$$\begin{eqnarray}&&{I}_{\nu }\simeq {B}_{\nu }(T)(1-{e}^{-{\tau }_{\nu }})={B}_{\nu }(T)(1-{e}^{-{\rm{\Sigma }}{\kappa }_{\nu }}),\end{eqnarray} \tag{ 1 }$$

where I_ν is the observed intensity of the corresponding band, B_ν(T) is the temperature-based filter-weighted blackbody radiation, τ_ν is the optical depth, Σ is the mass surface density, and κ_ν is the filter-weighted opacity. We adopted the thin ice mantle dust opacity model of Ossenkopf & Henning (1994) and a dust-to-gas mass ratio of 1/142 (Draine 2011) to estimate dust opacity, κ_ν.

The bolometric luminosity of each clump was then derived from the integrated intensities inside the given apertures (Table 3) through the following method. We found that the measured radius of any given extended source is similar at all wavelength bands ≤160 μm, and thus for each source we use the same aperture size for photometry at these wavelengths. We also found that we could use radii of approximately 2.0, 2.5, and 3.0 times the aperture size used at shorter wavelengths to perform the aperture photometry at 250, 350, and 500 μm, while we use the 20 μm aperture size for all Spitzer–IRAC bands. In order to reproduce the intrinsic fluxes of MIR-extended sources, which we assume are young stellar clusters embedded in the middle of dense molecular structures, we had to de-redden the observed ${F}_{\nu ,\mathrm{tot}}$. The 1D radiative transfer equation of absorption, ${F}_{\nu ,\mathrm{tot},1}\simeq {F}_{\nu ,\mathrm{tot},0}^{-\tau }$, was used to determine the intrinsic intensity, ${F}_{\nu ,\mathrm{tot},0}$, where ${F}_{\nu ,\mathrm{tot},1}$ is the observed flux (i.e., with extinction). Here we use the median mass surface density value, $\tilde{{\rm{\Sigma }}}$, of each extended source to derive the optical depth, τ_ν. Checking these values against those of previous studies (e.g., Kang et al. 2010), we find that our derived optical depth values are in agreement to within a factor of 2. We make the simple assumption that the young clusters are embedded at the center of the molecular clumps so that the dust structures in front and at the back of the cluster are symmetric along the line of sight. Therefore, because we assumed the material to be optically thin, and because the Σ values are derived for the total column density along the line of sight, we divided $\tilde{{\rm{\Sigma }}}$ by 2 to de-redden only the foreground extinction of the cluster so that τ_ν = 1/2 κ_ν$\tilde{{\rm{\Sigma }}}$. In addition to the photometric uncertainty levels of each band (Section 4.1), one needs to also consider the de-reddening effect, the contribution of different temperature components, and nearby source contamination as additional errors. With all these aspects, we assume ∼30% total uncertainty for 4.5 μm, ∼40% for 20 and 37 μm, and ∼50% for 70 and 160 μm. The 3.5, 5.8, and 8 μm bands are treated as upper limits due the the expectation of high PAH contributions. The 250, 350, and 500 μm bands are also assumed to be upper limits, due to the coarse angular resolution and possible contamination from extended emission of nearby sources.

When trying to fit the 3–500 μm photometry data of the extended sources with graybody fits, it was found that a single graybody was not sufficient, but that a two-component fit worked nicely for all sources (Figure 14). Therefore, following the work of Ma et al. (2013), we derived the bolometric luminosity via a best-fit graybody model with two temperature components, i.e., cold and warm dust components. Based on these SEDs, we discovered that the SOFIA–FORCAST 20 and 37 μm photometry points are crucial in distinguishing the presence of the different temperature components. Figure 14 shows an example that represents well the two-component nature of the SEDs of all of the sources in Table 4. Integrating under these SEDs allows us to derive the bolometric luminosity of each source.

Zoom In Zoom Out Reset image size

Table 4.  Virial Parameters of Extended Sources in W51A

Source	Mvir	M	L	Tcold	Twarm	L/M	αvir
	(M☉)	(M☉)	($\times {10}^{5}\,{L}_{\odot }$)	(K)	(K)	(L⊙/M⊙)
G49.4–0.3
a	580	2560	3.66	76.2	264.6	71.4	0.23
b	2490	9510	12.0	71.0	261.6	63.0	0.26
c	1370	1580	3.61	91.0	261.2	113.8	0.87
e	373	819	0.65	62.6	283.2	39.6	0.45
f	125	916	1.64	91.0	273.0	89.4	1.36
G49.5–0.4
a	1900	1430	2.77	88.6	253.2	96.8	1.21
b	1810	9930	5.14	69.9	247.6	25.9	0.18
c	3950	6870	10.3	66.1	270.7	81.0	0.58
d	947	⋯	⋯	⋯	⋯	⋯	⋯
e	5420	⋯	⋯	⋯	⋯	⋯	⋯
f	1800	432	1.98	93.0	256.7	228.5	4.16
g	1950	320	2.18	100.5	255.7	340.8	6.11
h	1520	122	1.93	116.6	250.7	790.1	12.50
i	511	107	0.57	116.4	254.7	266.4	4.77
j	1570	386	3.26	104.8	270.7	421.6	4.06

Note. Sources d and e of G49.5–0.4 are saturated in Herschel–SPIRE observations.

Download table as:  ASCIITypeset image

Table 5.  Observational Parameters of Subcomponents and Point Sources in W51A in the Spitzer–IRAC Bands

		3.6 μm		4.5 μm		5.8 μm		8.0 μm
Source	Rint	Fint	${F}_{\mathrm{int} \mbox{-} \mathrm{bg}}$	Fint	${F}_{\mathrm{int} \mbox{-} \mathrm{bg}}$	Fint	${F}_{\mathrm{int} \mbox{-} \mathrm{bg}}$	Fint	${F}_{\mathrm{int} \mbox{-} \mathrm{bg}}$
	('')	(mJy)	(mJy)	(mJy)	(mJy)	(Jy)	(Jy)	(Jy)	(Jy)
G49.4–0.3
a/#1	4.80	123	116	347	337	0.68	0.63	0.87	0.72
b/#1	6.00	91.0	43.2	117	45.4	0.75	0.24	2.25	0.40
b/#2	7.20	57.9	37.5	77.5	56.4	0.67	0.44	1.82	1.06
b/#3	9.60	158	95.3	247	130	1.59	1.14	4.80	2.52
b/#4	6.00	83.8	56.7	175	122	1.07	0.55	2.99	1.57
b/#5	6.00	40.0	14.1	49.6	16.3	0.26	0.02	0.80	0.05
b/#6	4.80	46.8	26.2	41.5	20.5	0.36	0.17	1.00	0.39
e/#1	7.20	51.9	25.2	66.6	35.3	0.70	0.30	1.92	0.74
SHA17 3	4.80	33.0	1.20	31.8	2.40	0.27	0.01	0.77	0.04
SHA17 4	6.00	48.9	17.5	54.6	18.7	0.25	0.02	0.75	0.04
G49.5–0.4
b1	21.6	831	590	788	533	7.23	4.13	20.5	10.3
b2	10.8	241	136	296	186	1.76	0.63	4.85	1.66
b2/#1	7.20	46.9	21.4	53.7	31.3	0.51	0.12	1.35	0.46
b3	12.0	134	45.1	160	56.2	1.49	0.53	4.53	1.34
d4e+d4w	4.80	48.7	9.70	74.1	22.1	0.62	0.19	1.43	0.41
d6	4.80	88.4	33.4	200	105	1.15	0.18	2.47	0.57
e7	9.60	246	116	322	168	2.12	0.74	5.64	1.95
e9	4.80	41.0	12.5	95.5	25.3	0.73	0.25	2.12	0.91
e15	4.80	43.7	16.2	72.1	26.1	0.48	0.10	1.34	0.30
f/#1	6.00	61.9	16.0	60.9	18.4	0.56	0.16	1.57	0.38
i	18.0	433	247	479	288	2.82	1.69	8.20	4.57
i/#1	7.20	43.9	9.90	55.5	17.9	0.26	0.03	0.99	0.08
IRS1/#1	4.80	352	284	1080	921	4.26	3.05	8.63	5.86
IRS1/#2	4.80	188	136	717	626	2.71	2.11	5.99	4.43
IRS1/#3	4.80	23.1	14.4	66.1	42.0	0.22	0.03	0.54a	⋯
IRS1/#4	3.60	4.70	0.50	9.60	2.20	0.06a	⋯	0.16a	⋯
IRS1/#5	3.60	3.40	0.20	8.10	2.70	0.06a	⋯	0.15a	⋯
IRS1/#6	3.60	56.6	27.7	96.1	36.4	0.74	0.23	1.87	0.76
IRS1/#7	9.60	178	53.6	258	69.0	2.25	0.48	5.89	1.08
IRS1/#8	9.60	165	37.5	247	76.7	1.93	0.43	5.21	0.86
IRS1/#9	6.00	401	296	1260	1100	4.63	2.70	12.7	4.71
IRS1/#10	9.60	237	49.7	278	38.7	2.61	0.38	6.98	1.10
IRS1/#11	3.60	11.0	3.50	20.2	13.1	0.09	0.02	0.22	0.02
IRS2/#1	3.07	312	302	765	751	2.85	2.73	3.09b	⋯
IRS2/#2	3.07	117	107	254	240	0.89	0.74	1.93	1.56
IRS2/#3	3.07	79.7	70.9	142	125	0.75	0.60	1.82	1.50
IRS2/#4	3.07	59.3	51.1	113	96.8	0.57	0.42	1.27	0.94
IRS2/#5	3.07	418	410	1040	1030	6.21	6.08	2.43b	⋯
IRS2/#6	3.60	17.4	5.00	28.5	9.80	0.22	0.06	0.50	0.13
IRS2/#7	4.80	49.0	29.5	58.9	38.7	0.48	0.27	1.19	0.57
IRS2/#8	3.60	25.2	11.4	27.6	12.8	0.22	0.07	0.56	0.12
IRS2/#9	3.60	30.0	18.7	37.4	25.5	0.25	0.12	0.62	0.27
IRS2/#10	3.84	86.2	72.6	201	184	1.03	0.85	3.05	2.60
IRS2E	3.84	1770b	⋯	2370b	⋯	13.3b	⋯	1.78b	⋯
IRS2W	3.84	1750b	⋯	2670b	⋯	12.1b	⋯	2.31b	⋯
IRS3	4.80	598	420	1460	1130	5.69	4.20	9.04	3.32
IRS4	9.60	234	138	442	289	2.49	1.32	7.84	3.97
LS1	4.80	740	717	782	759	0.93	0.84	0.87	0.64

Notes. Same as Table 1 but for Spitzer–IRAC bands. The center positions of the apertures are based on SOFIA observations in Table 1.

^aThe F_int value is used as the upper limit because the source is difficult to distinguish from the background, due to the relatively weak source emission. ^bThe F_int value is used as the lower limit because the source is partially/entirely saturated.

Download table as:  ASCIITypeset image

Table 4 shows the M, L, and L/M values of the extended sources in columns 3, 4, and 7, respectively. We did not retrieve the M and L of extended source d or e, because the Herschel images are mostly saturated in those regions. From the 13 remaining extended sources, we see a large variation of L/M: 25 ≲ (L/M)/$({L}_{\odot }/{M}_{\odot })$ ≲ 790. The G49.4–0.3 sources show typically smaller L/M than sources in G49.5–0.4, i.e., 40 ≲ (L/M)/(L_⊙/M_⊙) ≲ 110. The extended sources in G49.5–0.4 f–j show high L/M values (∼270–790 L_⊙/M_⊙), and the sources in the G49.5–0.4 a–c area have 25 ≲ (L/M)/(L_⊙/M_⊙) ≲ 100. One might assume that the relative ages of stellar clusters in W51A are high at G49.5–0.4 f–j while the sources in the G49.5–0.4 a–c and G49.4–0.3 regions are possibly in similar evolutionary stages. This may also be seen from the derived cold temperature components (column 5 of Table 4). We find that the T_cold of high L/M sources (i.e., G49.5–0.4 f–j) are ∼20–30 K higher than the other extended sources, which can indicate that the dust grains are internally heated by embedded young stellar clusters so that both T_cold and L/M should increase simultaneously.

Virial analysis is an effective tool for determining the importance of kinematic and gravitational energies of ISM structures, especially for molecular clumps and cores (Bertoldi & McKee 1992). As molecular clumps evolve, kinematic energy from internal sources (e.g., radiative pressure, outflow, and shock) are assumed to increase their influences on the system. This can be traced by comparing the estimated mass at virial equilibrium (i.e., virial mass), M_vir, and the mass of each clump. Bertoldi & McKee (1992) defined the virial mass of molecular clumps as ${M}_{\mathrm{vir}}=5{\sigma }^{2}R/G$, where M_vir is the mass of the structure if it were in virial equilibrium, σ is the velocity dispersion (σ = Δv/(8 ln 2)^1/2) for the Gaussian line profile of a corresponding clump with Δv as the FWHM of the molecular line profiles), R is the radius of the clump, and G is the gravitational constant. The virial parameter, α_vir, is defined as α_vir = M_vir/M, where M is the intrinsic mass of the clump. For simplicity, the effect of magnetic fields is ignored even though the magnetic field is important in regulating the dynamics of molecular clumps (Bertoldi & McKee 1992; Tan et al. 2013). In this assumption, the virial status of a clump can be explained as self-gravitationally collapsing (α_vir < 1), in virial equilibrium (α_vir = 1; the clump is gravitationally stable, i.e., virialized), in quasivirial equilibrium (1 < α_vir ≤ 2; the clump is slightly expanding but still gravitationally bound), or gravitationally unbound (α_vir > 2). In order to inspect the virial state of the extended sources in W51A (i.e., the regions in Table 3), we utilized the public ¹³CO (2–1) data cube from 10 m Heinrich Hertz Telescope (Kang et al. 2010). Note that we measure the ratio of the internal kinetic energy to gravitational binding energy in the extended sources, ignoring surface pressure terms and effects of magnetic fields. We used the integrated ¹³CO line profile of each clump to fit a Gaussian. In order to determine the central gas ¹³CO velocity of each extended source, we used the literature values of the velocity ranges defined in Kang et al. (2010) and Ginsburg et al. (2015).

We derive the virial parameter, α_vir, assuming constant density for the extended sources so that

$$\begin{eqnarray}&&{\alpha }_{\mathrm{vir}}=\displaystyle \frac{{M}_{\mathrm{vir}}}{M}\sim 210\,\times {\left(\displaystyle \frac{\sigma }{\mathrm{km}{{\rm{s}}}^{-1}}\right)}^{2}\times \left(\displaystyle \frac{R}{\mathrm{pc}}\right)\times \left(\displaystyle \frac{{M}_{\odot }}{M}\right),\end{eqnarray} \tag{ 2 }$$

where R is the radius of the clump in parsec scale, σ is the FWHM of the ¹³CO (2–1) line in km s⁻¹, and M is derived from submillimeter dust emission-based Σ in units of M_☉. If we assume the density profile falls off as 1/r, the α values will be ∼10% smaller than in the constant-density case (MacLaren et al. 1988). As we can see from Equation (2), the uncertainty of α_vir is derived from the errors of the gas velocity width, derived clump mass, and distance estimation so that a conservative total uncertainty of α_vir is about factor of 2 (e.g., Kauffmann et al. 2013).

We present these parameters in Table 4. The extended sources in the G49.4–0.3 region show a mean α_vir ∼ 0.63, while the median α_vir ∼ 0.45. The sources in G49.5–0.4 a–c show mean and median α_vir values of 0.66 and 0.58, respectively. The G49.5–0.4 f–j sources show a mean α_vir ∼ 6.32 and median α_vir ∼ 4.77. These derived values of α_vir show that the sources in G49.4–0.3 are mostly subvirial, which indicates these sources are probably undergoing self-gravitational collapse. The extended sources G49.5–0.4 f–j are all supervirial with α_vir ≳ 4, indicating the sources are gravitationally unbound and expanding. The source G49.5–0.4 a is close to the virial equilibrium status (i.e., gravitationally stable). Source G49.5–0.4 b shows the lowest virial parameter, α_vir ∼ 0.18, which is unique from any other extended source in this study. We do not detect any individual MYSO sources within G49.5–0.4 b. From the lowest α_vir value and the absence of significant YSOs, we assume G49.5–0.4 b is the youngest molecular clump in the W51A area.

Figure 15 shows the L/M versus α_vir of all extended sources in W51A. The correlation shows that both evolutionary tracers of stellar cluster formation are in good agreement, indicating G49.5–0.4 f–j are relatively older than the G49.5–0.4 a–c sources, while the entire G49.04–0.3 region is likely younger on average than the entire G49.5–0.4 region. The result of L/M versus α_vir is not only consistent with previous studies, but as we will now discuss, the result helps to further clarify our understanding of the star formation history of the W51 area.

Zoom In Zoom Out Reset image size

Elmegreen (1992) suggested that several Myr of age difference in between two nearby (∼10–50 pc away) sources may be evidence for triggered star (cluster) formation. Using this logic, Okumura et al. (2000) hypothesized that G49.5–0.4 had undergone triggered sequential star formation. They suggest that the stellar clusters in the region around G49.5–0.4 h and j (which they call "Region 1") are the oldest, while the region around G49.5–0.4 a–e (their "Region 3") are the youngest, and that the star cluster formation in Region 3 is triggered by the stellar wind from the evolved stars in Region 1 and the expansion of the G49.5–0.4 f and g H ii regions (their "Region 2").

We derived the total mass ratio between NIR-revealed stars (from Okumura et al. 2000) to MIR-revealed stars (from Kang et al. 2009 and this study), ${{\rm{\Sigma }}}_{M* ,\mathrm{MIR}}$/${{\rm{\Sigma }}}_{M* ,\mathrm{NIR}}$, in Regions 1, 2, and 3. We assume that the NIR-detected stars are less embedded and therefore relatively older than the deeply embedded MIR-detected stars, meaning that the ratio ${{\rm{\Sigma }}}_{M* ,\mathrm{MIR}}$/${{\rm{\Sigma }}}_{M* ,\mathrm{NIR}}$ should get smaller with cluster age. We find that ${{\rm{\Sigma }}}_{M* ,\mathrm{MIR}}$/${{\rm{\Sigma }}}_{M* ,\mathrm{NIR}}$ ∼ 0, 0.01, and 0.10 for Regions 1, 2, and 3, respectively. This relative decrease in evolutionary state from Regions 1 to 2 to 3 is basically consistent with the relative ages that are claimed by Okumura et al. (2000), and consistent with the trends we see from L/M and the virial parameter (Figure 15).

In contrast, Clark et al. (2009) claimed that they could not find any evidence of sequential triggered star formation in the W51A area but found indication of independent star formation activities in multiple positions that could be generated by external triggering effects as Nanda Kumar et al. (2004) suggested. While Clark et al. (2009) and Nanda Kumar et al. (2004) investigated the star formation history of the W51 region via NIR observations, Ginsburg et al. (2015) pointed out the absence of signs of triggered star formation from expanding H ii regions toward G49.5–0.4 a–e based on Karl G. Jansky VLA (JVLA) centimeter observations. Kang et al. (2010) studied the overall structure of the W51 area by observing CO isotopologues. They suggested that the W51A region underwent cloud–cloud collision to produce the stellar clusters possessing high-mass stars, and the extended source G49.5–0.4 b is possibly at the colliding location between two molecular clouds that could be distinguished with one at ∼58 km s⁻¹ (encompassing G49.5–0.4 a–e) and another at 68 km (a.k.a. the High Velocity Stream; Carpenter & Sanders 1998). They found a "bridge" in the position–velocity diagram which connects two different CO velocity components (<61 and ∼68 km s⁻¹) as well as the self-absorption line at the location of the G49.5–0.4 b region. They insisted that the self-absorption line was due to the colliding interface (G49.5–0.4 b) having been heated while the surroundings were still cold while the "bridge" showed the interaction between two different clouds.

The comparison of two relative evolutionary tracers, L/M and α_vir, could be a more accurate way to determine the evolutionary states of star cluster formation in the W51A area than the absolute age calculations that were performed in previous NIR studies (e.g., Okumura et al. 2000). In general, these previous studies focused on analyzing the morphologies of IR and/or millimeter bubbles around YSOs, and they estimated the ages of YSOs as ∼0.5–3 Myr. The ages were typically determined from the isochrones (e.g., Schaller et al. 1992; Meynet & Maeder 2000), sometimes comparing them to the expansion ages of H ii regions (Wood & Churchwell 1989). However, these calculations can have relatively large errors (up to 100%; Vacca & Sandell 2011), and thus assuming star-forming history based on these values and the morphologies of molecular bubble structures is likely to be highly uncertain. An example of this is Clark et al. (2009) estimating the age of LS1 to be at least 3 Myr (possibly 6 Myr or older), while Okumura et al. (2000) estimated the age as ∼2.3 Myr. The difference in derived ages between Okumura et al. (2000) and Clark et al. (2009) led to the different interpretations of star-forming history in W51A.

From Figure 15, we can see that the evolutionary states of the G49.5–0.4 f–j sources are clearly separate from sources on G49.5–0.4 a–c and G49.4–0.3 regions. If the internal feedback of the G49.5–0.4 f–j regions could affect the star formation in G49.5–0.4 a–c, one would expect to find a smoothly continuous trend of L/M versus α_vir among all G49.5–0.4 sources. This might support the scenario sketched by Ginsburg et al. (2015) that the location of the GMC possessing the G49.5–0.4 f–j regions is different from either G49.5–0.4 a–e or G49.4–0.3. In this case, the independent formation of stellar clusters, as suggested in Clark et al. (2009), would mean that the non-interacting (and thus separated) clouds induce their own star formation history.

Given its very low α_vir and L/M, the source G49.5–0.4 b might be the youngest clump in the W51A area. The low α_vir and L/M can be due to the lack of internal heating sources (i.e., young stars) so that the internal gas motion is not strong enough to overcome the gravitational pressure of the molecular clump. The evolutionary state of G49.5–0.4 b as the uniquely young molecular clump in the W51A region can be explained by the recent cloud–cloud collision that is suggested by Kang et al. (2010). Comparing the CO observational results of Kang et al. (2010) to synthetic CO emission lines from a theoretical simulation of a cloud–cloud collision scenario (e.g., Wu et al. 2017; Bisbas et al. 2018) can be helpful to address the effect and evidence of cloud–cloud collision on the molecular clump formation and evolution.