The Gaia-ESO Survey: dynamics of ionized and neutral gas in the Lagoon nebula (M8) ††thanks: Based on observations collected with the FLAMES spectrograph at VLT/UT2 telescope (Paranal Observatory, ESO, Chile), for the Gaia-ESO Large Public Survey (program 188.B-3002). Tables 2 and 3 are only available in electronic form at the CDS via anonymous ftp to cdsarc.u-strasbg.fr (22.214.171.124) or via http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/
Key Words.:ISM: individual objects: (Lagoon nebula) – ISM: general – HII regions
Aims:We present a spectroscopic study of the dynamics of the ionized and neutral gas throughout the Lagoon nebula (M8), using VLT/FLAMES data from the Gaia-ESO Survey. The new data permit to explore the physical connections between the nebular gas and the stellar population of the associated star cluster NGC6530.
Methods:We characterize through spectral fitting emission lines of H, [N II] and [S II] doublets, [O III], and absorption lines of sodium D doublet, using data from the FLAMES/Giraffe and UVES spectrographs, on more than 1000 sightlines towards the entire face of the Lagoon nebula. Gas temperatures are derived from line-width comparisons, densities from the [S II] doublet ratio, and ionization parameter from H/[N II] ratio. Although doubly-peaked emission profiles are rarely found, line asymmetries often imply multiple velocity components along the same line of sight. This is especially true for the sodium absorption, and for the [O III] lines.
Results:Spatial maps for density and ionization are derived, and compared to other known properties of the nebula and of its massive stars 9 Sgr, Herschel 36 and HD 165052 which are confirmed to provide most of the ionizing flux. The detailed velocity fields across the nebula show several expanding shells, related to the cluster NGC6530, the O stars 9 Sgr and Herschel 36, and the massive protostar M8East-IR. The origins of kinematical expansion and ionization of the NGC6530 shell appear to be different. We are able to put constrains on the line-of-sight (relative or absolute) distances between some of these objects and the molecular cloud. The data show that the large obscuring band running through the middle of the nebula is being compressed by both sides, which might explain its enhanced density. We also find an unexplained large-scale velocity gradient across the entire nebula. At larger distances, the transition from ionized to neutral gas is studied using the sodium lines.
The Lagoon nebula (M8, NGC6523) is one of the brightest H regions in the solar neighborhood, and has been the subject of many observational studies (e.g., Lada et al. 1976 in CO and optical lines, Tothill et al. 2002 in CO and sub-mm, Takeuchi et al. 2010 in CO). It harbors the young cluster NGC6530, only a few Myrs old, and also intensively studied especially in recent years at optical, infrared and X-ray wavelengths (e.g. Walker 1957, van den Ancker et al. 1997, Sung et al. 2000, Damiani et al. 2004, 2006, Prisinzano et al. 2005, 2007, Kumar and Anandarao 2010, Povich et al. 2013). The H region is illuminated by many massive stars of O and B spectral types, the hottest one being 9 Sgr (HD 164794, type O4V((f))z); a few other late-O/B type stars are also found in the region. The optically brightest part of the Lagoon nebula is the so-called Hourglass nebula, which surrounds and partially oscures the O7:V star Herschel 36 (Woodward et al. 1986); stars in the Hourglass are thought to be younger than the NGC6530 cluster. Also noteworthy is the presence, to the East of NGC6530, of the embedded massive protostar M8E-IR (Wright et al. 1977, Simon et al. 1984, Henning and Gürtler 1986), indicating that star formation in the region has also taken place recently. The most recent determination of the distance of the NGC6530 cluster (and by inference of the H region as well) is 1250 pc (Prisinzano et al. 2005). The properties of the whole region were reviewed by Tothill et al. (2008).
The Lagoon nebula and its stellar population show a well-defined spatial organization. The brightest nebular region (Hourglass) does not lie near the geometrical center of the whole nebula, but several arcmin ( pc) to the West. Closer to the nebula center lies instead the bulk of low-mass cluster stars (Damiani et al. 2004), cospatial with the B stars. The most massive member 9 Sgr is also offset with respect to the B stars, but appears not directly related to any localized bright nebulosity, leading Lada et al. (1976) to suggest that it actually lies several parsecs in front of the nebula, not within it. All around the central region, many bright-rimmed dark clouds are found, which being located along the outer border of the nebula suggest the “blister” nature of the H region. Behind one of them to the South-East of the main cluster, the mentioned young, massive star M8E-IR is found. This and other low-mass stars in the same neighborhood show indications of being younger than the main NGC6530 cluster (Damiani et al. 2004). The bright Hourglass region has received much more attention than the outer nebula parts, which remain relatively little studied.
In this work we study the kinematics of the ionized and neutral gas giving rise to the strong optical emission lines (H, [N II], [S II], [O III]), and sodium D absorption lines, across the whole nebula, using new spectroscopic data from the Gaia-ESO Survey (Gilmore et al. 2012, Randich et al. 2013). In Section 2 we describe the observations, while in Section 3 we present our results. In Section 4 we discuss the main implications of our work for the structure and dynamics of the Lagoon nebula, including issues that deserve further research.
2 Observational data
The observations of the Lagoon nebula were obtained as part of the Gaia-ESO Survey (internal release iDR4), targeting Milky Way stars and clusters across a wide range of parameters with the ESO VLT/FLAMES multi-fibre spectrograph (Pasquini et al. 2002). Twenty-seven observing blocks (OBs) were devoted to NGC6530 as part of the Survey program, down to a limiting stellar magnitude . Observations were spread over 17 nights, in September 2012 and June-September 2013. Fifteen fibres per OB are targeted at star-free sky positions, interspersed between stellar position, with the intent of obtaining purely nebular spectra needed to correct the stellar spectra through subtraction. Spectrographs used were Giraffe (setup HR15N, , wavelength range 6444-6818Å) and UVES (setup 580, , range 4768-6830Å; and setup 520, , range 4180-6212Å). The number of sky spectra obtained with Giraffe is 647 (on 197 individual sky positions), 4 spectra with UVES setup 580 (3 positions), and 25 spectra with UVES setup 520 (21 positions). In addition, being so bright, the nebula dominates clearly the stellar spectra of faint stars, and the strongest nebular lines could be studied with good results whenever the ratio between H peak and stellar continuum is larger than 100 (henceforth “faint stars”), so that the H line cannot be dominated by the star, even in strongly accreting T Tauri stars. The same method was successfully used by us in a previous work on the emission of the Carina nebula using Gaia-ESO data (Damiani et al. 2016). The number of faint star spectra in the NGC6530 Gaia-ESO dataset is 980 (892 individual stars, with some having been observed more than once). Therefore, only considering Giraffe spectra, we have nebular spectra for 1089 positions across the Lagoon nebula. The Giraffe HR15N wavelength range includes several strong nebular lines: H, the neighboring [N II] lines at 6548, 6584Å, and the two [S II] lines at 6716, 6731Å. Also the He I line at 6678Å falls in the range, but is found to be too weak in our spectra for a proper study. Spectra for the same sky position were coadded to improve the S/N ratio. All the studied nebular lines are orders of magnitude stronger than atmospheric lines, as estimated from other Gaia-ESO spectra in clusters free of nebulosity.
The UVES sky spectra are very few as mentioned, and add little information despite the much wider wavelength range; they are therefore not considered further. On the other hand, the UVES spectra of stellar targets in NGC6530 are more numerous (16 stars using setup 580, and 44 using setup 520, mostly of early type), and were used to study two important lines: the [O III] line at 5007Å, clearly detected in most spectra, and the Na I D1, D2 absorption lines at 5895.92, 5889.95Å. The latter enable us to perform a comparative study of the ionized and the neutral gas in the region. Like the ionic lines, also these sodium absorption lines are much stronger than typical atmospheric sodium absorption. An H image of the Lagoon nebula from the VPHAS survey (Drew et al. 2014) is shown in Figure 1, together with all sky positions considered here, and positions of several of the most massive stars.
Since the exposure times (20-50 min) were determined by the requirements dictated by the faintest stars, the signal-to-noise ratio (S/N) in the main nebular lines is usually very high (but lines remain unsaturated). The high S/N and the large number of nebular positions comprised in our dataset make it one of the richest datasets ever available for the study being performed.
3.1 H, [N II] and [S II] lines from Giraffe data
In order to study the dynamics of the nebular gas the optical emission lines were analyzed and modelled with analytical functions, starting with lines from Giraffe spectra. Figures 2 to 4 show several examples of the observed lines of H, [N II] 6584Å, and [S II] 6731Å (panels labeled with coordinates), and the residuals after subtraction of single-Gaussian (panels labeled with ‘1-g’) or double-Gaussian (labeled ‘2-g’) best-fit models. In nearly all cases the optical lines are found single-peaked: among all 197 pure-sky fibres only two cases of doubly-peaked lines are found (Fig. 4, right panels), and only in the [N II] and [S II] lines while not in H. In general, lines are found to be nearly Gaussian in shape, suggesting a single kinematical component along each line of sight for each emitting ion, as the representative examples of Figs. 2 and 3 show. However, while a single Gaussian (henceforth “1-g” model) accounts well for the bulk of the emission in each line, examination of the fitting residuals (panels ‘1-g’) reveals that the detailed line shape is systematically misreproduced by a simple Gaussian function, by a small but significant amount in terms of the available S/N (note that the ordinate scale in each of the 1-g residuals panels is 1/15 that of the panel above it). The systematic pattern of the 1-g model residuals is suggestive of at least two unresolved velocity components along the same sightline; only in rare cases the 1-g residuals show no such pattern (Fig. 2, spectrum #12). In order to model the line profiles in detail, double-Gaussian (“2-g”) models were therefore attempted, whose residuals are also shown in Figs. 2-4 (panels ‘2-g’): these show in nearly all cases no systematic patterns, indicating that two Gaussian components provide a sufficient characterization of emission lines in the Lagoon nebula. It should be remarked that each emission line (H, [N II] 6584 and [S II] 6717, 6731Å) was fitted independently; as the first three panels of Fig. 3 show particularly well, the H emission peak may lie at velocities significantly different than the peaks of [N II] and [S II] lines. All velocities here are heliocentric. Results from our 1-g best fits are reported in Table 2.
While the 2-g models can be considered “better” than the 1-g ones because of the smaller residuals, the corresponding best-fit parameters must be treated with caution. The “blue” and “red” best-fit radial velocities (RVs) of each modelled line are in most cases only a few km/s from one another, i.e. much closer together than the line widths () themselves: this makes the relative intensities of the two Gaussian components much more uncertain than their total value, which is instead robustly modelled by 1-g fits (this is especially true of low S/N spectra). Therefore, the choice of considering one set of model parameters or the other will depend on the specific problem. We find that diagnostics involving two lines simultaneously (e.g., line-intensity ratios, or linewidths comparisons) are derived much more robustly from 1-g fitting parameters; on the other hand, the description of the gas dynamics considering RVs alone benefits also from the 2-g model velocities, whenever S/N is sufficient. In this respect, we remark that the smaller thermal widths of the [N II] and [S II] lines permit often a more accurate derivation of components’ RVs, despite these lines being weaker than H.
In several cases among those shown in Figures 2 to 4, residuals from 1-g fits show a blue component. This might be indicative of an approaching ionized layer, blueshifted with respect to the bulk of ionized gas, and reminiscent of the layer in the outer part of the Orion nebula, known as the Veil (see the reviews by O’ Dell 2001, O’ Dell et al. 2008). More detailed indications on the dynamics of the neutral gas probably associated with this layer in the M8 nebula are derived from the sodium absorption lines in Section 3.2 below.
Some general properties of the ionized gas may be derived from the best-fit parameters. Figure 5 shows a comparison between RVs derived from 1-g fits to H and [N II] 6584Å lines. The same pattern is shown by measurements from pure-sky fibres (black dots) and from faint stars (gray), confirming that the usage of faint-star spectra introduces no biases in the derived nebular properties. The datapoints scatter is not caused by errors ( km/s as a rule) but is real: in many cases the [N II] RV is less negative than the H RV. The velocity of the low-mass stars in the NGC6530 cluster is km/s (Prisinzano et al. 2007), which is assumed to coincide with the center-of-mass velocity of all cluster stars. The Figure then shows than the bulk of ionized gas emitting in these lines has negative velocities (approaching us) from the standpoint of the cluster center of mass. This is unlike the case of the Carina nebula, where the double emission line peaks bracket usually the center-of-mass RV, indicating expansion towards both the near and far sides (Damiani et al. 2016). In NGC6530, the ionized gas seems instead to expand predominantly towards our side (but more details will be studied in Sections 3.3 and 3.4 below). Figure 5 however shows that along some sightlines the [N II] moves away from us, while hydrogen moves towards us: clearly, the different lines arise from dynamically distinct gas layers, a feature which cannot be understood from narrow-band images, which mix emission from all velocity layers (e.g. Tothill et al. 2008, fig.4). The comparison between H and [S II] RVs in Figure 5 shows a pattern very similar to Figure 5. This suggests that the [N II] and [S II] lines originate from gas layers which are more closely related mutually than with H. This is confirmed by the good correlation shown in Figure 5, between the [N II] and [S II] RVs, the small systematic shifts being attributable to the uncertainty in the adopted line wavelengths111Wavelengths adopted here are 6583.43Å for [N II], and 6716.44, 6730.815Å for [S II], the latter two from Kaufman and Martin (1993).. This figure also permits to estimate empirically the maximum errors in the best-fit RVs from the scatter of datapoints in the correlation, on the order of km/s. We remark that the [S II] and [N II] lines are the weakest being studied, the H line being much stronger; therefore, the scatter of datapoints in both Figures 5 and is certainly dominated by real effects.
The above results are reminiscent of those found in the Orion nebula, a well studied blister H region, where ionized gas flows away from the ionization front, and ionized hydrogen is found at more negative speeds with respect to [N II] and [S II], the ionization level gradually increasing as the gas acquires larger and larger speeds in a champagne-flow geometry (e.g., Balick et al. 1974, O’ Dell et al. 1993). The analogy with the Orion nebula, although very interesting, cannot however be pushed too far, since there are also important differences between it and M8: this latter is a much larger region, with more than one ionizing O-type star, and is probably also a more evolved blister, where at least the most massive star 9 Sgr has excavated a larger cavity in the parent cloud compared to that excavated by the Orion most massive member, Ori C. We will examine in more detail the relative geometry of 9 Sgr and the M8 nebula in Section 3.3.2.
A comparison between the 1-g model intensities of H and [N II] lines is shown in Figure 5. The intensity ratio is significantly non-uniform; this, under the typical conditions found in H regions, suggests significant differences in the ionization parameter across the region (e.g., Viironen et al. 2007), which will be studied in Section 3.3. In very general terms, in regions with high Lyman-continuum flux ionization will be highest, and hydrogen lines dominate over [N II] lines; in the same region, the diagram suggests that the highest densities and largest surfact brightnesses are also found. In Section 3.3 we will examine in much better detail how these quantities depend on position across the nebula. Figure 6 shows the best-fit linewidths of H and [N II], whose comparison provides a measure of temperature (since turbulent and instrumental broadenings are the same for the two lines). To avoid mixing unrelated gaseous layers, we only show datapoints having maximum absolute RV differences of 3 km/s between the two lines. In the Figure, dotted lines indicate loci for fixed temperatures of 5000, 10000, and 15000 K, and a range of combined turbulent+instrumental broadening (between [8-18] km/s). Most datapoints lie between 5000-10000 K, however with considerable spread.
We then examine the details of line profiles, as provided by our 2-g best-fit models. In Figure 7 the RV of the “blue” component is compared with that of the respective “red” component in the same spectrum; the orange point indicates the cluster . Also these more detailed RVs show motions predominantly towards us (as seen from center of mass); only a small number of spectra show blue and red RV components lying on opposite sides of (i.e., km/s and km/s), as it would be expected for an expansion originated from center of mass velocity. In the large majority of cases, the two components, both approaching us, have RVs differing by only few km/s, whose physical origin is not completely clear. One possibility is that the two components are actually an oversimplification of reality, and that they only represent the approximate RV range found in a rapidly decelerating layer emitting H. This accounts well for the significant correlation found in Fig. 7 between and in the H line222Alternatively, this might be an instrumental effect arising from the non-gaussianity of the line-spread-function, a systematic effect that becomes observable at the highest signal levels, see Damiani et al. (2016, Appendix).. The median intensity of the blue component is only slightly higher than that of the red component (1.3 to 1.5 times, from pure-sky and faint-star fibres respectively), thus backscattering from dust is unlikely to account for the bulk of the red component.
Still different is the picture derived from the corresponding diagram involving the [N II] line (Figure 8). Motion receding from us is much more frequently found, and the correlation between and is much less tight, if existing at all. The blue and red components of [N II] are therefore in many cases indicative of dynamically distinct gas layers, whose spatial characteristics will be examined in detail in Section 3.3. It is interesting to compare the RVs for the H and [N II] lines, as given by the 2-g fits, analogously to the 1-g RV comparison of Fig. 5. This is done in Figure 9: here we see that, unlike the 1-g fit RVs, there is in most cases a fairly good match between RVs of the two lines, apart from a minority of strong outliers. In order to reconcile this agreement with the disagreement of 1-g RVs shown by Fig. 5, one may expect that the relative intensities of the blue and red components are different between H and [N II], with the red component being dominant in [N II] over H, and viceversa for the blue component. This expectation is confirmed by the diagram of Figure 10, where the intensities of the different lines are compared (for each blue/red component separately). Only components with an absolute RV difference less than 3 km/s were plotted. It is clear that red components tend to have a larger [N II]/H intensity ratio than blue components, and therefore lower ionization, as mentioned above. Recalling the above result that the two components are diagnosing different layers from gas moving in the same direction, we infer that the gas moving faster (larger negative velocities: blue component) is more ionized than the slower red component.
3.2 [O III] and sodium D lines from UVES data
A representative selection of line profiles of the [O III] 5007Å line333The exact wavelength used here is 5006.843Å, from NIST (http://physics.nist.gov/asd, Kramida et al. 2015). from UVES spectra of several stars is shown in Figures 11 and 12. Although all spectra are from stars and not sky fibres, the nebular line is clearly evident; no stellar spectrum subtraction was performed (apart from a constant continuum level) for any of the spectra shown. This line is frequently the second strongest nebular line in our spectra, after H. In the [O III] line doubly-peaked profiles are much more frequently found (as already reported by Elliot and Meaburn 1975) than in lower-ionization lines like those of [N II]. Even if no double peaks are found, line distortions are evidently present in most cases, and require 2-g fits. In only one case (labeled star 520-5, upper left panel in Fig. 12) two Gaussians provide a bad fit (as shown), and three components are suggested by the line profile: this is the O7 star Herschel 36 in the Hourglass nebula, having also the strongest [O III] line among our sample. Shown with green lines in Figures 11-12 are our 2-g model fits, which reproduce the [O III] profiles very well, except for Herschel 36.
The UVES spectra of stars in NGC6530 also enable us to study the neutral gas along the line of sight to these stars, using the sodium D doublet. The selection of UVES targets in the Gaia-ESO Survey is such as to maximize the probability of their cluster membership (Bragaglia et al. , in preparation); therefore, most (not all) UVES spectra are likely to sample the entire column of neutral sodium between us and NGC6530. A wide selection of sodium absorption line profiles for these stars is presented in Figure 13. For not-too-hot stars, the stellar sodium lines were also evident in the spectra: these were divided out, using suitable template spectra chosen among the UVES-POP library (Bagnulo et al. 2003), such that Fig. 13 shows only non-stellar absorption components. Both doublet lines are shown, the D1 line in black, and the stronger D2 line in red. The occasional features at and 30 km/s in the D1 line (only) are telluric. The intensity ratios of the two lines are closely related, being originated from the same (ground) level, and having oscillator strengths of (D2) and (D1). The low ionization energy of sodium (5.139 eV) implies that these lines must originate from layers more distant from the OB stars (nearer to us) than the ionic lines studied in Section 3.1. In basically all cases, more than one absorbing layer, each with a distinct RV, is needed to model the sodium absorption profiles. The most complex line profiles is modelled with five Gaussians.
The fitting function was chosen as follows: we rewrite eq.(1) from Hobbs (1974) using instead of as (for a single component)
and the term containing the residual intensity , for a combination of absorbing layers , as
where is the central wavelength of absorbing component , at velocity , and is the line rest wavelength. is the component intrinsic width in wavelength units. The explicit form for is therefore:
Since the lines from cold neutral sodium may be very narrow, and saturated absorption profiles may become highly non-Gaussian, the instrumental resolution of UVES was separately introduced as a fixed Gaussian broadening through convolution, and the final functional form used to fit the observed, normalized line profiles is:
where is the convolution operator and ( km/s) is the UVES instrumental linewidth. The chosen function for corresponds to absorption from cold gas, whose emissivity is approximated as zero. Using this function, non-linear fits were attempted to the D2 line profiles (the strongest of the doublet) with two to five components; the optimal number of fitting components was determined by inspecting the results visually. The chosen best-fit models of D2 lines are superimposed to the observed profiles in Fig. 13 as orange curves: these match so well the observed profiles as to result often indistinguishable from them. Velocities of the individual best-fit components are indicated in the Figure with vertical red dashed lines. They range from km/s to km/s; however, positive-velocity components are found only in two cases, the most evident example being that of star 520-14 in the Figure. As it should be clear from eq.1, the profiles of the D1 line can be obtained from those of the D2 line (eqs.3 and 4) by multiplying the intensity of each component by the constant factor . Model profiles of the D1 line obtained in this way are also shown in Fig. 13 as blue lines. In most cases, they match well the observed D1 profiles (black), providing a test of the goodness of the adopted best-fit models. Best-fit parameter values are reported in Table 3. Intensity values correspond to saturated components and are highly uncertain.
However, a number of cases are found where the observed D1 profile and its model do not match (see e.g. stars 520-13, 520-14, 520-28, 520-36, 520-39), to a significant degree for the given S/N. The discrepancy occurs always near velocities of to km/s. It is always in the same sense, with the observed D1 profile being deeper and closer to the respective D2 profile than the model would predict: this is suggestive of that particular component saturating towards a finite, non-zero intensity, contrary to the assumption above of zero emissivity. If this latter condition is not met, the profile modeling becomes extremely more complex, the contributions from different layers not being writable as a linear sum in ; also the order of layers along the line of sight becomes important (the emission of one layer can be absorbed only by the layers closer to us), which is not the case when pure absorption is modelled. We have therefore not modelled those few cases in quantitative detail. Qualitatively, it is suggested that the layers at velocities to km/s are sometimes hotter than those at lower and higher velocities, their source function being much higher than zero compared to the photospheres of the UVES targets. Temperatures of several thousands K are therefore likely for those particular sodium layers. We consider it likely that these layers are spatially adjacent to those where recombination has recently taken place, as also supported by their similar velocities. The higher temperatures in the sodium layers at to km/s are also supported by their increased line widths , as shown by Figure 14. We remark that this figure shows intrinsic line widths , not the observed values, degraded by the instrumental resolution (as in Eq. 4). For this reason, their lower bound was set to zero in the fitting; however, values of lower than km/s should be regarded as upper limits.
The wide variety of absorption profiles found within small angular scales on the sky suggests strongly that the associated neutral gas lies in the immediate vicinity of the Lagoon nebula, and not all along the line of sight. Also the observed large radial velocities of sodium absorption are not expected for the general ISM gas in nearly circular orbits and observed almost in the direction of Galactic center, and further support this argument. The same cannot be said for the strong, saturated component near km/s, which although not strictly identical in all profiles is so widespread (and consistently narrow) that we cannot rule out a line-of-sight origin. A puzzling feature from the sodium spectra is the clear existence of components at much larger negative velocities than found from the ionic lines in Section 3.1: at face value, this would imply that the expanding gas, after recombining at a typical approaching velocity of to km/s (Fig. 5), keeps accelerating towards us to produce the neutral layers around to km/s. Also the origin of the discrete velocity components (as opposed to a continuous velocity distribution) in the sodium profiles is unclear, i.e. whether it is related to a episodic energy input or to different pre-existing layers of neutral gas, swept by the expanding envelope. In the former case, a single star would be driving the phenomenon, or otherwise the uncorrelated input from many stars would produce a smoother velocity distribution. In the second, a continuously (non-episodically) expanding shell would encounter discrete gas layers, and decelerate as more and more mass is being pushed; this seems contradictory with the above suggestion that neutral gas is accelerated outwards, in the sodium-line formation region.
We have also computed equivalent widths for the D1 line (labeled D1 EW in Figure 13), which fall in the range Å for the bulk of stars. Using the relation given by Munari and Zwitter (1997), this range would imply a range of , in good agreement with that found by Sung et al. (2000) for the NGC6530 massive stars.
3.3 Spatial maps
Most of the properties characterizing the nebular emission, as derived from the line modelization discussed in Section 3.1, show distinct spatial patterns, which we examine in detail here. Figure 15 shows a map of the [N II]/H intensity ratio (proportional to symbol size), superimposed on a VPHAS image of the nebula as in Fig. 1. As mentioned in Section 3.1, a smaller intensity ratio corresponds to higher ionization parameter. While one would expect the latter to increase in the vicinity of the hottest stars (red triangles in the Figure), this is not always found in our data. Not considering the two O stars near the East (HD 165246) and West (HD 164536) edges of the nebula, where our spatial coverage is sparse, we observe a clear ionization increase around HD 165052 (in South-East region), while the bright nebula surrounding the Hourglass shows no clear ionization pattern around its three O stars (North to South: HD 164816, 9 Sgr, and Herschel 36). The darker nebula parts (the “Great Rift” in Lada et al. 1976) around this brightest region shows distinctly higher ionization than near the Hourglass. The relatively low ionization near 9 Sgr supports further the suggestion by Lada et al. (1976) that this star should lie several pc in front of the nebular material, not in its immediate vicinity. The irregular ionization near Herschel 36, instead, can be attributed to the patchy distribution of dense gas and dust all around this star, as suggested by the HST image of the Hourglass (Tothill et al. 2008). Near the central NGC6530 cluster, to the S-E of the Great Rift, ionization is often higher than in the Hourglass region; however, there seems to be a gradient across the cluster, not a peak near its center, so that the source of ionization is probably not internal to the cluster, with the best candidate remaining the O4 star 9 Sgr, despite being a few parsecs away. The issue will be examined in more detail in Section 3.3.2 below. The bright blue supergiant HD 164865 (B9Iab) probably contributes to ionization locally, but not predominantly since there is no associated ionization peak near it.
We next consider the nebular electron density, as measured from the ratio of [S II] 6717, 6731Å lines (Figure 15). The spatial pattern is here very different than the ionization pattern: a strong and distinct increase in density ( cm, in agreement with Bohuski 1973) is found throughout the vicinity of the Hourglass, with the peak coinciding with the Hourglass proper. The agreement between line ratios from pure-sky fibres and faint-star spectra is very good, which is especially crucial here because of the near coincidence between the [S II] 6717 line and a Ca I photospheric line. A very localized enhancement in density is also found in the immediate vicinity of the bright rim near M8E-IR (green triangle S-E of NGC6530). Near the stellar cluster core, instead, the density is not particularly high ( cm), not higher than in the neighboring Great Rift, despite the large difference in nebular brightness. The density decreases very smoothly towards the nebula edges, to cm. The density was derived from the doublet ratio using analytic expressions in Weedman (1968) and Saraph and Seaton (1970).
The radial velocity map for H, shown in Figure 16 provides a vast amount of information. Velocities (proportional to circle sizes) are here those derived from 1-g fits. Since , absolute velocities are nearly the same as , that is referred to the NGC6530 center of mass. There is no central symmetry in the velocity field. The cluster core (red plus sign) nearly coincides with a (negative) velocity maximum: absolute velocities decrease towards both S-E and N-W from this position. The location of these low-absolute-velocity datapoints defines a reference direction, indicated with a red arrow in the Figure; this is almost coincident with the normal to the galactic plane. The galactic plane itself is just off the figure region to the right. The Hourglass region is also characterized by large negative velocities, but no velocity minimum is detected West of it, probably also because of the incomplete spatial coverage. In the outermost nebula regions to the East the velocity smoothly decreases towards . However, this does not happen in the western edge, where velocities remain at large negative values, a surprising fact which will be discussed in more depth in Section 3.4.3.
The velocity map obtained from the 1-g [N II] line fits is instead shown in Figure 17: although it presents many similarities to the analogous map for H of Fig. 16, there are also important differences: in the Hourglass region the negative velocity maximum is much less pronounced; on the contrary, in the Great Rift even positive velocities are found (orange in the Figure). That is, the H and the [N II] lines in this region indicate gas moving in opposite directions with respect to the cluster center-of-mass velocity km/s. In the N-W part, the puzzling large negative velocities are again found.
The [S II] velocity map of Figure 18 confirms these trends even more, with slightly positive velocities being found also to S-E of NGC6530 cluster core, almost parallel to the Great Rift. The positive-velocity datapoints in this region follow closely the inner border of the bright-rimmed dark cloud hosting the massive protostar M8E-IR (green triangle), while velocities just outside the bright rim (i.e., projected against the most obscured part) become suddenly negative. Again, the sharpest velocity gradients occur along a line joining M8E-IR with the Hourglass region (arrow in Fig. 16), so that a more detailed understanding can be achieved from considering position-velocity diagrams along this direction. Before doing that, however, we consider the spatial maps obtained from results of the 2-g fits.
A 2-g velocity map for [N II] is shown in Figure 18. We prefer to study this line with respect to the brighter H because the narrower line widths permit a better determination of individual component velocities. We have omitted the weakest components, which contribute more to the noise than to show a clear pattern. The most important features shown are the crowdings of positive-velocity red/orange datapoints near M8E-IR and along a vertical strip passing through 9 Sgr/Hourglass. With big green dots are indicated the positions of the sub-mm knots found by Tothill et al. (2002; they did not explore the northern half of the nebula): the orange circles fill almost exactly the arc-shaped region delimited to the South by those knots, and ending with M8E-IR to the East. Also the orange circles in the Hourglass region tend to fill the interior of a region delimited by the sub-mm knots to the South. Almost everywhere else, no receding component (with respect to ) is observed in the ionized gas, even considering the 2-g line profile models.
3.3.1 Peculiar locations
The image of the nebula shows several dark globules or “elephant trunks” projected against the bright nebular background. For some of these dark nebulae we have fibre spectra, enabling us to discriminate the properties of the foreground gas against that of the brighter background. Two such examples are shown in Figure 19, left panels. The one in the lower left panel (called the “Dragon” by Brand and Zealey 1978 and Tothill et al. 2008) is one of the most evident, and its spectrum was shown in Figure 4, labeled as nr. 124. This latter reveals that the nebular emission, although attenuated by a factor with respect to adjacent unobscured positions, is still substantial, and images show it as dark only because of the sharp contrast with the surrounding bright emission; nevertheless, this spectrum shows a peculiar component, as a significant emission residual at velocity km/s. This is not seen, at least not as clearly as here, in any other spectrum, including those of other obscured patches like that in the upper left panel of Fig. 19. Therefore, at least in the direction of the “Dragon”, we see that ionized gas with velocity km/s exists above a distance from the nebula enclosing 90% of the bulk emission, from simple aperture photometry with respect to nearby unobscured positions. The existence of faster gas component at large distances agrees with the results from the sodium lines presented in Section 3.2, and with those of Meaburn (1971) on [O III] lines.
The right panels of Fig. 19 show instead the locations, in the extreme West of the entire nebula, where the most asymmetric or even splitted line profiles are found in our dataset (spectra labeled as nr. 72, 92, 93 both here and in Fig. 4); here again the nebular emission is so weak as to appear nonexistent in the narrow-band image, but is enough to be detected and studied in our spectra. Figures 4 and 19 together show that the nebular line profiles vary smoothly with spatial position, which confirms that their peculiarities do not arise from some random effect, or isolated cloudlet. We will discuss therefore the implied large-scale motion of this part of the nebula in Section 3.4.3.
3.3.2 The position of 9 Sgr
Another peculiar location is that of the most massive star, 9 Sgr, which as mentioned is suspected to lie at some distance (5-10 pc according to Lada et al. 1976) in front of the cloud. Assuming like these authors that it is the dominant ionizing source for the entire nebula (i.e., except in the vicinity of Herschel 36 or HD 165052), its line-of-sight distance from the nebula may actually be estimated from the decay of measured ionization with sky-projected distance. We are able to estimate the ionization parameter from the H/[N II] ratio, using e.g. the curves shown by Viironen et al. (2007), and the electron density from the [S II] doublet ratio. The product is proportional to ionizing flux ; this will follow a spatial decay like , where is the sky projected distance from 9 Sgr, and is the line-of-sight distance of 9 Sgr from the nebula, assumed flat. Figure 20 shows the result of this experiment: a well defined peak is indeed found, with additional, local enhancements in ionizing flux near and arcmin due to Herschel 36 and HD 165052, respectively; a minor enhancement near arcmin corresponds instead to the B stars in the NGC6530 core. The proposed functional form for is shown by the lines, with the green one corresponding to 3 arcmin, or pc at the nebula distance. If the adopted curve maximum is lowered, to account for the fact that the actual maximum in the datapoints is due to Herschel 36, a 4-arcmin curve (blue dashed) is also satisfactory, corresponding to pc. These normal distances are much smaller than the estimates by Lada et al. (1976); one possible explanation is the assumed planar geometry for the illuminated nebula: a slightly concave geometry would decrease center-to-edge differences, and require larger to produce the same observed effect. However, lacking these detailed geometrical informations on the nebula itself, we cannot derive better estimates, and may only consider pc as lower limits to the 9 Sgr distance from the part of the nebula immediately behind it.
This distance is much larger than the distance between the Orion-nebula ionizing star Ori C and its molecular cloud, of pc (Wen and O’ Dell 1995, O’ Dell 2001). This difference is undoubtedly an important factor and may explain many of the differences we find between the properties of M8 and the Orion nebula (see Section 3.4.1). It also suggests that 9 Sgr has excavated a larger cavity in its parent cloud compared to Ori C, and in turn than the M8 H region as a whole is probably in a later evolutionary stage than the Orion nebula.
Both Figs. 20 and 15 suggest that HD 165052 (O7Vz+O7.5Vz binary, Arias et al. 2002) is instead the dominant ionizing source in its neighborhood, despite being almost irrelevant to the nebular dynamics (Figs. 16 and 17). This is confirmed by the dependence of ionizing flux from distance to this star, shown in Figure 21, analogous of Fig. 20; five arcmin to the West of HD 165052, the flux from 9 Sgr still dominates, but nearer to HD 165052 the 9 Sgr contribution becomes unimportant. We can therefore fit (although with higher uncertainties) a profile depending on the normal distance from HD 165052 to the cloud, as above: it turns out that also this star, like 9 Sgr, is likely to be found distinctly above the cloud, at a distance in the range 1.5-1.8 pc.
3.4 Position-velocity diagrams
3.4.1 The NGC6530 region
Having discussed in Section 3.3 the existence of a reference direction for the nebular velocity field (red arrow in Fig. 16), we study here position-velocity diagrams along this direction. This symmetry properties pertaining only to the region around the NGC6530 cluster core, we consider here only the region (24 arcmin in diameter) within the red circle in Fig. 16. The relevant position-velocity diagram for H (with velocities from 1-g fits) is shown in Figure 22. Positive projected distances are towards the arrow head of Fig. 16, i.e. towards the galactic plane (to N-W). The origin of distances is at the position of the red plus sign in Fig. 16. As mentioned, the low-mass cluster stars have a well-defined peak in their RV distribution, at km/s (Prisinzano et al. 2007); it is reasonable to assume that also the B stars in the same spatial region have the same mean RV. The most massive star in the cluster core region is HD 164906 (MWC280; type B0Ve, Levenhagen and Leister 2006). The cluster center coincides also with a CO bright spot (nr. 3 in Lada et al. 1976), at velocity km/s (heliocentric, corresponding to km/s as reported in Lada et al. ), whose position is also shown in Fig. 17. As Fig. 22 shows, the cluster core corresponds to the expansion center, in both position and velocity, of a shell-like structure in the ionized gas, reaching maximum negative speeds of km/s; no gas is found at in the vicinity of the cluster center (distance origin). Interestingly, the colder CO molecular gas moves in the opposite direction with respect to , and Fig.4 of Lada et al. (1976) also shows that it possesses a velocity gradient along the N-S direction, suggesting a shell-like geometry as well. The resulting picture is that of a localized expanding gaseous bubble, pushed by the cumulative effect of winds from massive B stars in the NGC6530 cluster core (there are no O stars inside it). Therefore, in this part of the Lagoon nebula the nebular emission arises in front of the star cluster, not in its background. As the velocity map of Fig. 16 shows, such expansion is not spherical, with near-zero radial velocity reached at small distances only along of the direction of the arrow. While dust extinction in front of the cluster stars is relatively low, it rises considerably behind them (e.g. Damiani et al. 2006), implying large amounts of dust just behind the cluster. Interestingly, there is no trace in Fig. 22 of any receding ionized shell, which we interpret with the ionizing radiation being absorbed by the dust on the rear side. For comparison, in the Orion nebula a rather regular sequence of velocities is found, with some of the ionized gas layers having speeds within a few km/s relative to the background CO ( km/s for [S II], km/s for [O III], km/s for H; O’ Dell et al. 1993, O’ Dell 2001). Here instead (and see also Figure 26 below) we see little or no ionized gas at velocities so close to that of background CO, which points to important differences between the structure of the ionized regions in M8 and in the Orion nebula. Also the emission from the PDR on the molecular cloud surface is not clearly recognizable in the position-velocity diagrams.
The stellar wind push of the massive stars in the NGC 6530 core, on the other hand, may be responsible for the peculiar positive-velocity displacement of the CO emitting gas (note that the other two CO spots found by Lada et al. 1976, the brightest one coincident with the Hourglass nebula, have velocities km/s, heliocentric). The current view that the star cluster did form on the near-side surface of the molecular cloud is in good agreement with the proposed interpretation: the far side of the star cluster faces regions with more dust, molecular gas, and higher-density gas in general, than the near side, facing the outer, more rarefied parts of the cloud. Therefore, also the absolute speed reached by the near-side ionized diffuse gas ( km/s in H) is larger than than of the far-side denser and colder gas ( km/s in CO).
At projected distances larger than arcmin, Fig. 22 shows a wide scatter in the RV distribution, with no clearly defined geometry; this region corresponds to the Hourglass nebula, as also clear from the size of the symbols in the Figure (proportional to intensity). In the Figure are also shown the massive stars M8E-IR (green triangle) and HD 164816, 9 Sgr, and Herschel 36 (red triangles, left to right). Interpreting such a complex velocity field in term of expansion driven by one or more massive stars requires to know the radial velocity of these stars with sufficient accuracy. This information is not always available for massive stars, which have few lines broadened by very fast rotation, are found very often in binary systems, and are studied mostly with single-epoch observations. For convenience we compile in Table 1 the literature radial-velocity data on the most massive stars discussed here, from the SIMBAD database, except for the 9 Sgr velocity taken from Williams et al. (2011); note however that this star is a long-period SB2 binary, and the systemic velocity is subject to large errors (Rauw et al. 2012). No Gaia-ESO velocity measurements are available for these massive stars. Literature velocities are missing for four stars, including the massive young object M8E-IR. These are plotted in our position-velocity diagrams at a velocity for reference. The two stars in the western regions have similar velocities ( km/s), but strongly discrepant with ; yet, they agree more with the approaching velocity of the H emission in the same region, discussed in Section 3.4.3 below. The positive velocity of 9 Sgr is surprising at first sight, being so different from that of the CO clouds; however, after considering that this star lies pc above a massive molecular cloud ( for each of the CO clouds in the region according to Takeuchi et al. 2010, with the cloud associated with the Hourglass being one of the most massive), it becomes plausible that this star has gained a considerable speed toward the cloud during the last few Myrs.
From Fig. 22 there is no apparent connection between the position and motion of 9 Sgr and the ionized gas, despite this star being the most massive of the region. This supports further the arguments of Lada et al. (1976) on its lying at some distance from the cloud, in its foreground. In those outermost nebular regions, the local gas density is likely so low that the ionized front approaching us becomes undetectable. Outside the ionized regions, there is nevertheless neutral gas associated with the Lagoon nebula, as discussed in connection with sodium absorption in Section 3.2. Therefore, Figure 23 shows the same H velocities as in Fig. 22, but on an expanded velocity scale, with the addition of the Na I D velocities: except for the dominant component near km/s, the neutral gas moves at much larger negative velocities than the H-emitting gas. No clear pattern is seen, indicating that the geometry of the neutral-gas expansion is different from that of the ionized gas. It is interesting to remark that despite 9 Sgr is distant from the nebula, the sodium absorption is still closer to us, since several absorption components at negative velocities up to km/s or more are evident also in the 9 Sgr UVES spectra shown in Figure 24. In this Figure a definite time variability of the sodium absorption components is seen, especially near velocities km/s, in both velocity and line width; this agrees with our arguments of Section 3.2 that the sodium layer at km/s is the one most subject to dynamical changes.
Consideration of the 2-g model fits to the H line in the same cluster core region provides us with only a marginally clearer picture (Figure 25): the velocity splitting between blue and red components is here mostly small compared to the absolute velocity values. Again, near the shell center essentially no near-zero-velocity gas is detected, while maximum negative velocities attain km/s. In the Hourglass region the velocity spread is highest, again without clear geometrical pattern; we recall that this latter property agrees with the highly anisotropic brightness distribution of the Hourglass nebula itself, whose oscuring material lets the radiation from Herschel 36 leak only through irregularly-distributed “windows”.
More illuminating is the examination of the position-velocity diagram involving 2-g fits to [N II] and [O III] lines, and shown in Figure 26. While the emission at velocities between 0 to km/s has not greatly changed, new features are seen at both positive velocities (as in the map of Fig. 18), and at velocities km/s. In the projected distance range from to arcmin a weak but significant positive-velocity component is found in both [N II] and [O III]. The velocity is found to be largest near the projected position of M8E-IR (green triangle), and to decrease gradually towards position arcmin. There is no corresponding structure in the position-velocity plane at negative velocity. This suggests strongly the existence of a shell of ionized gas, expanding away from M8E-IR (or its immediate vicinity), of which only the receding component is visible to us, and limitedly to the part unobscured by the dense bright-rimmed cloud. This latter characteristics is easily explained assuming that any approaching gas is blocked or hidden by the dark dusty structures seen as bright-rimmed clouds, which also occult M8E-IR from our direct view. While all the literature on this object (see the review in Tothill et al. 2008) agrees that it must be very young and surrounded by thick layers of dust, the existence of a emispheric shell ionized by this object implies that the dust thickness between M8E-IR and the nebula behind it is much less than the dust thickness in the direction towards us. Alternatively, M8E-IR, a known outflow source (Mitchell, Maillard and Hasegawa 1991), might be only the source of the mechanical push exerted on the receding gas, which is instead ionized by another UV source, maybe 9 Sgr further away. Even in this latter case, the total column density of matter on our side of M8E-IR must be much larger than on its rear side, in order to block any approaching gas expanding from it.
Considering now the region around 9 Sgr (red triangle at km/s in Fig. 26), we observe that positive-velocity emission is found near its position, up to km/s, and declining away from the star position. Already in Fig. 18 the positive-velocity datapoints were seen to follow an almost half-arch around this star’s position. While no (or very little) positive-velocity gas is found in correspondence of the cluster core, blocked by the dense molecular material behind, this blocking effect does not operate for the gas receding from 9 Sgr, if this star lies at large distances in front of the cluster as already discussed. Therefore, we obtain a coherent picture by assuming that the positive-velocity gas is pushed by 9 Sgr towards the nebula, counteracting its expansion locally.
The region around the Hourglass/Herschel 36 continues to show a rather chaotic position-velocity pattern even using the 2-g model fits in Fig. 26 as it was using 1-g models in Fig. 22 above. The largest green circle in Fig. 26 represents the strong negative-velocity [O III] emission in the Herschel 36 spectrum; a corresponding, much weaker positive-velocity component is also found near km/s, demonstrating again asymmetric expansion in the immediate vicinity of this star, where the high-ionization [O III] line arises.
3.4.2 The Hourglass nebula
We next discuss position-velocity diagrams in the Hourglass region (orange circle in Fig. 16); since the positive-velocity datapoints align along approximately the N-S direction in Fig. 18, we take here the reference direction along RA for the position axis. The position-velocity diagram of Figure 27 shows together the 2-g fit results from four lines: [N II] (black/gray), H (blue/cyan), [O III] (green), and Na I absorption (red), with circle size proportional to intensity. The position origin is taken coincident with the RA of 9 Sgr (triangle at km/s). Herschel 36 is the triangle at km/s. The diagram, although very complex, helps us to appreciate better several effects. The [N II] emission (black/gray symbols), although found at both positive and negative velocities in the neighborhood of 9 Sgr, is on average stronger at positive RVs, whereas H is largely absent at positive RVs. This explains the discrepancy of bulk velocities between the two lines, found in Figure 5. To the right of position origin, the velocity splitting in H increases regularly until the position of Herschel 36: there, H shows both a negative-velocity component (reaching km/s), and a slower component, apparently at rest with respect to Herschel 36. On the other hand, [N II] shows both a rest-frame component and another one with slightly receding velocities with respect to Herschel 36. Also in this case as for 9 Sgr, the H to [N II] intensity ratio is different between the near and far side of the massive star. We cannot make more quantitative studies since the intensity ratios for the (unresolved) 2-g fit components are affected by too large errors individually, as remarked in Section 3.1. The existence of positive-velocity [N II] emission from the envelope around Herschel 36, unlike the cluster core region, suggests that this star like 9 Sgr lies at some distance above the dust-rich molecular cloud, otherwise an inward-directed flow at positive velocities would have been blocked.
Much fainter than the positive-velocity [N II] emission, but still clearly detected is negative-velocity [N II] at km/s, overlapping some of the Na I absorption layers. As also discussed in Section 3.2, the transition between ionized and neutral gas should take place near this velocity range. Although the spatial coverage of the UVES data used for the study of the sodium line is much less dense than that of the Giraffe data, we may tentatively identify in the Figure a regular pattern in the sodium absorption, as indicated by the two dotted red curves. These are centered to the 9 Sgr position, and suggest that the large-scale expansion, at distances where the gas is neutral, is driven by this star. The radii of the circle segments shown in Fig. 27 are respectively 13 and 15 arcmin, corresponding to 4.74 and 5.47 pc. The star 9 Sgr, showing sodium absorption at least around km/s, must be interior to at least the lower-velocity shell, in agreement with its line-of-sight position derived in Section 3.3.2. By dividing the inferred sodium-shell radii by their maximum velocities we obtain timescales of order of yr. Note however that even stars farther away from 9 Sgr than 15 arcmin show sodium absorption, so that the proposed geometry for the absorbing layers must be only considered as tentative, and more complex neutral-gas structures are certainly present across the entire face of the nebula. These layers, well above the ionized gas, and at negative velocities with respect to it, are very similar to the Orion Veil, already mentioned in Section 3.1. The existence of a blueshifted, neutral layer is also in very good agreement with predictions of champagne-flow models of blister H regions (Tenorio-Tagle 1979), fully appropriate to a region like M8.
The results from the 2-g fits to [S II] lines, shown in Figure 28, provide independent confirmations of the findings just discussed. In particular the brightest [S II] emission near Herschel 36 is found near the stellar rest velocity (as for H), but clear [S II] emission near the Hourglass is found also at positive velocities (while H shows large negative velocities in the same region).
3.4.3 The outer parts of the Lagoon nebula
Finally, we examine the large-scale velocity patterns of the nebula, outside the central parts examined above. We have already remarked that the Western regions do not show the same dynamics as the Eastern ones, and therefore a radial coordinate does not prove useful. Instead, we consider separately an East and a West radial coordinate, from the same NGC6530 cluster center as in Section 3.4.1. The dependence of 1-g H velocity on these radial distances is shown in Figure 29. Towards East the gradual decrease in absolute velocity (towards ) agrees well with a global-expansion pattern, with velocity vectors becoming orthogonal to the line of sight at the largest radii. To the West, there is no sign of this, with velocities remaining at values from to km/s even where the nebula becomes very faint. Our spatial sampling in the S-W outer regions is almost nonexistent, so these result pertain essentially to the N-W parts, i.e. those closest to the galactic plane. Adding complexity to the puzzle, the two massive stars in the West (HD 164536 and 7 Sgr) both have negative velocities, similar to the neighboring gas but contrasting with .
One possible explanation of the East-West large-scale velocity gradient in the Lagoon is rotation. In order to keep in bound keplerian rotation matter at a speed of 5 km/s at a radius of 10-11 pc, and in the absence of internal pressure support, a mass of is required, which is not unreasonable compared to the mass estimates for the CO clouds in M8 given by Takeuchi et al. (2010) using NANTEN, or the value of given for the M8E region alone by Tothill et al. (2008). However, one obvious difficulty of this hypothesis is that, if this was the case, then Fig. 29 would suggest for the cloud center of mass a km/s, in strong disagreement with the RVs of both the CO bright spots of Lada et al. (1976) and the NGC6530 .
The adoption of 2-g fits does not clarify the issue of large-scale dynamics, even using [N II] and [O III] having smaller linewidths (Figure 30): not only the datapoint scatter is increased, but the overall East-West velocity gradient becomes barely observable in these lines, which therefore originate in layers well distinct from H. For some locations, as discussed in Section 3.3.1, the [N II] lines are split, with central RV very close to that of the nearby O star HD 164536: these add to the scatter seen in the rightmost datapoints in Fig. 30, where local expansion adds to the average local cloud velocity. The splitting center being so close to the O-star RV might also be seen as a confirmation that the (poorly studied) star HD 164536 lies within the Lagoon itself. The figure also shows velocities of sodium absorption components, which may be of some usefulness: if we consider the lowest-found (absolute) sodium velocities at various radial distances, we may observe a regular gradient from km/s at the East extreme, towards km/s at the opposite one. As discussed in Section 3.2, this sodium layer, being the most uniform of all found here, is probably the outermost one, and therefore least affected by local phenomena. If it is really associated with the Lagoon nebula, it might be considered as the best indicator of a global rotation of the nebula. A much better spatial coverage in the sodium absorption data would be needed, however, before accepting this possibility.
Alternatively, the observed velocity gradient might reflect a shear motion, caused by interaction between the parts of M8 closest to the galactic plane (to the Northwest) and other dense clouds.
4 Discussion and summary
The various pieces of evidence described in Section 3 enable us to draw a complex picture of the ionized and neutral gas in the Lagoon nebula, and of its physical connection to massive stars and molecular material in the cloud. Perhaps the most effective way of building a coherent and understandable picture is by means of a drawing. We show therefore a graphical summary of most of the results obtained in Figure 31, which refers to a section through the nebula along a line joining M8E-IR with the Hourglass region, until the western nebula parts. The Figure represents most of our findings with some level of detail. The black solid line represents the approximate boundary of the molecular cloud; dots indicate dusty regions. O stars are shown as bigger cyan stars.
The red stars indicate the core of the NGC6530 cluster, where most of its B stars are found; the entire cluster would fill most of the region shown. The B stars drive a strong expanding shell towards us (blue dashed half-circles) visible in the ionized-gas lines. On the opposite (far) side of the cluster, no redshifted expansion is detected in the ionized lines, suggesting that the cluster lies very close to the denser, dusty molecular cloud, as shown. The CO clump in the same direction is redshifted (Lada et al. 1976, Takeuchi et al. 2010), probably pushed by the same cluster stars producing the blueshifted optical lines. We represent this with the magenta dashed arcs. Note that the other two strong CO clumps in this region (also labeled ’CO’ in magenta) are found at rest with respect to the NGC6530 .
The Hourglass region around Herschel 36 lies probably at some distance from the background molecular cloud, since some high-velocity redshifted emission is seen around it, as well as blueshifted emission. The geometry of any material expanding away from Herschel 36 is however very irregular, because of the non-isotropic distribution of dense clouds all around this star: we represent this as discontinuous shell fragments (dashed), both redshifted and blueshifted. We recall that the nebular densities found in this region are the highest of the whole Lagoon nebula.
The M8-East region is also characterized by a partial-shell geometry, since we detect the unobscured portion of a redshifted expanding arc, centered on M8E-IR. This star, or another sufficiently massive star near to it, must be heavily obscured on our side (tens of magnitudes visually), but very little on the “inner” side, in order to be able to drive an inward flow. The outward flow is instead blocked by the dense cloud, and its blueshifted emission not detected accordingly.
A crucial role is played by the most massive star, 9 Sgr, which we find to lie well separated (at least 1 pc, or more) from the cloud surface, in agreement with previous works, but still inside the large blister concavity. Being found in a low-density environment, its radiative and mechanical push becomes detectable only against the higher-density gas on the inward side of the star, as redshifted emission in [N II] and [S II] (triple red dashed arcs); on its outward side, very faint or undetectable blueshifted emission is present (dotted blue arc). A combination of factors may therefore explain the characteristics found in the “Great Rift” region: it is denser than its surroundings, being compressed by both sides (NGC6530 shell and Herschel 36 shell), and pushed towards the cloud by the 9 Sgr wind and radiation, hence the positive velocities on its ionized surface. It would be interesting to examine whether such compression is able to trigger new star formation. There is no contradiction between the enhanced dust density in the Great Rift, responsible for obscuration of background stars, and the inconspicuous electron density found there from the [S II] doublet ratio (Figure 15), since this latter only refers to its ionized surface and not to its colder, inner parts.
Completing the picture, the western parts of the cloud are found to be approaching us, as is the O star HD 164536, whose wind is probably responsible of some line splitting in its vicinity, indicative of a bidirectional expansion, as shown by the blue and red arcs. At large distances in front of the whole cloud, discrete neutral layers are found, approaching us over a range of (negative) velocities and distances. They might be named the “Lagoon Veil”, by analogy with the Orion Veil.
In addition to these results, at least another one deserves some discussion. The velocity profile around NGC6530, whatever the diagnostic line and the modeling approach (1-g or 2-g) chosen, leaves little doubt about an expanding shell being driven by stars in the cluster core. At the same time, both the ionization parameter (Fig. 15) and the ionizing flux (Fig. 20) show a gradient across the cluster face, in the direction of 9 Sgr. The latter Figure shows that, with respect to the 9 Sgr ionizing flux, only a small excess (less than a factor of 2) is found at the NGC6530 position, attributable to the NGC6530 B stars themselves. Therefore we reach the conclusion, on solid observational grounds, that the NGC6530 shell is mechanically driven from inside, but ionized from outside. This geometry is very unlike classical Strömgren spheres. A deeper treatment of the problem is clearly outside our scopes here. The biggest problem is, since we observe recombination in the H line, where the recombined neutral gas lies: gas outside of the shell is ionized by 9 Sgr, inside it is ionized by the NGC6530 B stars. Perhaps a double shell develops, with an intermediate sheet of neutral gas: this might account for the small velocity splitting indicated by our 2-g fits to H lines. In the absence of a detailed modeling, we cannot however derive any firm conclusion on this issue.
Another interesting issue is that related to rotation of the entire cloud, as suggested by the blueshifted lines in the N-W parts. It is worth noting in this respect that in the molecular CO lines the appearance of the Lagoon nebula is very different than in the optical, and splits in three main condensations, coincident with M8-East, NGC6530 core, and Hourglass regions respectively, with little in between (Fig. 8 of Takeuchi et al. 2010). The Lagoon N-W regions appear as an extension of the Hourglass CO cloud, whose core is at rest with respect to . It is therefore possible that only the Hourglass molecular cloud is rotating, not the entire Lagoon nebula; this motion does not involve NGC6530 and is in better agreement with the other existing dynamical data. Alternatively, the CO maps of Takeuchi et al. (2010) make it clear that the N-W region, the closest to the galactic plane, is also near other CO clouds, and might be interacting with them. Therefore, while the bulk motion of approach of the N-W Lagoon nebula region is an established observational result, its interpretation is not unambiguous.
Acknowledgements.We wish to thank an anonymous referee for his/her many interesting comments and suggestions. Based on data products from observations made with ESO Telescopes at the La Silla Paranal Observatory under programme ID 188.B-3002. These data products have been processed by the Cambridge Astronomy Survey Unit (CASU) at the Institute of Astronomy, University of Cambridge, and by the FLAMES/UVES reduction team at INAF/Osservatorio Astrofisico di Arcetri. These data have been obtained from the Gaia-ESO Survey Data Archive, prepared and hosted by the Wide Field Astronomy Unit, Institute for Astronomy, University of Edinburgh, which is funded by the UK Science and Technology Facilities Council. This work was partly supported by the European Union FP7 programme through ERC grant number 320360 and by the Leverhulme Trust through grant RPG-2012-541. We acknowledge the support from INAF and Ministero dell’ Istruzione, dell’ Università’ e della Ricerca (MIUR) in the form of the grant ”Premiale VLT 2012”. T. Z. acknowledges support from the Slovenian Research Agency (research core funding No. P1-0188). F. J.-E. acknowledges financial support from the Spacetec-CM project (S2013/ICE-2822). M. T. C. acknowledges the financial support from the Spanish Ministerio de Economía y Competitividad, through grant AYA2016-75931-C2-1-P. The results presented here benefit from discussions held during the Gaia-ESO workshops and conferences supported by the ESF (European Science Foundation) through the GREAT Research Network Programme. This work is also based on data products from observations made with ESO Telescopes at the La Silla Paranal Observatory under programme ID 177.D-3023, as part of the VST Photometric H Survey of the Southern Galactic Plane and Bulge (VPHAS, www.vphas.eu). This research has made use of the SIMBAD database, operated at CDS, Strasbourg, France.
- Arias et al. (2002) Arias, J. I., Morrell, N. I., Barbá, R. H., et al. 2002, MNRAS, 333, 202
- Bagnulo et al. (2003) Bagnulo, S., Jehin, E., Ledoux, C., Cabanac, R., Melo, C., Gilmozzi, R., & ESO Paranal Science Operations Team 2003, The Messenger, 114, 10
- Balick et al. (1974) Balick, B., Gammon, R. H., & Hjellming, R. M. 1974, PASP, 86, 616
- Bohuski (1973) Bohuski, T. J. 1973, ApJ, 184, 93
- Brand & Zealey (1978) Brand, P. W. J. L., & Zealey, W. J. 1978, A&A, 63, 345
- Damiani et al. (2004) Damiani, F., Flaccomio, E., Micela, G., Sciortino, S., Harnden, F. R., & Murray, S. S. 2004, ApJ, 608, 781
- Damiani et al. (2006) Damiani, F., Prisinzano, L., Micela, G., & Sciortino, S. 2006, A&A, 459, 477
- Damiani et al. (2016) Damiani, F., Bonito, R., Magrini, L., et al. 2016, A&A, 591, A74
- Drew et al. (2014) Drew, J. E., Gonzalez-Solares, E., Greimel, R., et al. 2014, MNRAS, 440, 2036
- Elliott & Meaburn (1975) Elliott, K. H., & Meaburn, J. 1975, MNRAS, 172, 427
- Gilmore et al. (2012) Gilmore, G., Randich, S., Asplund, M., et al. 2012, The Messenger, 147, 25
- Henning & Gürtler (1986) Henning, T., & Gürtler, J. 1986, Ap&SS, 128, 199
- Hobbs (1974) Hobbs, L. M. 1974, ApJ, 191, 381
- Kaufman & Martin (1993) Kaufman, U., & Martin, W. C. 1993, Journal of Physics Conference Series, 22, 279
- Kramida et al. (2015) Kramida, A., Ralchenko, Yu., Reader, J., and NIST ASD Team 2015, NIST Atomic Spectra Database (ver. 5.3), National Institute of Standards and Technology, Gaithersburg, MD
- Kumar & Anandarao (2010) Kumar, D. L., & Anandarao, B. G. 2010, MNRAS, 407, 1170
- Lada et al. (1976) Lada, C. J., Gottlieb, C. A., Gottlieb, E. W., & Gull, T. R. 1976, ApJ, 203, 159
- Levenhagen & Leister (2006) Levenhagen, R. S., & Leister, N. V. 2006, MNRAS, 371, 252
- Meaburn (1971) Meaburn, J. 1971, Ap&SS, 13, 110
- Mitchell et al. (1991) Mitchell, G. F., Maillard, J.-P., & Hasegawa, T. I. 1991, ApJ, 371, 342
- Munari & Zwitter (1997) Munari, U., & Zwitter, T. 1997, A&A, 318, 269
- O’dell et al. (1993) O’dell, C. R., Valk, J. H., Wen, Z., & Meyer, D. M. 1993, ApJ, 403, 678
- O’dell (2001) O’dell, C. R. 2001, ARA&A, 39, 99
- O’Dell et al. (2008) O’Dell, C. R., Muench, A., Smith, N., & Zapata, L. 2008, Handbook of Star Forming Regions, Volume I, 4, 544
- Pasquini et al. (2002) Pasquini, L., Avila, G., Blecha, A., et al. 2002, The Messenger, 110, 1
- Povich et al. (2013) Povich, M. S., Kuhn, M. A., Getman, K. V., et al. 2013, ApJS, 209, 31
- Prisinzano et al. (2005) Prisinzano, L., Damiani, F., Micela, G., Sciortino, S. 2005, A&A, 430, 941
- Prisinzano et al. (2007) Prisinzano, L., Damiani, F., Micela, G., & Pillitteri, I. 2007, A&A, 462, 123
- Randich et al. (2013) Randich, S., Gilmore, G., & Gaia-ESO Consortium 2013, The Messenger, 154, 47
- Rauw et al. (2012) Rauw, G., Sana, H., Spano, M., et al. 2012, A&A, 542, A95
- Saraph & Seaton (1970) Saraph, H. E., & Seaton, M. J. 1970, MNRAS, 148, 367
- Simon et al. (1984) Simon, M., Cassar, L., Felli, M., Fischer, J., Massi, M., & Sanders, D. 1984, ApJ, 278, 170
- Sung et al. (2000) Sung, H., Chun, M.-Y., & Bessell, M. S. 2000, AJ, 120, 333
- Takeuchi et al. (2010) Takeuchi, T., Yamamoto, H., Torii, K., et al. 2010, PASJ, 62, 557
- Tenorio-Tagle (1979) Tenorio-Tagle, G. 1979, A&A, 71, 59
- Tothill et al. (2002) Tothill, N. F. H., White, G. J., Matthews, H. E., et al. 2002, ApJ, 580, 285
- Tothill et al. (2008) Tothill, N. F. H., Gagné, M., Stecklum, B., & Kenworthy, M. A. 2008, Handbook of Star Forming Regions, Volume II, 5, 533
- van den Ancker et al. (1997) van den Ancker, M. E., The, P. S., Feinstein, A., et al. 1997, A&AS, 123, 63
- Viironen et al. (2007) Viironen, K., Delgado-Inglada, G., Mampaso, A., Magrini, L., & Corradi, R. L. M. 2007, MNRAS, 381, 1719
- Walker (1957) Walker, M. F. 1957, ApJ, 125, 636
- Weedman (1968) Weedman, D. W. 1968, PASP, 80, 314
- Welsh (1983) Welsh, B. Y. 1983, MNRAS, 204, 1203
- Wen & O’dell (1995) Wen, Z., & O’dell, C. R. 1995, ApJ, 438, 784
- Williams et al. (2011) Williams, S. J., Gies, D. R., Hillwig, T. C., McSwain, M. V., & Huang, W. 2011, AJ, 142, 146
- Woodward et al. (1986) Woodward, C. E., Pipher, J. L., Helfer, H. L., et al. 1986, AJ, 91, 870
- Wright et al. (1977) Wright, E. L., Lada, C. J., Fazio, G. G., Low, F. J., & Kleinmann, D. E. 1977, AJ, 82, 132
|Id||RA||Dec||H||[N II] 6584||[S II] 6717||[S II] 6731|