Probing the structure of a massive filament: ArTeMiS 350 and 450 micron mapping of the integral-shaped filament in Orion A

(abridged) Within the Orion A molecular cloud, the integral-shaped filament (ISF) is a prominent, degree-long structure of dense gas and dust, with clear signs of recent and on-going high-mass star formation. We used the ArTeMiS bolometer camera at APEX to map a 0.6x0.2 deg^2 region covering OMC-1, OMC-2, OMC-3 at 350 and 450 micron. We combined these data with Herschel-SPIRE maps to recover extended emission. The combined Herschel-ArTeMiS maps provide details on the distribution of dense, cold material, with a high spatial dynamic range, from our 8'' resolution (0.016 pc) up to the size of the map ~10-15 deg. By combining Herschel and ArTeMiS data at 160, 250, 350 and 450 micron, we constructed high-resolution temperature and H2 column density maps. We extracted radial profiles from the column density map in several, representative portions of the ISF, that we fitted with Gaussian and Plummer models to derive their intrinsic widths. We also compared the distribution of material traced by ArTeMiS with that seen in the higher density tracer N2H+(1-0) recently observed with the ALMA interferometer. All the radial profiles that we extracted show clear deviation from a Gaussian, with evidence for an inner plateau, previously not seen using Herschel-only data. We measure intrinsic half-power widths in the range 0.06 to 0.11 pc. This is significantly larger than the Gaussian widths measured for fibers seen in N2H+, which probably traces only the dense innermost regions of the large-scale filament. These half-power widths are within a factor of two of the value of 0.1 pc found for a large sample of nearby filaments in various low-mass star-forming regions, which tends to indicate that the physical conditions governing the fragmentation of prestellar cores within transcritical or supercritical filaments are the same over a large range of masses per unit length.


Introduction
It is well established that star formation takes place in dense molecular clumps embedded in large molecular clouds. The evo-This publication is based on data acquired with the Atacama Pathfinder Experiment (APEX), project 098.F-9304. APEX is a collaboration between the Max-Planck-Institut für Radioastronomie, the European Southern Observatory, and the Onsala Space Observatory. The temperature and column density maps are available in electronic form at the CDS via anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via http://cdsweb.u-strasbg.fr/cgi-bin/qcat?J/A+A/ lution of these dense clumps towards the onset of gravitational collapse and the formation of solar-type stars is relatively well understood in outline, but several key questions remain unanswered, including how the star formation rate and efficiency depend on the environmental conditions and whether there are notable differences in the physical processes leading to the formation of low-mass and high-mass stars.
The Herschel Space Observatory has revealed ubiquitous filamentary structures in the cold interstellar medium (ISM; e.g. André et al. 2010;Molinari et al. 2010;Arzoumanian et al. 2011Arzoumanian et al. , 2019Schisano et al. 2020). Many subsequent studies, in-Article number, page 1 of 17 arXiv:2104.03717v2 [astro-ph.GA] 26 May 2021 A&A proofs: manuscript no. orion_artemis cluding theoretical work (e.g. Federrath 2016), have confirmed how important filaments are to the evolution of interstellar matter and star formation activity, and they have supported a filament paradigm for star formation (see the review by André et al. 2014). However, most detailed (resolved) studies to date have concentrated on a few, typically nearby clouds, where mostly low-mass and solar-type stars are forming (Arzoumanian et al. 2011;Hacar et al. 2013;Könyves et al. 2015, among others).
The Orion A giant molecular cloud is the nearest site of active star formation that includes high-mass stars. At a distance of only 410 pc (Menten et al. 2007;Stutz et al. 2018), it is an ideal target for investigating the initial phases of high-mass star formation in great detail. It is located within the Barnard's Loop, which is part of a large complex of bubbles and filaments in a superbubble (Ochsendorf et al. 2015). A large filamentary structure with a length of several tens of parsecs is seen along the north-south direction across Orion A. The densest part of this filament is a ∼1.5 deg long structure (∼10 pc if in the plane of the sky) known as the integral-shaped filament (ISF; Bally et al. 1987). This structure is clearly visible in dust continuum emission (Nutter & Ward-Thompson 2007;Shimajiri et al. 2011) or in CO molecular lines (Kong et al. 2018;Suri et al. 2019). The Orion Nebula Cluster (ONC) appears approximately projected on the middle of this dense filament.
On large physical scales (0.05 to 8.5 pc), Stutz & Gould (2016) concluded from the analysis of Herschel data at ∼36 angular resolution that the radial distribution of the gas surface density near the ISF follows a power law. At much higher angular resolution (4.5 ), using data in the high-density tracer N 2 H + (1-0) line obtained with the Atacama Large Millimeter/Submillimter Array (ALMA), Hacar et al. (2018) argued that the ISF contains a large number of fibres organised in a complex network. They measured emission profiles characterised by full widths at half maximum (FWHMs) in the range 0.015-0.065 pc, with a median value of 0.035 pc, which is significantly smaller than the typical 0.1 pc half-power width derived from Herschel column density maps in a number of nearby molecular complexes (Arzoumanian et al. 2011. Since these column density maps were derived from dust thermal emission in the sub-millimetre, they also trace lower-density material than the N 2 H + (1-0) transition. Therefore, the question remains open as to whether the observed difference in filament widths can be explained by the difference in the angular resolution of the data, or if they trace different material with different physical conditions.
With an angular resolution of 8 at 350 µm, more than three times better than Herschel at the same wavelength, the ArTéMiS camera installed at the Atacama Pathfinder Experiment (APEX) 12 m telescope (Revéret et al. 2014;André et al. 2016) is an ideal tool for probing the cold, dusty ISM at <0.1 pc scale up to distances of ∼3 kpc. We observed a ∼0.1 deg 2 region covering the northern end of the ISF with this instrument at 350 and 450 µm. The ArTéMiS maps, combined with Herschel data, provide a detailed view, with a large spatial-dynamic range, of the distribution of cold gas and dust, allowing us to draw an accurate picture of the ISF and its substructures. We describe the data that we use in Sect. 2. We then present our measurements of the radial intensity profiles along several portions of the main filament in Sect. 3. We discuss our results and compare them with previous findings in Sect. 4. Finally, we summarise our conclusions in Sect. 5.

Herschel data
As part of the Herschel Gould Belt Survey (HGBS) key project (André et al. 2010), a large map of 22 deg 2 covering the Orion A cloud and nebula was observed at 70, 160, 250, 350, and 500 µm. Details regarding the HGBS standard data reduction and calibration procedure can be found in Könyves et al. (2015Könyves et al. ( , 2020. In particular, the emission seen on the largest scales should be accurate since zero-level offsets as estimated from Planck and IRAS data were applied (cf. Bernard et al. 2010).
Initial results from the HGBS observations towards Orion A were published by: Roy et al. (2013), who investigated the physical properties of the dust in this region; Stutz et al. (2013), who made a systematic census of the reddest protostars; and Polychroni et al. (2013), who found two different core mass functions for cores within and out of the filaments. Here, we focus on the northern part of the Orion A cloud, which covers a large fraction of the ISF, including OMC 1, OMC 2, and OMC 3.
The Herschel-SPIRE parallel-mode data are affected by saturation in the vicinity of the bright source Orion-KL: 55 pixels cannot be used in the original (10 /pixel) map at 350 µm, nor can 14 pixels be used in the original (14 /pixel) map at 500 µm. We corrected for this by using dedicated SPIRE-only data taken towards Orion-KL and interpolating the values measured on neighbouring pixels. This leads to an extra uncertainty in the flux calibration for this region, which covers ∼0.7 × 2 , a scale well probed by the ArTéMiS observations (see below).

ArTéMiS observations
New observations were conducted with ArTéMiS, which is installed at the APEX telescope (Güsten et al. 2006) in Chile. The first maps were acquired in July 2013, during the commissioning run of the first incarnation of the camera, with only one focal plane at 350 µm. Most observations were done with the current version of ArTéMiS (Talvard et al. 2018), which covers the same ∼5×2.5 arcmin 2 field of view at 350 and 450 µm, in several runs between October and December 2016. All observations consist of large on-the-fly maps, with scanning speeds of 20 /s and 60 /s and steps between lines of 5 and 30 for the data taken in 2016 and in 2013, respectively. In total, about 16 hr were spent on-source.
The weather conditions were generally very good, with a total amount of precipitable water vapour (PWV) between 0.25 and 0.7 mm. The zenith opacity at 350 and 450 µm was frequently measured with skydip observations and found to be between 0.8 and 1.2 at 350 µm and between 0.6 and 1.1 at 450 µm. The absolute flux calibration was checked and corrected for using maps of the primary calibrator Uranus and several secondary calibrators (NGC 253, NGC 2071, and CRL 618). The overall calibration uncertainty is estimated to be ∼30% at 350 µm and ∼20% at 450 µm.

ArTéMiS data reduction
The ArTéMiS data at 350 and 450 µm were processed with the dedicated IDL pipeline, which takes care of converting the raw data to IDL-friendly data structures, applying flux calibration, and rejecting unusable pixels. Special care was taken to correct for possible pointing errors between individual maps before a combined map could be computed. Then, the data were processed with a tailored version of the Scanamorphos software adapted for ArTéMiS data (Roussel 2013(Roussel , 2018. This pipeline makes use of the redundancy in the data to properly separate the various sources of noise (especially sky noise) from the real signal. This allows us to recover emission at scales that are usually filtered out by the typical data reduction of ground-based bolometer observations because such emission mimics the sky emission.
In order to recover extended emission at even larger scales, we combined the ArTéMiS maps with Herschel-SPIRE maps at similar wavelengths using the immerge task in MIRIAD (Sault et al. 1995). The immerge algorithm combines the available data in the Fourier domain after determining an optimal calibration factor to align the flux scales of the input images (here from ArTéMiS and SPIRE) in a common annulus of the uv plane. Here, we adopted an annulus corresponding to the range of baselines from 0.5 m (baseline b, sensitive to angular scales λ/b ∼ 2.4 at λ = 350 µm) to 3.5 m (the diameter of the Herschel telescope) to align the flux scale of the ArTéMiS 350 µm map to the flux scale of the SPIRE 350 µm map. We used the same method to combine the ArTéMiS 450 µm data with Herschel-SPIRE maps at 500 µm, which is the nearest wavelength observed with SPIRE.
The combined maps at 350 µm and 450 µm, covering about 30×10 arcmin 2 , are shown in Fig. 1. The effective resolution is ∼8 and ∼10 (FWHM) at 350 and 450 µm, respectively. The typical rms noise in this mosaic is between 0.3 and 0.5 Jy/8beam in regions with no strong emission at 350 µm. The discussion in the rest of the article is based on this combined Scanamor-phos-processed ArTéMiS and SPIRE map. We stress that the ArTéMiS observations allow us not only to improve the angular resolution compared to Herschel, but also to properly measure the peak of the emission towards Orion-KL, where the SPIRE emission maps are saturated.
2.4. Archival ALMA observations in N 2 H + Hacar et al. (2018) combined newly obtained ALMA data with data from the IRAM 30 m telescope to build a map in N 2 H + (1-0) at 4.5 resolution. They report the detection of 55 fibre-like structures, with typical FWHMs (as derived from Gaussian fitting) in the range 0.02 to 0.06 pc. We discuss the associations of these fibre-like structures with the large-scale filament seen in sub-millimetre continuum in Sect. 4.2.

High-resolution temperature and column density maps
In order to characterise the physical structures present in our maps, we first derived a map of H 2 column density at the highest possible angular resolution, using the following method. First, we re-projected the Herschel-PACS, SPIRE 250 µm, ArTéMiS-SPIRE 350 µm, and ArTéMiS 450 µm-SPIRE 500 µm maps to a common grid and smoothed all data to a common angular resolution of 18.2 (the lowest resolution of the above four maps, set by the SPIRE 250 µm data). We then generated a dust temperature map at 18.2 resolution by fitting a modified black- body of the form B ν (T d ) × κ ν to the four measurements between 160 µm and 500 µm on a pixel-by-pixel basis, where B ν (T d ) is the Planck blackbody function at frequency ν (or wavelength λ) for a dust temperature T d and κ ν (or κ λ ) is the dust opacity. For simplicity and easier comparison with Herschel work, our nominal assumption for the dust opacity law is the same as that adopted in HGBS papers: κ λ = 0.1 × (λ/300 µm) −β cm 2 per g of gas + dust (1) with an emissivity index β = 2 (Hildebrand 1983;Roy et al. 2014). The resulting dust temperature map is shown in Fig. 2. The dust temperature values range from 14 K to 25 K in the northern part of our map, and from 22 K to 52 K in the southern part, with the largest values (above 45 K) only found towards the Orion-KL Hii region. The general trends seen in our map agree well with the temperature map published by Lombardi et al. (2014) based on Herschel data only, but our data provide a better spatial resolution and allow us to properly estimate the temperature around the Orion-KL region, where the Herschel data are affected by saturation. In a second step, we converted the ArTéMiS-SPIRE 350 µm intensity map to a high-resolution (∼8 ) column density map, assuming a fixed dust temperature at each position given by the temperature map obtained in the first step at somewhat lower resolution. Assuming optically thin dust emission, the specific intensity I 350 (in MJy/sr) at λ = 350 µm was converted to H 2 column density as follows: where µ H 2 = 2.8 is the mean molecular weight. The resulting column density map is shown in Fig. 3 and is publicly available from the Centre de Données astronomiques de Strasbourg (CDS). Uncertainties in the dust opacity law induce uncertainties in the derived temperature and column density maps. The temperature map does not depend on the normalisation of the dust opacity law but is somewhat sensitive to the assumed value of the emissivity index (β), although the effect is rather small. For instance, changing β from 2 to 1.5 would only increase the dust temperatures by ∼ 20% in the northern part and by ∼ 35% on average in the southern part of the main filament. This would in turn decrease the derived column densities by ∼ 50% and ∼ 70% on average in the northern and southern parts, respectively. Observationally, the spectral index map obtained between 450 µm and 850 µm by Johnstone & Bally (1999) with SCUBA is consistent with 1.5 < ∼ β < ∼ 2 in the bulk of the ISF region. In their multi-wavelength study of the OMC-2/3 subregion, Sadavoy et al. (2016) find β = 1.7-1.8. Lower, more extreme values of β may apply locally to compact dense cores or very bright objects, such as OMC-3 MM6 or Orion KL, but these values have only a limited impact on the present study, which is focused on the main filament.
On the other hand, the column density map scales directly with the adopted normalisation of the dust opacity law at 300 µm, κ 300 . Based on a comparison of the Herschel results with a near-infrared extinction map of Orion A from 2MASS, Fig. 4. Skeleton map with the filament crests identified with DisPerSE overlaid on the ArTéMiS +SPIRE 350 µm map (logarithmic colour scale). The numbers mark filament segments roughly corresponding to the N 2 H + fibres discussed by Hacar et al. (2018), as listed in Col. 1 of Table 1. The yellow crest corresponds to the ISF portion referred to as 'ISF-OMC-2/3' in the text, and the red one corresponds to 'ISF-OMC-1'. Roy et al. (2013) find evidence of a weak trend between dust opacity and column density, κ 300 ∝ N 0.28 H 2 , in the regime 1 < ∼ A V < ∼ 10 mag, which they interpreted as a signature of dust grain evolution. Assuming this correlation between κ 300 and N H 2 continues at higher A V , the default value assumed here for κ 300 should remain valid within 60% accuracy for 3 × 10 21 cm −2 < ∼ N H 2 < ∼ 10 23 cm −2 (see Roy et al. 2014). Along the spine of the integral filament where N H 2 exceeds 10 23 cm −2 , the actual value of κ 300 may be a factor of 2 to 2.5 higher than our default value. In turn, the map shown in Fig. 3 may overestimate the actual column density along the spine of the integral filament by a factor of ∼2 to 2.5. Significant changes in dust opacity may also occur in the Orion bar and the photodissociation region around the Hii region (cf. Salgado et al. 2016), but these changes do not affect our analysis of the structure of the ISF away from OMC-1.
To summarise and illustrate the influence of dust opacity uncertainties, we derived alternative temperature and column density maps from a two-parameter fitting similar to that employed for Figs. 2 and 3 but using the following alternative opacity law instead of the nominal HGBS opacity law: where N 0 H 2 = 1.7 × 10 22 cm −2 and β(N H 2 ) = {2 × [100 − (N H 2 /10 21 )] + 1.5 × [(N H 2 /10 21 ) − 1]}/99 (i.e. a dust emissiv-ity index decreasing linearly between β = 2 at N H 2 = 10 21 cm −2 and β = 1.5 at N H 2 = 10 23 cm −2 ). This alternative opacity law combines the trend found by Roy et al. (2013) with a plausible variation in β with column density. We show the results in Appendix A, in the form of maps of relative difference in temperature, (T alt d − T std d )/T std d (Fig. A.1), and column density, (N alt H 2 − N std H 2 )/N std H 2 (Fig. A.2), between the alternative maps (T alt d , N alt H 2 ) derived using Eq.
(3) and the nominal maps (T std d , N std H 2 ) derived using Eq. (1). As expected, the alternative dust temperature map has slightly higher values than the nominal dust temperature map (Fig. 2), by ∼ 10-20% in the bulk of the integral filament and up to ∼ 50% towards Orion KL (see Fig. A.1). In the OMC-1 subregion, and especially in the immediate vicinity of Orion KL, the spectral energy distributions (SEDs) peak at wavelengths shorter than 160 µm and the dust temperature derived from measurements made longward of 160 µm is more uncertain. Accordingly, the alternative column density map has lower values than the nominal column density map (Fig. 3) by up to a factor of ∼ 3-4 towards Orion KL and only a factor of ∼ 2 along the bulk of the integral filament (see Fig. A.2).

Extraction of filamentary structures
We applied the DisPerSE algorithm (Sousbie 2011) to a version of the ArTéMiS+SPIRE 350 µm image smoothed to 16 resolution to trace the crests of the most prominent filamentary structures seen in our data, including the spine of the ISF. Apart from the main filament (i.e. the ISF), DisPerSE identified a network of fainter filamentary structures, most of which correspond to subfilaments or branches directly connected to the ISF, especially in the OMC-1 region (see Fig. 4) In the present paper, we mostly focus our analysis on the density structure of the main filament, whose DisPerSE crest is highlighted in red and yellow in Fig. 4. We stress that the definition of this main crest depends only weakly on the algorithm used to trace filamentary structures. In particular, we verified that the crests obtained with the alternative algorithms FilFinder (Koch & Rosolowsky 2015) and getsf (Men'shchikov 2021) are very similar to that shown in Fig. 4 from DisPerSE (cf. Fig. B.1 in Appendix B). Indeed, filament-finding algorithms tend to yield consistent results for high-contrast filaments such as the ISF.
We constructed radial density profiles by taking perpendicular cuts at each pixel of the 8 resolution column density map along each filament crest, using a similar method to that used by André et al. (2016) for NGC 6334. The column density profiles derived for the northern (OMC-2 and OMC-3) and southern (OMC-1) parts of the main filament (the ISF) are shown in Figs. 5, 6, and 7. We also investigated a smaller, relatively isolated filament (marked with the number 9 in Fig. 4) in the southwestern part of our map (Figs. 9 and 10).

Fits to the radial profiles
To estimate the physical width of each filament, we fitted the median column density profiles with both Gaussian and Plummerlike functions. We fitted the two sides of each filament separately to account for strongly varying background emission. We also cut the main filament into three portions, corresponding to the OMC-1, OMC-2, and OMC-3 subregions as marked in Fig. 1, to account for the varying physical conditions between the northern and southern part of our map. While it is clear that other nearby features can affect individual measurements, the use of median A&A proofs: manuscript no. orion_artemis  profiles allows us to derive the average global properties of each portion of the main filament.
Most of the filamentary structures observed here exhibit power-law emission wings, such that Gaussian functions do not reproduce their entire radial profiles well, as is evident in Figs. 5, 6, and 7. Only the inner part of the radial profiles was thus fitted with a Gaussian function. To do so, the width of each column density profile at half power above the background emission was first roughly estimated using the closest point to the filament crest for which the logarithmic derivative of the profile, d ln N H 2 /d ln r, showed a significant switch from negative to positive (or negligible) values. The observed profile was then fitted with a Gaussian function over a range of radii corresponding to twice this initial width estimate.
The Plummer-like model function that was fitted to the full radial profile of H 2 column density (N(r), after convolution with the approximately Gaussian ∼ 8 beam of the observations) is the following: where R flat is the radius of the flat inner region with roughly constant emission in the profile, p is the power-law exponent of the profile at radii larger than R flat , N 0 is the central column density, and N bg is the column density of the background cloud. A more detailed description of the fitting procedure is given in Sect. 3 of Arzoumanian et al. (2019). For a cylindrical model filament with a Plummer-like column density profile following Eq. (4), the underlying volume density profile has a similar form:  − Unless otherwise specified, the Plummer fits on either side of each filament segment were performed over the whole range of radii shown in Figs. 5 to 7, i.e. from 0 to ∼ 0.45 pc, and the Gaussian fits from 0 to twice the simple half-power radius derived without any fitting (see text), i.e. typically ∼ 0.08-0.1 pc. The western side of the OMC-2 segment was fitted with a Plummer model only out to a radius of 0.08 pc due to the presence of a shoulder in the background beyond that radius (see Fig. 6b); the eastern side of the OMC-1 segment was fitted out to a radius of 0.3 pc, and the western side of fibre 9 out to a radius of 0.1 pc.
− The values shown in parentheses denote cases where the column density profile does not exhibit clear evidence of a power-law wing (for example fibre 9; see Fig. 10), or where nearby emission strongly affects the measured profile, resulting in higher uncertainties in the fitted values.
Article number, page 7 of 17 A&A proofs: manuscript no. orion_artemis where n 0 is the central volume density of the filament. Moreover, n 0 is related to the projected central column density (N 0 ) by the simple relation n 0 = N 0 /(A p R flat ), where A p = 1 cos i × B 1 2 , p−1 2 is a constant factor that depends on the inclination angle (i) of the filament to the plane of sky and B is the Euler beta function (cf. Palmeirim et al. 2013).

Results
We report the best-fit values of D flat ≡ 2 × R flat and p from Plummer fitting, as well as the deconvolved FWHMs from Gaussian fitting, for several segments of the ISF in Table 1. As pointed out by several authors, the errors in D flat and p are significantly correlated, leading to some degeneracy in the derivation of these two parameters (e.g. Malinen et al. 2012;Smith et al. 2014;Arzoumanian et al. 2019). For this reason, we also provide, in the fifth column of Table 1, the half-power diameter derived from Plummer fitting: defined by The advantage of D Plummer HP is that it is a more robust estimator of the inner width of a Plummer-like profile than D flat because its derivation is not as strongly correlated to that of p. It is worth pointing out that for high-contrast filaments (N bg << N 0 ) such as the ISF, D HP depends only weakly on the background column density (N bg ). We further show in Appendix B that D HP is a robust quantity, affected by only small systematic errors ( < ∼ 30%) due to uncertainties in the dust opacity and the presence of temperature gradients along the line of sight.
The inner widths reported in Table 1 represent median values for each ISF segment, derived from a detailed fitting analysis of median column density profiles such as those shown in Figs. 5, 6, and 7. As is evident from the yellow error bars in these figures, there are significant fluctuations in the profiles along the length of the ISF. To quantify the impact of these fluctuations on our filament width estimates, we display a histogram of 221 individual width measurements made at beam-spaced positions along the whole crest of the ISF in Fig. 8. The individual inner widths used in Fig. 8 do not result from any Gaussian or Plummer fitting but correspond to simpler half-maximum diameters estimated as twice the radius from the filament crest at which the background-subtracted amplitude of the observed radial profile is half of the value on the crest (see 'hd widths' in Arzoumanian et al. 2019). Such simple hd width estimates are directly comparable to Gaussian FWHM or Plummer D HP estimates, but they are more robust and less affected by outliers. (When checked and reliable, Gaussian and Plummer fits are nevertheless preferable since they provide a better characterisation of the filament profiles and additional information such as line mass and density estimates.) Figure 8 also compares the distribution of individual widths obtained here for the ISF (black histogram) with the distribution of widths found by Arzoumanian et al. 2019 (blue histogram). Although the latter is somewhat broader than the former, which is not surprising since it is based on a much broader sample of 599 filaments in eight nearby clouds at different distances, it can be seen that the two distributions have very similar overall shapes and peak at the same value, just below 0.1 pc.
Overall, we measure inner D Plummer HP widths in the range of 0.06 to 0.11 pc, consistent within better than a factor of two with  Table 1 are only statistical fitting uncertainties. To assess the importance of systematic errors due to uncertainties in the dust properties, we also performed a similar radial profile analysis on the alternative column density map derived using Eq. (3) instead of Eq. (1) for the dust opacity law. The results are provided in Appendix A (Table A.1). Generally speaking, the measured radial profiles tend to be slightly shallower, with slightly lower p index values (by δp ∼ 0.2), in the alternative column density map. The inner widths derived from the alternative column density map range from 0.07 to 0.16 pc and are typically ∼ 30% higher than those found with the default column density map. They remain consistent with the filament widths found by Arzoumanian et al. (2019). Overall, uncertainties in the dust opacity law have a stronger influence on the absolute level of the column density profiles than on the profile shapes and the derived filament widths.
We also verified that our results on the widths do not depend on the choice of the algorithm employed to trace the crest of the ISF (see Appendix B). In particular, a Kolmogorov-Smirnov test reveals that the three distributions of individual widths measured along the ISF using the DisPerSE, FilFinder, and getsf crests, respectively, are statistically indistinguishable (see Fig. B.2).
The inner widths that we measure are almost always larger than the 0.02-0.06 pc range of values derived by Hacar et al. (2018) from combined ALMA+IRAM 30m observations in the high-density tracer N 2 H + (1-0). Because the N 2 H + (1-0) line is For the northern part of the ISF, they are also roughly consistent with the power-law exponent p = 1.5 ± 0.5 reported by Mattern et al. (2018) for another wide sample of almost 200 larger-scale, velocity-coherent filaments in the inner Galaxy from the AT-LASGAL 870 µm and SEDIGISM 13 CO(2-1) surveys (Schuller et al. 2009(Schuller et al. , 2017. These relatively shallow exponents indicate that the main filament features pronounced non-Gaussian wings of lower-density material that extend well beyond 0.1 pc from the ridge, as evident in Figs. 5 to 7. The profile of ISF-OMC-1 is strongly affected by the Orion-KL star-forming region, resulting in a somewhat steeper exponent. The mean central H 2 densities inferred from our Plummer fits range from ∼ 4×10 5 cm −3 for OMC-3 and OMC-2 to ∼ 10 6 cm −3 for OMC-1, assuming the ISF is in the plane of the sky (i.e. i = 0). The mean central density we infer for the OMC-1 portion of the ISF is a factor of three lower than the nominal volume density of 3 × 10 6 cm −3 found by Teng & Hirano (2020) for N 2 H + filamentary structures in OMC-1 (outside of cores) through non-LTE modelling of the N 2 H + (3-2) and (1-0) lines. Our dust continuum estimate of the volume density is, however, affected by a rather large uncertainty of a factor of ∼2-3 or more in the OMC-1 subregion, which is due to particularly uncertain dust opacity and temperature gradient effects in this unusually dense and strongly irradiated area (see e.g. Figs. A.1 and A.2 in Appendix A). The N 2 H + density estimates of Teng & Hirano (2020) are also uncertain by a factor of ∼3 (see their Table 3). Given these uncertainties, the two sets of density estimates are broadly consistent.
By integrating the radial profiles measured in the column density map of Fig. 3 over radii up to D HP (i.e. twice the halfpower radius), we also derived robust estimates of the mean mass per unit length of the inner portion of each ISF segment, excluding low-density material at large radii. The HGBS re- sults suggest that most stars form within this inner portion of molecular filaments (e.g. Könyves et al. 2015Könyves et al. , 2020. We obtain average values of ∼ 200 M /pc for OMC-3, ∼ 250 M /pc for OMC-2, ∼ 500 M /pc for the OMC-1 portion, and ∼ 300 M /pc for the whole northern ISF (i.e. OMC-1+OMC-2+OMC-3). For the filament shown in Fig. 9, which can be associated with fibre 9 of Hacar et al. (2018), we estimate a mean line mass of ∼ 100 M /pc (see the column density profile in Fig. 10). These line mass estimates were validated via comparison with the lowresolution (5 ) dust optical depth map of Planck Collaboration et al. (2014). They are more than one order of magnitude higher than the line masses of the 'fertile fibres' observed in the nearby Taurus cloud (Hacar et al. 2013) and of the many transcritical filaments that dominate the filament line mass function (FLMF) in the Arzoumanian et al. (2019) HGBS sample (see Fig. 1 of André et al. 2019), but they are consistent with the bulk of confirmed Galactic filaments discussed by Mattern et al. (2018). Our measurements are also in agreement with the results of Schisano et al. (2020), who find a mean value of 250 M /pc for a sample of 18,000 Galactic filaments with reliable distance estimates. It is not surprising to find such high line mass values in the Orion ISF since it is known to be undergoing fragmentation and ongoing embedded star formation (e.g. Takahashi et al. 2013;Teixeira et al. 2016).

Density structure
The large-scale distribution of gas and dust in the northern part of Orion A is dominated by the ISF. Based on a column density map at 36 resolution built from HGBS data, Stutz & Gould (2016) find that the radial distribution of gas near the ISF is well represented by a power law. They reported a line mass profile (λ(w), in M /pc) as a function of enclosing radius (w) that follows the following function well: (1994), which roughly corresponds to the same dust opacity parameterisation as adopted in Sect. 3 but with β = 1.8 instead of β = 2. More precisely, the OH5 dust opacity at ArTéMiS wavelengths is a factor of ∼30% higher than the HGBS opacity law of Eq. (1). Equation (8) also corresponds to a line mass profile averaged over the full 1 • length of the ISF, including a 0.4 • long portion south of the field mapped with ArTéMiS, whose average column density is a factor of two or more lower than the northern part of the ISF discussed here, according to Herschel and Planck data. Considering these differences, the average line mass of ∼ 300 M /pc that we derived in Sect. 3.4 for the northern part of the ISF, integrated over a transverse enclosing radius of D HP ∼ 0.07-0.1 pc, is consistent with the value of K reported by Stutz & Gould (2016) for a radius of w = 1 pc.
The column density map used by Stutz & Gould (2016) is sensitive to scales between 0.05 pc and 8.5 pc. Based on our higher-resolution data, we find an inner plateau with a total halfpower width of order ∼ 0.06 pc to 0.11 pc (see Table 1). Therefore, while our results are consistent with the analysis of Stutz & Gould (2016), they allow us to resolve an inner break in the radial density and line mass profiles of the ISF that could not be seen with Herschel data only. From their analysis at 36 resolution, Stutz & Gould (2016) did not find any significant variation of the column density profile along the extent of the ISF, but this is not surprising since they did not resolve the inner plateau in the density profile.

Comparison with high-density fibres
Although it is nearly impossible to associate each N 2 H + fibre reported by Hacar et al. (2018) with the ArTéMiS/SPIRE filaments that we extracted with DisPerSE, we will now discuss a few examples.

ISF-OMC-2
First, we focused on the northern part of the ISF, where the main filament shows a relatively clear structure as compared to the highly complex network of filaments around the extreme region Orion-KL. The map published by Hacar et al. (2018) covers only −05 • 25 ≤ δ ≤ −05 • 05 ; therefore, we considered only the southern part of what we call ISF-OMC-2/3, which roughly corresponds to OMC-2. We analysed the radial column density profiles of different segments of the ISF, which correspond to different ranges in declination, as shown in Fig. 11. For each segment, we give the parameters of the best-fit Plummer and Gaussian models in Table 1, together with the range of FWHMs measured in N 2 H + by Hacar et al. (2018) for each fibre that we tentatively associate with the ArTéMiS filaments (see also Fig. 4 for the numbering of the fibres).
The segment from δ = −05 • 11 to δ = −05 • 09 30 , which roughly corresponds to fibre 43 from Hacar et al. (2018), contains the bright infrared source OMC-2 FIR 4. The values of the inner width that we measure for this segment (D Plummer HP = 0.07 ± 0.025 pc) are in the lowest boundary of our sample. However, we caution the reader that this measurement is strongly affected by the presence of the centrally condensed protostellar source OMC-2 FIR 4 within this short segment (Shimajiri et al. 2008;López-Sepulcre et al. 2013). Assessing the intrinsic width of the underlying parent filament at this location is more uncertain. Nevertheless, such peculiar sources have very little effect on our estimates of the global properties of the ISF since we evaluated them from median radial profiles.

ISF-OMC-1
In the southern part of our ArTéMiS map, we also split the ISF into several segments, but only north of the Hii region Orion-KL. We tentatively associated these segments with fibres seen in N 2 H + (see numbering in Fig. 4). The inner widths as measured from Gaussian and Plummer model fitting are also given in Table 1, together with the FWHMs measured in the N 2 H + data. Here also we derive inner widths significantly smaller than 0.1 pc for some portions of the main filament (e.g. the segment associated with fibre 21), where a dense protostellar core is embedded and locally affects our measurements.

Comparison between ArTéMiS and N 2 H +
We show in Fig. 12 the comparison between the D Plummer HP diameters measured on our column density map and the FWHM of the N 2 H + fibres that we associated with the different segments of the ISF, as listed in Table 1. This figure shows that there is a trend of increasing FWHM for the N 2 H + fibres with increasing D Plummer HP : A linear fit to the data results in a slope of 0.53 (±0.24) and an intercept of -0.005 (±0.020); the Pearson correlation coefficient is equal to 0.66. Although the correlation is only marginal, we interpret this as an indication that N 2 H + emission generally traces the densest portion along the central axis of the main filament.
Slight but significant deviations between the peaks in N 2 H + emission and the filament crests traced by our sub-millimetre dust continuum data are visible in some cases, such as the filament associated with fibre 9 of Hacar et al. (2018, see our Fig. 9a). In that particular case, it is noteworthy that there is better agreement between the ArTéMiS-SPIRE filament and the filamentary structure detected in NH 3 by Monsch et al. (2018) at 6 resolution (see Fig. 9b). At least in projection, fibre 9 lies between two large-scale H 2 'fingers' associated with the explosive CO outflow emanating from the immediate vicinity of Orion KL (cf. Zapata et al. 2009;Bally et al. 2017). The southern end of fibre 9 also overlaps with the area where high-velocity CO(6-5) emission was detected in the APEX observations of Peng et al. (2012). As CO is known to be a destroyer of N 2 H + (e.g. Aikawa et al. 2001), it is possible that energetic finger-like outflow features may significantly reduce the abundance of N 2 H + in the lateral edges of the dense filamentary structure. Strong anisotropic irradiation of fibre 9 by the Orion nebula (see e.g. the dust temperature map shown in Fig. 2) may also play a role via the desorption of CO molecules from grain mantles. In any event, a comparison of the N 2 H + map of Hacar et al. (2018) with the NH 3 map of Monsch et al. (2018) clearly indicates that the abundance of N 2 H + is far from uniform in the OMC-1 region. The most extreme difference between the two maps is that no N 2 H + emission is detected in the immediate vicinity of Orion KL (see also Teng & Hirano 2020), presumably due to the destruction of N 2 H + by CO, while strong NH 3 (1,1) and NH 3 (2,2) emission is seen in the Monsch et al. (2018) data and a prominent column density peak is observed in our data (cf. Figs. 1 and 3).

Conclusions
We have presented 350 and 450 µm observations with ArTéMiS of the ISF in Orion, the nearest site of active high-and intermediate-mass star formation. With angular resolutions of 8 and 10 , the ArTéMiS data, combined with Herschel-SPIRE maps to recover large-scale emission, probe scales from 0.015 pc to a few pc. By combining the new ArTéMiS data with Herschel data at shorter wavelengths, we were able to build, for the first time, high-resolution temperature and column density maps covering a large region around OMC-1 that are not affected by saturation; these maps are publicly available from the CDS 1 .
We extracted the radial profiles and intrinsic widths of several segments of the Orion ISF from the high-resolution column density map. Our conclusions can be summarised as follows.
-We resolve an inner plateau with a typical half-power width in the range 0.06 to 0.11 pc along the northern and southern 1 Link to A&A online or to the CDS. parts of the main filament. This plateau could not be seen in the Herschel-SPIRE data because of their limited spatial resolution of ∼0.06 pc.
-These values of the inner width are consistent within better than a factor of two with the values measured in several nearby molecular clouds of the Gould Belt (Arzoumanian et al. 2011Palmeirim et al. 2013) and in the massive star-forming complex NGC 6334 (André et al. 2016). -The mean line masses that we derive from our data (between ∼100 and ∼ 500 M /pc) are in the extreme upper range of those found for Herschel filaments in the Gould Belt André et al. 2019), as well as in the upper range of those measured for more than 18,000 filaments found in the Hi-GAL survey by Schisano et al. (2020). Our values are comparable to the line masses measured for larger-scale filaments throughout the Galaxy in the APEX/SEDIGISM survey (Mattern et al. 2018); this is consistent with the picture in which filament fragmentation and star formation occurs within the densest transcritical or supercritical portions of long, roughly ∼ 0.1 pc wide filaments.
The present study also highlights the importance of combining data that cover a wide range of spatial scales and of probing material at various densities. In particular, large high-resolution dust continuum maps with a high spatial dynamic range can efficiently trace a broad range of column densities, while complementary spectroscopic data are sensitive either to the densest inner regions of filaments (e.g. N 2 H + ) or to the surrounding diffuse medium (e.g. CO and isotopologues). (1). Contours are +5%, +20%, and +35%.

Appendix B: Influence of the filament-finding algorithm
To check whether the choice of the algorithm employed to trace filament crests had any significant effect on our results, we used two alternative algorithms, FilFinder (Koch & Rosolowsky 2015) and getsf (Men'shchikov 2021), in addition to DisPerSE (Sousbie 2011) to define the spine of the ISF. Figure B.1 shows the results. Although getsf tends to break up filamentary structures into smaller pieces than the other two algorithms, it can be seen that the three methods agree quite well in all three portions (OMC-1, 2, and 3) of the ISF. Moreover, the three distributions of individual widths measured along the whole ISF using the three sets of crests are very similar (see Fig. B.2). A Kolmogorov-Smirnov test confirms that the distribution of widths found with the DisPerSE crests (Fig. 8) is statistically indistinguishable from the other two distributions.

Appendix C: Influence of temperature variations along the line of sight
To evaluate the effect of temperature gradients along the line of sight on our results, we constructed synthetic images for a model filament with a Plummer-type density distribution (Eq. 4 with p = 2) and a realistic temperature distribution, assuming an ambient interstellar radiation field (ISRF) typical of the Orion region (G 0 = 1000) 2 . The cylindrical model filament was assumed to have a central column density of N 0 = 2 × 10 23 cm −2 on its crest, an intrinsic half-power diameter of D HP = 0.075 pc, and an outer diameter of 0.55 pc. It was also assumed to be embedded in a background cloud of column density N bg = 9 × 10 21 cm −2 and, for simplicity, to lie in the plane of sky. The properties of this model filament are roughly similar to those derived for the ISF (see Sect. 3 and Table 1). The synthetic dust temperature profile within the model filament was calculated for G 0 = 1000 using an analytic approximation formula on a grid of radiative transfer models performed with the MODUST code (see Bouwman et al. 2001 andAndré et al. 2003). Two sets of synthetic emission maps for the model filament were generated at all observed Herschel and ArTéMiS wavelengths, assuming optically thin dust emission with the nominal dust opacity law given by Eq.
(1) on one hand ('B.1 test') and the alternative dust opacity law given by Eq.
(3) on the other hand ('B.2 test'). Column density and dust temperature maps were then produced from these synthetic emission maps in the same manner as for the real data, 2 The radiation field in the OMC-1 subregion is somewhat higher than the value adopted in the model described here. Accordingly, the internal dust temperature gradient within the OMC-1 portion of the ISF and the corresponding effects on the OMC-1 column density profiles may be somewhat stronger than shown in Figs  by fitting a modified blackbody to the observed SEDs on a pixelby-pixel basis and assuming the same nominal dust opacity law in both cases (cf. Sect. 3.1). The purpose of the B.2 test is to assess the combined effect of line-of-sight temperature gradients and dust opacity uncertainties. The results of this experiment and a comparison with the input model are provided in Figs. C.1 and C.2, in the form of radial temperature and column density profiles, and in Table C.1. As expected, it can be seen in Figs. C.1a and C.2b that the temperature profiles derived from SED fitting (green curve) overestimate the intrinsic dust temperature profile of the model (black curve). This is due to the fact that the model filament is colder in its inner interior and that the temperatures derived from SED fitting represent line-of-sight averages that are significantly affected by the warmer outer layers of the filament. Accordingly, the column density profile derived assuming the correct dust opacity (green curve in Fig. C.1b) underestimates the intrinsic column density profile of the model (black curve in Fig. C.1b) by a factor of 2.3 at small radii. In the B.2 case, on the other hand, the temperature effect does not quite compensate for the incorrect assumption about the dust opacity, and the derived column density profile (green curve in Fig. C.2b) still exceeds the intrinsic column density profile (black curve in Fig. C.2b) by a factor of two at the centre of the filament.
The shape of the derived column density profiles is nevertheless very similar to the shape of the model column density profile. Consequently, the half-power diameters derived from Plummer fits to the SED-based column density profiles (red curves in Figs. C.1b and C.2b) remain close to (within 30% of) the true half-power diameter of the input model (see Table C.1). More precisely, the derived D HP diameter overestimates the intrinsic half-power diameter of the model (0.075 pc) by only ∼ 30% in test B.1 and underestimates it by only ∼ 20% in test B.2. By comparison, the derived central column density and p index values are more uncertain and affected by larger systematic errors (see Table C.1). These tests demonstrate that line-of-sight averaging effects have only a little impact on the filament width measurements reported in Sect. 3.

a) b)
Fig. C.1. Synthetic dust temperature (a) and column density (b) profiles of a model filament with Plummer-type density structure, illustrating the effect of temperature gradients along the line of sight. The true dust opacity law and the dust opacity law adopted to derive temperature and column density maps from synthetic emission maps were identical and given by Eq.
(1). a) Comparison between the intrinsic model temperature profile as a function of radius (black curve) and the temperature profile as a function of projected radius resulting from SED fitting and thus averaging along the line of sight (green curve). b) Comparison between the corresponding model column density profile (black curve) and the column density profile resulting from SED fitting (green curve). The dashed and dash-dotted red curves represent two Plummer fits to the SED column density profile, which differ only from each other and from the model in the outer regions.

a) b)
Fig. C.2. Same as Fig. C.1, for the same model filament but assuming the true dust opacity law is given by Eq.
(3) while the dust opacity law adopted to derive temperature and column density maps from SED fitting is given by Eq. (1).