In this paper
AbstractIntroductionObservationsResultsSpectraEmission morphologies of CO, [C I] and [C II]Position-Velocity (PV) maps[C II] high-velocity (HV) emission and comparison to radio continuumAnalysisSpatial structure analysis via constrained diffusion decompositionMulti-Gaussian decomposition of [C II] , [C I] and COFlux recovery and kinematic association with CO peak componentsCorrelation analysis of the four speciesLine ratios and physical conditionsInside vs outside the [C II] bright regionK-Means clustering with ATLASGAL surface brightnessDiscussionStratification of the carbon phases in NGC 6334Kinematic signatures and dynamical (de-)couplingLine ratios and physical stratification[CI] and [CII] as dynamical tracer of flowsSummary and conclusionsAdditional plotsPosition velocity maps of [Cii] emission associated with Hii regionsFigures and tablesReferencesWorksPhysics and Astronomy
Multi-phase Carbon Structure and Dynamics in NGC 6334: Insights from CO, [C I], and [C II]
Abstract
We present high-angular and velocity resolution observations of the emission lines of [CII] at 158 micron, [CI] at 609 micron, and the 12CO and 13CO (J=3-2) transitions toward the central region of NGC 6334, providing a detailed multi-phase view of carbon in this massive star-forming complex. By combining morphology and spatial scales information, kinematics, line profiles, and line ratio analyses, we quantify the distribution and physical state of ionized and atomic carbon, and carbon in its molecular form in CO. [CII] emission is the most extended, permeates the mapped region and is dynamically complex, while 13CO is compact and quiescent, and 12CO and [CI] occupy intermediate regimes connecting dense regions to transitional diffuse layers. We interpret these emission distributions such that the brightest [CII] emission traces photodissociation regions (PDRs) close to the H II region bubbles located along the central filament, while the extended [CII] emission arises from the more diffuse warm gas, indicating either cloud formation or dispersal processes. Remarkably, the [CI] is more extended than the 13CO, but less than [CII] and may probe the gas velocity structure resulting from colliding flows or cloud-cloud collisions. Line ratios involving [CII], such as [CII]/[CI] or [CII]/CO are systematically high, revealing widespread diffuse emission and localized UV-illuminated surfaces. Line ratio analysis and K-means clustering of [CI]/13CO versus dust continuum (cold 870 micron and warm 70 micron) emission reveal stratified chemical structure that UV photons penetrate deeply into the clumpy ridge (Av~18), suggesting a porous PDR. These results together depict that NGC 6334 is a clumpy, feedback-dominated PDR ridge with coexisting carbon phases, where [CI] and [CII] also trace dynamical processing through multiple velocity components linked to cloud-cloud collisions.
Introduction
NGC 6334 is a star forming complex (of extent pc), harbouring a central dense filament ( pc length) and a large number of OB-type stars, H II regions and bubbles in the extended region Figure 1 (e.g., [47, 55, 65, 35, 41]). This complex is located at a distance of 1.7 kpc (e.g., [56]). The central dense filament contains sites of high-mass star formation in a sequence of evolutionary stages (e.g., [65]) and is proposed to undergo longitudinal collapse, possibly triggered by a past high-mass star forming event ([68, 55, 41]). NGC 6334 has been reported to be undergoing large scale cloud-cloud collision by Fukui et al. [22], supported by recent high-resolution observations of CO and CO that identified multiple velocity components and ‘bridge’ features ([41]). It has been proposed that the cloud-cloud interaction is happening over the past few million years triggering sequential star formation in the filamentary complex ([22]). Owing to the complex interplay of dense molecular cores, extended photon-dominated regions (PDRs), active feedback and dynamic interactions, NGC 6334 provides an ideal environment to investigate the origins of molecular, neutral, and ionized carbon emission.

Figure 1. Three color image of the NGC 6334 central filament. Red: , Green: , Blue: ATLASGAL 870 m emission. The H ii regions from Anderson et al. [3] are indicated by the dashed magenta circles. Six far-infrared (FIR) sources (I[N], I to V from left to right; for more details see Kraemer & Jackson [30]) located in the central filament are indicated by the black plus signs. White star symbols indicate the OB–type stars from Persi & Tapia [47]. The cyan and yellow rectangular boxes outline regions covered in the [C i] and [C ii] lines, respectively.
The molecular, neutral and ionized carbon1 probe different layers of molecular clouds and PDRs. The CO and CO lines trace the colder, denser molecular interior, with CO providing a more optically thin probe of the cold and dense molecular gas (e.g., [41]). Neutral carbon with its [C I] – line at GHz (609 m) is thought to probe both molecular and neutral medium tracing the transitional layer (e.g., [21, 16]). Ionized carbon with the [C ii] – line at GHz (158 m) is dominant in the warm, UV-illuminated outer PDR layers of molecular clouds (e.g., [64, 62, 37]). Disentangling these multiple origins is challenging, as [C ii] emission traces a wide range of physical conditions in PDRs, from H ii region interiors to moderately dense cloud surfaces to extended diffuse atomic envelopes (e.g., [48, 67, 32, 36, 57]).
High-resolution observations of CO, C, and C are therefore essential to isolate spatial and kinematic components, constrain excitation conditions, and build a coherent understanding of carbon chemistry and thermal balance in the interstellar medium. Our aim in this paper is to investigate the spatial and velocity structure of these tracers in the central NGC 6334 filament (see Figure 1) using high-resolution data from the SOFIA and APEX telescopes, convolved to a common angular resolution of 20 ( pc at 1.7 kpc). The NGC 6334 central region was selected in the SOFIA FEEDBACK legacy project (Schneider et al. [59]) because of its distinct environment as it harbors a large number of embedded and optical H ii regions in and around the central dense filament. NGC 6334 does not exhibit clearly defined large-scale expanding [C ii] shells (Faerber et al. [19]), that are commonly detected in H ii region bubbles such as Veil bubble in Orion, RCW120, RCW49, NGC7538, and RCW36 ([44, 66, 9, 13]).
This paper is organized as follows. Section 2 describes the observations and data reduction. Section 3 presents the data, including representative spectra, emission morphology, and position–velocity diagrams. In Section 4 we analyze the spatial structure using constrained diffusion decomposition, perform multi–Gaussian line fittings, examine line correlations, and derive key line ratios. Section 5 discusses the physical implications of these findings, and Section 6 summarizes our conclusions.
Observations
We mapped the extended region of NGC 6334 in the and molecular lines using the 7-pixel heterodyne array receiver, the Large APEX sub-Millimetre Array (LAsMA) (Güsten et al. [25]), on the 12-meter single-dish Atacama Pathfinder EXperiment (APEX) (Güsten et al. [26]). The data reduction procedure is outlined in Neupane et al. [41], where we presented the molecular line emission properties and velocity structure of the region, covering an area of 2.5 1.2 square degrees. The spectral cubes have a resolution of 0.25 km s, a typical rms of 0.40 K in one channel, and a full width at half maximum (FWHM) beam size of 20.
The [C i] observations at 492 GHz were conducted using the nFLASH460 instrument at the APEX telescope. We obtained On-The-Fly (OTF) maps covering the region of size 0.567 0.372 square degrees in the central NGC 6334 filament. While the native FWHM beam width of the data is 14, we sampled the data to a spatial resolution of 20 to match the resolution of the molecular CO data from the LAsMA observations. The spectra were smoothed to a resolution of 0.25 km s, with a typical rms level of 0.25 K in one channel.
The ionized carbon [C ii] observations at 1.9 THz (or 158 m) toward the central NGC 6334 filament were carried out using the upGREAT instrument (Risacher et al. [53]) on SOFIA (Stratospheric Observatory for Infrared Astronomy) as part of the FEEDBACK project (Schneider et al. [59]). Due to the shutdown of the observatory in 2022, only 36% of the planned area could be observed, covering approximately an area of square degrees. The observations were done as OTF maps with several coverages. For more observational details, see Schneider et al. [59]. The data reduction and smoothing were performed using the GILDAS² software package. The native FWHM beam width of the observations is . We resampled the spectra to a common resolution of km s, achieving a typical rms of K in antenna temperature, and smoothed the data to a spatial resolution of to match the resolution of the CO and [C I] line datasets mentioned above.
The mapped regions in [C I] (using nFLASH/APEX) and [C II] (using upGREAT/SOFIA) are indicated in Figure 1 by cyan and yellow rectangles, respectively. In this work, we study the central NGC 6334 region of size square degrees ( pc pc) where [C II] observations are available. To obtain the corrected main beam temperature () from the antenna temperature (), we used beam efficiencies of 0.67, 0.60, and 0.72 for upGREAT/SOFIA, nFLASH, and LAsMA/APEX observations, respectively. We also complement these observations with continuum images, primarily using ATLASGAL 870 m maps ([61]), Spitzer Space telescope ([20]) emission maps at 8 m ([7]) and at 70 m ([14]). For a quantitative comparison of the fluxes, these emission maps were smoothed to a common resolution of and re-gridded to a pixel size of .
Results
Spectra
Figure 2 presents the average spectra of the four transitions in the common area (Figure 1). The [C II] line is the brightest and exhibits pronounced high-velocity emission. The CO line is the second brightest, whereas CO and the [C I] line display comparable brightness temperatures and similar spectral profiles. Emission near km s is detected in CO and [C I], consistent with a local velocity component, characterizing the foreground interstellar medium (historically identified by Heeschen [27]; Riegel & Crutcher [51], and mapped globally by Dame et al. [17]).

Figure 2. Spectra of four observed species. Left: Average spectra toward the common mapped region. The systemic velocity range to km s is indicated by dashed black lines. Right: Average [C II] and [C II] -0 spectra toward the region indicated by the white dashed circle in Figure 3. Peak values of [C II] and [C II] intensity are 54.77 K and 1.03 K, respectively, and the rms in the average spectrum is 0.12 K.
In the same velocity range, [C II] shows an absorption feature that is likewise associated with the local emission. The right panel in the figure presents spectra of the [C II] hyperfine component -0. The region where this line is detected is indicated by a white dotted circle in Figure 3 (middle panel of [C II]) and is not particularly associated with the location of the H II regions. Outside of this area, [C II] was not or only weakly detected (see below). Since [C II] is widespread in velocity, the [C II] -1 component is blended with the [C II] emission and the [C II] -1 component is not detected as it has the lowest relative line strength. Relative line strengths of the lines are 0.250, 0.625 and 0.125, respectively, for the -0, 2-1 and 1-1 hyperfine lines ([43]). We estimate the optical depth of [C II] using the [C II] and [C II] -0 peak intensity ratio and obtain a value of , assuming C/C ratio of 67, e.g. [31, 40, 28]. In this area the [C II] line is thus optically thick and line profiles show indications of self-absorption, pointing toward temperature gradients along the line-of-sight ([24, 28]). Interestingly, in the regions where [C II] is very bright, the [C II] -0 is not detected and we cannot derive the opacity of the [C II] line (upper limit on for K). However, in the central ridge is mostly similar or higher than the value of the averaged spectrum, so that [C II] likely stays bright because of the high kinetic temperature. The line possibly remains optically thick in the central dense ridge and optically thin in the outskirts. This assumption is particularly valid if significant emission in [C II] is coupled with dense gas tracers in the inside region producing optically thick emissions. From the observed average main beam brightness temperature of K (Figure 2), we obtain a Rayleigh-Jeans corrected temperature of 93 K using,

Figure 3. Moment-0 maps of the observed four lines. The maps display the main velocity ranges found in NGC6334. Contour levels are at 10, 30, 50, 70, 90 percent of the peak emission. In the [C II] map, the region where the [C II] F=1-0 line is detected in an averaged spectrum is outlined by a white dashed circle. Green plus markers on the maps indicate far-infrared sources I to V [30] from left to right and white star symbols give the location of OB stars. Magenta dotted circles show WISE H II regions from Anderson et al. [3].
where and denote the Planck and the Boltzman constants, respectively, and is the frequency. This is the excitation temperature for the [C II] line for regions of densities well above the critical density, below which it is the lower limit to the kinetic temperature (). We assume here that the densities are above the critical density of [C II] for collisions with atomic and molecular hydrogen ( cm and - cm at K, respectively ([24]). [C II] emission however can also arise from electron collision, particularly relevant for regions inside the H II regions. Theoretical models suggest that the electron collision contribution in [C II] emission is largely defined by the electron density, the size of H II regions relative to the [C II] emitting PDR layer and the stellar continuum ([1]). We briefly come back to the electron collision contribution in § 3.2.
In various parts of the analysis section presented below, we separate regions inside and outside of the 55 K contour (Figure A.1) to investigate properties of [C II] bright vs less bright regions.
Emission morphologies of CO, [C I] and [C II]
To illustrate the PDR-dominated nature of the region, three color composite maps of [C I] (blue), [C II] (green), and dust emissions (red) are presented in Figure A.2. While 8 m traces PAH molecules excited by FUV radiation on cloud surfaces, 70 m emission highlights warm dust associated with internal heating in PDR interfaces and denser regions. The widespread [C II] is strongly correlated with both dust tracers on logarithmic scale (Pearson coefficient for 8 μm and for 70 μm, see Fig. A.3), confirming that it primarily arises from warm, FUV-illuminated atomic gas in the PDRs. Bright 8 μm emission appears as arc and rims toward the H II regions, whereas bright 70 μm emission follows the filamentary ridge. [C I] shows only moderate correlation with the dust tracers (Pearson coefficient for 8 μm and for 70 μm), but stands out near FIR-I and in the filamentary structure extending north-south toward FIR-V. The emission morphologies of all four species are examined in detail in the following.
We present moment 0 maps of the observed four species in Figure 3, separating the velocity ranges of −30 to −15 km s, −15 to +5 km s and +5 to +30 km s. The intermediate range of to km s represents the systemic velocity, as most CO emission is found within this velocity range [41]. The other velocity ranges indicate high-velocity features that are clearly seen in the [C II] and [C I] lines, though not always covering the same areas, and marginally in the CO line (Figure 2).
The CO emission is found to trace both dense and diffuse gas along the filament while CO emission mostly outlines the dense ridge regions (c.f. [41]). CO also outlines both, blue-shifted and red-shifted wing emissions, tracing outflowing gas toward some FIR sources.
Within the systemic velocity range, the [C I] emission distribution closely follows that of CO. In the blue-shifted interval ( to km s), this correspondence weakens, and [C I] traces more extended emission, particularly around the FIR sources III and IV and along the western edge of the map. The red-shifted interval shows less prominent extended [C I] emission, which is again confined to the western map boundary. In both high-velocity regimes, no spatial correspondence with [C II] emission is observed.
On the other hand, [C II] emission prevails in all three velocity ranges. However, in the blue velocity range ( to km s), [C II] is tracing patchy emission morphologies in and around the H II regions. Most [C II] emission is seen within the systemic velocity ranges, though the morphologies are quite different in comparison to all other tracers. At redder velocities ( to km s), [C II] shows extended emissions above the central filament where there are no known H II regions, while it shows bright and clumpy emission in and around the H II regions along the filament. We also identify some ring-like structures that become better visible in channel maps and position-velocity cuts shown in Figure A.4. As a preliminary interpretation, we propose that the [C II] emission beyond the systemic velocities is tracing infalling or expanding gas and large-scale flows as well as turbulent ionized gas due to feedback from OB stars and H II regions.
Comparing [C I] and CO emission further in the common region, we find that some seemingly extended regions exhibit [C I] emission above the level, while CO is below for the integrated intensity maps at the systemic velocity range to km s, and in the ratio maps in the wider velocity range to km s. We present the maps of this region in Fig. A.5. Certainly the emission excess in [C I] is not due to contamination from the local emission around km s. Regions where [C I] is detected but CO is below level, despite strong CO emission, likely correspond to low density or moderately shielded gas or [C I] is additionally probing a dynamical processed gas such as inflows or cloud-collisions.
The critical density of the CO transition (– cm at K) is an order of magnitude higher than that of the [C I] line (– cm). The critical densities are derived following Mangum & Shirley [38] where the ranges reflect consideration of the ortho- and para-H collision rate coefficients from LAMDA database [60]. In gas with density below this threshold, CO level populations are sub-thermally excited, and the line is faint. Additionally, [C I] is more abundant in atomic to molecular transition zones and remains emissive where CO is partially photodissociated or insufficiently excited [64]. The detection of [C I] emission that is correlated with CO but not with CO indicates a diffuse component in which [C I] has not yet been ionized and traces lower density quiescent gas that has not yet been converted into molecular gas in star-forming regions.
On the other hand, [C I] excess may indicate dynamically processed gas, potentially probing converging flows and cloud-cloud collisions, where shock compression, turbulence and partial CO dissociation together enhance relative abundance of atomic carbon (e.g. [16, 22, 10, 8]). To investigate the dynamical scenarios, we produced spectra averaged over of the four species toward selected positions (P1 to P6), where [C I] emission is above while CO is below , presented in the Fig. A.5. We identify multiple velocity components; a clear component in CO and [C I] is observed in the velocity range to km s (P1, P3, P6). The km s component is detected in 5 of 6 positions, with three of those (P4, P5, P6) showing detection in all tracers except CO. In one position (P2) the km s emission is seen in CO and [C I], while in another position (P3) it is seen only visible in [C II]. The km s component, corresponding to the main molecular cloud velocity, is clearly detected in all positions, together with km s in three positions. The coexistence of multiple velocity components and [C I] emission excess, with respect to the CO, indicates dynamical interaction zone, possibly tracing converging flows or cloud-cloud collisions.
In Figure 4, we further plot a 1.3 GHz MeerKAT radio continuum map [23] toward NGC 6334 and overlaid [C II] emission contours. The radio continuum emissions correlate nicely with the H II regions. Overall, the bright [C II] emission (contour at 400 K km s) also shows morphological similarity with the bright radio continuum emission region. This hints that the bright emission in [C II] at least partly results from the feedback due to embedded H II regions in the central filament. Fraction of [C II] emission may arise from collision with electrons inside the H II regions and not from the PDRs surrounding it. Deriving an accurate fraction of [C II] emission from collision with electrons is not straightforward. Observations of both [N II] and [C II] lines (e.g. Oberst et al. [42]) is one way to constrain this which is beyond the scope of this work. However, using the 1.3 GHz radio continuum emission we can make qualitative assessments. Following Ramaila et al. [50] (their equation 2) and assuming an electron temperature of K and using the 1.3 GHz flux density within the H II region radii (adopted from Anderson et al. [3]), we derive an rms electron density ranging from to 2000 cm toward the H II regions. Consequently, [C II] emission originating from within the H II region is thermalized. At these moderate densities, theoretical models predict 10 to 30% contribution of electron collision in the total [C II] emissions (Abel [1]), lower values for higher electron density. Comparable relative fractional contribution of ionized gas has been observed toward Carina Nebula (27%: Oberst et al. [42]), while 5–15% toward M43 region and other sub-regions in the Orion complex (Pabst et al. [45]).

Figure 4. MeerKAT 1.3 GHz radio continuum map overlayed with [C II] contours (at levels 200, 400 and 800 K km s) in white dotted lines. The rectangle plotted in red color indicates our mapped region. Also shown are the H II regions, OB stars and FIR sources as in Figure 3.
Position-Velocity (PV) maps
To evaluate the velocity extent of the emission, we first constructed global position-velocity (PV) maps of the four species, as presented in Figure 5. The PV maps clearly illustrate that [C II] and CO exhibit the most extended velocity distributions, while [C I] and CO are confined to narrower velocity ranges relative to the systemic velocity. However, [C I] emission is clearly broader than CO. Notably, [C II] displays prominent blue-shifted and red-shifted velocity extensions, whereas CO predominantly traces blue-shifted gas, including bipolar outflows emanating from dense sources (e.g., FIR sources IV and V) along the central ridge. In CO and [C I], we also observe local emission at +7 km s, while it is seen in absorption in the [C II] map. From these PV maps, together with the emission morphologies presented in Figure 3, it becomes obvious that [C II] emission is clearly distinct to the other tracers. Unlike many FEEDBACK survey targets (Schneider et al. [59]), the NGC 6334 region does not contain a single, spherical and dominant H II region, but harbors multiple embedded H II regions along the dense filament. We thus performed PV-cuts crossing the individual H II regions (Fig. B.1). We do not observe any clear elliptical [C II] emission distribution in the PV maps that are characteristic of fast ( km s) expanding [C II] bubbles (e.g., Luisi et al. [36]) or ring-like emission with low-velocity ( km s) expansion signatures (Dannhauer et al. [18]). The only example of a fragmented ring is seen for G351.170+00.704. We will come back to this point in the next section. The CO PV-maps show some high-velocity features due to outflow emission from protostellar sources. This emission arises from molecular clumps along the line-of-sight towards the H II regions.
!Position-Velocity maps of [C II], [C I], CO, and CO toward the central filament bbox=0.09,0.08,0.91,0.34
Figure 5. Position-Velocity (PV) maps of the four species toward the central filament. The slice is extracted along the Galactic longitude direction, centered at , spanning to , with a slice width of . Contour levels on each panel are at , , , , and . Vertical black dotted lines mark the Galactic longitudes of FIR sources.
[C II] high-velocity (HV) emission and comparison to radio continuum
Average spectra toward the five H II regions (Figure 4), that are mostly covered in [C II] emission maps, are shown in Figure 6. The line emission profiles, in particular in [C II] emission, reveal broad velocity ranges, with significant high-velocity (HV) components extending up to km s beyond the systemic velocity ( to km s in most cases (see Figure 6). All regions show HV red wing emissions. Except for G351.170+00.704 and G351.246+00.673, all regions also show HV blue wing emissions. The presence of HV wing emission [C II] in the average spectra is not found for the other species, even with CO emission which is known to trace HV molecular outflows. This alone already indicates that [C II] is tracing highly dynamic ionized gas toward these H II regions.

Figure 6. Average spectra toward H II regions. Dotted blue lines indicate the systemic velocity range of to km s toward NGC 6334.
To investigate the spatial distribution of the HV gas traced by [C II], we constructed moment-0 maps for the blue ( to km s) and red ( to km s) wings (Figure 7). We overlaid the contours of the MeerKAT 1.3 GHz radio continuum emission on top of these maps. The emission morphologies reveal varying patterns across the five H II regions that we briefly summarize in the following.
!Maps of HV blue and red wing emission in [C II] toward H II regions bbox=0.09,0.31,0.94,0.52
Figure 7. Maps of HV blue ( to km s) and red ( to km s) wing emission in [C II] toward H II regions (top and bottom panels, respectively). The radio continuum contours in white color are overlayed on both red and blue emission maps at contour levels [0.01, 0.03, 0.05, 0.075, 0.1, 0.15] Jy/beam. The cyan dashed circles indicate the H II regions, names labeled on top panels.
G351.170+00.704 exhibits filled radio emission without compact structure. The blue HV emission matches arc-like emission along the edge of the H II region, at least on one side. The red HV emission is filled and diffuse, but brighter than the blue HV emission.
G351.246+00.673 also reveals a bipolar structure in radio continuum (though more narrow compared to G351.311+00.663). The HV blue wing emission in [C II] is faint and filled, with some signatures of extended emission beyond the H II region radii. Smaller, clumpy emission in blue seems to correlate with the bipolar emission seen in radio. The HV red wing emission is mostly filled within the H II region radius, with diffuse extensions perpendicular to the bipolar radio emission.
G351.311+00.663 exhibits bipolar radio continuum emission. The HV blue emission in [C II] is more diffuse and filled, while red emission shows bright arc-like emission at the edge of the H II region radius. This hints at an asymmetric PDR structure in red-shifted gas toward the H II region.
G351.367+00.640 displays bright and filled radio emission. The blue HV emission shows two compact bright spots inside the edges of the H II region radii. The red HV emission is clumpy and brighter, possibly indicating a partial shell of the H II region, and extends outside the radius.
G351.383+00.737 exhibits diffuse and irregular radio continuum emission. Blue HV emission in [C II] is irregular, faint, and mostly diffuse. Red HV emission is relatively brighter (particularly around the bottom edge of the H II region) and exhibits diffuse enhancements beyond the radius and corresponds well with the compact radio continuum structures.
Summarizing, we find hints of arc-like HV blue emission only toward the H II region G351.170+00.704 and red HV emission is generally brighter than the blue one. In G351.311+00.663 and G351.367+00.640, HV red emission forms irregular or ring-like structures, tracing the edges of the H II regions. Overall, [C II] emission does not seem to trace coherent ring-like structures or expanding bubbles associated with the H II regions as was seen in many other studies (Luisi et al. [36]; Tiwari et al. [66]; Bonne et al. [13]; Keilmann et al. [29]; Dannhauer et al. [18]) and as we already outlined in Sect. 3.3 for the PV-maps. Instead, it traces asymmetric and irregular diffuse enhancements, predominantly in HV red wing emissions, suggesting complex geometry and feedback interactions rather than coherent symmetric expansions.
The finding that HV [C II] emission is not always correlated with expanding bubbles of fragmented rings (Dannhauer et al. [18]), that are driven by the radiation and winds of the central OB-stars of the H II regions, points toward a different interpretation of the emission. As proposed by Bonne et al. [12], it may indicate molecular cloud dispersal, in which the PDR surfaces of the associated molecular clouds are eroded by stellar feedback and give rise to HV [C II] emission. The molecular cloud is clearly highly filamentary and fragmented [5, 41, 4], so that holes and channels allow radiation and winds to impact the cloud.
Analysis
In this section, we investigate the spatial structure of the emissions of different forms of carbon, perform quantitative analysis of velocity components, and analyze the line ratios from the integrated intensities. For the analysis based on multi-Gaussian fitting and line ratios we only use the common region where all data is available (box size centered at and , see Figure 1). To understand the correlation with the dust properties we make use of cold dust emission in ATLASGAL 870 m maps and warm/hot emission from Spitzer 8 m maps from the archive.
Spatial structure analysis via constrained diffusion decomposition
To investigate the spatial structure of the emission from different gas tracers, we applied the constrained diffusion decomposition (CDD) model [34] to the moment-0 maps. CDD decomposes the integrated intensity maps by solving modified nonlinear diffusion equation that separates structures into different scales (characteristic sizes pixels) while preserving positivity and avoiding negative artifacts common in wavelet based decomposition methods. The cumulative flux recovery at each scale quantifies the contribution of structure of different sizes and the residual fraction indicates unresolved or truly diffuse emissions.
We applied CDD to the moment-0 maps (pixel scales of or equivalent physical size of pc) in two velocity ranges: a broader range covering the whole velocity space ( to km s) and a more narrow range of the systemic velocity ( to km s). The objective is to investigate if inclusion of high velocity emissions, such as outflows or expanding gas, gives a different outcome from the analysis. The results, visualized in Fig. C.1, illustrate the flux recovery fraction at different scales ( pixels, to ). Table 1 summarizes the results.
| Tracer | \multicolumn{4}{c}{Flux Recovery} | |||
| [−30, +30] km s | [−15, +5] km s | |||
| (pixels) | Scale (pc) | (pixels) | Scale (pc) | |
| CO | 2.82 | ∼0.35 | 2.89 | ∼0.37 |
| CO | 3.42 | ∼0.53 | 3.36 | ∼0.51 |
| [C I] | 3.56 | ∼0.58 | 3.48 | ∼0.55 |
| [C II] | 4.87 | ∼1.47 | 4.96 | ∼1.53 |
Table 1. Flux recovery scales (50%) for different tracers.
The molecular tracers CO, CO and [C I] consistently exhibit compact structures on – pc with residual fractions ranging from –, indicating small variation between different velocity ranges. This illustrates that their emission is predominantly coming from dense cores and envelopes. In contrast, [C II] requires pc scales for 50% flux recovery, with high residual fractions (–) in both velocity ranges. The CDD model of broader emission ( to km s) is only marginally different with respect to the narrower velocity range (−15 to +5 km s), demonstrating that [C II] traces extended, diffuse emission on cloud wide scales, even at systemic velocities, reflecting widespread UV illumination throughout the complex.
Multi-Gaussian decomposition of [C II] , [C I] and CO
We performed multi-Gaussian fitting to the [C II] , CO, and [C I] spectral lines using gausspy+ algorithm Riener et al. [52] to derive the central velocity (km s), the velocity dispersion (km s) and the amplitude (K). Table Table C.1 presents a summary of the statistics from the Gaussian fit results toward the common observational region and Fig. C.2(a) presents velocity probability density functions (PDFs) of the Gaussian components for each species. The velocity PDF clearly illustrates that [C II] spans a wider velocity range (standard deviation of 5.45 km s) than [C I] or CO, though their centroid velocities are comparable.
[C II] exhibits the broadest velocity dispersions and highest mean brightness temperatures. The broad dispersion, even from the Gaussian fitted individual components, suggests that it is tracing turbulently mixed ionized gas or directed flows, arising from photoevaporation of gas from PDR clump surfaces or wind-driven gas dynamic. In contrast, CO and [C I] show narrower line profiles with generally negative velocity skewness, consistent with denser, comparatively quiescent molecular gas possibly also affected by infall motions. Elevated kurtosis values for all three tracers point to multiple overlapping components and complex kinematics within the NGC 6334 central ridge.
Inside the bright [C II] region ( K, Fig. A.1), [C II] components are stronger and often require multiple Gaussian per line of sight, supporting moderate optical thickness or multiple velocity layers along the line of sight. CO and [C I] lines in this region are also slightly brighter and narrower than the map average, reflecting denser, more compact molecular gas. Skewness and kurtosis are enhanced, especially for [C II] and CO, highlighting the increased complexity and turbulence within the central ridge. Outside the bright region ( K), all three lines show weaker and somewhat simpler profiles. [C II] retains a similar velocity width but with fewer bright peaks, tracing more diffuse ionized gas, while CO and [C I] exhibit slightly broader but less skewed distributions, consistent with relatively quiescent molecular environments.
Comparison of linewidth distribution of [C II] toward the H II regions and ambient cloud environment is presented in Fig. C.3. The ambient region is chosen at and westward to the FIR-V, with dynamic size matching the size of each H II regions. The bulk [C II] linewidth distribution toward H II regions remains comparable to the ambient cloud environment, but H II regions exhibit a systematically enhanced high-linewidth tail, indicating localized feedback-driven dynamical broadening at irradiated cloud interfaces.
Flux recovery and kinematic association with CO peak components
To understand kinematic association of the different tracers, we analyzed the Gaussian fit parameters by selecting the peak CO component in each pixel and matching it with the nearest [C I] and [C II] components in velocity space. The aim here is to disentangle velocity coupled gas traced by each species. Using a strict velocity tolerance of ±0.75 km s (3 channels), only a limited match was achieved, with flux recovery fractions of 75% for CO, 44% for [C I] , and 21% for [C II] , calculated by comparing Gaussian area maps to moment-0 maps. To better capture emission dynamically coupled with molecular gas, we adopted a wider velocity tolerance of ±1.5 km s (6 channels) and selected the brightest [C I] and [C II] components where multiple were present.
This broader matching improved both spatial coverage and flux recovery reaching 77% for CO, 62% for [C I] , and 44% for [C II] . The line width distributions of the velocity coupled gas show that CO and [C I] share similar velocity dispersions (Fig. C.2(b)), while [C II] remains broader, consistent with the results from the full component analysis (Table C.1). These trends confirm that [C I] closely follows the molecular gas traced by CO, while [C II] partially overlaps but exhibits more turbulent components.
Even with the wider tolerance, [C II] recovers a smaller fraction of the total flux (21% to 44%) and indicates that about one third of the [C II] emission is kinematically linked to molecular gas. However, its lower recovery compared to [C I] and CO and its broader line widths suggest that much of the [C II] emission arises from more diffuse or extended ionized layers, consistent with the predictions from the CDD model (see § 4.1). Optical depth effects may contribute but likely play a secondary role.
Correlation analysis of the four species
To investigate how the intensities of CO, CO, [C II] , and [C I] relate across NGC 6334, we conducted a pixel-by-pixel correlation analysis using their moment-0 maps (pairwise threshold above ) toward the common region. This approach allows us to examine how closely the different tracers probe common gas phases or distinct environments. Though CO is optically thick for most of the map and [C II] is likely optically thick toward the brightest regions, we however perform a line ratio analysis from the integrated intensities. Accordingly, this demands caution in interpreting the results.
We evaluated three cases: (1) the full map (i.e., the common mapped region), (2) the region inside the [C II] K contour, and (3) the region outside of this contour (see Fig. A.1).
First, Pearson correlation coefficients were computed directly from the moment-0 maps, as shown in Figure 8. We also calculated correlation coefficients using standardized and arcsinh transformed data, which reduce the influence of heavy-tailed intensity distributions. The resulting coefficients differ only slightly, indicating that the observed correlations are robust against non-Gaussian intensity effects.

Figure 8. Heat map showing correlation matrices (Pearson coefficients) between the pair of species from moment-0 maps () toward the common region.
Across the full map, CO, CO, and [C I] show very strong correlations (), reflecting their common origin in the molecular gas reservoir. [C II] exhibits more moderate correlations with CO () and CO (), and a weaker link with [C I] ().
Within the bright [C II] region ( K), the molecular tracers remain tightly correlated (), while correlations between [C II] and the molecular lines weaken ().
This indicates that although [C II] emission is intense here, a significant fraction originates from gas not directly coupled to the dense molecular component, likely from PDR surfaces or partially ionized layers. This supports our earlier result that [C II] exhibits broad velocity dispersions and requires multiple Gaussian components to reproduce its line profiles.
Outside the 55 K contour, molecular tracers remain extremely well correlated (–), whereas [C II] becomes largely decoupled (–). This contrast reinforces the dual nature of [C II] emission: it partly traces dense PDR gas within the bright ridge but is increasingly dominated by diffuse ionized material in fainter regions.
Line ratios and physical conditions
We analyzed the pixel-by-pixel line ratios of [C II], [C I], CO, and CO to explore spatial variations in the physical conditions. Ratios were computed from moment-0 maps for pixels above in both species. Figure 9 shows the distributions of the line ratios, and Table C.2 summarizes their statistics.

Figure 9. Line ratios of the species shown in violin plots. Left: full map, middle: inside vs outside, right: Cluster1 vs Cluster2. The black box and white horizontal line inside the violins (left panel) indicate the interquartile range and median value, respectively. In the split-violin plots (middle and right panels), the internal horizontal lines display the individual quartiles and medians for each group.
Across the common mapped region, ratios involving [C II] are systematically higher and more broadly distributed than those among the molecular tracers. The [C II]/CO and [C II]/CO ratios show extended high ratio tails, reflecting bright [C II] emission relative to CO in regions of diffuse or ionized gas. The [C II]/[C I] ratio is similarly large but with a more narrow spread, indicating partial spatial and physical overlap between the ionized and atomic carbon layers. Ratios between molecular tracers, particularly [C I]/CO and CO/CO, remain more uniform, consistent with their shared molecular origin.
We note that opacity correction in [C II] can only increase the ratio values in [C II]/CO and [C II]/[C I]. Since both, [C I] and CO are likely optically thin, opacity effects in these ratios are negligible to minor. Considering only velocity coupled components, the overall trends derived from moment zero based line ratios are preserved (Table C.4), including the relative ordering of the ratios and their positive skewness. Thus, the line ratio analysis from the integrated intensities is valid to understand their general trends in dense and diffuse regions while derivation of physical parameters based on these may lead to larger uncertainties.
Inside vs outside the [C II] bright region
To examine environmental variations, we investigate in the following regions inside and outside the [C II] intensity contour at K. Inside the 55 K contour, the mean [C II]/CO and [C II]/CO ratios ( and , respectively) are slightly lower than the full map averages, while molecular ratios such as [C I]/CO and CO/CO remain nearly constant ( and ) (see Table C.2). This indicates that, although [C II] is bright, its emission here arises from a mixture of dense and diffuse UV illuminated PDR gas.
Outside the contour, mean [C II] ratios increase modestly ([C II]/CO , [C II]/CO ), reflecting a higher relative contribution of diffuse [C II] to fainter CO. In contrast, [C I]/CO rises from inside to outside, while CO/CO grows from to . These trends point to enhanced atomic carbon in the more diffuse outer regions, consistent with the morphological trends discussed in § 3.2.
K-Means clustering with ATLASGAL surface brightness
To complement the contour based analysis, we examined the [C I]/CO ratio as a function of ATLASGAL 870 m surface brightness (Figure 10, left panel). Using pixels above in both lines, we applied K-means clustering (using the scikit-learn package [46]), which separated the data into two populations: a low-brightness cluster (Cluster 1) and a high-brightness cluster (Cluster 2), with 870 m brightness transition near 1.6–1.8 Jy beam.

Figure 10. [C ii] /CO and [C i] /CO ratio as a function of ATLASGAL 870 m brightness in a log-log scale. The blue and red color data points are Cluster 1 and Cluster 2 from K-means clustering. The black crosses represent mean positions of each clusters.
Cluster 1 shows systematically higher ratios involving both [C II] and [C I] (e.g., [C II]/CO , [C I]/CO ) (Table C.3), suggesting that carbon is more readily in ionized or atomic form in lower density and more UV exposed environments. Cluster 2, coinciding with the bright [C II] region, exhibits lower ratios ([C II]/CO , [C I]/CO ) characteristic of predominantly molecular gas. Morphologically, the contour of [C I]/CO closely follows the T_{\rm peak}(\[C II\]) = 55 K boundary, except toward the western ridge near FIR V (Fig. C.6).
We investigated the [C II] to [C I] ratio of the two clusters versus the 8 m brightness, presented in Figure 11. While the mean value of 8 m flux for Cluster 1 is about half the value of Cluster 2, [C II] to [C I] ratio is systematically higher in Cluster 1 for ranges of 8 m brightness. This suggests diffuse regions favor [C II] dominated warm PDR layers with weaker PAH excitation while denser regions show relative dominance of [C I] despite stronger PAH emission from the exteriors revealing density stratified chemical structure.

Figure 11. [C ii] to [C i] ratios with respect to the 8 m flux. The blue and red color data points represent Cluster 1 (low ATLASGAL brightness) and Cluster 2 (high ATLASGAL brightness), respectively. Cyan and magenta contours show kernel density estimate (KDE) iso-density levels for the two clusters.
We further performed similar clustering analysis of the [C I] to CO ratio with respect to the H column density (from PPMAP, Marsh et al. [39]) and 70 m emission (Carey et al. [14]). It turns out that the transitional column density between the high-brightness and low-brightness clusters is approximately at cm (; derived following Bohlin et al. [11]), a consistent result from all four methods investigated here (Fig. C.6). At this transitional layer, the corresponding dust temperature is approximately 28 K (also from PPMAP, Marsh et al. [39]). This depth is well within the dense gravity dominated column density regime in molecular clouds (e.g., Schneider et al. [58]). These results imply that a significant fraction of carbon may reside in “CO-dark” molecular gas in NGC 6334. Models of molecular cloud formation via colliding flows also predict that “CO-dark” gas may dominate the early gas for substantial fraction of their assembly history (e.g.,
Clark et al. [15]. Additionally, the deep penetration of the UV field may suggest a porous and inhomogeneous cloud geometry.
Discussion
Most of PDR theories and models (e.g., Tielens & Hollenbach [64]; Le Petit et al. [33]; Röllig & Ossenkopf-Okada [54]; Pound & Wolfire [49]) consider a one dimensional layered geometry for UV-irradiated molecular clouds in which the spatial distribution of [C ii] , [C i] , and CO emission are interpreted. These tracers are expected to probe different chemical and thermal layers as an effect of progressive attenuation of the radiation field into the cloud interior. In this simplified picture, C is expected to dominate the outermost layer, C traces the intermediate layer, and CO is found in the denser interior part. The [C ii] emitting surface layers arise from regions of –3, while [C i] and CO originate from the regions where the gas becomes shielded from the incident FUV radiation. However, molecular clouds exhibit more complex geometry, and assessing the validity of this stratified picture requires multi-wavelength data sets at high angular and velocity resolution. In this work, we obtained state of art data set for each species toward NGC 6334 central region. We discuss our findings in the following.
Stratification of the carbon phases in NGC 6334
The morphology of the emission distribution (Figure 3) highlights clear differences among the tracers (see Sect. 3.2). [C ii] is widespread and extends far beyond the dense ridge, while CO is compact and confined to the dense ridges. CO and [C i] occupy an intermediate regime, tracing both dense regions and surrounding envelopes. This hierarchical distribution is further supported by the spatial scale analysis (Sect. 4.1), where both CO, and [C I] recover a significant fraction of their emission on scales of ∼0.3–0.5 pc, characteristic of dense clumps and filamentary interfaces, whereas [C II] remains dominated by larger-scale emission. Thus, at first order, the overall large scale picture is consistent with classical stratified carbon phases, in which carbon transitions from ionized to atomic to molecular form with increasing shielding and density.
Looking into the details, we observe the spatial distributions of the tracers with significant overlap, with [C II] extending far beyond classical surface layers and [C I] bridging both diffuse and dense PDR gas components. A significant fraction of the [C II] emission arises from large scale (≥ 1 pc), diffuse PDR gas that is dynamically and spatially decoupled from the molecular phase, as only ∼20–40% of the [C II] emission is velocity coupled to the dense components (Sect. 4.2.1). The positive correlation with radio continuum emission and [C II] bright regions associated with the H II regions suggest that significant [C II] emission arises from local feedback. These results together indicate that, while a global stratification of carbon phases is preserved, the carbon phases are also locally mixed and distributed within NGC 6334 by density inhomogeneities, stellar feedback, and possible inflows or cloud-cloud collision.
Kinematic signatures and dynamical (de-)coupling
The velocity structure of the carbon tracers in NGC 6334 reveals a clear dynamical differentiation among the phases (see Sect. 3.3). [C II] exhibits the broadest velocity spread and often requires multiple components to fit its line profiles, indicative of multi-layered, turbulent, inflows or feedback-processed gas. shows second broadest emission, including high-velocity wings associated with some of the FIR sources located in the dense filamentary ridge. In contrast, remains narrow and confined, and [C I] displays an intermediate behavior, sometimes showing blue-shifted emission not present in but seen in . Median linewidth of [C I] in logarithmic ATLASGAL 870 µm bins show larger values in comparison to for brightness bins up to ≲5 Jy/beam, that is more prominently observed in the 95th percentile (P95) values. At higher brightness bins both tracers show comparable median linewidth while exhibits broader P95 values (see Fig. C.4). The linewidth distribution presented in Fig. C.5 further illustrate the trend for outside vs inside regions. The high-velocity [C II] emission does not spatially match with the outflow features in the radio continuum, implying its minor role, whereas the positive association of [C II] with overall radio continuum emission morphologies around H II regions points to stellar feedback as the dominant driver. Gaussian decomposition confirms that [C II] has significantly broader line widths than and [C I], even when only velocity-coupled components are considered, reflecting a combination of high turbulence and multiple unresolved line-of-sight layers, including the possible existence of converging flows, components related to cloud-cloud collisions, and components related to the cloud dispersal from the local feedback. Overall, these results indicate that [C II] traces dynamically complex, diffuse PDR gas shaped by large scale feedback and flows, whereas CO and [C I] probe progressively denser and quiescent structures, highlighting a pronounced dynamical decoupling between the ionized, atomic, and molecular carbon phases.
Line ratios and physical stratification
The line ratio and clustering analyses (Sect. 4.4) highlight the chemical differentiation across NGC 6334, complementing the spatial stratification revealed by the morphology. The [C I]/ ratio increases outward from the dense ridge to the diffuse envelope, showing that atomic carbon extends beyond molecular ridges and clumps. Ratios involving [C II] are systematically higher and positively skewed, with enhanced values outside the dense central ridge where diffuse emission dominates (see Figure 11). The transition in [C I]/ around H column densities of ∼1.7 × 10 cm ( ∼ 18) indicates that FUV photons penetrate deep into the cloud along cavities and clump interfaces, enhancing atomic carbon abundance in the PDR interiors (Sect. 4.4). Taken together with the spatial and kinematic results, these chemical diagnostics reinforce a coherent picture of NGC 6334 as a globally stratified but locally complex medium. The stratification is preserved on large scales, while turbulence, flows, stellar feedback, and cloud porosity mix the layers locally, producing chemically distinct but dynamically (de-)coupled phases.
[CI] and [CII] as dynamical tracer of flows
In addition to the stratified and chemically differentiated structure, the carbon tracers reveal signatures of dynamically interacting gas in NGC 6334 (see § 3.2). We observe regions that exhibit [C I] emission above 3σ where is below the noise level (see Fig. A.5), despite strong emission. These could indicate low density or moderately shielded gas in which [C I] remains emissive while is sub-thermally excited or partially photo-dissociated. However, we find multiple velocity components in the averaged spectra (see Fig. A.5), particularly around −20 to −25 km s and −9 km s in addition to the systemic velocity of ∼ −4 km s, suggesting overlapping gas velocity components. In some regions, the −9 km s component is visible only in [C II], while in most regions it is clearly identifiable in and [C I].
This highlights that [C I] and [C II] can also be tracing atomic layers that are dynamically distinct. Such features are consistent with predictions of converging flows or cloud–cloud collisions, where shock compression, turbulence, and partial CO dissociation together enhance the atomic carbon abundance ([15, 16, 22]). NGC 6334 has been reported to be undergoing large scale cloud-cloud collision by Fukui et al. [22]. In the extended NGC 6334 filamentary complex (extent of ∼ 80 pc), we identified multiple velocity components ([25, −15] km s feature only in and −9.2 and −3.9 km s components in ) and ‘bridge’ features in velocity ranges −12 to −7 km s associated with the cloud-cloud collision (Neupane et al. [41]). Collision time scale is about 0.4 Myr, considering 10 pc spatial scale of the central filament (c.f. Fukui et al. [22]). Time dependent chemical models suggest that full conversion of CI to CO requires time scale of the order of Myr (e.g., [63, 2]). In young cloud like NGC 6334, the dynamical interaction can dominate the observed chemical behavior of the different carbon phases, particularly toward the interaction regions. Together, these observations imply that a fraction of the [C I] and [C II] emission probes dynamically processed regions in which dense and diffuse layers interact. Thus these tracers can also be effective probes of kinematic features in converging flows or cloud–cloud collisions.
Summary and conclusions
We used high-resolution observations of the three primary carbon phases, ionized carbon (C), neutral carbon (C), and carbon monoxide (CO and CO 3→2), to investigate the multi-phase structure of the NGC 6334 central ridge and its environment. By combining morphological, kinematic, spatial scale, and line ratio analyses, we examined how ionized, atomic, and molecular gas coexist and interact under the influence of strong stellar feedback in the region.
– Our observations show that ionized, atomic and molecular gas co-exist in the NGC 6334 central ridge, however, they are only partially coupled. [C ii] is most extended and spread beyond the dense molecular ridge, tracing both warm PDR surfaces (well correlated with 8 and 70 μm continuum) and feedback from H ii regions (correlated with radio continuum). CO is compact and confined to the dense ridge while CO and [C i] occupy an intermediate regime. CO additionally traces outflowing material linked to protostellar activity toward FIR sources in the central ridge.
– [C ii] shows broadest velocity coverage that includes diffuse PDR to the feedback driven components around the H ii regions. Only 20-40% of the [C ii] is velocity coupled to CO suggesting significant gas (de-)coupling. CO and [C i] stay similar except that [C i] is slightly extended in morphology more in bluer velocities with respect to systemic velocity and shows broader high line widths tail both in the dense and diffuse regions.
– Spatially [C ii] remains largely diffuse and does not reach total flux recovery of 50% even at scales of ∼1.5 pc, while CO and [C i] recover half their total flux on 0.3–0.5 pc scales (clumps/filaments). These results indicate [C ii] traces predominantly the larger scale diffuse PDR gas while CO and [C i] are associated with progressively denser, more localized regions. H ii regions contribute up to 30% of the local [C ii] emission tracing the ionized gas.
– CO and [C i] are generally described by narrower components, while [C ii] often exhibits multiple and broader components highlighting the kinematic complexity of the region. Comparison of [C ii] line widths toward H ii regions versus ambient medium further highlights the thermal plus dynamical broadening of the line due to local feedback.
– The line ratios [C ii] /[C i] and [C ii] /CO are higher and skewed, signaling extended UV-irradiated diffuse gas. [C i] /CO on the other hand shows a transition around H column densities of ∼1.7 × 10 cm ( ∼ 18) deeper into the cloud separating the dense ridge (moderate ratios) from the outer diffuse envelope (higher ratios ≥1), indicating deep FUV penetration and a porous cloud geometry.
– We observe [C i] and [C ii] also tracing distinct velocity components as in CO (∼ −20, −9, and systemic −4 km s) that are associated with cloud-cloud collisions and conclude that [C i] and [C ii] also probe dynamically processed gas.
Together, these results depict the NGC 6334 central ridge as a strongly layered, feedback influenced photon-dominated region. Dense molecular gas traced by CO, CO, and [C i] is embedded within a widespread, dynamically complex envelope of diffuse [C ii] emission that is only partially coupled to the molecular gas. FUV radiation penetrates deeply into the molecular ridge, enhancing atomic and ionized carbon even at high column densities. This implies that a significant fraction of carbon may reside in ‘CO-dark’ gas, shaped by feedback and the porous, clumpy cloud geometry. Overall, this study highlights that NGC 6334 is a complex, prototypical example of a clumpy, multi-phase PDR in which stellar feedback governs chemical layering, energy balance, and dynamical structure and evolution of the molecular cloud.
Acknowledgements. This study was based on observations made with the NASA/DLR Stratospheric Observatory for Infrared Astronomy (SOFIA). SOFIA is jointly operated by the Universities Space Research Association Inc. (USRA), under NASA contract NNA17BF53C and the Deutsches SOFIA Institut (DSI), under DLR contract 50 OK 0901 to the University of Stuttgart. upGREAT is a development by the MPlfR and the KOSMA/University Cologne, in cooperation with the DLR Institut für Optische Sensorsysteme. This work was carried out in part at the JPL, California Institute of Technology, under contract with NASA.
Financial support for FEEDBACK at the University of Maryland was provided by NASA through award SOF070077 issued by USRA. The FEEDBACK project was supported by the BMBI via DLR, project number 500R2217.
This work was supported by the CRC 1601 (SFB 1601 sub-projects B1, B2) funded by the Deutsche Forschungsgemeinschaft (DFG) – 500700252 and by DYNAVERSE. The CO and [C I] data were collected under the Atacama Pathfinder Experiment (APEX) Project, led by the Max Planck Institute for Radio Astronomy. S.N. also acknowledges continuous support from the APEX staffs and telescope operators during the observations. G.G. gratefully acknowledges support by the ANID BASAL project FB210003.
This work made use of Astropy (https://www.astropy.org) a community-developed core Python package and an ecosystem of tools and resources for astronomy [6].
Additional plots
!Peak emission map of [C II] bbox=0.27,0.11,0.73,0.36
Figure A.1. [C II] peak emission map. The contour is drawn at of 55 K.

Figure A.2. Three color images of the NGC 6334 central filament. Left: 8 m (red), [C II] (green), [C I] (blue). Right: 70 m (red), [C II] (green), [C I] (blue). The symbols and markers are same as in Figure 1, except that the OB–type stars are shown in white color.

Figure A.3. [C II] versus dust correlations (left: with 8 m, right: with 70 m).
!Channel maps of [C II] emission across velocity channels bbox=0.17,0.07,0.89,0.87
Figure A.4. Channel maps of [C II] emission from to km s with a step of km s. Other symbols are same as in Figure 3.

Figure A.5. Top: Map of [C I] integrated intensity (K km s) for pixels where C I and CO , in velocity ranges -15 to +5 km s. Middle: C I/CO ratio map in velocity ranges -20 to +20 km s. Contour level at the ratio 1 is shown in black color and mask region where C I and CO (created from the integrated intensity maps) is shown by the black dotted contour lines. Bottom: Average spectra toward regions P1 to P6 indicated by green circles (of size 60′′) in the top and middle panel. Black dotted lines indicate two main velocity components (-3.9 and -9.2 km s) of CO toward the region [41]
Position velocity maps of [Cii] emission associated with Hii regions

Figure B.1. Position-velocity maps along different angles passing through the center of the H II regions. On each figure, left panels show moment-0 maps of and [C II]. On the right panels, PV-maps are constructed along the path shown in the moment-0 maps. Contours are drawn at noise level on each maps. Movies of the PV-maps for rotating angular slices are available in the online versions.
Figures and tables

Figure C.1. Flux recovery at different scales obtained from the constrained diffusion decomposition (CDD) model. The moment-0 maps were used for the velocity ranges -30 to +30 km s (left) and -15 to +5 km s (right). Residual fractions are also noted for each species.
Figure C.2. (a) Velocity centroid PDF made using the Gauss fit components. (b) Line widths distribution of the three species from Gauss fit peak component analysis. The color lines over plotted on the histograms are kernel density estimation (KDE) lines.

Figure C.3. Linewidth comparison of H ii regions (in red) with respect to ambient region (in blue). The ambient region is centered at and westward to the FIR-V, with dynamical size matching the size of each H ii regions. Though average values of the linewidth are similar to ambient region, all H ii regions exhibit high linewidth tail.

Figure C.4. [C I] and linewidth in logarithmic ATLASGAL 870 brightness bins. P95 indicate 95th percentile linewidth for each bins.

Figure C.5. [C I] and linewidth distribution for Gauss components in different velocity ranges toward Outside vs Inside region.
!ATLASGAL map with [C II] intensity contours bbox=0.11,0.57,0.60,0.81
Figure C.6. ATLASGAL map overlayed with [C II] intensity contour at 55 K showing the bright [C II] region. High flux cluster boundaries based on N(H), ATLASGAL, and 70 emission are shown in red, yellow and blue color, respectively.
| Tracer | Parameter | Mean | Median | Std. Dev. | Skewness | Kurtosis |
| Full Map | ||||||
| [C II] | (km s) | -3.47 | -3.45 | 5.45 | 0.40 | 1.19 |
| (K) | 18.10 | 13.21 | 15.55 | 1.84 | 4.23 | |
| (km s) | 5.42 | 4.22 | 4.22 | 2.45 | 8.02 | |
| CO | (km s) | -4.03 | -4.01 | 2.80 | -1.97 | 15.11 |
| (K) | 8.71 | 6.96 | 6.41 | 1.39 | 2.10 | |
| (km s) | 2.60 | 2.42 | 1.19 | 1.64 | 9.33 | |
| [C I] | (km s) | -3.94 | -4.09 | 3.61 | -1.57 | 19.65 |
| (K) | 5.17 | 3.80 | 4.06 | 1.68 | 3.14 | |
| (km s) | 3.06 | 2.73 | 1.73 | 1.20 | 2.32 | |
| Inside Region ([C II] K) | ||||||
| [C II] | (km s) | -1.92 | -2.41 | 6.07 | 0.21 | 0.56 |
| (K) | 24.01 | 16.66 | 20.78 | 1.14 | 0.93 | |
| (km s) | 5.51 | 3.88 | 4.79 | 2.15 | 5.42 | |
| CO | (km s) | -3.26 | -3.17 | 2.61 | -1.08 | 8.80 |
| (K) | 10.80 | 8.49 | 7.99 | 0.99 | 0.39 | |
| (km s) | 2.69 | 2.50 | 1.18 | 1.28 | 4.45 | |
| [C I] | (km s) | -3.44 | -3.40 | 3.40 | -3.27 | 34.41 |
| (K) | 5.78 | 4.38 | 4.35 | 1.36 | 1.65 | |
| (km s) | 3.20 | 2.81 | 1.84 | 1.42 | 3.13 | |
| Outside Region ([C II] K) | ||||||
| [C II] | (km s) | -3.98 | -3.94 | 5.06 | 0.58 | 2.22 |
| (K) | 16.17 | 14.18 | 10.51 | 0.87 | 0.22 | |
| (km s) | 5.50 | 4.11 | 4.42 | 2.47 | 7.72 | |
| CO | (km s) | -4.73 | -4.56 | 2.73 | -2.64 | 19.04 |
| (K) | 7.24 | 6.15 | 4.71 | 1.07 | 0.91 | |
| (km s) | 2.56 | 2.38 | 1.26 | 2.47 | 17.77 | |
| [C I] | (km s) | -4.33 | -4.58 | 4.55 | -1.49 | 14.51 |
| (K) | 4.05 | 2.93 | 3.01 | 1.78 | 3.99 | |
| (km s) | 2.91 | 2.57 | 1.69 | 1.03 | 1.17 |
Table C.1. Statistics of [C II], CO, and [C I] Gaussian components.
| Ratio | Mean | Median | Std. Dev. | Skewness | Kurtosis |
| Full Map | |||||
| [C II]/CO | 4.79 | 3.28 | 3.86 | 1.69 | 3.37 |
| [C II]/CO | 18.25 | 12.34 | 15.95 | 1.61 | 2.85 |
| [C II]/[C I] | 20.05 | 16.47 | 13.77 | 1.28 | 1.68 |
| [C I]/CO | 0.97 | 0.85 | 0.46 | 1.95 | 7.02 |
| [C I]/CO | 0.24 | 0.22 | 0.10 | 6.43 | 105.63 |
| CO/CO | 4.57 | 4.22 | 1.73 | 1.10 | 1.42 |
| Inside [C II] K | |||||
| Ratio | Mean | Median | Std | Skewness | Kurtosis |
| [C II]/CO | 3.97 | 3.00 | 2.76 | 2.26 | 6.54 |
| [C II]/CO | 16.78 | 10.71 | 15.51 | 2.17 | 5.93 |
| [C II]/[C I] | 20.51 | 16.71 | 12.83 | 1.28 | 1.48 |
| [C I]/CO | 0.76 | 0.70 | 0.28 | 1.89 | 5.35 |
| [C I]/CO | 0.20 | 0.19 | 0.05 | 0.69 | 0.24 |
| CO/CO | 3.88 | 3.57 | 1.27 | 1.16 | 1.62 |
| Outside [C II] K | |||||
| Ratio | Mean | Median | Std | Skewness | Kurtosis |
| [C II]/CO | 5.28 | 3.69 | 4.31 | 1.38 | 2.08 |
| [C II]/CO | 19.45 | 14.54 | 16.20 | 1.21 | 1.00 |
| [C II]/[C I] | 19.76 | 16.26 | 14.33 | 1.28 | 1.73 |
| [C I]/CO | 1.15 | 1.05 | 0.50 | 1.73 | 6.01 |
| [C I]/CO | 0.26 | 0.23 | 0.12 | 6.16 | 88.35 |
| CO/CO | 5.13 | 4.82 | 1.84 | 0.87 | 0.85 |
Table C.2. Statistics of line ratios derived from moment-0 maps.
| Ratio | Cluster 1 (Low ATLASGAL flux) | Cluster 2 (High ATLASGAL flux) | ||||||||
| Mean | Median | Std | Skew | Kurtosis | Mean | Median | Std | Skew | Kurtosis | |
| [C II]/CO | 4.45 | 3.43 | 3.24 | 1.84 | 4.59 | 2.23 | 2.19 | 0.90 | 1.00 | 2.96 |
| [C II]/CO | 23.33 | 18.93 | 16.71 | 1.27 | 1.83 | 6.96 | 6.27 | 3.67 | 1.68 | 5.06 |
| [C II]/[C I] | 21.22 | 18.19 | 13.66 | 1.08 | 1.18 | 11.50 | 10.53 | 5.99 | 1.10 | 1.80 |
| [C I]/CO | 1.13 | 1.03 | 0.47 | 1.89 | 7.03 | 0.62 | 0.61 | 0.13 | 0.44 | 0.59 |
| [C I]/CO | 0.217 | 0.209 | 0.063 | 1.11 | 2.93 | 0.210 | 0.204 | 0.052 | 0.49 | -0.18 |
| CO/CO | 5.26 | 4.93 | 1.61 | 1.06 | 1.60 | 3.03 | 2.91 | 0.56 | 1.06 | 1.15 |
Table C.3. Line ratio statistics for the two K–means clusters.
| Ratio | Mean | Median | Std | Skew | Kurtosis |
| CII/CO | 0.63 | 0.52 | 0.51 | 5.51 | 76.84 |
| CII/CI | 17.64 | 10.47 | 25.06 | 5.32 | 52.98 |
| CII/CO | 9.68 | 5.12 | 13.61 | 4.40 | 37.24 |
Table C.4. Line Intensity Ratio Statistics (Peak Component Analysis)
References
- [1]Abel, N. P. 2006, MNRAS, 368, 1949
- [2]Aikawa, Y., Jiménez-Serra, I., & Caselli, P. 2026, ARA&A, 64, 673
- [3]Anderson, L. D., Bania, T. M., Balser, D. S., et al. 2014, ApJS, 212, 1
- [4]André, P., Mattern, M., Arzoumanian, D., et al. 2025, ApJ, 984, L59
- [5]Arzoumanian, D., Rüssel, D., Zavagno, A., et al. 2022, A&A, 660, A56
- [6]Astropy Collaboration, Price-Whelan, A. M., Lim, P. L., et al. 2022, ApJ, 935, 167
- [7]Benjamin, R. A., Churchwell, E., Babler, B. L., et al. 2003, PASP, 115, 953
- [8]Beuther, H., Gieser, C., Linz, H., et al. 2026, A&A, 706, A329
- [9]Beuther, H., Schneider, N., Simon, R., et al. 2022, A&A, 659, A77
- [10]Beuther, H., Wang, Y., Soler, J., et al. 2020, A&A, 638, A44
- [11]Bohlin, R. C., Savage, B. D., & Drake, J. F. 1978, ApJ, 224, 132
- [12]Bonne, L., Kabanovic, S., Schneider, N., et al. 2023, A&A, 679, L5
- [13]Bonne, L., Schneider, N., García, P., et al. 2022, ApJ, 935, 171
- [14]Carey, S. J., Noriega-Crespo, A., Mizuno, D. R., et al. 2009, PASP, 121, 76
- [15]Clark, P. C., Glover, S. C. O., Klessen, R. S., & Bonnell, I. A. 2012, MNRAS, 424, 2599
- [16]Clark, P. C., Glover, S. C. O., Ragan, S. E., & Duarte-Cabral, A. 2019, MNRAS, 486, 4622
- [17]Dame, T. M., Hartmann, D., & Thaddeus, P. 2001, ApJ, 547, 792
- [18]Dannhauer, S. M., Vider, S., Schneider, N., et al. 2025, A&A, 703, A197
- [19]Faerber, T., Anderson, L. D., Luisi, M., et al. 2025, ApJ, 990, 30
- [20]Fazio, G. G., Hora, J. L., Allen, L. E., et al. 2004, ApJS, 154, 10
- [21]Frerking, M. A., Keene, J., Blake, G. A., & Phillips, T. G. 1989, ApJ, 344, 311
- [22]Fukui, Y., Kohno, M., Yokoyama, K., et al. 2018, PASJ, 70, S41
- [23]Goedhart, S., Cotton, W. D., Camilo, F., et al. 2024, MNRAS, 531, 649
- [24]Goldsmith, P. F., Langer, W. D., Pineda, J. L., & Velusamy, T. 2012, ApJS, 203, 13
- [25]Güsten, R., Bayarshy, A., Bell, A., et al. 2008, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7020, Millimeter and Submillimeter Detectors and Instrumentation for Astronomy IV, ed. W. D. Duncan, W. S. Holland, S. Withington, & J. Zmuidzinas, 72000
- [26]Güsten, R., Nyman, L. Å., Schilke, P., et al. 2006, A&A, 454, L13
- [27]Heeschen, D. S. 1955, ApJ, 121, 569
- [28]Kabanovic, S., Schneider, N., Ossenkopf-Okada, V., et al. 2022, A&A, 659, A36
- [29]Keilmann, E., Dannhauer, S., Kabanovic, S., et al. 2025, A&A, 697, L2
- [30]Kraemer, K. E. & Jackson, J. M. 1999, ApJS, 124, 439
- [31]Langer, W. D. & Penzias, A. A. 1990, ApJ, 357, 477
- [32]Langer, W. D., Pineda, J. L., Goldsmith, P. F., et al. 2021, A&A, 651, A59
- [33]Le Petit, F., Nehmé, C., Le Bourlot, J., & Roueff, E. 2006, ApJS, 164, 506
- [34]Li, G.-X. 2022, ApJS, 259, 59
- [35]Louvet, F., Neufpane, S., Garay, G., et al. 2019, A&A, 622, A99
- [36]Luisi, M., Anderson, L. D., Schneider, N., et al. 2021, Science Advances, 7, eabe9511
- [37]Madden, S. C., Cormier, D., Hony, S., et al. 2020, A&A, 643, A141
- [38]Mangum, J. G. & Shirley, Y. L. 2015, PASP, 127, 266
- [39]Marsh, K. A., Whitworth, A. P., & Lomax, O. 2015, MNRAS, 454, 4282
- [40]Milam, S. N., Savage, C., Brewster, M. A., Ziurys, L. M., & Wyckoff, S. 2005, ApJ, 634, 1126
- [41]Neupane, S., Wyrowski, F., Menten, K. M., et al. 2024, A&A, 692, A114
- [42]Oberst, T. E., Parshley, S. C., Stacey, G. J., et al. 2006, ApJ, 652, L125
- [43]Ossenkopf, V., Röllig, M., Neufeld, D. A., et al. 2013, A&A, 550, A57
- [44]Pabst, C., Higgins, R., Goicoechea, J. R., et al. 2019, Nature, 565, 618
- [45]Pabst, C. H. M., Goicoechea, J. R., Hacar, A., et al. 2022, A&A, 658, A98
- [46]Pedregosa, F., Varoquaux, G., Gramfort, A., et al. 2011, Journal of Machine Learning Research, 12, 2825
- [47]Persi, P. & Tapia, M. 2008, in Handbook of Star Forming Regions, Volume II, ed. B. Reipurth, Vol. 5, 456
- [48]Pineda, J. L., Langer, W. D., Velusamy, T., & Goldsmith, P. F. 2013, A&A, 554, A103
- [49]Pound, M. W. & Wolfire, M. G. 2023, AJ, 165, 25
- [50]Ramaila, A. J. T., Thompson, M. A., Smirnov, O. M., et al. 2026, MNRAS, 550, stag1194
- [51]Riegel, K. W. & Crutcher, R. M. 1972, A&A, 18, 55
- [52]Riener, M., Kainulainen, J., Henshaw, J. D., et al. 2019, A&A, 628, A78
- [53]Risacher, C., Güsten, R., Stutzki, J., et al. 2016, A&A, 595, A34
- [54]Röllig, M. & Ossenkopf-Okada, V. 2022, A&A, 664, A67
- [55]Russeil, D., Schneider, N., Anderson, L. D., et al. 2013, A&A, 554, A42
- [56]Russeil, D., Zavagno, A., Adami, C., et al. 2012, A&A, 538, A142
- [57]Schneider, N., Bonne, L., Bontemps, S., et al. 2023, Nature Astronomy, 7, 546
- [58]Schneider, N., Bontemps, S., Girichidis, P., et al. 2015, MNRAS, 453, L41
- [59]Schneider, N., Simon, R., Guevara, C., et al. 2020, PASP, 132, 104301
- [60]Schöier, F. L., van der Tak, F. F. S., van Dishoeck, E. F., & Black, J. H. 2005, A&A, 432, 369
- [61]Schuller, F., Menten, K. M., Contreras, Y., et al. 2009, A&A, 504, 415
- [62]Stacey, G. J., Geis, N., Genzel, R., et al. 1991, ApJ, 373, 423
- [63]Suzuki, H., Yamamoto, S., Ohishi, M., et al. 1992, ApJ, 392, 551
- [64]Tielens, A. G. G. M. & Hollenbach, D. 1985, ApJ, 291, 722
- [65]Tigé, J., Motte, F., Russeil, D., et al. 2017, A&A, 602, A77
- [66]Tiwari, M., Wolfire, M., Pound, M. W., et al. 2022, AJ, 164, 150
- [67]Velusamy, T., Langer, W. D., Goldsmith, P. F., & Pineda, J. L. 2015, A&A, 578, A135
- [68]Zernickel, A., Schilke, P., & Smith, R. J. 2013, A&A, 554, L2
