1. Introduction

Massive (O and B-type) stars have an enormous impact on their surroundings due to their ultraviolet (UV) ionizing radiation and energetic winds. Ionized gas in HII regions produces strong infrared (IR), optical, and thermal radio continuum emission. In addition, ionized gas mixed with heated dust makes an HII region bright in thermal IR emission.

A common spatial shape reported for individual Galactic HII regions at IR wavelengths is the ring morphology, or "bubble" seen in projection. A visual examination of the images at 8.0 µm from the Galactic Legacy Infrared Survey Extraordinaire (GLIMPSE; ^{Benjamin et al. 2003}) allowed the identification of about 600 full or partial IR dust bubbles (IRDBs) in the inner Galactic plane (^{Churchwell et al. 2006}, ²⁰⁰⁷) between longitudes from -60° to +60°. Presently, more than 5000 IRDBs have been identified (^{Simpson et al. 2012}). The main charac- teristics of many of these bubbles were investigated by several authors (see for example ^{Deharveng et al. 2010}; ^{Alexander et al. 2013}). These bubbles are about 1′ - 3′ in size, show filamentary appearence, and many of them lie close to massive stars and coincide (or enclose) classical and ultracompact HII regions.

At 8 µm, most of the emission originates in strong features of polycyclic aromatic hydrocarbons (PAH) molecules, which are considered to be good tracers of warm UV-irradiated photodissociation regions (PDR; ^{Hollenbach & Tielens 1997}). Since these complex molecules are destroyed in the ionized gas (^{Povich et al. 2007}; ^{Lebouteiller et al. 2007}), they de- lineate the ionization front and indicate the presence of substantial amounts of molecular gas surrounding the bubbles. Therefore, these bubbles provide a good insight of the sculpting influences of the UV photons of massive stars on the molecular clouds where they are born.

The geometry of the IR bubbles is also important for understanding triggered star formation scenarios. Classical models, like the collect and collapse mechanism (C&C; ^{Elmegreen & Lada 1977}) and the radiative driven implosion process (RDI; ^{Lefloch & Lazareff 1994}), suggest that the formation of stars can be triggered by the action of HII regions over their parental molecular environment. Molecular condensations lying at the border of many galactic bubble-shaped HII regions are then among the most likely sites for stellar births and for early stages of star formation (e.g. ^{Deharveng et al. 2008}; ^{Zavagno et al. 2010}; ^{Brand et al. 2011}; ^{Samal et al. 2014}). Detailed studies of IR bubbles have shown the presence of young stellar objects (YSOs) in their environments, although the triggered formation scenario cannot always be proven (e.g. ^{Dewangan & Ojha 2013}; ^{Alexander et al. 2013}).

As part of a project aimed to study and characterize galactic bubbles, we have selected S 21 from the sample of bubbles reported by ^{Churchwell et al. (2006)} to perform a study of its molecular and dust environment, and to search for candidates to YSOs in their vecinity. S 21 is located 10′ east from S 24 (^{Cappa et al. 2016}).

S 21 is an almost complete circular bubble (see Figure 1) of ≈ 0.′75 in radius centered on RA, Dec.(J2000) = 16^h53^m7.9^s, −44◦23′13′.′1. The dusty nebula appears filamentary at 8 µm, with a rather steep inner border and a more diffuse outer one (Figure 1). The point source IRAS 16495-4418 (RA, Dec.(J2000) = 16h53m8.5s, −44◦23′21′′) coincides with the bubble. ^{Watson et al. (2010)} determined the dust temperature inside the bubble using the emission at 24 and 70 µm from Spitzer-MIPS and a modified blackbody. They found a temperature gradient with the highest temperatures (85 K) close to the center of the bubble and the lowest ones (71 K) close to the border.

Fig. 1 Composite IRAC image of S 21. The emission at 3.6 μm is in blue, at 4.5 μm is in green, and at 8.0 μm is in red. The positions of IRAS 16495-4418, G341.3553-00.2885, and the catalogued B3-star are indicated with white, yellow, and green crosses, respectively. The color figure can be viewed online. 

Figure 1 reveals a bright point source detected in the IRAC bands, named G341.3553-00.2885 (RA, Dec.(J2000) = 16^h53^m7.12^s, −44◦23′24''.1), located close to the center of the bubble. It has counterparts in several optical (GSC, NOMAD, and DENIS) and IR catalogues (2MASS, Spitzer, WISE). Projected close to the northern border of the bubble, the B3-type star HD 329056 (Simbad, 16^h53^m7.63^s, −44◦22′9′.′28) has been identified.

The distance to S 21 is matter of some debate. Adopting a mean velocity of −44 km s⁻¹ for the gas linked to S 21 (see § 3.1), circular galactic rotation models predict near and far kinematical distances of 3.7 and 12-13 kpc (e.g. ^{Brand & Blitz 1993}). Bearing in mind that most of the IRDBs are closer than 8 kpc (^{Churchwell et al. 2006}), we will adopt for S 21 the near kinematical distance of 3.7 kpc. A distance un- certainty of 0.5 kpc results after taking into account a velocity dispersion of 6 km s⁻¹ for the interstellar gas.

In this study, we present a molecular line and dust continuum analysis toward S 21 and its environs, with the aim of studying the distribution and physical properties (mass, densities, temperature, kinematics, etc.) of the molecular gas and dust associated with the bubble. We based our study on ¹²CO(2-1) and ¹³CO(2-1) data obtained with the APEX telescope¹, and complementary archival IR, optical, and radiocontinuum data.

The simple morphology of S21, along with the strong evidence of star formation in its environs (see below), make this object an excellent laboratory for the investigation of possible scenarios of triggered star formation. With that aim, we also analyze the spatial distribution of the YSO candidates in their vicinity and their relation to the bubble, and search for probable exciting stars.

2. Data

3. Molecular gas linked to the bubble

3.1 Molecular gas distribution

In Figure 2 we show the ¹²CO(2-1), ¹³CO(2-1), and C¹⁸O(2-1) spectra averaged over a region ≈ 4′ × 4′ around the central position of S 21. The bulk of the molecular emission appears concentrated between −55 km s⁻¹ and −20 km s⁻¹, with three components peaking approximately at −45 km s⁻¹,−38 km s⁻¹, and −30 km s⁻¹. The spatial distribution of these components reveals that only the molecular component peaking at −45 km s−1 shows a morphological correspondence with S 21. The emission distribution in the velocity interval from −45.8 to −42.6 km s⁻¹ shows a bubble fully inmersed in its parental molecular cloud. CO shells have been reported for many other IRDBs (see for example ^{Arce et al. 2011}). Figure 3 displays the integrated ¹³CO(2-1) emission distribution in the mentioned velocity range in contours overlayed onto the emission at 8 μm. A clumpy circular structure coincident with the infrared bubble with a minimum in the emission projected onto its center can be discerned. The mean radius of the molecular shell is 1'.3. The molecular correspondence between the IR emission at 8 μm and the molecular emission within the quoted velocity intervalindicates that the shell is the molecular counterpart of the IR bubble, and that they are physically associated.

Fig. 2 ¹²CO(2-1) (red line), ¹³CO(2-1) (green line), and C¹⁸O(2-1) (blue line) averaged molecular line spectra obtained toward S 21. The color figure can be viewed online. 

Fig. 3 Overlay of the 8 μm emission (colorscale) and the ¹³CO emission (in contours) integrated in the velocity range from −45.8 to −42.6 km s⁻¹. Contour levels go from 3 K ( ≈ 45 rms) to 6.9 K in steps of 0.3 K and from 6.9 in steps of 1 K. The white crosses show the position of candidate YSOs identified in the VVV database projected on the molecular shell (see § 7). The color figure can be viewed online. 

Using Clumpfind we have identified several condensations in the molecular structure around S 21, which will be referred as "clumps" (^{Blitz 1993}; ^{Williams et al. 2000}). The location of the clumps, labeled from 1 to 5, is depicted in Figure 4a. We were left only with clumps adjacent to the bubble, which were very likely formed in the collected layers of the molecular gas. The coordinates and peak temperatures of the clumps are listed in Table 1.

Table 1 Parameters of the molecular clumps 

Clump	R. A. (J2000)	Decl.(J2000)	Tpeak12	∆v12	Tpeak13	∆v13	Texc	τ13	τ12
	(h m s)	(° ' '')	(K)	(Km s-1)	(K)	(Km s-1)	(K)
1	16:53:03	−44:22:16.3	11.5	3.6	7.5	1.6	16.6	0.96	34.3
2	16:53:00	−44:22:57.3	8.5	2.5	7.3	2.1	13.5	1.83	79.5
3	16:53:01	−44:22:09.2	13.0	4.3	10.2	1.7	18.1	1.50	33.5
4	16:53:08	−44:22:01.3	8.8	3.8	6.5	1.8	13.1	1.28	33.3
5	16:53:12	−44:23:13.7	10.5	4.0	7.8	1.9	15.6	1.30	33.1

In order to unambiguously ascertain the relation- ship of each clump with the nebula and to provide a better visual display, in Panels b to f of Figure 4 we show the channel maps of ¹³CO(2-1) in velocity intervals of 0.6 km s⁻¹, overlaid on the GLIMPSE 8.0 µm emission image. Between −45.8 to −45.2 km s⁻¹ (Figure 4b) Clumps 1 and 2 become noticeable, bordering the bubble from its northwestern side. In this velocity range, two bright molecular features can be also detected to the south of S 21, unconnected to this bubble. Clumps 1 and 2 achieve their maximum brightness temperature in the velocity interval from −45.2 to −44.6 km s⁻¹ (Figure 4c). Clump 2 is still detected between −44.6 to −44.0 km s⁻¹ (Figure 4d), where Clump 3 becomes first noticeable. In this velocity interval, Clump 5 appears as an extension of an elongated feature that borders the eastern side of S 21. From −44.0 to −43.4 km s⁻¹ (Figure 4e) Clumps 3, 4, and 5 attain their peak temperature. They surround the nebula from south to east (Clump 3) and from south to north (Clumps 4 and 5). Finally, between −43.4 to −42.8 km s⁻¹ (Figure 4f), Clumps 4 and 5 are barely detected but Clump 3 is very bright. Clump 3 dissapears at ≈ −42.1 km s⁻¹ (not shown here). At a distance of 3.7 kpc, the radius of the molecular shell is ≈ 1.4 pc.

Fig. 4 Panel a: ¹³CO emission in the velocity range from −45.8 to −42.6 km s⁻¹. Contour levels go from 3 K (≈ 45 rms) to 6.9 K in steps of 0.3 K and from 6.9 K in steps of 1 K. White dotted lines indicate the limits chosen to define the clumps (see text). The beam size is indicated in the lower left corner of this panel. Panels b to f: Channel maps of ¹³CO in intervals of 0.6 km s⁻¹ (white contours) superimposed on the GLIMPSE 8.0 μm red colorscale. Contours go from 4.5 K (≈ 32 K rms) to 10 K in steps of 0.5 K, and from 10 K in steps of 1 K. The color figure can be viewed online. 

The eastern section of the molecular shell (Clumps 4 and 5) coincides with the photodissociation region (PDR), while the western section (Clumps 1, 2, and 3) is projected onto regions of low emission at 8 μm close to the PDR.

4. The interstellar dust linked to the bubble

4.1. Dust distribution

Dust associated with the bubble can be analyzed from the distribution of the emission at different wavelengths in the infrared.

Each panel of Figure 5 shows a composite image of the emissions at 8 µm (in red) and 3.6 µm (in blue) from IRAC-GLIMPSE, and the emissions at 24, 70, 160, 250, 500, and 870 µm (in green) from MIPS, Herschel-PACS and -SPIRE, and ATLASGAL.

Fig. 5 Composite images showing the emission in the near-, mid-, and far-infrared. All the images display the emissions at 8 μm (in red) and at 3.6 μm (in blue) from IRAC-GLIMPSE. The emissions in green correspond to different IR wavelengths indicated in the upper left corner of each image. The angular resolution of the IR images is shown in the lower left corner. The color figure can be viewed online. 

The emission at 24 µm is seen projected onto the inner part of the bubble, with its maximum near the central point source. The emission at 70 µm coincides fairly well with that at 8 µm. The emission differs significantly at larger wavelengths. At 160 µm it encircles externally the bubble completely and ap- pears more extended than at 70 µm. At 250 and 500 µm, the emission is brighter than at lower wavelengths and extends toward the western and northern parts of the bubble. The SPIRE emission distribution at 350 µm (not shown here) is similar to that at 250 and 500 µm. Similarly to the case of S 24, cold dust emission detected at 870 µm resembles that at 250 and 500 µm, although the emission at 250 and 500 µm, which also shows the distribution of cold dust, appears more extended. We have to bear in mind that large scale dust continuum emission might be filtered out in LABOCA bolometric observations.

Two facts can be concluded from these images. On the one hand the spatial emission distribution in the mid- and far-IR seems to show a gradient in dust temperature, with lower values in the outer regions of the bubble where continuum emission at larger wave- lengths dominates. Indeed, emission at λ > 160µm is present well outside the PDR. On the other hand, the emission at 24 µm inside S 21 is indicative of the presence of exciting sources. Both statements will be analyzed in some detail in the next sections.

A comparison with the molecular gas distribution around the bubble (see Figure 6) shows that Clumps 1 to 4 partially coincide with the cold dust counterpart identified in the Herschel-SPIRE and in the LABOCA images.

Fig. 6 Composite image showing the emission at 8 μm (in red) and at 870 μm (in blue), and the same ¹³CO contours of Figure 3. The color figure can be viewedonline. 

The observed dust continuum emission at different wavelengths in the IR coincides with previous findings toward other IR dust bubbles.

The molecular emission from the ¹²CO(2-1) line may contribute to the thermal emission at 870 µm. In our case, this contribution is less than about 1% of the emission at 870 µm, and consequently, within the calibration uncertainties. The other process that contributes to the emission at 870 µm is the free-free emission from ionized gas. This contribution, bearing in mind the flux density at 1.4 GHz (see § 5), is less than 0.1% of the emission at 870 µm, and again, within the calibration uncertainties.

4.2. Dust temperatures and mass

Dust temperatures T_dust in the environs of the S 21 region can be obtained using the SPIRE images at 250 µm and 350 µm. To perform this we convolved the image at 250 µm down to the angular resolution at 350 µm and assumed that the emission is optically thin. The color-temperature map was constructed maps, i.e., T _dust= f(T)^-1, where f(T) is:

fT=S250S350= Bv(250 μm,T)Bv(350 μm, T)250350βd (7)

In this expresion S250 and S350 are the flux densi- ties in Jy beam⁻¹, B _ν (ν, T) is the blackbody Planck function and β_d, the spectral index of the thermal dust emission. The pixel-to-pixel temperature was calculated assuming β_d = 2. This is a typical value adopted for irradiated regions.

The dust temperature map is shown in Figure 7. The highest dust temperatures (33 K) are present at the NE extreme of the 8 µm bubble. Values in the range 24-33 K coincide with the eastern section of the bubble, while lower values (21 K) are obtained for the western side. Low dust temperatures coincide with molecular gas and with regions with faint emission at 8 µm. ^{Watson et al. (2010)} derived dust temperatures for the interior of S 21 based on images from MIPS at 24 µm (whose emission is detected inside the bubble) and 70 µm (detected up to the border of the bubble). Our estimates, based on images in the far IR, sample colder dust present in the outkirts of the bubble.

Fig. 7 Dust temperature map derived from the Herschel emission at 250 and 350 μm superimposed on the image at 8 μm. Blue color scale goes from 15 to 33 K. Brighter blue regions indicate higher dust temperatures. Contour levels correspond to 15 to 30 K, in steps of 3 K. The color figure can be viewed online. 

Dust masses can be estimated from the expression (^{Hildebrand 1983}):

Mdust=S870d2k870B870(Tdust) (8)

where S870 is the flux density at 870 μm, d = 3.7 ± 0.5 kpc, κ = 1.0 cm²/gr is the dust opacity per unit mass (^{Ossenkopf & Henning 1994}), and B₈₇₀(T_dust) is the Planck function for a temperature T_dust.

The flux density S₈₇₀ obtained by integrating the emission over the observed emitting area linked to S 21 at this wavelength (see Figure 5f) amounts to 1.9 ± 0.4 Jy. Adopting a mean value T_dust = 30 K for S 21, a dust mass M_dust = 1.53 ± 0.90 M⊙ can be estimated. For gas-to-dust ratios in the range 100-186 (^{Beuther et al. 2011}), the gas mass amounts to 153-285 M⊙.

5. The ionized gas

Figure 8 shows an overlay of the SUMSS image at 843 MHz (in white contours) and the emissions at 8 μm and 24 μm (in colorscale). The 843 MHz image (synthesized beam = 43’’×62’’) shows a radio source coincident with S 21 catalogued by ^{Murphy et al. (2007)} with a size of 75’’.7×69’’.6 and a flux density S_0.843 = 49.2 mJy. The elongated shape of the source is due to the synthesized beam of the data. The source is also detected at 1.4 GHz (SGPS, ^{Haverkorn et al. 2006}) with an estimated flux density of S_1.4 = 52 mJy. With these values, the derived spectral index α ≃ +0.1 (S_ν ∝ ν^α). Within uncertainties, this value is consistent with thermal emission of an HII region optically thick at 843 MHz.

Fig. 8 Overlay of the radio continuum emission at 843 MHz (white contours), the IRAC emission at 8μm (in green), and the MIPSGAL emission at 24μm (in blue). Contours correspond to 10, 15, 20, and 25 mJy beam⁻¹. The color figure can be viewed online. 

Thus, these results indicate the presence of ionized gas, and consequently, the existence of at least one exciting source, compatible with warm dust inside the bubble as shown by the 24 μm image. The emission at 8 μm due to PAHs reveals a PDR bordering the ionized region and the presence of molecular gas in its exterior. The radius of the ionized region inside the 8 μm dust bubble is 38’’ or 0.7 pc at 3.7 kpc. The derived spectral index confirms the classification of S 21 as an HII region by ^{Anderson et al. (2015)}. The characteristics of the region, visible in the IR and in the radio continuum, as well as the large mean original ambient density of 2100 cm⁻³ (see § 3.2) suggest a compact HII region (^{Urquhart et al. 2013}).

To produce the observed flux at 1.4 GHz we require an UV photon flux that can be obtained from ^{Matsakis et al. (1976)}

where S_1.4 is in units of mJy, ν in GHz, d in kpc, and the electron temperature T_e in 104 K. Considering the galactic electron temperatures for HII regions (^{Quireza et al. 2006}), we assumed T_e = 7000 K. The required UV flux amounts to 6.5×10⁴⁶ s⁻¹. This ionizing flux is underestimated since part of the stellar UV photons are used to heat the dust. Considering that half of the stellar photons are absorbed by dust (^{Inoue 2001}), the UV photon flux would be 1.3×10⁴⁷ s⁻¹. According to ^{Martins et al. (2005)}, this value indicates that the ionization of the gas could be produced by an O9.5V or earlier type star.

6. Search for an exciting star

Identifying the exciting star of this HII region is not an easy task. The extinction towards the inner part of the IR bubble can help to identity this star since we expect for it a similar extinction. This extinction can be deduced from the expresion by Bohlin et al. (1978):

Considering only the H ₂ column density toward the IR source and taking into account N(H ₂) = 2.9 × 10²¹ cm⁻² (estimated from the ¹³CO emission), we calculate a visual absorption of 30 mag

To search for exciting stars the color-magnitude diagrams were built using the VVV DR4 catalog (Figure 8). The foreground main sequence stars are situated around J −K _s = 0.7 mag, and the mean reddening is estimated as E(J − K _s) = 0.5 mag; some field red giants can be also identified. The 76 extremely red stars ((J − K _s) > 4 mag) are projected in the field (see also the lower panel of Figure 9); seven of them are projected within the inner radius of the bubble and could be candidates for the exciting source. However, all of them are too faint and a spectroscopic follow up is necessary to reveal their nature.

Fig. 9 Upper panel. (J − K _s) vs. K _s and (H − K _s ) vs. K_s VVV color magnitude diagrams within a radius of 1’.5 centered on S 21. The vertical red line is the zero age main sequence, reddened with E(J −K _s) = 0.5 mag (the mean reddening of the field stars). The candidate YSOs projected on the molecular shell are overplotted as red squares. The red arrow shows the reddening vector corresponding to Av=10 mag. The G341.3553-00.2885 and HD329056 sources are labeled. Bottom panel. The (H − K _s) vs. (J − H) color-color diagram. The continuous and dashed lines represent the sequence of the zero-reddening stars of luminosity classes I (Koornneff et al. 1983) and V (^{Schmidt-Kaler 1982}). The color figure can be viewed online. 

As pointed out in § 1, G341.3553-00.2885, located close to the center of the IR bubble, is detected both in the optical and at IR wavelengths, appearing saturated in the VVV images. The analysis of the emission of this IR source using the VOSA⁵ and TLUSTY⁶ tools and all the available detections at different wavelengths (taking into account stellar atmosphere models for O and B stars) suggests that the star might be a O8V or a B0I star. From the comparison of the observed colors taken from the 2MASS catalogue (source 2MASSJ 16530711-4423239) with the intrinsic magnitudes and colors taken from the 2MASS catalogue (source 2MASSJ 16530711-4423239) with the intrinsic magnitudes and colors taken from ^{Martins & Plez (2006)} for O8V stars and from ^{Bibby et al. (2008)} for B0I stars, we estimate visual absorptions and distances of 21 mag and 0.58±0.9 kpc for an O8V type star and 20.9 mag and 1.7±0.8 kpc for a B0I type star, using the standard reddening law and Bessel et al. (1998) transformations. The calculated distance is not comparable with the kinematical distance of 3.7 kpc to the complex. Thus, G341.3553-00.2885 is most probably a foreground object. Spectroscopic data are necessary to verify this suggestion and to unambiguously identify the exciting star of this bubble, but this is beyond the scope of this paper.

7. Star formation toward S 21

As pointed out in § 1, star formation can be favored in the molecular and cold dust clumps in the environs of HII regions. To help testing if this is the case for the S 21 bubble, we can analyze the stability of the molecular clumps by comparing virial to LTE masses. Following ^{MacLaren et al. (1988)}, the virial mass can be obtained as:

where R and ΔV are the radius of the region and the velocity width measured from the ¹³CO(2-1) emission emission, and k ₂ depends on the density distribution in the clump, being 190 or 126 according to ρ ∝ r⁻¹ or ρ ∝ r⁻², respectively. Virial masses are included in Column 9 of Table 2 for the two density profiles. The ratio MvirM(H2)=γ is listed in Column 10.

As classical virial equilibrium analysis establishes, a ratio γ > 1 would imply that the clumps could be stable against collapse, while lower values indicate that collapse is possible. The derived ratio for Clump 3 is consistent with collapse, while results for the other four clumps are not conclusive.

Both virial and LTE masses have large uncertainties due to the distance indetermination (30% for M(H ₂) and 15% for M _VIR ). Minor errors are due to uncertainties in the boundaries of the clumps resulting in errors in their areas and in the LTE masses.

Virial masses are not free of additional errors, since the existence of magnetic fields might overestimate by up to a factor of 2 the derived values (^{MacLaren et al. 1988}). Another source of error is the density profile of the clump, which is unknown. Uncertainties in the distance in the LTE mass and those due to the density profile in virial masses were taken into account in the value of γ.

To investigate if YSOs are detected toward the molecular clumps we performed a search for candidates in the available point source catalogues within a region of 1’.5 centered on the bubble. In the region of the bubble and the surrounding molecular shell, no sources with characteristics of YSOs were identified in the Spitzer and WISE databases by applying color criteria (^{Allen et al. 2004} for Spitzer data; ^{Koenig et al. 2012} for WISE data), mainly because the WISE data are not deep enough to measure the fainter stars.

To select new candidate YSOs in the region we used the VVV database and applied photometric criteria. First, from the near-infrared (J − H)/(H − K) color - color diagram we selected all stars which were at least 3σ distant from the reddening line that marks the colors of dwarf stars. The list thereby obtained was cross-matched with GLIMPSE measurements. They are shown in Figure 10. The objects with [K _S − [3.6]] > 0.5 or [[3.6] − [4.5]] > 0.5 magnitudes are considered to be most probable Class I and Class II YSOs. These limits are set in order to avoid selecting objects that are more likely Class III objects or normal stars (dashed red line in Figure 10). Our final list contains 71 YSO candidates.

Fig. 10 Upper panel. The K _S − [3.6],[3.6] − [4.5] color-color plot of stars that are detected in GLIMPSE I. The red dashed lines represent the limits used to select Class I and Class II YSOs. Middle panel. The (J − H) vs. (H − K _S) color-color diagram of the sample. The continuous and dashed lines represent the sequence of the zero-reddening stars of luminosity classes I (^{Koornneef et al. 1983}) and V (^{Schmidt-Kaler 1982}). Lower panel. The extremely red stars detected in the field. The color figure can be viewed online. 

The location of the sources in the color magnitude and color-color diagrams is shown in Figure 9, while Figure 3 shows their spatial correlation with the molecular clumps. The presence of candidate YSOs projected on the molecular shell suggests that star formation has been active recently. However, it is not possible to determine if all candidate YSOs are linked to the molecular shell around S 21.

The widely known collect-and-collapse mechanism (C&C; ^{Elmegreen & Lada 1977}) proposes that the expansion of an ionization front over its parental molecular cloud can trigger the star formation process. The molecular gas may fragment alongside the ionization front as it expands and the fragments may become inestable, giving rise to a new generation of stars. To test whether the C&C mechanism has triggered star formation in the molecular shell around S 21 we applied the analytical model by ^{Whitworth et al. (1994)}. For the case of HII regions, the model predicts the age of the HII region at which the fragmentation occurs (the fragmentation time scale), t _frag, the size of the HII region at that moment, R _frag, the mass of the fragments, M _frag, and their separation along the compressed layer, r _frag. The parameters required to derive these quantities are the UV photon flux of the exciting star, N _Ly, the ambient density of the surrounding medium into which the HII region is evolving, n₀, and the isothermal sound speed in the shocked gas, a_s.

To estimate these parameters we took into account a large range of spectral types, i.e. from O3V to O9.5V stars, with UV fluxes in the range N _Ly = (43−0.4)×10⁴⁸ s⁻¹ (^{Martins et al. 2005}). Using the mean H ₂ ambient density n _H2 = 2100 cm⁻³ (see § 3.2), and a_s = 0.2-0.6 km s⁻¹, we obtained t _frag = (1.0 − 1.5) × 10⁶ yr, R _frag = 4.3 − 3.0 pc, M _frag = 20-29 M⊙, and, r _frag = 0.5-0.3 pc.

The dynamical age of the Hii region can be estimated using the equation (^{Dyson & Williams 1997}):

where R_S is the original Strömgren radius, equal to 0.13-0.63 pc for the adopted spectral types, and c_s is the sound velocity in the ionized gas. Derived dynamical ages span the range (1.4-33)×10⁴ yr.

We found that the dynamical age is significantly smaller than the fragmentation time scale tfrag for the adopted ambient density, and hence, the C&C process does not seem to be responsible for the triggering of star formation in the envelope. An RDI scenario could be investigated. However, evidences of this process (such the presence of pillars) appear to be absent. The rest of the parameters, R _frag, M _frag, and r _frag, confirmed that the HII region is too young to start triggering.

8. Conclusions

We performed a multiwavelength study of the IR dust bubble S 21 using APEX observations of the ¹²CO(2-1) and ¹³CO(2-1) lines and complementary images in the near-, mid-, and far-IR from IRACGlimpse, MIPSGAL, Herschel, and ATLASGAL.

The molecular emission in the ¹²CO(2-1) and ¹³CO(2-1) lines obtained with the APEX telescope toward the IR dust bubble S 21 revealed a molecular shell encircling the bubble and partially coincident with the PDR shown by the IRAC emission at 8 μm. With a mean radius of 1.4 pc, the molecular shell is larger than the 8 μm bubble. This shell is detected in the velocity interval from -45.8 to -42.6 km s⁻¹. The velocity of the shell confirms that S 21 belongs to the same complex than S 24. Five clumps were identified in the molecular shell, with radii in the range 0.64-0.75 pc, LTE masses of 350-880 M _⊙, and volume densities of (4.5-7)×10³ cm⁻³. Virial masses for the clumps suggest that at least one of them can collapse. The original ambient density in the region was about 2100 cm⁻³.

Complementary images in the near-, mid-, and far-IR from IRAC-Glimpse, Herschel, and ATLASGAL were used to characterize the dust linked to the ubble. The emission at 24 μm coincided with the inner part of the bubble, indicating warm dust inside. The spatial distribution of the emission in the far-IR from 70 to 160 μm coincided with the 8 μm bubble and the molecular emission, while the emission at 500 and 870 μm resembled that at 250 μm. The spatial distribution of the Herschel-PACS and Spire, and ATLASGAL emissions showed a cold dust component coincident with the molecular gas. Dust temperature determinations using the emissions at 250 and 350 μm allowed us to estimate dust temperatures in the range 21-33 K for the cold dust component linked to the 8 μm bubble.

Thermal radio continuum emission at 843 MHz and 1.4 GHz was detected from inside the bubble, indicating the existence of ionized gas and excitation sources, in agreement with the presence of warm dust. We conclude that a compact HII region has developed. However, the identification of the exciting star is a difficult task and deserves additional studies.

A search for candidate YSOs was performed. We were able to identify many candidates in the VVV database projected on the molecular clumps, although it is not clear whether all these candidates are linked to the molecular shell. This result suggests that star formation has been active recently. The HII region is probably very young for the C&C process to be active.