Molecular clouds toward three Spitzer bubbles S116, S117, and S118: Evidence for a cloud–cloud collision which formed the three H ii regions and a 10 pc scale molecular cavity

Abstract

We carried out a molecular-line study toward the three Spitzer bubbles S116, S117, and S118, which show active formation of high-mass stars. We found molecular gas consisting of two components with a velocity difference of ∼5 km s⁻¹. One of them, the small cloud, has a typical velocity of −63 km s⁻¹ and the other, the large cloud, has one of −58 km s⁻¹. The large cloud has a nearly circular intensity depression, the size of which is similar to that of the small cloud. We present an interpretation that its cavity was created by a collision between the two clouds and that this collision compressed the gas into a dense layer elongating along the western rim of the small cloud. In this scenario, the O stars including those in the three Spitzer bubbles were formed in the interface layer compressed by the collision. Assuming that the relative motion of the clouds has a tilt of 45° to the line of sight, we estimate that the collision continued for the last 1 Myr at a relative velocity of ∼10 km s⁻¹. In the S116-S117-S118 system the H ii regions are located outside of the cavity. This morphology is ascribed to the density-bound distribution of the large cloud which caused the H ii regions to expand more easily toward the outer part of the large cloud than towards the inside of the cavity. The present case proves that a cloud–cloud collision creates a cavity without the action of O-star feedback, and suggests that the collision-compressed layer is highly filamentary.

1 Introduction

O stars have a mass range of from 20 to 300 |$M_{\odot }$| in the Local Group (Walborn et al. 2002). They influence interstellar space considerably by dynamically disturbing the interstellar medium (ISM) over their whole lifetime and by enriching heavy elements at the end of their lives with supernova explosions. These actions have a significant influence on galactic evolution. It is therefore of crucial importance to understand the formation mechanism of O stars better. The mass accretion rate required for O star formation is as high as 10⁻⁴ |$M_{\odot }\:$|yr⁻¹ ; otherwise the high stellar radiation pressure halts the mass accretion and higher-mass stars do not grow in mass to become mature O stars (e.g., Wolfire & Cassinelli 1987; Zinnecker & Yorke 2007).

It is often discussed that dense and massive clouds, like infrared dark clouds, are plausible sites of high-mass star formation. This seems to be a natural scenario, whereas we often do not have any H ii regions associated with infrared dark clouds: regions where O stars are formed unambiguously are H ii regions which are ionized by ultraviolet photons emitted by O or early B stars. H ii regions in the Galaxy show a variety of morphology and evolutionary stages, from ultra-compact H ii regions to extended H ii regions (e.g., for review Churchwell 2002), and more careful studies of H ii regions have a potential to shed a new light on O star formation. In particular, a detailed study of molecular gas, the raw material which forms stars, is still not thoroughly made in H ii regions, due to a limited angular resolution or a low dynamic range in molecular line observations, in spite of the previous efforts in molecular observations toward H ii regions (e.g., Andreson et al. 2009).

Spitzer bubbles are seen at 8 μm with a ring-like shape and usually harbor an O star and an H ii region in the ring. In total 322 bubbles are cataloged in Galactic longitude from −65° to 65° and in Galactic latitude from |${-1{^{\circ}_{.}}5}$| to |${1{^{\circ}_{.}}5}$|⁠, except the Galactic center (l < ±10°) (Churchwell et al. 2006). It is usually considered that the ultraviolet photons and stellar winds accelerate and compress the surrounding gas to form an expanding bubble and trigger second-generation star formation (the collect-collapse scenario) (Deharveng et al. 2005). A CO survey of Spitzer bubbles was made by Beaumont and Williams (2010), who observed 43 Spitzer bubbles with James Clerk Maxwell Telescope (JCMT) in the CO J = 3–2 transition. Their results indicate that the gas is ring-like and flattened with no significant sign of expansion, as opposed to the wind-blown scenario, raising a puzzle which is not immediately explained by the conventional picture.

Torii et al. (2015) recently presented the novel scenario that the bubble in a most typical Spitzer bubble RCW 120 (Zavagno et al. 2007; Deharveng et al. 2009) was formed by cloud–cloud collision, via the Habe and Ohta (1992) model, which numerically simulates a head-on collision between small and large spherical clouds. These simulations showed that such uneven collisions can create a cavity in the large cloud that is the size of the small cloud and the compressed layer between them becomes dense and self-gravitating, triggering star formation. If the formed star is capable of ionizing the surroundings, i.e., the inside of the cavity, the model provides an alternative explanation for the ring-like shape of RCW 120, the inside of which is ionized (Torii et al. 2015). Torii et al. (2015) did not find an observational signature of expansion of the bubble in the molecular gas, which led to the cloud–cloud collision scenario. RCW 120 has two molecular components having velocities of −8 km s⁻¹ and −28 km s⁻¹: one of which shows a good correspondence with the bubble, and the other associated with the opening of the bubble. These properties are consistent with the results found by Beaumont and Williams (2010). The two clouds show an enhanced temperature toward the ring in spite of their large velocity separation, ∼20 km s⁻¹, which provides robust verification of their physical association with RCW 120. Since the velocity separation is too large to be bound by the cloud gravity, Torii et al. (2015) concluded that RCW 120 is a case of cloud–cloud collision which triggered the formation of a single O7 star. It is noteworthy that the new model offers an explanation of the elongated horseshoe morphology of the bubble along with the off-center position of the exciting star close to the bubble bottom: the simple stellar-wind bubble does not offer a natural explanation for such asymmetry.

Figure 1 shows the region of three Spitzer bubbles S116, S117, and S118. The distance of S116 is estimated kinematically to be 5.9 ± 0.9 kpc (Churchwell et al. 2006). The green and red colors in figure 1 show 8 μm and 24 μm images, respectively. The 8 μm radiation traces the polycyclic aromatic hydrocarbon (PAH) emission and the 24 μm radiation the heated dust by the ultraviolet radiation of the O stars. The bubbles are not perfectly circular and are significantly elongated. Spitzer bubbles are often isolated, and such a case with three lined-up bubbles are rare.

In order to have a better understanding of formation of multiple Spitzer bubbles, we undertook new CO observations of molecular gas toward S116, S117, and S118 with three mm/sub-mm telescopes, NANTEN 2, Atacama Submillimeter Telescope Experiment (ASTE), and the Mopra 22 m telescope, in the CO rotational transitions. Section 2 gives descriptions of the observations and section 3 presents the observational results. Section 4 gives a discussion of the results and tests a cloud–cloud collision scenario, and section 5 concludes the paper.

2 Observations

2.1 ¹²CO J = 1–0 with NANTEN2

We performed ¹²CO J = 1–0 observations with the NANTEN2 4 m millimeter/sub-millimeter radio telescope (Mizuno & Fukui 2004). The observations were carried out in 2011 October. The front end was a 4 K cooled Superconductor-Insulator-Superconductor (SIS) mixer receiver. The typical system noise temperature including the atmosphere was ∼250 K in the double-side band (DSB) during observations. The back-end was a digital-Fourier transform spectrometer (DFS) with 16384 channels of 1 GHz bandwidth, which corresponds to the velocity coverage of ∼2600 km s⁻¹ and to the resolution of 0.16 km s⁻¹. We used the on-the-fly (OTF) mapping mode to cover an area of 1° × 1°. The pointing accuracy was confirmed to be better than |${15^{\prime \prime }}$| with daily observations toward the Sun. The absolute intensity was calibrated by observing Orion KL Nebular every day. The final half-power beam width (HPBW) is 180″. The final root-mean-square (rms) noise fluctuation of the data was ∼1.0 K at a velocity resolution of 0.16 km s⁻¹.

2.2 ¹²CO J = 1–0 and ¹³CO J = 1–0 with Mopra

Detailed ¹²CO J = 1–0 and ¹³CO J = 1–0 distributions around the Spitzer bubbles S116, S117, and S118 were obtained using the 22 m ATNF (Australia Telescope National Facility) Mopra mm telescope at a high angular resolution of |${33^{\prime \prime }}$| in two periods: 2013 June–October and 2014 July–August. For the field shown in figure 1 by a dashed box we simultaneously observed ¹²CO J = 1–0 and ¹³CO J = 1–0 in the OTF mode with a unit field of |${4^{\prime }}\times {4^{\prime }}$|⁠. The typical system noise temperature including the atmosphere was between 400 K and 600 K in the single-side band (SSB) during observations. The Mopra back-end system “MOPS,” which provides 4096 channels across 137.5 MHz in each of the two orthogonal polarizations, was used in the observations. The velocity resolution was 0.088 km s⁻¹ and the velocity coverage was 360 km s⁻¹ at 115 GHz. The pointing accuracy was checked every 90 minutes to keep within |${4^{\prime \prime }}$| with observations of 86 GHz SiO masers. The absolute intensity was calibrated in comparison with CO J = 1–0 data observed with NANTEN2. The obtained data were smoothed to a HPBW of |${40^{\prime \prime }}$| with a 2D Gaussian function and to a 0.6 km s⁻¹ velocity resolution. The final data cube had an rms noise fluctuation of ∼0.6 K at a velocity resolution of 0.088 km s⁻¹.

2.3 ¹²CO J = 3–2 with ASTE

Observations of the ¹²CO J = 3–2 emission were performed using the ASTE 10 m telescope located in Chile in three periods: 2013 September, 2014 June, and 2015 November. The waveguide-type sideband-separating SIS mixer receiver for the SSB “CAT345,” with a system noise temperature of 300 K, and the digital spectrometer “MAC” in the narrow-band mode, which provides 128 MHz bandwidth and 0.125 MHz resolution corresponding to 450 km s⁻¹ velocity coverage and 0.43 km s⁻¹ velocity resolution at 345 GHz, were used. The observations were made in the OTF mode at a grid spacing of |${7{^{\prime\prime}_{.}}5}$|⁠, and the HPBW was |${24^{\prime \prime }}$| at the ¹²CO J = 3–2 frequency. The observed area is the same as that of the Mopra observations. The pointing accuracy was checked every 90 minutes to keep within |${2^{\prime \prime }}$| with observations of RAFGL 4202 (α_J2000.0, δ_J2000.0) = |$({14^{\rm h}52^{\rm m}23{^{\rm s}_{.}}82},\ {-62^{\circ}04^{\prime }19{^{\prime\prime}_{.}}2})$|⁠. The absolute intensity was calibrated by observing W 44 and IRC +10216 every 90 minutes. The obtained data was smoothed to a HPBW of |${22^{\prime \prime }}$| with a 2D Gaussian function and to a 1 km s⁻¹ velocity resolution. The final data cube had an rms noise fluctuation of ∼0.38 K at a velocity resolution of 0.43 km s⁻¹.

3 Results

3.1 CO distributions

Figure 1 shows an infrared image of the present region taken with Spitzer (Benjamin et al. 2003; Carey et al. 2009). The three Spitzer bubbles S116, S117, and S118 are distributed over 20 pc in the north–south direction. We find nearly 10 smaller H ii regions, which are bright warm dust traced by the Spitzer 24 μm emission. The area indicated by the dashed-line box was observed with Mopra and ASTE, while the whole area was mapped with NANTEN2. Figure 2 shows a large-scale view of the ¹²CO J = 1–0 emission observed with NANTEN2. We find two clouds with distributions that are distinctly different. The −63 km s⁻¹ cloud has a peak at (l, b) = |$({314{^{\circ}_{.}}22},\ {0{^{\circ}_{.}}33})$|⁠. The −58 km s⁻¹ cloud extends along the plane, and has a peak at (l, b) = |$({314{^{\circ}_{.}}21},\ {0{^{\circ}_{.}}25})$| and an intensity depression at (l, b) = |$({314{^{\circ}_{.}}3},\ {0{^{\circ}_{.}}36})$| in addition to several surrounding intensity peaks. We hereafter call the −63 km s⁻¹ cloud and the −58 km s⁻¹ cloud the small cloud and the large cloud, respectively, because of their sizes. The small cloud has a sharp intensity decrease in every direction. The large cloud shows a sharp intensity gradient toward the west.

Figure 3 shows velocity channel distributions every 1.3 km s⁻¹ in the ¹²CO J = 1–0 emission taken with the Mopra telescope, and indicates that the small cloud becomes large in size with increase in velocity from −69.3 km s⁻¹ to −62.7 km s⁻¹. We also find that the CO distribution is extremely filamentary in the velocity range of from −66.6 km s⁻¹ to −61.3 km s⁻¹, particularly toward the small cloud. The filamentary structure is described in more detail in subsection 3.3.

Figure 4 shows the detailed distribution of ¹²CO J =1–0, ¹³CO J = 1–0, and ¹²CO J = 3–2 images of the small cloud (figure 4a, 4b, and 4c) and the large cloud (figures 4d, 4e, and 4f) overlaid with infrared contours. The bubbles S116 and S118 delineate the northern and southern boundaries of the small cloud, and S117 is located toward the small cloud. In the large cloud, the cavity is clearly seen with a sharp, nearly circular boundary. We find another intensity depression in the north at (l, b) = |$({314{^{\circ}_{.}}27}{-}{314{^{\circ}_{.}}32},\ {0{^{\circ}_{.}}41}{-}{0{^{\circ}_{.}}43})$|⁠. The western edge of the large cloud shows a distribution similar to the small cloud at |$b = {0{^{\circ}_{.}}3}$|–|${0{^{\circ}_{.}}45}$|⁠, and S116 and S118 are also located along the edge of the large cloud. In figures 4b and 4e the distribution of the ¹³CO J = 1–0 emission, which is likely optically thin, shows a correspondence to the intense part of the ¹²CO emission. We find that the effect of self-absorption is small in this region from the similarity between the ¹²CO and ¹³CO distributions. The physical parameters of the two clouds are as follows. The size of the cavity hole is ∼5 pc in radius for an assumed distance of 5.9 kpc. The masses of the small and large clouds in the area shown in figure 2 and their peak column densities are (2.0 × 10⁴ |$M_{\odot }$|⁠, 0.7 × 10⁵ |$M_{\odot }$|⁠) and (1 × 10²² cm⁻², 2 × 10²² cm⁻²), respectively, where an X(CO) factor of 1.0 × 10²⁰ cm⁻² K⁻¹ km⁻¹ s (Okamoto et al. 2017) was assumed.

3.2 Radio continuum distribution and the properties of the O stars

In figure 5 the ASTE ¹²CO J =3–2 images of the two components are superposed on the SUMMS 843 MHz radio continuum distribution (Bock et al. 1999; Mauch et al. 2003) (figures 5a and 5b) and the Spitzer image (figures 5c and 5d). We find that the radio continuum distribution coincides well with the Spitzer bubbles. S117 is located toward the center of the small cloud, and it fits the intensity depression in the large cloud (figure 5c and 5d). The heavy obscuration does not allow us to observe the exciting stars directly, and instead we used the radio continuum radiation as a measure of the stellar ultraviolet radiation. We estimate the spectral types of the exciting O stars in each H ii region from the radio flux by using the relationship given by Panagia (1973) and Kurtz, Churchwell, and Wood (1994), as shown in table 1, where the exciting star is assumed to be a single star in each H ii region. The ultraviolet photon flux was used to estimate stellar spectral types and corresponding stellar masses for ZAMS (zero-age main sequence) stars. As a result, the spectral types are estimated as follows; O6–06.5 for S116, O9 for S117, and O7.5–O8 for S118, and the masses are inferred to be 30 |$M_{\odot }$|⁠, 20 |$M_{\odot }$|⁠, and 23 |$M_{\odot }$|⁠, respectively, from stellar parameters for the luminosity class V stars (Martins et al. 2005).

3.3 The intensity ratio of the ¹²CO J = 3–2 to the ¹²CO J = 1–0

Figures 6a and 6d show the intensity ratio of ¹²CO J = 3–2 to ¹²CO J = 1–0 (R_3−2/1−0) for the two clouds. R_3−2/1−0 is mainly affected by the ¹²CO J = 3–2 distribution. The typical R_3−2/1−0 of molecular clouds in our galaxy without an extra heat source is R_3−2/1−0 ∼ 0.4 (Fukui et al. 2016). R_3−2/1−0 is enhanced to 1.0–1.4 outside of the cavity toward S116 and S118, while R_3−2/1−0 is 0.6–1.0 around the cavity. The enhanced ratio toward S116 and S118 suggests that the stellar radiation is heatig the gas. Red and blue lines in figures 6b and 6e represent the filamentary structures of R_3−2/1−0, where red and blue colors show that the filament is overlapping with the H ii regions (red), or not (blue). These filamentary structures were identified by eye on a criteria of R_3−2/1−0 > 0.9 and length ≳2 pc. The filamentary structures have a typical size of ∼1 × 3 pc. Multiple filamentary structures are seen not only along the H ii region but also in the cavity without H ii regions.

4 Discussion

We have carried out large-scale CO observations with NANTEN2 in the region of the three Spitzer bubbles S116, S117, and S118, which include at least several O stars in addition to a few tens of low-mass stars as Akari point sources (figure 1). We made follow-up high-resolution observations with the ASTE and Mopra telescopes. We found that the molecular gas comprises at least two velocity components of different morphology: the −63 km s⁻¹ cloud (the small cloud) and the −58 km s⁻¹ cloud (the large cloud), which extend along the plane over 40 pc if the weak-extended CO emission is included (figure 2). The large cloud has an intensity peak and an intensity depression of ∼10 pc in size. Higher-resolution observations with Mopra and ASTE show that the small cloud grows in size with increase in velocity (figure 3).

4.1 A cloud–cloud collision scenario in S116-S117-S118

In order to explain the velocity distribution we propose the hypothesis that a cloud–cloud collision took place between the two clouds and that the small cloud pushed the surface of the large cloud to produce the cavity in the large cloud. Figure 7a presents a schematic view of the collision seen from the direction perpendicular to the cloud’s relative motion. In the plane of figure 7a, θ is the angle of the line of sight to the relative motion of the cloud. The small cloud was seen to be separated in the upper left-hand corner of figure 7a prior to collision, and then it moved along a straight line towards the large cloud. The two clouds collided with each other and the small cloud created a cavity in the large cloud. The layer between the two clouds is compressed to form an interface layer, which is denoted by dark blue in front of the small cloud in figure 7a. The two clouds observed on the sky are divided into three sections, A, B, and C, as shown in figure 7a. A shows the large cloud and the cavity; B shows the interface layer, the small cloud, and the large cloud with the cavity; C shows part of the large cloud prior to collision.

In order to gain an insight into observed cloud properties, we describe the physical states of colliding clouds using the hydrodynamical numerical simulations of Takahira, Tasker, and Habe (2014). Their simulations deal with a head-on collision between a small cloud and a large cloud, which are idealized to be spherically symmetric. We adopt the model listed in table 2 for this discussion: the radius of the small cloud is 3.5 pc and that of the large cloud 7.2 pc. The two are currently colliding at 7 km s⁻¹ and have internal turbulence on the order of 1–2 km s⁻¹ with highly inhomogeneous density distribution. For more details see Takahira, Tasker, and Habe (2014). The cloud parameters do not correspond exactly to the present cloud, although the difference does not critically affect the qualitative comparison below.

We assume θ = 45° in the following and an epoch of 1.6 Myr after the onset of the collision, where cloud signatures that are typical of collisions are seen. The assumption on θ is not so critical as long as θ is not very close to 0° or 90°. The small cloud is producing a cavity in the large cloud by the collisional interaction. The density of the interface layer of the two clouds is enhanced by the collision, where the internal turbulence and the momentum exchange between the clouds mix the gas distribution in space and velocity. The gas in the two clouds is continuously merging into the layer during the collision. Figures 7b–7i show the velocity channel distributions every 1.0 km s⁻¹. In cloud–cloud collisions it might be expected to see two distinct clouds of a narrow linewidth, whereas the actual distribution in the simulations present a merged single cloud which is continuous in velocity and space. The small cloud is seen at a velocity range of from −5.3 to −2.3 km s⁻¹, and the large cloud at that of from −1.3 to 1.7 km s⁻¹. The velocity range of from −2.3 to −1.3 km s⁻¹ corresponds to the interface layer which was created by merging. Note that the velocity ranges of each panel in figure 7 do not fit the velocity ranges of the two clouds exactly, due to the mixing in velocity. We find that the small cloud becomes large with increase in velocity via merging with the large cloud, as it is consistent with the observations. The cavity is produced in the large cloud by the small cloud, while the boundary of the cavity is less clear in the model than in observations, reflecting that turbulence is more enhanced in the model than in the observed cloud.

In S116-117-118, there is displacement by ∼10 pc in the sky between the small cloud and the cavity (figure 3). We interpret that the displacement is caused by projection of a tilt of the cloud relative motion to the line of sight. The collision velocity is estimated to be ∼7 km s⁻¹ if we tentatively assume the angle between the relative motion and the line of sight to be θ = 45°.

Figure 8 shows a comparison of the position–velocity diagram taken in the direction of the relative motion of the two clouds. Figure 8a is an overlay of the small and large clouds taken from figures 7e and 7h. Figure 8a shows distributions of the two clouds. Figure 8b shows a synthetic position–velocity diagram produced from the same numerical simulations in the plane of the two cloud centers. The clouds as a whole show a “V” shape as traced by the solid lines, and the main peak is found at X = 5–7 pc and V = −4–−1 km s⁻¹ in figure 8b. This peak is formed by a combination of the small cloud and the interface layer which merged together. There is an intensity depression at X = 4–6 pc and V = −2–0 km s⁻¹, which correspond to the blueshifted part of the cavity in the large cloud. Figure 8c is the observed two velocity components and figure 8d the observed position–velocity diagram. The “V” shape which is typical of collisions is visible in figure 8d. Correspondence is seen between the model and the observation qualitatively: the observations reproduce the cavity at |$l={314{^{\circ}_{.}}3}$|v_LSR = −58 km s⁻¹ and the small cloud at |$l={314{^{\circ}_{.}}2}$|v_LSR = −63 km s⁻¹, except for the sharp cut of the large cloud at |$l={314{^{\circ}_{.}}2}$| which is not taken into account in the model. This correspondence lends support to applying the cloud–cloud collision model to the S116-S117-S118 system.

The interface layer is strongly compressed by collision, and the O stars ionizing the three Spitzer bubbles S116-117-118 were formed in the layer due to gravitational instability. This explains the distribution of the bubbles along the western surface of the small cloud which is interacting with the large cloud. The timescale of the collision is roughly estimated to be ∼1.3 Myr (= 10 pc/7 km s⁻¹). An O star of 30 |$M_{\odot }$| is formed in 10⁵ yr within the timescale for a mass accretion rate 3 × 10⁻⁴ |$M_{\odot }\:$|yr⁻¹ , which is adopted from a typical value in the compressed layer of cloud–cloud collision in magnetohydrodynamical numerical simulations (Inoue & Fukui 2013). This satisfies the mass accretion rate required to form O stars by overcoming the stellar radiation feedback (Wolfire & Cassinelli 1987).

The collisional compression possibly extends to 40 pc beyond the size of the cavity in the north–south direction vertical to the projected motion of the two clouds. The layer is observed as a north–south elongated molecular ridge at l = 314.2 from 0|${{^{\circ}_{.}}}$|1 to 0|${{^{\circ}_{.}}}$|6 in b, and we find a possible sign of further triggered formation of lower-mass stars along the compressed layer (figure 1). Thus, it is possible that triggering is extensive in space, while O-star formation is probably limited to the region of high molecular column density, (1–2) × 10²² cm⁻².

4.2 Comparison with RCW 120

In S116-117-118, the H ii regions are located outside the collision-created molecular cavity, although in RCW 120 the H ii region is located inside the cavity. In both cases, the collision formed the interface layer in which O star(s) form, and the morphological difference between the two cases are to be explained. Figure 9 shows the schematic drawings of the two colliding clouds in a position--velocity diagram of the two cases. For RCW 120, Torii et al. (2015) presented a cloud–cloud collision model in order to explain the formation of the O star within the bubble. The diameter of the RCW 120 bubble is ∼3 pc, whereas that of the S116-S117-S118 cavity is ∼10 pc. Therefore, the volume of the S116-S117-S118 cavity is larger than that of the RCW 120 cavity by a factor of ∼30 if a spherical cavity is assumed, and the molecular mass inside the cavity is significantly larger in S116-S117-S118 than in RCW 120. This offers a possible explanation for the larger number of O stars in the present case. Figure 10 shows the schematic images of cloud–cloud collision in RCW 120 and S116-S117-S118. According to the Habe and Ohta (1992) model of cloud–cloud collision, we expect that two clouds, one of which is delineating a Spitzer bubble and the other localized inside the bubble, exhibit a complementary distribution in the early phase after the collision (phase 1 in figure 10). Later, the small cloud is destroyed by ionization due to the formed star and by collisional merging into the interface layer in RCW 120. The difference between the two cases is the location of the O stars. In RCW 120 the O star is inside the cavity (figure 10a, phase 3), whereas in S116-S117-S118 the O stars are located outside of the cavity (figure 10b, phase 3). In S116-117-118, we suggest that the collision happened by chance close to the edge of the large cloud. This situation caused the shocked layer to be exposed to the outside of the large cloud where its density drops suddenly. We infer that an ad hoc geometry in S116-117-118 caused the O-star formation outside the collision-created molecular cavity.

4.3 The physical properties of the collisional interface layer

The S116-S117-S118 cavity demonstrates clearly the role of cloud–cloud collision in creating a cavity, which shows a well-ordered nearly circular boundary. Because no O star exists inside the cavity, there is no room for the cavity to be produced by an O star. In RCW 120 the inside wall of the cavity is partially ionized and the physical conditions are affected by the ionization, implying that the collisional interface does not keep the physical conditions just after the collision. The present case is different because it provides physical conditions in the shock-compressed layer unaffected by the ionization. Although S117 seems to lie toward the peak of the small cloud, it does not affect the small cloud and the inside of the cavity, as shown by the lack of enhanced R_3−2/1−0 ratio toward S117 in figure 6d. In this context, we pay attention to the interface layer toward the small cloud and the cavity exhibiting highly filamentary distributions, which are obvious in the R_3−2/1−0 distribution in figure 6a. Inoue et al. (2018) showed that molecular filaments are formed in the shocked interface in a cloud–cloud collision, and similar filament formation is seen in the simulations by Takahira, Tasker, and Habe (2014) and Balfour, Whitworth, and Hubber (2017). The width and length of the filaments in these simulations deserve a further detailed comparison with observations in order that the physical processes in collision may be better understood.

Figures 6c and 6f show the distribution of ¹³CO J =1–0 emission of the small cloud and the large cloud overlaid on the filamentary distribution of R_3−2/1−0. The high ¹³CO J = 1–0 intensity part and the R_3−2/1−0 filamentary structures in the collision interface do not correspond with each other. Rather, the high R_3−2/1−0 regions correspond to the edge of the ¹³CO J = 1–0 intensity peaks. Since there is no H ii region toward the cavity, the high R_3−2/1−0 is possibly due to heating by the collisional shock but not by the enhancement of density or irradiation by ultraviolet photons of O stars. In the density regime concerned, the molecular cooling timescale is ∼10⁴ yr and collisional shock heating may be responsible for the high line intensity ratio on such a small timescale. S116-S117-S118 may be a rare case where heating by ultraviolet radiation is not important in a collision-produced cavity, allowing us to test the isolated contribution of the shock heating.

5 Conclusion

We carried out a molecular-line study toward the three Spitzer bubbles S116, S117, and S118. The region is associated with nearly 10 smaller H ii regions, indicating the active formation of high-mass stars over the length of 50 pc, although the region did not attract much attention until now. The detailed molecular data in the present work lead to the following conclusions, which offer a novel insight into the formation of the Spitzer bubbles:

1. The molecular clouds in the region of the three Spitzer bubbles S116-S117-S118 include two velocity components; one of them, the small cloud, has −63 km s⁻¹; and the other, the large cloud, has −58 km s⁻¹. The two clouds appear to be continuously distributed in a position–velocity diagram. The large cloud has a cavity, an intensity depression, which is apparently correlated with the morphology of the small cloud. Two of the Spitzer bubbles, S116 and S118, and additional small H ii regions are distributed along the northwestern and southwestern edges of the small cloud, and the other bubble, S117, is distributed toward the peak of the small cloud.

2. We present an interpretation that the cavity was created by a collision between the two clouds ∼1 Myr ago and that this collision compressed the gas into a dense layer elongating in the north–south direction to an extent of ∼20 pc at a kinematic distance of 5.9 kpc. In the compressed layer produced by the collision, the O stars exciting the H ii regions, including the three Spitzer bubbles, were formed. We show that a position–velocity diagram including the small cloud and the cavity exhibits a pattern characteristic of cloud–cloud collisions as numerically simulated by Takahira, Tasker, and Habe (2014). By assuming that the angle between the relative motion of the clouds is and the line of sight 45°, we estimate that the collision velocity and the collision timescale are ∼7 km s⁻¹ and ∼1 Myr, respectively.

3. Our morphology is different from the collision-induced star formation in RCW 120 where the H ii region was formed within the cavity created by the collision. We suggest the difference is due to the density-bound distribution of the large cloud in S116-117-118, which made the H ii regions expand toward the less dense outer part of the cloud rather than the more dense inner part. The present case is important in two ways. One is that it unambiguously demonstrates the formation of a molecular cavity by cloud–cloud collision without O stars. Another is that it allows us to watch details of the collision-compressed layer with no influence of ultraviolet photons of O stars; in particular, the highly filamentary distribution is seen as it is consistent with the numerical simulations of cloud–cloud collision that predicted filamentary distributions in the shocked gas (Takahira et al. 2014; Balfour et al. 2017; Inoue et al. 2018).

To summarize, we conclude that the three H ii regions are ionized by O stars formed by triggering in the cloud–cloud collision. This is a case where H ii regions expanded outside the collision-created cavity. The collision timescale is estimated to be ∼1 Myr. A mass accretion rate of 3 × 10⁻⁴ |$M_{\odot }\:$|yr⁻¹, based on MHD simulations of cloud–cloud collision (Inoue & Fukui 2013), explains O-star formation in 10⁵ yr, significantly smaller than the collision duration, which is consistent with the O-star formation in the late phase of the collision.

Acknowledgements

We are grateful to Ken Takahira to supply the numerical simulation data. This work is based on observations made with the Spitzer Space Telescope, which is operated by the Jet Propulsion Laboratory, California Institute of Technology under a contract with NASA. The MOST is operated by the University of Sydney with support from the Australian Research Council and the Science Foundation for Physics within the University of Sydney. NANTEN2 is an international collaboration of 11 universities: Nagoya University, Osaka Prefecture University, University of Bonn, University of Cologne, Seoul National University, University of Chile, University of New South Wales, Macquarie University, University of Sydney, University of Adelaide, and University of ETH Zurich. The Mopra radio telescope is part of the Australia Telescope National Facility. The University of New South Wales, the University of Adelaide, and the National Astronomical Observatory of Japan (NAOJ) Chile Observatory supported operations. The ASTE telescope is operated by NAOJ. This work was financially supported by Grant-in-Aid for Scientific Research (KAKENHI) of Japanese Society for the Promotion of Science (JSPS, Nos. 15K17607, H15H05694). Finally, we are grateful to the referee for his/her thoughtful comments.

References