UNVEILING A NETWORK OF PARALLEL FILAMENTS IN THE INFRARED DARK CLOUD G14.225−0.506

Filaments are ubiquitous structures in star-forming complexes (e.g., Schneider & Elmegreen 1979; Wiseman & Ho 1998; Hatchell et al. 2005; Goldsmith et al. 2008; Wang et al. 2008; Jackson et al. 2010; Schneider et al. 2010; Molinari et al. 2010; André et al. 2010; Arzoumanian et al. 2011), and often intersect in high-density regions of low aspect ratio and associated with star formation, known as hub–filament systems (e.g., Myers 2009; Liu et al. 2012). However, their formation and their role in the star formation process are not yet well understood.

In nearby (d ∼ 200–500 pc) molecular clouds, recent photometric results from Herschel suggest that large-scale turbulence might be responsible for the formation of filaments (Arzoumanian et al. 2011), while spectroscopic studies, sensitive to smaller scales, show that filaments present subsonic non-thermal motions (Hacar & Tafalla 2011; Pineda et al. 2011), indicative of a dissipation of turbulence at smaller scales. While this is consistent with observations in more distant and massive star-forming regions, such as G28.34+0.06 (Wang et al. 2008), a number of studies reveal supersonic non-thermal motions and suggest the formation of filaments by the convergence of flows or by filament–filament collisions on large scales (Schneider et al. 2010; Csengeri et al. 2011; Heitsch et al. 2008; Jiménez-Serra et al. 2010; Henshaw et al. 2013; Miettinen 2012; Nakamura et al. 2012). In addition, theoretical studies propose that magnetic fields could play a role in the formation of filaments (e.g., Nagai et al. 1998; Nakamura & Li 2008). It is clear, then, that several formation mechanisms have been invoked to explain the formation and alignment of filaments and therefore further spectroscopic studies of filamentary regions are essential to investigate the origin and evolution of such structures.

Filaments are prevailing structures in infrared dark clouds (IRDCs; cf. Rathborne et al. 2006). In this Letter we present combined Very Large Array (VLA) and Effelsberg 100 m telescope observations of the NH₃ (1,1) and (2,2) transitions toward the IRDC G14.225−0.506 (hereafter G14.2). Most of the studies performed so far toward this region focus on the brightest infrared sources, IRAS 18153−1651 (hereafter I18153) and IRAS 18152−1658 (hereafter I18152) with a luminosity of ∼1.1× 10⁴ L_☉ and ∼4× 10³ L_☉, respectively, and located at a distance of 2.3 kpc (Jaffe et al. 1981, 1982). Single-dish observations show that I18153 is associated with H₂O maser emission (Jaffe et al. 1981; Palagi et al. 1993), and dense gas emission (Plume et al. 1992; Anglada et al. 1996; Bronfman et al. 1996). More recent VLA observations reveal H₂O maser emission in nine different positions (Wang et al. 2006), which indicates that star formation is already ongoing in some parts of the cloud. IRDC G14.2 has been identified, using Spitzer data, by Peretto & Fuller (2009) as a cloud containing an important amount (∼100) of density enhancements or fragments displaying a filamentary morphology. The filamentary appearance of G14.2 (Figure 1) and its relatively nearby distance make this region a good selection to investigate the physical properties of filaments and their formation mechanism at high spatial resolution.

Zoom In Zoom Out Reset image size

The observations of the NH₃ (1,1) and (2,2) transitions were conducted using the VLA¹² in the D configuration on 2005 November 12 (project AW666). We performed a 34-pointing mosaic covering an area of 7' × 13'. The integration time was ∼4.5 minutes per pointing. The adopted flux density of the flux calibrator 3C 286 was 2.41 Jy at a wavelength of 1.3 cm. The time variation of the gains was calibrated using J1832−105, with a bootstrapped flux of 0.97 ± 0.01 Jy, and the bandpass calibrator used was 3C 273. We used the 4IF spectral line mode, which allows simultaneous observations of the NH₃ (1,1) and (2,2) lines with two polarizations for each line. The bandwidth used was 3.12 MHz, divided into 63 channels with a channel spacing of 48.8 kHz (∼0.6 km s⁻¹ at 1.3 cm), centered at ∼21 km s⁻¹. The visibility data sets were calibrated using the AIPS software package of the NRAO.

To recover extended structures filtered out by the interferometer, we performed NH₃ observations with the Effelsberg 100 m telescope (project 101-07). The observations were carried out between 2008 April 4 and 7. We used the 18–26 GHz HEMT receiver tuned to a frequency of 23.7 GHz with the 16384 channel fast Fourier transform spectrometer, allowing simultaneous observations of the NH₃ (1,1) and (2,2) lines. The total bandwidth used was 100 MHz, which provides a velocity resolution of 0.075 km s⁻¹. The observations were conducted in frequency switching mode with a frequency throw of 7.5 MHz. At the observed wavelength, the half-power beamwidth of the telescope is ∼40''. The map covered an area of 8' × 13' and was made by observing the positions of a grid with half-beam spacing. The pointing was checked at hourly intervals, with a pointing accuracy better than 8''. To convert the arbitrary noise tube units of the Effelsberg data to main beam brightness temperature, we observed as a primary flux calibrator NGC 7027 and a nearby quasar as a secondary flux calibrator. Data reduction was performed using the CLASS package, which is part of the GILDAS¹³ software. We combined the visibility data from the VLA and Effelsberg 100 m telescope for both NH₃ (1,1) and NH₃ (2,2) lines following the MIRIAD procedure outlined in Vogel et al. (1984). We applied a uv-taper function of 23 kλ during imaging. The resulting synthesized beams were 82 × 70 (P.A. =−15°) for NH₃ (1,1) and 80 × 69 (P.A. =−16°) for NH₃ (2,2). The rms was ∼8 mJy beam⁻¹ per 0.6 km s⁻¹ spectral channel.

Figure 2 (top left) shows the combined (VLA+Effelsberg) zero-order moment map of the NH₃ (1,1) emission overlaid on the 8 μm Spitzer image. The overall morphology of the NH₃ (1,1) dense gas consists of extended and clumpy filamentary structures, strikingly mimicking the extinction feature seen in the Spitzer image. While the NH₃ (1,1) emission is spatially extended, the NH₃ (2,2) emission is compact (Figure 2, top right), suggesting that the extended emission is at lower temperatures. We identified the most prominent filaments based on the morphology of the NH₃ (1,1) together with the fact that these structures are coherent in velocity. We used the following criteria: (1) filaments must have aspect ratio larger than 6; (2) the signal-to-noise ratio should be larger than 9;¹⁴ and (3) they must appear in at least two velocity channels and spanning a maximum velocity range of 3 km s⁻¹. Figure 2 (top right) shows, for comparison, the 870 μm continuum emission from the LABOCA bolometer at the APEX telescope (Busquet 2010), supporting our identification.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

We identified a network of eight filaments and two hubs (named hub-N and hub-S in Figure 2), which were recognized using the NH₃ (2,2) emission as denser regions in which some filaments intersect. The NH₃ filaments, which cover a total area of 4.7 × 8.7 pc, appear approximately parallel, in projection, in two preferred directions, at P.A. of 10° and 60°, and they contain chains of dense cores¹⁵ aligned along the filament axis and distributed at somewhat regular spacings of about ∼30'' or 0.33 ± 0.09 pc at the distance of the cloud. The averaged projected separation between adjacent filaments is between 0.5 pc and 1 pc. In Table 1 we report on the length and width at FWHM of each filament obtained from NH₃ (1,1) data. On average, we found that the aspect ratio is ∼15:1, with a typical FWHM width of ∼0.12 pc. This value is close to the filament width of 0.1 pc reported for the IC 5146, Aquila, and Polaris molecular clouds from Herschel observations (Arzoumanian et al. 2011).

In Figure 2 (bottom left) we present the first-order moment map of the NH₃ (1,1) main line. Within each filament the velocity variations are small, in the range of 1–2 km s⁻¹ (see Table 1), similar to other filamentary IRDCs (e.g., Jackson et al. 2010). This network of filaments seems to be separated into two main velocity components, one at v_LSR ∼18.3–20.8 km s⁻¹ and another one at v_LSR ∼20.8–23.2 km s⁻¹, which overlap in the hubs. The second-order moment map is presented in Figure 2 (bottom right), and shows that the velocity dispersion is locally enhanced (σ ∼ 1 km s⁻¹) toward hubs. Additionally, a high velocity dispersion (σ ∼ 1.6 km s⁻¹) is seen toward an arc-shaped structure connecting filament F10-E with the southern filaments, in a small region intersecting filament F60-C1 and labeled as position "e" in Figure 2. In this region the large values of the velocity dispersion are due to the presence of two velocity components separated by ∼3 km s⁻¹ (see Figure 3(e)). The presence of two velocity components is also found in regions where filaments intersect hubs (see Figures 3(d) and (f)). In contrast, all the other filaments appear more quiescent, with a typical velocity dispersion of ∼0.4–0.6 km s⁻¹.

To obtain the main physical properties (rotational temperature T_rot, total velocity dispersion σ_TOT, and mass per unit length M/l) of each filament, we extracted an averaged spectrum of NH₃ (1,1) and (2,2) over the filament area at FWHM. The results are reported in Table 1. The rotational temperature ranges between 10 K and 16 K. The total velocity dispersion of the gas ranges from ∼0.5 km s⁻¹ up to 1.1 km s⁻¹, and the non-thermal velocity dispersion over the isothermal sound speed, σ_NT/c_s, ranges between 2 and 5, implying that filaments in G14.2 are characterized by supersonic non-thermal motions. In Table 1 we also list the mass and virial mass per unit length, the observed separation between cores, and the number of cores in each filament. Finally, the total surface density estimated by taking the spectrum averaged over all NH₃ filaments is Σ ≃ 1.9 g cm⁻².

In the previous section we presented the main properties of the two hub–filament systems in G14.225−0.506. The hubs are more compact (aspect ratio 5 versus 15), warmer (T_rot ≃ 15 K versus 11 K), and show larger velocity dispersion and larger masses per unit length than filaments. Interestingly, the hubs are associated with H₂O maser emission (Wang et al. 2006) and mid-infrared sources (see Figures 1 and 2), and they are the main sites of stellar activity within the cloud.

The stability of the filaments can be studied by estimating the virial parameter α_vir = M_vir/M (Bertoldi & McKee 1992), which is ≲2 for all the filaments and hubs except for F10-E, and five out of eight filaments are near virial equilibrium (α_vir ≃ 1). This indicates that most of the filaments are unstable (collapsing) and probably undergoing fragmentation, compatible with the clumpy structure of G14.2. It is worth noting that filament F10-E has T_rot and velocity dispersion values similar to the hub properties. Filament F10-E presents some striations converging toward it. However, while hub-N and hub-S seem to be places where two different velocity structures converge, F10-E shows only one velocity component. We speculate that I18153, an UC H ii region with L ∼ 10⁴ L_☉ (A. Sánchez-Monge 2012, private communication), may compress the gas, heating and injecting turbulence to this filament (α_vir ≃ 5). The interaction of this UC H ii region with the dense gas is also seen in hub-N, where the NH₃ (2,2)/NH₃ (1,1) map shows a local heating (Figure 4). The position–velocity (PV) plot along this hub (see Figure 4) reveals an inverted C-like structure, consistent with an expanding shell (Arce et al. 2011).

Zoom In Zoom Out Reset image size

We investigated the fragmentation of filaments in the magnetohydrodynamic "sausage"-type instability scenario (Chandrasekhar & Fermi 1953; see also Jackson et al. 2010), which predicts periodic separation between fragments (or cores) for a given density and isothermal sound speed. For an isothermal gas cylinder of finite radius R, the core separation can be expressed as λ_f = 22H for R ≫ H, where H is the scale height (see Table 1 for the formal expression). This is the case of G14.2, since R and H are 0.12 pc and 0.04 pc, respectively. Adopting a density of 10⁵ cm⁻³, we estimated the predicted core separation using first the isothermal sound speed, yielding λ_f ∼ 0.16–0.21 pc, and then replacing c_s by the total velocity dispersion σ_TOT, which gives λ_f ∼ 0.4–0.8 pc (see Table 1). The observed separation, ∼0.33 pc, is in agreement with these two extreme cases. It is noteworthy that most of the cores appear to be elongated along the direction of the filament, which could imply the possibility of further fragmentation at smaller scales as observed in the IRDC G28.34+0.06-P1 (Zhang et al. 2009; Wang et al. 2011).

One of the most intriguing features of G14.2 is the network of filaments that are aligned in parallel. The filaments appear to take two preferred directions, one group at a P.A. of 10°, and the others at a P.A. of 60°. This network of filaments may arise from a layer of self-gravitating gas. Instability analysis has been performed for a layer of an isothermal infinite sheet under dynamical perturbation (Ledoux 1951; Schmid-Burgk 1967; Elmegreen & Elmegreen 1978; Larson 1985; Nagai et al. 1998; Myers 2009). The gas is unstable to perturbations, which leads to high-density columns in the plane as a result of gravitational instability. The spacings between the high-density columns of gas correspond to the wavelength of the fastest growth mode. In the absence of magnetic fields, the growth of instability does not have a preferred direction in the plane. As a result, a grid of connected filaments appears in the gas layer. If the gas layer is threaded by magnetic fields, the growth of the instability develops unrestricted in one direction and is suppressed along the orthogonal direction. Nagai et al. (1998) analyzed a pressure-confined isothermal gas layer threaded by uniform magnetic fields. They found that in the regime of smaller external pressure (i.e., the scale height H ≪ Z_b, where Z_b is the thickness of the gas layer) the instability grows faster along the field lines. As a consequence, high-density columns or filaments develop with their longitudinal axis perpendicular to the field lines. In the high-pressure regime (i.e., H ≫ Z_b), the fastest growth of instability is perpendicular to the field lines and gives rise to filaments parallel to the magnetic fields.

Recent numerical simulations (S. Van Loo 2012, private communication) confirm the linear analysis in Nagai et al. (1998). The simulations show that an array of high-density columns develop in the gas layer with magnetic fields. In addition, lower density filamentary structures are also present, interconnecting the main filaments. The highest density structures are found at the intersections of major and minor filaments, as in this work. Furthermore, the simulations show that gravitational instability develops within a filament during the filament formation. This means that fragmentation of a filament into cores occurs simultaneously with the fragmentation of the sheet, but according to Toalá et al. (2012) with different free-fall times.

The array of filamentary structures in G14.225−0.506 may arise from gravitational instability of a thin gas layer with magnetic fields. In fact, preliminary near-infrared polarimetric observations around hub-N (G. Busquet et al. in preparation) reveal that the magnetic field is perpendicular to filaments at P.A. ∼ 60° (see Figure 2). Therefore, according to Nagai et al. (1998), G14.2 would be in the regime of small external pressure (H ≪ Z_b). Using the total surface density (see Section 3), the scale height H of the initial gas layer is H ∼ 0.09 pc. This value should be regarded with caution, and to definitively assess its validity one would need observations of a low-density gas tracer to be sensitive to the gas layer. The wavelength of the fastest mode can be expressed as λ_fastest = 4πH (Equation (60) in Nagai et al. 1998). Using our estimation of H, the predicted separation is ∼1.1 pc, in agreement with the observed filament separation (between 0.5 and 1 pc). It is not clear how such a large gas layer (4.7 × 8.7 pc) may form initially. The convergence of dynamic flows could be responsible for the formation of such a large gas layer that subsequently could fragment into parallel filaments as a result of magnetic modulation. Our NH₃ data, although showing two velocity components, do not reveal evidence of converging/interacting flows and further observations of low-velocity shock tracers, like SiO or CH₃CN (Jiménez-Serra et al. 2010; Csengeri et al. 2011), and a tracer of low-density material are required to definitely identify signatures of converging flows. Overall, our data suggest that magnetic fields might play an important role in the alignment of filaments, and polarization measurements in the entire cloud would lend further support to this scenario.

The authors are grateful to the anonymous referee for valuable comments. G.B. is deeply grateful to Eugenio Schisano for very fruitful discussion on filaments. G.B. is funded by an Italian Space Agency (ASI) fellowship under contract number I/005/07/0. A.P., R.E., and I.d.G.-M. are supported by the Spanish MICINN grant AYA2011-30228-C03 (co-funded with FEDER funds). A.P. is supported by a JAE-Doc CSIC fellowship co-funded with the European Social Fund, under the program "Junta para la Ampliación de Estudios," and by the AGAUR grant 2009SGR1172 (Catalonia). F.P.S. and G.A.P.F. are partially supported by CNPq and FAPEMIG. This work is partially based on observations with the 100 m telescope of the MPIfR (Max-Planck-Institut für Radioastronomie) at Effelsberg.