An ALMA View of Molecular Filaments in the Large Magellanic Cloud. I. The Formation of High-mass Stars and Pillars in the N159E-Papillon Nebula Triggered by a Cloud–Cloud Collision

Figure 1 shows integrated intensities of ¹³CO(J = 2–1) toward the N159E-Papillion region obtained in the ALMA Cycle 4 and 1 (Paper I). We found overall agreement with each other, while the Cycle 4 data clearly resolve significantly more details of the filaments. Figure 2 shows the velocity-channel map of the region, which also shows the multiple filamentary distributions. Most of the filaments are aligned roughly in the north–south direction. The first-moment map of ¹³CO(J = 2–1) in Figure 3(a) shows the velocity gradient along that direction. Note that a similar figure was also presented in the previous lower-resolution study (Figure 6(a) in Paper I). The averaged ¹²CO, ¹³CO, and C¹⁸O spectra shown in Figure 3(b) demonstrate that there are no velocity components significantly over the blueshifted (≲225 km s⁻¹) and redshifted (≳245 km s⁻¹) ranges. This means that there are no external components affecting the determination of the first-moment map (Figure 3(a)). Figure 3(c) shows the ¹³CO position–velocity (PV) diagram along the dashed rectangle in Figure 3(a). We estimated the centroid velocity in each X-axis bin with Gaussian fittings, and then the velocity gradient was derived by the least-squares fitting with a linear function within the X-range of −1.8 to 1.0 pc. The velocity gradient is ∼1.5 km s⁻¹ pc⁻¹ and its possible origin will be discussed in Section 4.1.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The previous observations in the Cycle 1 with an angular resolution of ∼1'' (∼0.24 pc) did not resolve the thin filaments; however, the Cycle 4 data clearly identified much narrower filaments that were previously unresolved. To identify the filamentary structures from the ¹³CO data quantitatively, we applied the DisPerSE algorithm (Sousbie 2011), which has been developed to extract structures such as filaments and voids in astrophysical data. This algorithm was applied to the dust continuum data obtained by the Herschel Gould Belt survey (e.g., André et al. 2010; Arzoumanian et al. 2011) and molecular line emission in Taurus (Panopoulou et al. 2014). To extract topological components from the molecular line cube, we set a "persistence" threshold, and the difference in value between the two points in the pair lower than the given threshold is cancelled (see Sousbie 2011), 20 K in brightness temperature of ¹³CO(J = 2–1), which roughly corresponds to the typical peak temperature in the high-column density (≳5 × 10²² cm⁻²) regions. We defined structures with an aspect ratio >3:1 as the filaments after the measurements of the width (see below) and length. For this reason, some of the pillars (Section 3.2) and high-column density regions around the Papillon Nebula are not identified in this criterion. In total, we identified ∼50 filaments that are overlaid on the ¹³CO velocity-integrated intensity map in Figure 4(a). To estimate filamentary widths, we performed Gaussian fits to the normalized radial intensity profile to the filament along the identified skeleton shown in Figure 4(a). An example of the fitted profile is shown in panel (b). The resultant median FWHM widths along the filaments (W_fil), shown in Figure 4(b), has a remarkable peak at ∼0.1 pc with the standard deviation of 0.08 pc. The widths of most of the filaments are significant at the present resolution (∼0.07 pc). Furthermore, the derived width is not an effect of the averaging of the radial profile across all the identified filaments, as shown in Figure 4(d). Such a small filamentary width is similar to the Galactic star-forming regions reported by Arzoumanian et al. (2011, 2019) and others, and is the first detection of such narrow filaments outside our Galaxy (see also THF18). We estimated the column density and mass of the filamentary clouds using the ¹²CO and¹³CO data by assuming the local thermodynamical equilibrium. We obtained the excitation temperature (T_ex) map from the ¹²CO data assuming that the line is optically thick. Then we got the optical depth of ¹³CO using the T_ex map, and finally, the ¹³CO column density was derived using the previous calculations. We adopted a ¹³CO/H₂ abundance ratio of 3.1 × 10⁻⁷, taking ¹²CO/H₂ as 1.6 × 10⁻⁵ and ¹²CO/¹³CO as 50 (see also Mizuno et al. 2010 and the references therein), to obtain the H₂ column density. The cumulated mass of all structures shown in Figure 1 using the mean molecular weight, μ, of (Kauffmann et al. 2008) is ∼1 × 10⁴ M_⊙, which corresponds to ∼10%–20% of that of the entire molecular cloud in the N159E region (Paper I). The averaged column density of the N159E-Papillon region shown in Figure 1 is calculated to be ∼5 × 10²² cm⁻². The central column densities at the filament crests (Σ₀) are ∼(5–30) × 10²² cm⁻² and the resultant line masses (M_line) are calculated to be ∼(1–7) × 10² M_⊙ pc⁻¹ using the equation ${M}_{\mathrm{line}}\approx {{\rm{\Sigma }}}_{0}\times {W}_{\mathrm{fil}}$ (e.g., André et al. 2014). This is consistent with filaments found in the Galactic high-mass star-forming regions (Hill et al. 2012 for Vela C; André et al. 2016 for NGC 6334). Such filaments with very large line-mass are supposed to be quite unstable against the gravitational collapse unless we assume that there is an additional support force, such as strong magnetic fields (see also discussions in André et al. 2016).

Zoom In Zoom Out Reset image size

Figure 5 shows the ¹²CO (panels (a)–(c)), ¹³CO, C¹⁸O(J = 2–1) (panel (d)), and the Hα (panel (b)) distributions around the Papillon Nebula. Our previous observations confirmed that there is a CO cavity centered on the Papillon Nebula YSO (Paper I). The present high-resolution data clearly resolved several elongated structures along the Hα bright region, the Papillon Nebula, as shown in Figure 5(b). These structures roughly extend toward the position of the Papillon Nebula YSO and are considered to be heated by the protostars and the H ii region because they show high-brightness temperature (60–80 K) in ¹²CO (Figure 5(c)). We define these structures as "CO pillars" by eye based on the abovementioned features, independently of the filament identification (Section 3.1), and the CO pillars are shown by white lines in Figure 5(c). This is the first detection of "pillars of creation" in an external galaxy (see Hester et al. 1996 for the Eagle Nebula). The properties of the CO pillars are listed in Table 1. The averaged column densities of the CO pillars are (0.9–3.0) × 10²³ cm⁻². We derived the width (FWHM) of the pillars by Gaussian fitting to the averaged column density profile perpendicular to the white lines in Figure 5. The resultant width of the pillars is ∼0.1–0.2 pc, which is similar to that of the filaments in the region. The average volume density, n(H₂), is ∼(1–7) × 10⁵ cm⁻³. These results show that the density and temperature are significantly higher than those of the surrounding filaments. The high-density nature is also supported by the detection of the C¹⁸O(J = 2–1) emission along the CO pillars (Figure 5(d)). We do not see clear extinctions of the Hα emission along the CO pillars, except for the central part of the Papillon Nebula, despite its high-column density. This indicates that the pillars are embedded in the ionized region or located behind it.

Zoom In Zoom Out Reset image size

Table 1.  Properties of CO Pillars around the Papillon Nebula

Name	${T}_{\max }^{{}^{12}\mathrm{CO}}$ (K)a	$N{({{\rm{H}}}_{2})}_{\max }$ (cm−2)b	$N{({{\rm{H}}}_{2})}_{\mathrm{ave}}$ (cm−2)c	Width (pc)d	$n{({{\rm{H}}}_{2})}_{\mathrm{ave}}$ (cm−3)e
Pillar 1	77	4.0 × 1023	2.5 × 1023	0.12	6.7 × 105
Pillar 2	76	3.7 × 1023	2.6 × 1023	0.18	4.7 × 105
Pillar 3	80	4.1 × 1023	3.0 × 1023	0.22	4.4 × 105
Pillar 4	67	1.3 × 1023	1.0 × 1023	0.13	2.7 × 105
Pillar 5	61	1.4 × 1023	9.0 × 1022	0.09	3.2 × 105

Notes.

^aMaximum brightness temperature of ¹²CO(J = 2–1). ^bMaximum H₂ column density. ^cAveraged H₂ column density within certain regions. In order to separate from the surrounding structures, the boundaries of Pillar 1, 2, 3, and 5 were defined as 50%, 60%, 70%, 50%, respectively, of the maximum column densities of each pillar. The boundary of Pillar 4 is manually defined at the edge of the white line because we cannot divide the surrounding high-column density gas located at the northern part of the pillar with the similar criterion. ^dFWHM of the pillars (see the text). ^eAveraged H₂ volume density obtained by dividing the N_ave by the width.

Download table as:  ASCIITypeset image

Figure 6(a) shows the 1.3 mm continuum and C¹⁸O(J = 2–1) distribution around the Papillon Nebula. The extended 1.3 mm continuum emission around the Papillon Nebula YSO roughly agrees with the distribution of Hα emission and has no counterparts of molecular gas, indicating that the continuum emission seems to be dominated by free–free emission from the ionized gas (see also Paper I). On the other hand, the filamentary dust clumps along the C¹⁸O emission, a tracer of cold/dense molecular gas, are considered to be dominated by thermal dust emission. Two major local maxima, hereafter MMS-1 and MMS-2, are found in the 1.3 mm continuum image in Figures 6(a)–(c). Because we detected extended emission in 1.3 mm and C¹⁸O along the north–south direction, the two sources are considered to be dense cores embedded in the filamentary cloud. The peak column densities and total masses of both sources deduced from the 1.3 mm continuum emission are ∼1 × 10²⁴ cm⁻², and ∼2 × 10² M_⊙, respectively, with an assumption of the dust emissivity, ${\kappa }_{1.3\mathrm{mm}}$ of 1 cm² g⁻¹ for protostellar envelopes (e.g., Ossenkopf & Henning 1994), a dust-to-gas ratio of 3.0 × 10⁻³ (Herrera et al. 2013; Gordon et al. 2014), and a uniform dust temperature of 20 K, which is typically adapted/estimated for galactic high-mass star-forming dust clumps (e.g., Urquhart et al. 2014; Yuan et al. 2017). Note that the core boundaries deriving their total masses were set as the 30% intensity level of each continuum peak to avoid contaminations from the parental filamentary clouds. Although these sources were not cataloged as point sources in the infrared observations (e.g., Chen et al. 2010; Carlson et al. 2012), their detection in the 1.3 mm is strongly suggestive of their protostellar nature. We have identified outflow wings that have a velocity span of ∼30 km s⁻¹ in ¹²CO(J = 2–1) toward MMS-1 and MMS-2. Figures 6(b)–(e) show the distributions and the line profiles of the outflow wings. We extracted the spectra and chose the velocity ranges using the following procedure. We defined the outer boundaries of the outflow spectra (i.e., maximum relative velocity) above the 2σ level of the velocity-smoothed spectra with a smoothing kernel of 9 ch (=1.8 km s⁻¹). We extracted the spectra at positions ∼1'' away from the millimeter sources in the east direction as the references without the outflow contaminations and defined the velocities below 2σ intensity levels as the inner edge of the outflow spectra (see Figures 6(d) and (e)). The outflow and YSO properties are listed in Table 2. The size of the outflow is as small as ∼0.1 pc, and escaped detection with our lower-resolution study (Paper I). It is likely that the launching points of the outflows coincide with the 1.3 mm continuum peaks. Because the positional offsets between the blueshifted and redshifted lobes are quite small toward MMS-1 and MMS-2, the outflow orientations may be close to pole-on or very small. The dynamical time (t_d) of the outflows is roughly estimated to be <10⁴ yr from a ratio of 0.1 pc/20 km s⁻¹. We roughly estimated the mechanical forces of the outflow lobes (F_out) using the following equation: F_out = M_outV_out/t_d (e.g., Beuther et al. 2002), where V_out and M_out are the outflow velocity and mass, respectively. The estimated F_out is ∼10⁻³–10⁻² M_⊙ km s⁻¹ yr⁻¹, depending on the assumption of the outflow inclination angle (i = 30°–70°). The F_out and the envelope mass derived from millimeter dust continuum observations are consistent with those of Galactic high-mass protostars (e.g., Beuther et al. 2002) and the N159W-South region (TFH19). A large amount of the surrounding gas with a mass of ∼10² M_⊙ and the nondetection of infrared emission suggest that MMS-1 and MMS-2 are in an extremely young phase of high-mass star formation, with an age of ≲10⁴ yr. We note that the Papillon Nebula YSO is more evolved than the two sources; however, the age is as young as ∼0.1 Myr (Paper I) judging from the combination of the spectral energy distribution fitting using the Spitzer/Herschel data and the compactness of the H i region (see also Chen et al. 2010). Although certain time differences are found, the three high-mass protostellar systems with separations of ∼1–2 pc are formed within the order of ∼0.1 Myr. We discuss the possible processes of high-mass star formation taking place in the N159E-Papillon region in Section 4.1.

Zoom In Zoom Out Reset image size

We discuss the formation of the protostars and the filaments in the N159E-Papillon region. In Paper I, we discussed the possibility that collisions by three molecular filaments induced the star formation activity in the Papillon Nebula. However, such a filamentary cloud collision model may not be appropriate to explain the star formation activities in this region revealed by the present observations for some reasons. (1) The present observations clearly show that the parental molecular cloud is composed of many complex filamentary structures (Figures 1(a) and 4) rather than the previously identified several filaments (Figure 1(b)). Therefore, it is more likely that the molecular cloud was formed in an entangled filamentary configuration rather than by a coalescence process of many individual filaments. (2) The previous lower-resolution observations could not find the protostellar activities in MMS-1/2 (Section 3.3), thus it was thought that the high-mass star formation seen at the Papillon Nebula is a local event around this region. Despite this, the projected separations between the Papillon Nebula YSO and MMS-1/2 are as large as ∼1–2 pc, and these objects are independently formed within a few 0.1 Myr. In addition to this, our separate observations toward the N159W-South region (TFH19), which is located ∼50 pc apart from the N159E-Papillon cloud in projection, found that both regions show quite similar star formation activities and molecular gas distributions (see also Figure 7). In N159W-South, we detected multiple protostellar sources with a separation of ∼1–2 pc, along with the massive molecular filaments (TFH19). The physical properties of the outflows and molecular filaments are quite similar to those in the N159E-Papillon region and the global orientation of both filaments is roughly parallel to each other. These facts indicate that the star formation activity and the filament formation taking place in the N159 region are not only local phenomena with a scale of 1–2 pc, but also a more global event extending up to >50 pc. We thus consider the large-scale gas kinematics as an origin of the formation of massive protostars/filaments in this region.

Zoom In Zoom Out Reset image size

We calculated the filament width to be ∼0.1 pc, which is reported in a number of Galactic studies (e.g., André et al. 2014 and references therein) and its high line mass, ∼a few ×100 M_⊙ pc⁻¹. Such high line-mass filaments are unstable against the gravitational collapse unless some additional strong support forces emerge to regulate and slow down the collapse. There is a possibility that the massive filaments were formed by a large-scale converging flow quite recently with a timescale of ∼a few 0.1 Myr (see also Section 1). If we ignore the magnetic field, the virial mass per unit length, ${M}_{\mathrm{line},\mathrm{vir}}$ = $2{\sigma }_{{\rm{v}}}^{2}$/G (see Fiege & Pudritz 2000), is calculated to be ∼1.4 × 10² M_⊙ pc⁻¹, as the typical velocity dispersion, σ_v, is ∼0.6 km s⁻¹ measured from the present ¹³CO data. The ${M}_{\mathrm{line},\mathrm{vir}}$ is smaller than the observed line mass (Section 3.1) and it is likely that the filaments are gravitationally unstable. Nayak et al. (2018) also reported low virial parameters in molecular clumps in the N159 region, thus we suggest that the filaments may not be long-lived objects. Even if the strong magnetic field of ∼300 μG is penetrated perpendicular to a ∼0.1 pc width filament, a line mass of ∼70 M_⊙ pc⁻¹ is a limit that can be magnetically supported according to Inoue et al. (2018).

We summarize the early atomic/molecular gas studies around the N159 region. Figure 7 shows the large-scale gas distributions seen in H i (Kim et al. 2003) and CO (Fukui et al. 2008; Minamidani et al. 2008, see also Wong et al. 2011). Fukui et al. (2017) subtracted the galactic rotation from the H i data and found three major velocity features, L-, I-, and D-components (see also Luks & Rohlfs 1992; Tsuge et al. 2019). The largest giant molecular complex in the LMC, called "the molecular ridge," which is shown in green contours in Figures 7(a) and (b), is mainly located at the intermediate velocity range between the L- and D-components (Fukui et al. 2017). These authors proposed that the tidal interaction between the LMC and the Small Magellanic Cloud (SMC) drove a kpc H i flow, which originated in the past close encounter between the two galaxies, which was modeled in the numerical simulations by Bekki & Chiba (2007) following the original suggestion by Fujimoto & Noguchi (1990). After settling outside the LMC disk for 200 Myr, the gas could be now inducing the formation of the super star cluster R136 in 30 Dor as well as the molecular ridge containing the N159 region (Fukui et al. 2017). Because the GMC in the N159 region is the most massive molecular reservoir of the region (e.g., Fukui et al. 2008; Minamidani et al. 2008; Chen et al. 2010), this GMC may be the end product of a cloud strongly compressed by the H i flow and enriched to the large mass we observe currently (Fukui et al. 2017). Recent MHD simulations by Inoue et al. (2018) investigated detailed processes of filament/protostar formation driven by cloud–cloud collisions. These authors demonstrated that massive hub filaments, with their complex morphology that qualitatively resembles the filaments in the N159E-Papillon region, are formed in the compressed layer between the cloud and the flow, within a few ×0.1 Myr after the collisional event. The colliding flow or GMC is likely moving from the northwest to the southeast, as suggested by Fukui et al. (2017) (see also the white arrow in Figure 7(b)), and the global orientation of the molecular filaments in the N159E/W region roughly follows the flow (Figures 7(d) and (e)). These features are reproduced by numerical simulations of colliding clouds, as demonstrated by Inoue et al. (2018). The present case in the N159E-Papillon region shows that the typical lengths of the filaments are ∼6 pc, with a vertical angle of ∼90° and they are in a north–south projected direction, as schematically shown in Figure 8 (see also Figure 4(a)). This distribution can be explained as follows: the initial small cloud collided with the extended cloud as simulated by Inoue et al. (2018). As shown in Figures 7(a) and (b), the distribution of H i gas is inhomogeneous, thus the colliding flow is considered to have local high-density parts. Such a colliding flow can mimic a collision between a small cloud and a large one (Figure 8).

Zoom In Zoom Out Reset image size

The observed filamentary cloud has a velocity gradient of ∼1.5 km s⁻¹ pc⁻¹ in the north–south direction (Figures 2 and 3(a) and (b)). Given the filament formation scenario based on the H i colliding flow (Figures 7 and 8; see also Fukui et al. 2017), we have a clear picture of the gas flow in the N159 region. Assuming that the clump interacting with the D-component is falling from the northwest (Figure 8), it is likely that the clump comprises the hub of the N159E filamentary system. We expect that the filaments are decelerated via the frictional interaction with the ambient H i gas more than the dense hub. This deceleration will cause a velocity gradient, as the hub is moving toward us, and is more blueshifted than the less dense filaments, which explains the velocity gradient. The N159W-South also has a similar interaction with the disk H i, and shows a similar velocity gradient (TFH19), although the observed gradient is smaller than that of the N159E filamentary system, possibly due to the projection effect. It is thus a natural outcome that the two filamentary systems have a common velocity gradient in spite of a separation of ∼50 pc.

Assuming that the ionized shock front proceeds at 5 km s⁻¹ (e.g., Fukui et al. 2016 for RCW38), the dynamical timescale of the interaction is ∼1.2 Myr, as the size (length) of the molecular clouds is ∼6 pc. The interaction will result in the formation of an O-type star and the resulting ionization of a cloud in the following 0.1 Myr. This means that the lower limit of the CO filament lifetime after their formation is of the order of ∼1 Myr, which is consistent with our argument that the CO filaments were formed very recently, as discussed in the previous paragraphs. The collision took place between two velocities of H i flow (Fukui et al. 2017), whereas the currently available H i data with an angular resolution of ∼1' (Kim et al. 2003) are not high enough to distinguish the relation between the molecular filaments and the large-scale H i gas. The initial H i gas still needs to be observed at a resolution comparable to that of our ALMA observations to better understand its connection with the molecular reservoirs.

 It is also important to understand how feedback is affecting the environment of high-mass stars in disentangling the gas distribution that is usually a mixture of the initial conditions prior to high-mass star formation and the results of feedback effects. In that context, we consider the formation mechanism and the evolutionary state of the CO pillars around the Papillon Nebula. In this section, we compare our study with the early results for the Eagle Nebula in M16, the most famous "pillars of creation" in the Galaxy, and compare them with the present results in N159E-Papillon. Pound (1998) observed the Eagle Nebula in ¹²CO, ¹³CO, and C¹⁸O(J = 1–0) using the Berkeley–Illinois–Maryland Array with a resolution of ∼0.1 pc. The mean column density and volume density deduced from the CO observations are 2 × 10²² cm⁻² and ∼3 × 10³ cm⁻³, respectively. The total ionizing flux is S ∼ 1.2 × 10⁵⁰ s⁻¹, mostly from two O5 stars in M16 and the separation between the ionizing stars and the pillars is a few parsecs (Hillenbrand et al. 1993). The age of M16 is estimated to be 1.3 ± 0.3 Myr (Bonatto et al. 2006). The Papillon Nebula, with an age of ∼0.1 Myr (Paper I), is in a much younger stage than the Eagle Nebula and the separation of the pillars is ∼0.5 pc from the YSO. Although the total ionization fluxes of both regions are almost the same or higher in the Papillon Nebula (S ≫ 1.1 × 10⁵⁰ s⁻¹, Paper I), the column density and volume density of the CO pillars in the Papillon Nebula, (0.9–3.0) × 10²³ cm⁻² and ∼(3–7) × 10⁵ cm⁻³, respectively, are an order of magnitude higher than those of the Eagle Nebula. This indicates that dissociation of the molecular gas toward the CO pillars has not reached the innermost part. On the other hand, because the boundaries of the ¹²CO, ¹³CO, and C¹⁸O distributions facing the Hα emission are very close to each other (Figure 5), the surrounding low-density envelopes of the CO pillars have already been dissociated.

We investigate the velocity structure of the CO pillars to explore the formation mechanism. If the Rayleigh–Taylor instability (Spitzer 1954) is working to develop pillar-like objects, the velocity gradient should increase with the square root of the distance from the top of the pillar (Frieman 1954). Pound (1998) pointed out the Rayleigh–Taylor instability is unlikely, as the formation process of the pillars in the Eagle Nebula is based on the CO velocity fields. Our observations also show similar velocity gradients to what is shown in Figure 5 of Pound (1998) or no significant velocity gradient, as shown in Figure 3. We thus exclude the Rayleigh–Taylor instability as a candidate mechanism to form the CO pillars in the Papillon Nebula. Alternatively, we suggest that the radially distributed hub filaments were the initial configuration before the high-mass star formation (see also Figure 8). The present gas distribution of the Papillon Nebula (Figure 5) can be explained as follows: the high-mass star was formed at the intersection of the hub and the central part and the relatively low-density gases around the filaments were dissociated. Hub filaments without any ionization regions are also seen in the N159W-South clump, where ionization has not yet started (Paper I; TFH19) and it may be in an earlier stage than the Papillon Nebula. N159W-South may possibly evolve later into a state similar to N159E-Papillon when the protostar becomes a mature O star.

We estimated the missing flux of our Cycle 4 image in the N159E-Papillon region using the Cycle 1 data (Paper I). We converted the Cycle 4 data shown in Figure 1(a) to be the same resolution as the Cycle 1 data (Figure 9(a)), and produced the ratio map and histogram between the two data (Figures 9(b) and (c)). The distribution of the ratio map is roughly unity and the histogram has a peak around 0.8, indicating that the Cycle 4 image has no significant missing flux compared to the Cycle 1 data.

Zoom In Zoom Out Reset image size