Kinematics and Star Formation in the Hub–Filament System G6.55-0.1

Hub–filament systems (HFSs) being the potential sites of formation of star clusters and high-mass stars, provide a testbed for the current theories that attempt to explain star formation globally. It is thus important to study a large number of HFSs using both intensity and velocity information to constrain these objects better observationally. Here, we present a study of the HFS associated with G6.55-0.1 using newly obtained observations of the radio continuum and the J = 2–1 transition of CO, 13CO, and C18O. The radio continuum maps show multiple peaks that coincide with far-infrared dust continuum peaks, indicating the presence of more than one young massive star in the hub of the HFS. We used the velocity information from the C18O(2–1) map to (a) show that the source G6.55-0.1 is not physically associated with the supernova remnant W28 and (b) disentangle and identify the velocity components genuinely associated with G6.55-0.1. Among the velocity-coherent structures identified in the region, we conclude that only the two filaments at 13.8 and 17.3 km s−1 contribute a total mass accretion rate of 3000 M ⊙ Myr−1 to the hub. Both the filaments also show a V-shaped structure, characteristic of gravitational collapse, in their velocity profile at the location of the hub. The estimated mass per unit length of the segments of the filaments is smaller than the critical line masses derived from virial equilibrium considerations. This suggests that the filaments are not gravitationally collapsing as a whole, although their inner parts clearly show evidence of collapse in the form of young star-forming cores. We further conclude that the observed velocity gradients are consistent with the gravitational collapse of the main source in the region as estimated from its mass and size.


Introduction
The interstellar medium (ISM) is observed to be filamentary both in its atomic (Heiles 1979) and molecular (Schneider & Elmegreen 1979) phases.This view is firmly established by data obtained with modern observing facilities with everimproving angular resolution and sensitivities (André et al. 2014;Hacar et al. 2018;Arzoumanian et al. 2019).Observations have also indicated a direct connection between the filamentary structure of the ISM and the initial conditions for star formation (see André et al. 2014;Hacar et al. 2022;Pineda et al. 2023, for recent reviews).The large-scale Herschel far-infrared dust continuum emission surveys showed that most dense cores and young stellar objects (YSOs) are preferentially formed in association with dense filaments in both nearby (André et al. 2010) as well as in more distant Galactic plane (Molinari et al. 2010) molecular clouds.The low-mass cores found along parsec-scale filaments (Hartmann et al. 2001) likely form from their internal gravitational fragmentation (Schneider & Elmegreen 1979;Inutsuka & Miyama 1997).Most of the dense clumps harboring high-mass stars and clusters are found at the junction of massive filaments, forming the so-called hub-filament system (HFS; Myers 2009;Kumar et al. 2020).Filaments provide connection(s) to additional mass reservoir(s) and could funnel large amounts of material toward cores and clumps (e.g., Peretto et al. 2013).Yet, the connection between the existing star formation models and the new filamentary conditions of the ISM remains under debate, particularly in the case of high-mass stars (e.g., Motte et al. 2018).The multiscale dynamics leading to the movement of matter through the filaments to the cores is understood to be a continuous interplay between turbulence and gravity, with the former driving the nonthermal motions from filaments down to small scales, where gravity begins to dominate the dynamics once the regions reach a surface density above a critical value of ∼0.1 g cm −2 (Ohashi et al. 2016;Liu et al. 2023).
Currently, there exist two families of competing theories of star formation: in one, which has a fair amount of success in explaining low-mass star formation, supersonic turbulence is the one mechanism responsible for defining the mass reservoirs accessible to individual protostars, and as a result, for setting the stellar initial mass function (e.g., Krumholz et al. 2005;Padoan et al. 2020); the other models, which provide a more consistent picture of high-mass star formation, predict that the hierarchical gravitational collapse of molecular clouds is what drives their evolution (e.g., Peretto et al. 2007;Ballesteros-Paredes et al. 2011;Vázquez-Semadeni et al. 2017).Because of the HFSs' potential as the cradle of low-and high-mass star formation, they have recently been the targets of many observational studies (Peretto et al. 2013;Liu et al. 2023;Zhou et al. 2023, and references therein).Longitudinal flows along filaments with rates of ∼10 −4 -10 −3 M e yr −1 converging to clusters of stars suggest that such flows, triggered by a hierarchical global collapse, are adequate to form stars (Treviño-Morales et al. 2019).Stability analyses based on estimates of virial parameters of the filaments as well as of the embedded clumps reveal that a vast majority of the clouds are self-gravitating but stable on larger scales.On a parsec scale, however, the central clumps and/or cores in the same clouds are undergoing gravitational collapse (Mookerjea et al. 2023;Peretto et al. 2023).
The far-infrared continuum source IRAS 17577-2320 shows clear signatures of high-mass star formation, such as the multipeaked compact H II region G6.55-0.1 (Wood & Churchwell 1989) and methanol maser at 13.6 km s −1 (Walsh et al. 1998;Kim et al. 2019).Using Herschel continuum images and a dust emissivity exponent β = 2, Paradis et al. (2014) estimated dust temperatures of 30 and 25 K, respectively, for the central and the surrounding regions.The source IRAS 17577-2320 has also been identified in the literature as one of the star-forming regions near (in projection) the supernova remnant (SNR) W28.The farinfrared continuum images revealed an HFS with the hub coinciding with the IRAS source and four filaments with a total length of 12 pc and a total mass of 8.5 × 104 M e (Kumar et al. 2020).Considering the HFSs to be an important piece in the unsolved puzzle of massive star formation, here we use newly obtained radio continuum and molecular line (J = 2-1 transitions of CO and its isotopes) observations to study the nature of the massive protostar and mass accretion in the HFS associated with G6.55-0.1 located at a distance of 3 kpc (Wienen et al. 2012).This work particularly aims to identify the filaments associated with the HFS as velocity-coherent structures, study mass accretion through the filaments, and study their stability against gravitational collapse to understand their role in the formation of massive stars in the region.

Radio Continuum Observations and Data Reduction
We have mapped the low-frequency radio continuum emission at 750 and 1260 MHz toward G6.55-0.1 using the upgraded Giant Metrewave Radio Telescope (uGMRT; Gupta et al. 2017), India.The GMRT interferometer consists of 30 antennas, each with a diameter of 45 m, that are arranged in a Y-shaped configuration (Swarup et al. 1991).Of these, 12 antennas are located randomly within a central region of area 1 × 1 km 2 , and the remaining 18 antennas are placed along three arms, each with a length of 14 km.The shortest and longest baselines are 105 m and 25 km, respectively.The configuration enables us to map large-and small-scale structures simultaneously.The observations (Project code: 43_035) were carried out on 2022 August 21 and 22 in Band 4 (550-850 MHz) and Band 5 (1000-1460 MHz) with the GMRT Wideband Backend correlator configured to have a bandwidth of 400 MHz across 4096 channels.The radio source 3C286 was used as the primary flux calibrator and bandpass calibrator, and the sources 1911-201 (at 750 MHz) and 1822-096 (at 1260 MHz) were used as the phase calibrators.Primary calibrators were observed at the beginning and end of the observation for flux and bandpass calibration.The phase calibrators were observed after each scan (30 minutes) of the target to calibrate the phase and amplitude variations over the entire observing period.The map was centered at the position of α 2000 = 18 h 00 m 49.9 s and δ 2000 = −23°20′33″.The angular sizes of the largest structure observable with the uGMRT are 15′ and 9′ at 750 and 1260 MHz, respectively.Table 1 summarizes the details of the uGMRT observations and data.
The data reduction process of flagging, calibration, imaging, and self-calibration was done using the CASA 4 Pipeline-cum-Toolkit for Upgraded GMRT data REduction (CAPTURE)5 continuum imaging pipeline for uGMRT (Kale & Ishwara-Chandra 2021), which utilizes tasks from Common Astronomy Software Applications (CASA; McMullin et al. 2007).
The flux density calibration was done using the scale provided by Perley & Butler (2017).After the initial rounds of editing and calibration, we used the multiterm multifrequency synthesis (see Rau & Cornwell 2011) algorithm in the tclean task to account for possible deconvolution errors in wideband imaging.In the pipeline, five rounds of phase-only selfcalibration were performed before making the final image.These maps were then corrected for primary beam gain.

Molecular Line Observations with APEX
We used the dual-polarization NFLASH230 facility receiver of the APEX telescope6 (Güsten et al. 2006), connected to digital FFTS backends with a spectral resolution of 61 kHz (project ID M-0110.F-9518C-2022).The wide IF bandwidth of the spectrometer allows simultaneous recording of the J = 2-1 CO isotopologues 12 CO, 13 CO, and C 18 O.Atmospheric conditions during the observations (2022 November 11) were stable, with precipitable water vapor of 2.5 mm.The halfpower beam width of the APEX is 26 2 (at 230 GHz).Data are presented as main-beam temperatures, using a main-beam efficiency of η mb = 0.80.
Data were acquired in total power on-the-fly mode, sampling spectra on a 10″ grid while slewing the telescope in R.A.The observed field, 750″ × 750″ in size, was centered at R.A. =18 h 00 m 49.9 s decl.= −23°20′33 8 (2000).The reference position at R.A. = 17 h 47 m 51.8 s decl.= −21°19′58 8 (2000) was clean of 12 CO(2-1) emission at the level of T mb = −0.15K. CO line pointing was established on nearby RAFGL1922.For the APEX observations, the spectra were box smoothed to 0.25 km s −1 spectra resolution and the second-order baseline was removed.The reduced spectra have been binned to a spectral resolution of 0.5 km s −1 and the final CO data cubes with a pixel size of 10″ have an rms of 0.25 K.

Archival Infrared and Submillimeter Continuum Data
We have used archival infrared continuum emission maps of the region observed with Spitzer/IRAC, Spitzer/MIPS, Herschel/PACS, Herschel/SPIRE, and APEX/Laboca.The 3.6, 4.5, 5.8, and 8.0 μm maps at ∼2″ resolution were observed as part of the GLIMPSE program (Benjamin et al. 2003) and  3. Overview of the W28 Region and G6.55-0.1 The radio continuum source G6.55-0.1 is located in the vicinity of the SNR W28 and is associated with an HFS identified in the Herschel/SPIRE 250 μm image by Kumar et al. (2020).Figure 1 provides an overview of the region in 327 MHz radio continuum emission (Frail et al. 1993) and a close-up view of the emission at 8 μm arising from the farultraviolet irradiated polycyclic aromatic hydrocarbon as well as the dust continuum emission at 250 μm.The radio continuum shows an extended shell arising from the nonthermal and thermal emission, due to the SNR W28 that interacts with the dense molecular clouds in W28F to the east.The region G6.55-0.11stands out as a bright spot in the otherwise complex emission from W28.The dust continuum emission shows a centrally peaked structure that coincides with the peak of the radio continuum emission but also shows more extended filamentary emission features both to the north and the south.At 250 μm, the filaments are not so clearly discerned in this image; however, using dedicated feature-finding algorithms, Kumar et al. (2020) identified the source as an HFS.A total of 15 Hi-GAL continuum sources detected in at least two of the five Herschel wavebands have been identified in the region (Elia et al. 2017) and two additional sources were only detected in the 70 μm Herschel image.Figure 1 shows that of these 17 sources, nearly nine lie on the elongated structure extending to the south of the central bright source and seven of these sources have an associated source within 3″ in the GLIMPSE catalog.Four of the continuum sources are also detected at 870 μm in the ATLASGAL images (Schuller et al. 2009;Csengeri et al. 2016).Table 2 presents the properties of the Hi-GAL sources in the region as estimated by Elia et al. (2017), using graybody fitting of the Herschel flux densities.The derived masses of the continuum sources are typically less than 20 M e with only three sources, including the central source showing masses of 40 M e or more.Previous studies of CO emission from the region have focused on the molecular gas affected by the SNR W28 (Arikawa et al. 1999;Reach et al. 2005).These studies identify the shocked CO emitting gas centered at velocities less than 10 km s −1 and with CO(2-1) line widths between 10 and 30 km s −1 (Reach et al. 2005).
4. The Ionized Gas in G6.55-0.1 Figure 2 shows the radio continuum images at 750 and 1260 MHz observed using the uGMRT.At both frequencies, the emission shows a compact peak at the position of the H II region G6.55-0.1 with very low-intensity extended emission.
The emission from the central region, though compact, shows two more local peaks at the high resolution of the radio observations.In order to investigate whether there are any dust continuum peaks colocated with these radio continuum peaks, we have derived the intensity profile of the 750 MHz, 1260 MHz, and 70 μm PACS emission along the direction shown in Figure 2. The radio continuum intensity profiles show  three clear enhancements, while the 70 μm profile shows only the main and the secondary peaks (Figure 3) that match with the radio peaks.The vertical lines in Figure 3 show that the three intensity peaks in the radio continuum coincide with the location of the main source (S10) and two other Hi-GAL sources, and two other sources (S12 and S13), which are detected at 70 μm only (Molinari et al. 2016).
We have estimated the radius of the H II region to be 24″ by considering the area enclosed by the contour corresponding to the radio continuum intensity of 20% of the peak value.The integrated intensity within this region is 1.5 and 2.0 Jy at 750 and 1260 MHz, respectively.Assuming that the diffuse emission at 1260 MHz is optically thin, we estimate the Lyman continuum photon rate (N Lyc ) and the spectral type of the star that causes the ionized emission using the following equation (Mezger et al. 1974): where S ν is the flux density at frequency ν, which is 2.0 Jy at 1260 MHz, T e is the electron temperature assumed to be 10 4 K, and d is the distance to the source, which is 3 kpc (Urquhart et al. 2018).The N Lyc is 1.4 × 10 48 s −1 , and if it is assumed to be due to a single main-sequence star, then the spectral type of the zero-age main-sequence (ZAMS) star is O8.5V with a bolometric luminosity of 5.3 × 10 4 L e (Thompson 1984).The L bol estimated from the dust continuum flux densities at 70, 160, 250, 350, and 500 μm over the same region as the H II region is 4.9 × 10 4 L e .For this estimate, we have fitted a twocomponent graybody function to the dust continuum spectral energy distribution, assuming dust emissivity exponents of 1 and 2, respectively, for the warm and the cold dust.However, we also note that based on the radio continuum emission distribution, the massive stellar component creating the H II region is likely to be more than one early-type star.
A power-law fit to the observed intensities at 750 and 1260 MHz results in an index of 0.56.The positive spectral index is suggestive of the emission being primarily thermal and arising due to the free-free emission from the H II region created by the massive star(s) therein (Olnon 1975).At an angular resolution of >1′, Dubner et al. (2000) derived an  index of −0.35 for the entire W28 region, which likely corresponds to the extended nonthermal emission from the SNR W28.However, the positive spectral index also suggests that the emission, even at 1260 MHz, is unlikely to be optically thin, which could lead to an underestimate of N Lyc .

Column Density from Dust Continuum Emission
In order to obtain an overview of the distribution of cold molecular gas that is essentially the reservoir for material forming the stars, we used the far-infrared emission maps between 160 and 500 μm to obtain the distribution of column density of the cold dust in the region.The 70 μm data was not included in this analysis because at 70 μm there is still some contribution from smaller dust grains, which do not emit much at longer wavelengths, and also, the presence of hot dust necessitates the use of at least two dust temperature components thus making the fit less constrained.The Herschel images at the four wavelengths were first corrected for the zerolevel offsets.For this, we assume that the images probe the optically thin dust emission following the empirical fit (e.g., Planck Collaboration 2014), where ν 0 is the reference frequency at which the optical depth t n 0 is estimated, I ν is the specific intensity, and B ν (T) is the Planck function for the emission of dust at temperature T and frequency ν.All-sky maps of the dust optical depth at ν 0 = 353 GHz (850 μm) and the dust temperature maps generated by the Planck Collaboration XI (2014) at a resolution of 5′ were used.For β = 2, adopting ν 0 = 353 GHz as the reference frequency and substituting the optical depth at 353 GHz in Equation (2), we used the Planck dust temperature map to derive the images of G6.55-0.1 at 160, 250, 350, and 500 μm, corresponding to the Herschel wavelengths.By convolving each Herschel image to the Planck resolution, we obtained the offsets of each Herschel image by comparison with the derived intensity maps (Bernard et al. 2010).The offsets were found to be 200, 148, 75, and 20 MJy sr −1 at 160, 250, 350, and 500 μm, respectively.Dust temperature and column density maps of the region were derived by performing pixel-by-pixel graybody fitting of the offset-corrected continuum fluxes using = W - where F ν is the flux density measured in a pixel, Ω is the angular area of each pixel, B ν (T d ) is the blackbody function at dust temperature T d , μ = 2.86 is the mean molecular weight of H 2 , m H is the mass of a hydrogen atom, the absorption coefficient is given by κ ν = 0.1(ν/1000 GHz) β cm 2 g −1 , β = 2 is the dust emissivity index, and τ ν is the optical depth.We apply the technique described in Palmeirim et al. (2013) that considers flux measurements up to 500 μm but uses a multiresolution decomposition to generate the column density map at a resolution of 18 2 corresponding to the Herschel beam at 250 μm.The fitted dust temperature lies between 20 and 28 K and the column density ranges between 5 × 10 21 and 10 23 cm −2 (Figure 4), with the hub of the HFS showing the maximum column density.The dust temperature is maximum at the location of the H II region and does not show any enhancement toward the northeast where the SNR W28 lies.This suggests that the dust is primarily heated by the embedded high-mass star-forming region.

Molecular Line Emission and Column Density Estimates
The CO(2-1) emission from the mapped region is primarily dominated by the molecular gas associated and most likely interacting with the SNR W28 and is much brighter than the emission from the source G6.55-0.1 (Figure 5).In contrast, the 13 CO, and in particular, the C 18 O emission is dominated by molecular gas spatially overlapping with the H II region G6.55-0.1 and tracing the dust continuum features seen in the 250 μm map in Figure 1.This indicates that the higherintensity CO(2-1) emission in the map corresponds to lower column density high-temperature gas coinciding with the W28 shell, whereas the isotopes mainly trace the high column density gas in G6.55-0.1 and trace the northern feature only in a limited region.The CO, 13 CO, and C 18 O spectral profiles in the mapped region are complex, due to a combination of multiple velocity components as well as optical depth effects, particularly for CO and 13 CO.We compare the spectra of CO and its isotopes at three positions (marked in Figure 5) selected to demonstrate the diversity of velocity profiles detected in the region (Figure 6).We note that except for the CO spectrum at (30, 170), the spectra at the other two positions do not show any line broadening that might be suggestive of shock heating by the SNR W28.The C 18 O(2-1) spectra at the quiescent positions in the region show three distinct velocity peaks approximately at 15, 18, and 23 km s −1 .While the 23 km s −1 feature is also clearly detected in the 13 CO and CO, between 10 and 20 km s −1, the spectra for both these species are affected by moderate to severe selfabsorption.The two positions (30, 40) and (100, −300) are representative and capture the basic features of the three spectral features arising from G6.55-0.1 at most positions lying on the north-south extended emission feature that approximately terminates a little to the north of the radio continuum peak.We note that at these two positions, the 13 CO (2-1) emission also does not arise from gas less than 10 km s −1 .We use only the optically thin C 18 O(2-1) spectra for analyses of the velocity components and as well as for the derivation of other physical quantities.Wienen et al. (2012) observed NH 3 lines at the position of S10, also an ATLASGAL source, and identified two components at 12.8 and 16.1 km s −1 , both with a line width of 2 km s −1 .This suggests that the 18 km s −1 cloud may not correspond to high-density gas.Based on the C 18 O(2-1) spectra (Figure 6) and the channel map (Figure 7) it is difficult to separate out the extents of the emission from the different components since these spatially overlap.Thus, we have used an analysis involving the decomposition of the spectrum at each pixel into separate velocity components with Gaussian profiles, followed by the grouping of the velocity-coherent structures using a friend-of-friends (FoF) algorithm (Section 5.3). Figure 6.Comparison of CO (black), 13 CO (red), and C 18 O (filled histogram with green outline) spectra at selected positions in the mapped regions.Each spectrum is centered at an offset relative to the center of the map at 18:00:49.93 -23:20:33.89 (2000) marked in the panel and averaged over 30″.For better visibility, the 13 CO and C 18 O spectra are multiplied by factors mentioned in each panel.
Considering the C 18 O(2-1) emission seen in the channel map (Figure 7) as well as the individual spectra (Figure 6) we estimate the column density of the region around G6.55-0.1 from the optically thin C 18 O intensities integrated between 11 and 19 km s −1 , using the following equation: where Z is the partition function given by where B = 5.4891420 × 10 10 s −1 is the rotational constant for C 18 O, μ = 0.11079 D is its dipole moment, and J u is the upper level of the transition, equal to 2. The CO and 13 CO(2-1), though optically thick, are heavily affected by self-absorption, so an estimate of T ex is difficult using their spectra.We assume T ex to be 20 K, similar to the average dust temperature in the region and estimate a factor of 8.8 × 10 14 cm for conversion of the integrated C 18 O(2-1) intensities to N (CO).This factor increases by a factor of 2 for an assumed T ex of 10 K. We further adopt 16 O/ 18 O to be 500 and CO/ H 2 = 10 −4 to estimate the molecular H 2 column density, N(H 2 ), for the entire region (Figure 4).The N(H 2 ) distribution so obtained from the velocity-selected C 18 O(2-1) emission is similar to the distribution derived from dust continuum emission; this implies that the molecular line indeed traces the HFS detected in the 250 μm image by Kumar et al. (2020).We note that in the C 18 O column density map, an extra emission feature is detected toward the northwest of the hub, which is not seen in dust emission.The column density estimate from the C 18 O emission ranges between (0.5 and 6) × 10 22 cm −2 , which is less than the column density estimated from the dust emission by 60%.However, considering the fact that the C 18 O(2-1) intensity is for the velocity range of 11-19 km s −1 identified to be associated with G6.55-0.1, while the dust column density includes all components along the line of sight and also the uncertainty in the conversion factor, due to the assumed T ex , the column density estimates from dust and gas are consistent with each other.Based on the column density distribution we estimate the HFS G6.55-0.1 to be 7.6 pc long with a width of 2.3 pc at most places, with a total mass of 4520 M e .The mass estimated for the HFS is much smaller than the value of 8.5 × 10 4 M e estimated earlier for a larger structure with four filaments, as analyzed by Kumar et al. (2020).
With the C 18 O(2-1) emission being optically thin, the channel map primarily detects emission from the higher column density molecular cloud associated with G6.55-0.1 (Figure 7).The primary emission feature in C 18 O(2-1) extended approximately in the north-south direction is due to gas at velocities between 12.5 and 16 km s −1 , while the hub, which also corresponds to the maximum intensity, shows emission over the entire velocity range.The C 18 O(2-1) spectrum close to the center of the hub at an offset of (30″, 40″) shows that the emission from this position is due to at least three velocity components (Figure 6).The north-south extended feature is almost not seen in the channel map for the CO(2-1) emission, due to self-absorption (Figure 11), but is seen in 13 CO(2-1) with additional diffuse emission (Figure 12).The east-west extended emission feature lying to the north of the CO(2-1) map arises at smaller velocities, consistent with the molecular material associated and physically interacting with the SNR W28.The 13 CO(2-1) channel map detects the north-south feature clearly, but shows that while the peak of the emission around 13 km s −1 lies close to the position of the hub at 14-16 km s −1 , the emission peaks at a position about 50″ to the south of the hub, but still lying on the main feature.At velocities beyond 16.5 km s −1 , the north-south emission feature in C 18 O(2-1) appears to be extended toward the northwest going beyond the hub, and in 13 CO(2-1), it is connected to a feature around 18.5 km s −1 toward the south.The 13 CO(2-1) emission at velocities exceeding 17 km s −1 is distinctly fainter than the emission from the molecular gas at 13-15 km s −1 .This could explain the lack of connection between the features seen at velocities longer than 17 km s −1 in the C 18 O(2-1) channel map.Additionally, the feature seen to the northeast at velocities <13 km s −1 , though seen as detached peaks in C 18 O(2-1), lie on an extended filament-like structure in 13 CO(2-1), which appears to be connected with the northsouth feature at the top.
In the context of the star formation scenario, it is important to understand how and whether the emission features in the velocity ranges, i.e., <13, 13-15, 16-17, and 17-19 km s −1 in the CO channel maps are kinematically interacting and indeed are the filaments feeding the hub harboring G6.55-0.1.We have explored the position-velocity diagrams (Figure 8) along the direction shown in Figure 5.The CO(2-1) positionvelocity diagrams, possibly quite affected by self-absorption, show two narrow features centered around 12 and 20 km s −1 .The 13 CO and C 18 O position-velocity diagrams, on the other hand, show the 16 km s −1 emission feature to the south of the main emission feature around 13-14 km s −1 , coinciding with the hub in the HFS associated with G6.55-0.1.Exactly at the location of the hub, in both the C 18 O and the 13 CO positionvelocity diagrams, emission from the 17 km s −1 cloud is also visible, although only in CO and C 18 O it appears to show a bridge-like feature.However, it is not obvious that the two components are indeed interacting with each other or that the mass in the hub is being accreted through both of these components.

Gaussian Decomposition of Spectra
We have used the FUNStools.Decompose7 to decompose the C 18 O(2-1) data cubes into multiple Gaussian components.The tool is based on an algorithm that primarily decides the number of components and their velocity positions in the spectrum using the first, second, and third derivatives based on the idea that the velocity component in a filament is continuous with the surroundings.The results of the first fitting are given as the initial guess to a subsequent round of Gaussian fitting in an iterative manner.The outcome is the parameters such as the central velocity and the line width for each velocity component identified at each pixel in the C 18 O data cube.We identify coherent structures corresponding to each of the velocity components so identified by applying the FoF technique to the central velocity of each of the components.The algorithm works in an iterative manner in which it first selects the decomposed Gaussian seed component with the maximum amplitude and gives the structure a number.Subsequently, the other Gaussian components in the neighboring pixels whose pixel distance is less than 2 from the seed component are selected, and the velocity differences of the seed and other components in the neighboring pixels are checked.If the velocity difference between the seed and other components in the neighboring pixel is less than the velocity dispersion of the seed (s n seed ), the neighboring component is identified as a friend of the seed and assigned the same group number.If more than one velocity component in a neighboring pixel is within the range of velocity dispersion from the velocity of the seed, then only the closest one becomes the friend of the seed.In the next iteration, the assigned friends of the seed become the seeds of the structure and the same procedure is repeated until there are no more friends to assign.
Based on the analysis of the C 18 O(2-1) data, we identify a total of seven groups centered at velocities (in kilometers per second) of 11.8 ± 0.4, 12.6 ± 0.4, 13.8 ± 0.8, 14.3 ± 0.4, 16.1 ± 0.7, 17.3 ± 1.2, and 18.5 ± 0.2.Of these, the group centered at 13.8 km s −1 is the largest coherent structure identified in this region (Figure 9) and the 17.3 km s −1 component is also reasonably large, with emission from the northwestern part being quite faint.The 18.5 km s −1 feature to the south, which appears to be connected to the 17 km s −1 component in the 13 CO(2-1) channel map (Figure 12) is detected separately in the C 18 O data.While the groups are consistent with the features seen in the channel maps, exact Gaussian decomposition of the spectra into these components allows for a better quantitative estimate of the properties of the main components.

Velocity Gradient and Mass Accretion Rates along the Filaments
For clouds in gravitational collapse, the gas flow is expected to be accelerated toward the center of the potential well because of gravitational attraction (Gómez & Vázquez-Semadeni 2014).It is expected to produce a characteristic V shape on the gas velocity structure, which traces the accelerated gas motion around the central high-mass clumps or cores.Such features have been detected in other regions, including the OMC-1 cloud (Hacar et al. 2017) and G333 (Zhou et al. 2023).Figure 10 shows a plot of the velocity and intensity variation along the filaments at 13.8 and 17.3 km s −1 .For ease of discussion and quantitative evaluation, we have split up the filaments at 13.8 and 17.3 km s −1 into four and five segments, respectively.
For the 13.8 km s −1 filament, we find a very strong V-shaped variation of the velocity close to 4.8 pc from the southern tip of the filament between segments 3 and 4. The tip of the V in the velocity structure aligns exactly with the peak of the intensity profile corresponding to the continuum sources S9 and S10.We note here although S10 is the bright continuum source, the 13.8 km s −1 feature peaks at the location of S9.Although S9 and S10 being separated by 0.6 pc are spatially resolved we attribute the change in velocity gradient to a somewhat larger scale accretion onto both S9 and S10.There is a shallower V-shaped structure in the velocity profile that corresponds to the far-infrared continuum source S4 located right to the south of the hub.Thus, we clearly detect a signature of mass accretion from the 13.8 km s −1 filament onto three star-forming cores that are also detected in the far-infrared continuum.We have estimated the properties of molecular material in each of the segments assumed for both the filaments (Table 3).We calculated the mass accretion rate in each segment of the filament using  q = M v M tan grad (Kirk et al. 2013;Ma et al. 2023) where M is the mass of the segment and v grad is the velocity gradient across it, and θ is the angle made by the filamentary part with the plane of the sky which is assumed to be 45°.The four segments of the 13.8 km s −1 filament have masses ranging from 174-660 M e and show mass accretion rates between 152 and 1020 M e Myr −1 .The hub is located between segments 3 and 4 and shows a total mass accretion rate of ∼1780 M e Myr −1 from the 13.8 km s −1 filament.These mass accretion rates are comparable to the values observed in other filamentary star-forming systems (Yuan et al. 2018;Hu et al. 2021;Ma et al. 2023).We also performed the same analysis for the 17.3 km s −1 filament and found two V-shaped structures in its velocity profile (Figure 10), one at the location of the hub between segments 1 and 2 and the other around 1.5-2 pc to the northwest from the hub.The second V-shaped structure does not exactly align with a smaller peak in C 18 O (2-1) intensity that is seen slightly to the north of the V.The mass accretion rate to the hub from the 17.3 km s −1 filament is seen to be ∼1240 M e Myr −1 , resulting in a total accretion rate of ∼3000 M e Myr −1 from the two filaments.
The molecular line emission arising from the north-south extended HFS 6.55-0.1 lies between the velocities of 11 and 18 km s −1 , which kinematically corresponds to a distance of 3 kpc (Urquhart et al. 2018).This suggests that G6.55-0.1 lies well behind the SNR W28 along the line of sight, with the latter being at a distance of 1.9 kpc.Nevertheless, because of their proximity on the sky, we carefully searched our data for spectroscopic evidence of an interaction.The SNR W28 has extensive evidence for interaction with molecular clouds, Figure 8. Position-velocity diagrams for C 18 O, 13 CO, and CO (2-1) along the direction shown in Figure 5.The contour levels (in kelvin) for C 18 O, 13 CO, and CO, respectively, are 0.1-2.8(in steps of 0.2), 0.1-5.5 (in steps of 0.5), and 1-12 (in steps of 1).
particularly in the form of OH 1720 MHz maser emission (Claussen et al. 1997) and broad molecular line (BML) CO emission (Arikawa et al. 1999;Reach et al. 2005); however, these emission were observed from regions that do not overlap with the HFS G6.55-0.1.The CO(3-2) and (2-1) spectra from the BML regions are clearly marked by line widths of ∼20-30 km s −1 , with maximum values of up to 70 km s −1 and sharply peaked profiles that are typical of shocked gas (Reach et al. 2005).The spectral line observed in the main body of the HFS 6.55-0.1 for all three tracers typically have much smaller widths and do not have shapes that are indicative of shocked gas.The east-west extended feature lying to the north of the HFS is detected only in CO(2-1), with the main emission centered at velocities less than 7 km s −1 and with widths >5 km s −1 , which is in contrast to the velocity components identified in the HFS.Although the CO(2-1) emission is strongly affected by opacity issues, a clear discontinuity in emission, even between 9 and 11 km s −1 , can be seen in the channel map itself (Figure 11).Further, neither C 18 O nor 13 CO (2-1) spectra shows any signs of being affected by any external forces, particularly toward the north and the east.We thus conclude that HFS G6.55-0.1 shows no evidence of physical interaction with the SNR W28, and hence, the star Figure 9. Velocity-coherent structures identified in the region based on the C 18 O(2-1) spectra using Gaussian decomposition followed by the identification of groups using an FoF algorithm.The positions on the 13.8 and 17.3 km s −1 filaments used to extract the velocity and intensity profiles (Figure 10) are marked in black and red, respectively, in the right panel.The positions of selected Hi-Gal sources are also marked in the right panel for reference.formation in G6.55-0.1 is unlikely to be triggered by the impact of the SNR.

Gravitational Stability of the Filaments and Cores in
G6.55-0.1 The velocity distribution of the molecular material in the region around G6.55-0.1 being highly complex, Gaussian decomposition of the spectra was necessary to measure the typical line width corresponding to each of the velocity components.Due to the large velocity gradients, particularly at the location of the hub, part of the observed line width is due to the shift in velocity over the beam size of 26 2 of the C 18 O data.We have corrected for the effect of beam broadening on the line width by considering the velocities of all pixels within 27″ of a particular pixel.Following the method described by Stil & Israel (2002), we estimate the beam-deconvolved line width from the observed line width, using s ) , where V max and V min are the maximum and minimum values of V cen found within a beam centered at the particular pixel and the σs are related to the observed and deconvolved line widths through s D = V 8 ln 2 .We have assumed a constant intensity over the beam width resulting in the factor of 1/2 in the relation between observed and deconvolved σs The median deconvolved line width of each segment of the two filaments at 13.8 and 17.3 km s −1 are presented in Table 3.The deconvolved line width so obtained is a combination of contributions from both thermal and nonthermal gas motions.To estimate the nonthermal motions associated with gas turbulence or core formation motions, we have subtracted the thermal component from the beam-deconvolved line width, assuming that the two contributions are independent of each other so they add in quadrature (e.g., Myers 1983).We estimate the nonthermal velocity dispersion as follows: where k is the Boltzmann constant, T the gas kinetic temperature, and m mol the mass of the molecule under consideration.The velocity dispersion σ NT can be directly compared to the (isothermal) speed of sound of the gas, c s , which has a value of 0.27 km s −1 for ISM gas at 20 K. We use the median values of beam-deconvolved line widths for each of the segments identified in the two filaments and estimate the critical line mass of each segment for a temperature of 20 K following (André et al. 2014), For all the segments of the 13.8 and 17.3 km s −1 filaments, the mass per unit length estimated for the different parts of the two filaments is consistently smaller than the critical line masses derived for the segments (Table 3).This suggests that although the filaments show a clear kinematic signature of mass accretion, as a whole, these are not in a state of gravitational collapse.Considering typical velocity widths of 2.4 km s −1 , the critical line mass at 20 K is estimated to be 498.1 M e pc −1 , which is smaller than the mass per unit length of 594 M e pc −1 observed for the entire filament with a mass of 4520 M e and a length of 7.6 pc.In contrast to the values presented in Table 3 here we have considered the total mass of the HFS estimated over all the velocity components associated with it and comparison of this estimate with the critical line mass suggests that the HFS as a whole is likely to be self-gravitating.
We estimate the gravitational stability of the 16.1 km s −1 cloud by calculating the virial parameter defined as , such that for values of α ∼ <1, the cloud is gravitationally unstable.For the 16.1 km s −1 cloud, we use an effective radius of 0.8 pc, a total mass of 610 M e, and a beam-deconvolved line width of 1.5 km s −1 to obtain α vir ∼ 8.This suggests that the cloud at 16.1 km s −1 is supported against gravitational collapse by turbulent gas motions.
We note that due to ongoing high-mass star formation in the hub, the gas temperature of the 16.1 km s −1 cloud as well as the segments of the 13.8 and 17.3 km s −1 filaments, which are close to the hub, are likely to be significantly higher than the assumed value.This could lead to somewhat lower estimates for the nonthermal velocity component.
The velocity gradient arising due to freefall under gravity is related to the mass and size of the collapsing region by u = GM R grad 2 3 .From the H 2 column density estimated from the C 18 O(2-1) emission, we estimate the mass of the hub within a radius of 30″ (equal to 0.45 pc) in G6.55-0.1 to be 554 M e .The velocity gradient corresponding to the freefall of such a clump is estimated to be 3.6 km s −1 pc −1 , a value that is consistent with the range of velocity gradients seen in the segments of the 13.8 and 17.3 km s −1 filaments that are close to the hub.We note that the velocity gradients derived from the observations are subject to uncertainties due to the assumed inclination of the filament with the plane of the sky.Thus, the velocity gradients obtained around the hub can be consistently explained in terms of gravitational collapse.

Is G6.55-0.1 Really an HFS?
The source G6.55-0.1 was identified as an HFS based on dust continuum observations.We have used spectroscopic observations to identify the kinematically related structures (filaments) merging at the location of the associated H II region (hub).We find that the emission from molecular gas arises due to multiple velocity components, three of which at 13.8, 16.1, and 17.3 km s −1 , appear to be directly associated with the source G6.55-0.1.Of these, the components at 13.8 and 17.3 km s −1 appear to overlap at the hub, while the 16.1 km s −1 is seen a little to the south of the hub.The 13.8 km s −1 component shows two clear signatures of accelerated gas flow due to mass accretion, leading to the formation of the YSOs S9, S10 (G6.55-0.1),and S4.The 17.3 km s −1 component emission is strongly peaked at the location of S10 in the hub, while for the 13.8 km s −1 , although the intensity is enhanced in the hub, the absolute peak is slightly to the south of the hub, coinciding with the source S9.Both the 13.8 and 17.3 km s −1 features, however, show a strong change in the velocity gradient at the location of the hub (Figure 10).In order to ensure that the velocity gradients so derived for the filaments intersecting at the position of the hub do not suffer from any artifact due to the automated fitting procedure used in this work, we have checked the Gaussian fits within the hub manually as well, an example of which is shown in Figure 13.At the position of the center of the hub, which corresponds to an ATLASGAL source, NH 3 observations with a resolution of 40″ have identified two components at 12.8 and 16.1 km s −1 .This could be indicative of the fact that the 17.3 km s −1 component does not arise from gas at densities high enough to excite the NH 3 lines.For the 17.3 km s −1 component, based on the mass derived for the region around the peak, we estimate an average density of 900 cm −3 .
The C 18 O(2-1) spectra observed in the hub (Figure 13) are qualitatively similar to the profiles predicted by smooth particle hydrodynamic simulations of a Jeans-unstable dense prolate clump in the process of collapsing along its long axis on a nearfreefall dynamical timescale (Peretto et al. 2006(Peretto et al. , 2007)).For NGC 2264-C for a central core of 90 M e the velocity difference between the two peaks is 2 km s −1 , which is possibly consistent with the 3.6 km s −1 velocity gap that we observe for G6.55-0.1 where we do not resolve the central core but can only observe a bigger clump with a mass of 554 M e .The V-shaped velocity profiles with locations of the largest gradient coinciding with the intensity peak (Figure 3) are indicative of the accretion of mass from both directions.Additionally, Zhou et al. (2023) suggest that global hierarchical collapse under gravity would create a funnel-shaped structure in the position-positionvelocity space and the V-shaped velocity structure along the filament skeleton is likely to be a projection of this structure to the position-velocity plane.Though we do not spatially resolve the central accreting core, the spectral profiles, the V-shaped velocity profile and the far-infrared continuum sources together suggest that the hub is likely undergoing collapse.We note that the segment-wise virial analysis of the two main filaments suggests the filaments have masses smaller than their respective critical line masses (Table 3).This apparent contradiction could arise from an overestimate of σ NT , due to enhanced gas temperature and turbulence due to the high-mass stars already forming in the hub.On the other hand, the outcome of the selfgravitating clouds being stable on larger scales and the clumps on parsec scales being dynamically decoupled from their surrounding molecular cloud and collapsing has also been observed in several other sources (e.g., Mookerjea et al. 2023;Peretto et al. 2023).We thus conclude that the high-mass star formation in G6.55-0.1 occurs in a hub that is being fed by at least two filaments at 13.8 and 17.3 km s −1 .

Summary and Conclusions
We have explored the morphology and kinematics of the source G6.55-0.1, which is a massive star-forming region associated with an HFS.The newly obtained radio continuum maps detect multiple peaks that are associated with far-infrared sources, thus confirming the presence of multiple high-mass protostars in the hub region.The total radio continuum flux detected from the H II region can be explained by a single ZAMS O8.5V star or multiple B-type stars.We used the velocity information available from the J = 2-1 transitions of CO (and its isotopes) to identify velocity-coherent structures (filaments and cores) and investigate signatures of mass accretion along the filaments to the hubs.The molecular emission associated with G6.55-0.1 is strongly affected by the velocity crowding due to multiple overlapping components along the line of sight (including the SNR W28) as well as by the optical thickness of both CO and 13 CO(2-1) lines.The C 18 O(2-1) line, though optically thin, being fainter primarily traces the higher column density filaments and the hub.However, a comparison of the CO and C 18 O spectra, particularly toward the northeast, where the cloud overlaps with molecular gas affected by the SNR W28, enabled us to confirm that G6.55-0.1 is physically not associated with W28.The column density of molecular gas in the region derived from both dust continuum emission and C 18 O(2-1) intensities were found to be consistent within a factor of 2 with a median value of 7 × 10 22 cm −2 , a value typical of similar HFSs in which massive star formation is seen (Zhou et al. 2022).An automated Gaussian decomposition followed by identification of correlated structures using an FoF algorithm was used to disentangle the complex C 18 O(2-1) spectra.The components segregated using this procedure confirmed the interaction of only two of the filaments with median velocities of 13.8 and 17.3 km s − 1 at the location of the hub.The other velocitycoherent structures are either more core-like, or even if these are extended like the 11.8 km s −1 feature, the connection to the hub is tenuous, possibly because the C 18 O(2-1) emission is too faint to be detected.The use of 13 CO and CO spectra to confirm any interaction is unreliable, due to the contamination by the velocity components physically not associated with the source, as well as due to the optical thickness of the emission resulting in self-absorption.Both filaments at 13.8 and 17.3 km s −1 show increased velocity gradients and V-shaped structures in the hub that harbors the H II region that suggest gravitational collapse with mass accretion rates between 558 and 1020 M e Myr −1 .The mass accretion rates typically found in star-forming regions range from a few 10 to a few 100 M e Myr −1 (Hacar et al. 2022, and references therein), with the high-mass starforming HFSs showing up to 1500 M e Myr −1 (Hu et al. 2021;Ma et al. 2023).The combined accretion rate toward the hub is 3000 M e Myr −1 , which, though on the higher side, is consistent with the previous observations within the uncertainties of the inclination of the filaments on the plane of the sky.Segment-wise analysis of the filaments based on virial parameters suggests the filament to be stable, while the gas in the hub shows spectral and velocity profiles indicative of dynamical decoupling/collapse leading to the formation of high-mass stars.It is also likely that the filaments being identified at a resolution of 0.4 pc have coherent substructures that will require a similar analysis.Follow-up higher-resolution observations of high-density tracers, preferably without hyperfine structures, are needed to constrain the properties of the filaments and the flow of material through them.

Figure 1 .
Figure 1.Left: radio continuum image at 327 MHz at a resolution of 65″ adapted from Frail et al. (1993) with permission.The box shows the region around G6.55-0.1 studied in detail in this work.Right: dust continuum emission from the region G6.55-0.1 at 8 and 250 μm, respectively.Positions of the 17 Hi-GAL continuum sources and 7 GLIMPSE sources are marked with black "+" and purple-filled circles, respectively.

Figure 3 .
Figure 3.Comparison of the normalized radial profiles as a function of the offset from the southeast end of a cut shown in Figure (2) at 70 μm, 750 MHz, and 1260 MHz.The vertical-dashed lines show the positions of the three Hi-GAL sources along the cut.For better visibility, the 70 μm and 750 MHz profiles are shifted vertically by adding 0.25 and 0.75, respectively.

Figure 5 .
Figure5.Integrated intensity maps of J = 2-1 transitions of CO (color) with contours of (left) 13 CO and (right) C 18 O.The CO map is integrated between 4 and 25 km s −1 and the 13 CO and C 18 O is integrated by 6-25 km s −1 .The contour levels (in units of K km s −1 ) of the 13 CO(2-1) map are 7-45 (in steps of 2) and of the C 18 O(2-1) map are 1.8-15.4(in steps of 0.8).The "+" shows the selected positions for which the spectra are shown in Figure6.The blue line with an arrowhead shows the direction along which position-velocity diagrams for the CO, 13 CO, and C 18 O emission were studied.

Figure 10 .
Figure 10.Variation of velocity and intensity along the length of the velocity features at 13.8 (left) and 17.3 km s −1 (right).Multiple sections of the filament are marked to facilitate discussion in the text.The vertical-dashed line shows the positions of the Hi-Gal sources and molecular core identified in the region.

Figures
Figures 12 and 13 have been discussed in detail in Sections 5.3 and 6.3 respectively.

Table 1
Details of Radio Observations with GMRT

Table 2
Far-infrared Sources in the G6.55-0.1 Region Detected by the Band-merged Hi-GAL Catalog a (Molinari et al. 2016 is from the ATLASGAL catalog.Elia et al. (2017)estimated dust temperature (T dust ; column 9), mass, and luminosity of sources by fitting the SEDs with graybodies.aEliaetal.(2017)bD250 is the 250 μm FWHM.At the distance of 3 kpc 1″ corresponds to 0.015 pc.c Source not present in the band-merged catalog(Elia et al. 2017) but present in the 70 μm Hi-GAL catalog(Molinari et al. 2016).