The kinematical properties of superbubbles and H ii regions of the Large Magellanic Cloud derived from the 3D Hα Survey Free

Abstract

We report the results of a kinematical Hα survey of the Large Magellanic Cloud (LMC) presented in the form of a kinematical and photometric catalogue of 210 H ii regions. The observations have been obtained with a scanning Fabry–Perot interferometer that produced data cubes corresponding to 66 different pointings over this galaxy, each with a field of view of 38 arcmin, covering almost the whole extent of the LMC. We find a bimodal distribution of the Hα luminosity of LMC H ii regions. We also derive the local star formation and star formation rate (SFR) per unit area of the nebulae, concluding that star formation in the LMC has proceeded until the present time at an average rate of roughly 0.11 M_⊙ yr⁻¹. Also, we do not find any correlation between the SFR or Σ_SFR with ΔV (full width at half-maximum for a single Gaussian profile and the difference in velocities for multiple-components velocity profiles), the diameter, the distance to the kinematical centre of the LMC and age of the nebulae. Over most of the LMC ΔV appears to be of the order of 30 km s⁻¹. However, in a few regions the ΔV of the velocity profiles is as large as 50–100 kms⁻¹, corresponding to identified supernova remnants and superbubbles undergoing expansion motions.

1 INTRODUCTION

The Large Magellanic Cloud (LMC) is the nearest massive dwarf galaxy to the Milky Way. Furthermore, it is a galaxy viewed with an inclination that allows us to discriminate the nebulae and also to measure rotation motions. All of these properties make the LMC a privileged place to undertake global studies about the nature of its nebulae. Indeed, the special characteristics of the LMC distance of about 50 kpc from the sun (Westerlund 1997; Gibson 2000; Freedman et al. 2001) and the lack of interstellar absorption, make this galaxy very useful for studying the connection between stellar evolution and the structure and dynamics of the interstellar medium (ISM) in galaxies.

The LMC is rich in ring-shaped nebulae, such as bubbles, and superbubbles (e.g. Davies, Elliott & Meaburn 1976; Meaburn 1980; Chu & Kennicutt 1994; Kim et al. 1999) with special characteristics such as having intermediate [S ii]/Hα line-ratios (Davies et al. 1976; Georgelin et al. 1983). Bubbles are shell structures of small diameter (less than 50 pc, and they are presumably formed by only one star); superbubbles are similar but larger structures (diameter larger than 50 pc) and supershells are rings with radii larger than the LMC gas disc scaleheight (z_g ∼ 180 pc; Dawson et al. 2013).

Several surveys were carried out at different emission lines in order to know more about the morphology of these structures being one of the more recent the MCELS (Magellanic Clouds Emission Line Survey; Smith, Points & Winkler 2006). Several works concerning the nature of the bubbles, superbubbles and supershells have been published in the course of these years but the origin of superbubbles and supershells (Rosado 1986; Chu & Kennicutt 1994; Oey 1996; Reyes-Iturbide, Rosado & Velázquez 2008; Reyes-Iturbide et al. 2009; Rodríguez-González et al. 2011) still remains unclear, while it is thought that stellar winds with supernovae explosions are likely responsible. Moreover optical and X-ray observations can be used to obtain the kinematic of these structures.

An important way to shed some light on the study of bubbles and superbubbles is to correlate the optical emission and kinematics with surveys at other wavelengths (radio, X-rays, IR, UV). Another interesting and complementary possibility is to study the kinematics of these regions. With the aim of studying better these objects on a global (galactic) scale we undertook a kinematical survey of LMC nebulae and diffuse ionized gas (DIG) in the same way as we have done it for the Small Magellanic Cloud (Le Coarer et al. 1993). The LMC survey has been carried out in the Hα emission line, with a scanning Fabry–Perot interferometer, covering almost the whole extent of the LMC. With these observations we have generated a photometric and kinematical catalogue of H ii regions and nebulae in the LMC that we present in this work.

This work is organized as follows: in the next section, we give information about the observations and data reduction carried out in order to obtain the kinematical data of the LMC H ii regions. In Section 3, we explain the photometric and astrometric calibration process in order to obtain the Hα mosaic image. In Section 4, we present the LMC H ii regions catalogue and discuss in Section 5 the importance of this catalogue for understanding the nature of the nebulae and the star formation. In Section 6, we give the conclusion of this paper. In a coming paper, we will study the rotation curve constructed from the ionized hydrogen attached to young and massive stars and the kinematics of the DIG as well.

2 OBSERVATIONS AND DATA REDUCTIONS

The observations have been carried out in the frame of an Hα Survey of the Magellanic Clouds and the Milky Way with a 36 cm diameter telescope (located at La-Silla, European Southern Observatory) between 1990 and 1996. This telescope was equipped with a focal reducer, a scanning Fabry–Perot interferometer and a photon counting camera. A complete description of the instrumentation, including data acquisition and reduction techniques has been given in Amram et al. (1991) and Le Coarer et al. (1992a).

The observations (66 field of views, FoV) were made using a Fabry–Perot interferometer having an interference order (at Hα) p = 796, an effective Finesse F = 12.5, providing a spectral resolution R = pF ∼ 10 000 (or 30 km s⁻¹), with an accuracy 3 km s⁻¹ depending on the signal-to-noise ratio (S/N; Le Coarer et al. 1992a) and a free spectral range FSR ∼ 8.25 Å (or FSR ∼ 377 km s⁻¹). In order to follow the Shannon–Nersquist criteria, taking into account the finesse, the number of scanning steps was 24, providing a spectral sampling of ∼15 km s⁻¹. Each data cube covers a 38 arcmin × 38 arcmin( ∼ 1475 arc min²) FoV with a spatial sampling of 9 arcsec × 9 arcsec (equivalent to 2.2 pc at the LMC distance of 50 kpc; Feast 1991). The exposure time for FoV is 2 h, resulting in a total of 132 h of observations for the whole project at Hα. In addition, for only three FoVs, the observations have been made using a Fabry–Perot having an interference order p = 2600, but these data have only been used to complete the mosaic (see Fig. 4).

The data reduction was performed by means of the specialized software cigale (CInematique des GALaxiEs) that was designed specially (Boulesteix et al. 1983) for the acquisition and reduction of data obtained with a scanning Fabry–Perot interferometer (CIGALE is the name of our survey too).

For each typical observation we obtain the following informations.

• one wavelength calibration cube (neon lamp), from which we derive a phase map giving the wavelength origin at each pixel. Instrumental profile and flat-field are also obtained from this series of calibration rings.

• one science cube of the observed field around the Hα. From the calibration phase map we can compute a wavelength-calibrated cube that contains one spectrum for each of the 256 × 256 pixels.

The instrumental profile is determined by observing the narrow emission line of a Neon lamp at λ 6598.95 Å. The instrumental profile fits the classical Airy function characteristic of the instrumental response of the Fabry–Perot interferometer and will be used to determine the deconvolved line width of the profiles. For each filter used, flat-field correction is performed using the addition of the calibration interferograms at Hα obtained from a diffuse continuum source.

The filter used is centred on the Hα wavelength taking into account the LMC radial velocity (λ = 6569 Å); it has a width of Δλ = 16 Å. Several night sky emission lines are located within its band-pass, which are mainly the geocoronal Hα at 6562.78 Å and the OH night glow lines at 6553.62, 6568.72 and 6577.25 Å. The OH night sky lines at 6577.25 Å and 6553.62 Å, located in the wings of the interference filter, marginally affect the spectra. To minimize the number of night sky lines, the interferometer p = 796 has been designed on purpose, in order that the OH night sky lines at 6568.72 and 6577.25 Å match, modulo one FSR.

Fig. 1 shows a typical profile obtained from our radial velocity data located in regions where there is not so much nebulosity seen in order to identify the night sky lines as well as nebular radial velocity profiles from DEM 208, located in the south-east of the LMC, in order to compare them. From Fig. 1, we can point out that the main problem is that the OH night glow lines are close to the heliocentric systemic velocity of the LMC (274 km s⁻¹; Luks & Rohlfs 1992b). At faint level of galactic emission, night sky parasitic lines are hard to discriminate and subtract from the interferograms as it can been seen in the top right insert in Fig. 1 where the OH and Hα emission is by far stronger than the LMC Hα emission. Nevertheless, an accurate modelisation of the spectral point spread function allows the extraction of the signal of interest even in such faint regions.

Fig. 2 illustrates a typical Hα CIGALE data cube located at the edge of the mosaic, the FoV of this region is indicated by a green rectangle in Fig. 4. The trihedron drawn on the right-hand side of each of the six of the elementary projections indicates the orientation of the data cube (space and velocity): the right ascension direction (α) is represented by a red arrow while the declination (δ) is pointed by a green arrow and the radial velocity orientation (z) shown by a blue arrow. The intensity is given by the horizontal logarithmic rainbow scale in arbitrary units [in fact in Analogic Digital Unit (ADU) units per pixel] and range over three orders of magnitude (from 1 ADU – black; 10 – dark red; 1000 – white). The thin green lines materialize the limit of the cube and the inner grid; the green lines are parallel to the trihedron orientation; one elementary green square of the grid represents 230 × 230 arcsec² (1/10th of the linear FoV, 256/10 = 25.6 pixels) in the spatial direction and ∼25 km s⁻¹ in the velocity space. The first panel (top-left) is an orthographic projection while the five next panels (panels 2 to 6) are perspective projections as it can be followed by the parallel straight lines that correspond to foreground stellar continuum emission. From top-left to bottom-right panels: Panel (1) Line-of-View image similar to what can be seen in the mosaic on Fig. 4. Panels 2, 3 and 4 are position-velocity (PV) diagrams. Panel (2) shows a PV diagram integrated along the right ascension axis (rotation α = 90°, δ = 0°, z = 0°), we can clearly notice that the top-left Hα region in panel (1) has a lower radial velocity and a lower velocity dispersion than the main filamentary structure on the right side of panel (1) and we note a regular decrease of the radial velocity from the left to the right of this last structure. Panel (3) displays a PV diagram integrated along the declination axis (rotation α = 90°, δ = 0°, z = −90°); from this viewing angle the velocity gradient is smoother even if the same trend is visible. Panel (4) shows an edge-view PV diagram projected along both axis (α and δ), the data cube has been rotated by −45° (rotation α = 90°, δ = 0°, z = −45°), this gives access to a view of the cube at half way between panels (2) and (3) and allows us to emphasize the large range of velocity dispersions. Panels (5) and (6) display 3D-perspective of the nebulae (rotation α = −45°, δ = 0°, z = 90° and α = −45°, δ = 0°, z = 45°, respectively), allowing a perception of the ‘volume’ PV of the nebulae.

The normal procedure for the profile analysis is presented in Georgelin et al. (1994) and Amram et al. (1991). The Hα profiles observed are always very complex. They are composed of the ensemble of emission lines coming from each emitting layer along the line of sight, each with a potentially different velocity. When the velocity of the emissions is relatively constant over a large area, it is possible to decompose the observed profiles whatever their intensity variation. When, due to significant internal motions, the velocity field of a given H ii region is very complex, the profile decomposition is more difficult and can be done only if the intensity of the components is very different. Finally, in the case of very faint and extended regions, the S/N must be increased before profile decomposition. This is done by extracting the profile from a wider area. For more detailed explanation see Georgelin et al. (1994).

For each wavelength-calibrated data cube this analysis applied to each pixel results in the creation of:

• one monochromatic image computed by integrating the flux of each emission line profile.

• one radial velocity map derived from the Doppler shift of each profile. The software cigale computes the radial velocity field using the barycenter method and when it finds pixels with complex profiles (several components) automatically removes them from the map.

• one velocity dispersion map derived from the deconvolved width of each profile.

3 CALIBRATIONS

In order to build the Hα mosaic image and the LMC H ii regions catalogue, we have made photometric and astrometric calibrations.

3.1 Photometric calibration

We based the calibration process on the intensity-calibrated SHASSA survey (Gaustad et al. 2001) whose typical given uncertainty is 9 per cent. This survey provides wide-field narrow-band CCD Hα images of the southern sky taken with a robotic imaging camera situated at Cerro Tololo Observatory (CTIO). The camera used a small, fast, f/1.6 Cannon lens, which gave a very large (13°) FoV and a spatial resolution of 48 arcsec. Each SHASSA field was imaged through a narrow-band interference filter of 32 Å width centred at 6563 Å, as well as a continuum filter with two bands of 61 Å at 6440 Å and 6770 Å on either side of the Hα line. We have to keep in mind that the SHASSA's filter includes [N ii] lines at the edges. James et al. (2005) recall, for irregular galaxies, that [N ii]/Hα = 0.08 from individual extragalactic H ii regions while this ratio can reach 0.2 when considering integrated spectroscopy, more representative of the DIG. Combined with the filter transmission these translate into 3 and 7.5 per cent [N ii] possible contamination of the Hα emission for H ii regions and DIG, respectively. The SHASSA web site¹ makes available continuum subtracted, intensity-calibrated data. The LMC is covered by a single SHASSA field we retrieved from the web site.

For direct comparison, each night sky lines corrected Hα monochromatic image is re-binned and regrided to the SHASSA image (48 arcsec) resolution. For each individual field, the equivalent area of SHASSA data is then extracted and aligned to match our field. The images are then compared directly, pixel per pixel, to give a plot of SHASSA values, previously converted in erg cm⁻² s⁻¹, versus counts per re-binned pixel of our survey (Fig. 3). We adopt a 2 per cent typical uncertainty for the cigale data due to the statistical fluctuation in photon counts (Caplan & Deharveng 1985). We then fit the resulting comparison by a linear regression to determine the conversion factor to pass from our Hα counts to physical units (erg cm⁻² s⁻¹) for every of the 69 individual fields. In general (as in Fig. 3) our data follows a linear relation with the SHASSA values. The typical error on the calibration from the standard deviation of the slope is 4 per cent. Combining quadratically the different intensity uncertainty sources we estimate a typical error of 9.8 per cent on the CIGALE Hα intensity.

However, for few fields the correlation is not perfect at low intensities notably because of moon illumination and possible [N ii] contamination. As each cigale field was observed with variable moon-glow, some poor quality cubes (observed under strong moon-glow condition) are only used here to complete the mosaic. As the calibration is then based on the brightest Hα emission, the background in the Hα mosaic (Fig. 4), in some places, is not perfectly uniform.

3.2 Astrometric calibration

Our frames suffer from a distortion whose major part comes from a voltage dependency introduced by the photomultiplier photon counting camera. Another inhomogeneity was introduced during a long period of observations when the instrument was dismounted, introducing a misalignment of the north–south axis. To permit us to calibrate coordinates and to stitch fields in order to build the mosaic, a field-by-field correction has been applied. In order to correct the distortion, for each field, we have used the continuum image around the Hα emission where we identified an average of 40 stars brighter than magnitude 13.5 from online US Naval Observatory (USNO) catalogue. A specifically developed learning tool (kcicoor) permits us to adjust a bipolynomial quadratic correction to the image. The crossed polynomial terms include the correction of the field rotation. A typical 1.5 arcsec standard deviations in α and δ are obtained that must be compared to the 9.5 arcsec pixel size. The following astrometry corrections are then made: the first one gives a 1.5 arcsec accuracy determination of astrometry for each pixel, the second one provides a field-by-field bilinear interpolation of the tangent plane image centred on the individual field's centre, and finally a bilinear projection is made on our LMC centre of field (located at α = 05^h18^m10^s and δ = −69°13′28′). Fig. 4 shows the resulting Hα mosaic image, which is composed of the 69 different individual (38 arcmin × 38 arcmin) pointings.

4 THE KINEMATIC H ii REGION CATALOGUE

In this section, we present the kinematic and photometric catalogue. This catalogue allows us to identify the different motions for each region such as their expansion and rotation velocities, turbulence, or other smaller scale processes.

In Table 1, the columns are

• Column 1 and 2 give the identification number of the nebula in the Davies et al. (1976) and Henize (1956) catalogue, respectively.

• Column 3 gives the projected distance on the sky plane (in degrees) from the nebula to the kinematic centre found by Kim et al. (1998) (α = 05^h17^m54^s and δ =−69°5′4″).

• Column 4 gives the dimensions in arcminutes of the region measured in our Fabry–Perot data. Sizes of the major and minor axes are given for irregular nebulae.

• Column 5 gives the heliocentric radial systemic velocity in km s⁻¹. When the profile is complex, it is fitted using several Gaussians (as described in Section 2). We therefore obtain the heliocentric radial velocities of one or several components of any region and give the radial velocity for each component. The error in the value of the systemic velocity is given by |$(\sigma ^2_{\rm{inst}} +\sigma ^2_{\rm{gas}})^{1/2}$| weight by the S/N of each Gaussian. In this case the instrumental velocity resolution is σ_inst = 8 km s⁻¹ and the thermal width is σ_gas = 12 km s⁻¹.

• Column 6 gives ΔV, the velocity dispersion, in km s⁻¹. ΔV is a velocity parameter defined as follows: when the profile shows one component ΔV is the full width at half-maximum of the line (corrected for the instrumental function) and the error is half of the minimum width that we can adjust (1 km s⁻¹); when the profile shows multiple components, ΔV is the difference of the extreme velocity components and the error is the sum of errors in the velocities.

• Column 7 indicates whether the nebula belongs to the L or D component as defined by Luks & Rohlfs (1992b). To make this, we have compared the Hα systemic velocity with the H i velocity obtained by Luks & Rohlfs (1992b).

• Column 8 gives the Hα flux of each H ii region obtained from our photometric calibration.

• Column 9 gives the kinematic age. ΔV can be related at first order to the expansion velocity of the H ii regions that, combined with the size measurement of the regions, allows us to estimate their kinematic age (t_kin ∼ radius/ΔV).

• Column 10 gives the star formation rate (SFR). We derive the SFR using the relation given by Kennicutt, Edgar & Hodge (1989):

  \begin{equation*} \mathrm{SFR\ ({\mathrm{M}_{\odot}}\,yr^{-1})} = 7.9 \times 10^{-42}\ L(\mathrm{H}\,\alpha )\,(\mathrm{erg\ s}^{-1}). \end{equation*}

  We use here the L(Hα) without reddening correction because in the LMC the reddening is very small (A_v = 0.3; Imara & Blitz 2007).

• Column 11 gives the SFR per unit area in units of |${\mathrm{M}_{\odot}}$| yr⁻¹ kpc⁻² derived from Column 11 and dividing by the area calculated from the ionized gas emission radius.

• Column 12 gives comments describing mainly the morphology of the region.

Plotting the main physical parameters derived for every individual region (Fig. 5), we note that there is a correlation between the kinematic age and the diameter (Fig. 5 upper panel), which is not strange since the kinematic age depends on both the diameter and ΔV. Also we note that there is no clear correlation between ΔV and the luminosity (Fig. 5 middle panel) and ΔV and the diameter of the H ii regions (Fig. 5 lower panel).

5 DISCUSSION AND RESULTS

5.1 Global properties of H ii regions

From the Hα information on the catalogued H ii regions, we construct the luminosity and size functions (Fig. 6). The Hα luminosity and size functions of a galaxy are generally used to determine many of the global physical properties of H ii regions (e.g. Kennicutt et al. 1989). The LMC luminosity function we obtain is bimodal exhibiting two peaks, respectively, at ∼36.7 and ∼37.7, in agreement with Beckman et al. (2000) and Pellegrini et al. (2011). Beckman et al. (2000) suggest that the sharp peak at ∼37.7 can be attributed to the transition from ionizing to density-bounded regime. Pellegrini et al. (2011) demonstrate that the mean Hα luminosity of the density-bounded regions is ∼1 dex brighter than radiation-bounded regions.

More recently, Pellegrini et al. (2012) combined Hα, [S ii] and [O iii] maps to evaluate the Hα luminosity, optical depth and ionization structure of 401 H ii regions in the LMC. In particular they classify 271 H ii regions (while 130 were not classified because they are too faint or too diffuse) into four categories: optically thick (radiation-bounded), blister, optically thin (density-bounded) and shocked nebulae.

Thanks to our kinematic information the present catalogue embodies an essential physical complement to the 187 H ii regions we have in common. We then look at the statistical kinematic properties (Table 2) of the H ii regions classified by Pellegrini et al. (2012). Despite the large dispersion of the data and the low number of sources in some categories, we can notice an age gradient from the optically thick to the optically thin regions in agreement with H ii region's evolution scheme. The velocity dispersion shows only a small variation. The shocked nebulae exhibit a large velocity dispersion suggesting a strong interaction with their surrounding medium.

We cross-correlated our H ii region catalogue with the molecular cloud catalogue of Fukui et al. (2008). We find that 151 H ii regions (74 per cent of our sample) can be associated with a molecular cloud. We then determine the velocity difference dv (defined as V_CO–V_Hα) to find the H ii regions that are in the blister/champagne phase and to test, from the kinematic point of view, the classification of Pellegrini et al. (2012). The mean dv is −11 km s⁻¹ suggesting that most H ii regions are expanding in our direction. We selected H ii regions with a clear champagne flow as the ones with |dv| ≥ 9 km s⁻¹ and see that 26, 44, 21 and 9 per cent are Indeterminate, Optically thick, Blister and Optically thin, respectively, from the Pellegrini et al. (2012) classification. The clear decrease of the percentage for the optically thin category underlines the good efficiency of the Pellegrini et al. (2012) method to distinguish optically thin H ii regions from the other categories. The highest percentage corresponding to the optically thick category suggests that for these regions the flow is probably seen almost face on explaining why it is difficult for Pellegrini et al. (2012) to distinguish these blister H ii regions from the optically thick case.

5.2 Nebulae with peculiar kinematics and comparison with other kinematic works

From Table 1, we note that over most of the face of LMC the ΔV of the radial velocity profiles are of the order of 20–30 km s⁻¹. However, in 31 regions the ΔV of the velocity profiles is as large as 50–100 km s⁻¹. These regions are indicated in bold in Table 1. The most extreme cases are related to the expansion motions of supernova remnants (SNRs), bubbles and superbubbles.

It is important to remark that in Table 1 we are listing the values of ΔV of the radial velocity profile INTEGRATED over the whole extension of the region. In this sense, larger differences in velocities can be found in the velocity profiles integrated only over the central regions because we have expansion motions which are seen projected as more extreme in the centre. However, the velocity difference of velocity profiles integrated over the whole extension are also important when one wants to discriminate the effects of expansions and other non-circular motions in the extraction of rotation curves of more distant irregular galaxies, as it is one of the goals of this kinematic survey.

Let us start with some of the SNRs and their environments:

1. The kinematics of the SNRs N185 (DEML 25) and N70 (DEML 301), see Reyes-Iturbide et al. (2011), have been studied since a long time by Rosado et al. (1982) and Rosado et al. (1981) using a fixed gap Fabry–Perot interferometer attached to the European Southern Obseratory (ESO) 3.6 m telescope. The published expansion velocities of those objects (70 km s⁻¹) are the explanation of the high values measured for the ΔV (of 78 ± 18 km s⁻¹ for N185 and 48 ± 1 km s⁻¹ for N70) measured in this work. The detailed analysis of the spatial variations of the velocity profiles of those objects will be published elsewhere.

2. Small-scale complex motions at the centre of the nebular complex N11 (the second largest and brightest nebular complex after 30 Dor Nebula) have reported by Meaburn et al. (1989), as well as in the N11L SNR (SNR 0454-66.5) that shows expansion velocity larger than 350 km s⁻¹. The velocity profile integrated over the whole extension of N11 marginally shows large velocity differences (ΔV = 48 ± 1 km s⁻¹). The detailed analysis on the kinematics of this region has been studied by Rosado et al. (1996) using the data reported in this survey allowing to determine an expansion velocities of 45 km s⁻¹ for its central hole, implying a dynamical age of 2.5 × 10⁶ yr, and a possible product of the explosion of three SNe and shock-induced star formation. These authors found very complex velocity profiles in the faint emitting gas at the very centre of the central hole displaying different motions of smaller bubbles around single massive star. This study has been lately corroborated by the study of Nazé et al. (2001) as it can be seen in their fig. 6, ascertaining the complexity of the radial velocities very near the WC star HD32228.

3. A ΔV value of 64 ± 1 km s⁻¹ for the velocity profile INTEGRATED over the complete extension of the N186 (DEML 50) region, reveals the presence of the SNR N186D (SNR 0500-70.2) hosted in this nebular complex. A previous study carried out with another scanning Fabry–Perot interferometer attached to the ESO 1.5 m telescope (Laval et al. 1989) on this SNR, revealed a shock expansion of 100 km s⁻¹ quite consistent with the ΔV value determined in this work (velocity profile integrated over the whole extent of the SNR).

4. The velocity difference of 53 km s⁻¹ for the nebular complex N103A (DEML 84) reveals, on a larger scale, the internal motions mainly due to the SNR 0509-68.7 contained in this nebular complex and other bubbles studied in more detail in Ambrocio-Cruz et al. (1997), using the data of this survey and other data in the line of [O iii] (5007 Å).

5. High velocity differences in DEML 190, reaching 91 km s⁻¹, reveal the presence of SNR 0525-66.1 (N49), the brightest SNR in the LMC. From long-slit spectrophotometric data Melnik & Copetti (2013) obtain velocity differences reaching 270 km s⁻¹ at the very centre of this SNR. Our Fabry–Perot data reproduce those values when limiting the integration region to the central part of the SNR.

6. SNR N135 (DEML 316) is a region with a very high difference in velocities (165 km s⁻¹). This can be explained by the expansion motion of this SNR (SNR 0547-69.7), or pair of SNRs, seen probably superimposed in the line of sight as discussed in Toledo-Roy et al. (2009).

We find that most of the LMC SNRs included in the survey present velocity differences larger than those of typical H ii regions. Indeed, there are regions including SNRs whose integrated profile do not show atypical velocity differences. Such is the case of: DEML 66 (N23A or SNR 0506-68.0), DEML 134 (N120A, B, C, D, hosting the SNR 0519-69.7), DEML 152 (N44B,C hosting the SNR 0523-67.9) and DEML 221 (N206A, B hosting the SNR 0532-71.0). However, when we are limiting the integration region to the central part of the SNR we obtain complex profiles. In the case of SNR N120, a previous work with a scanning Fabry–Perot interferometer attached to the ESO 1.5 m telescope (Rosado et al. 1993) reported complex and intrinsically broad radial velocity profiles for that object. Thus, it is surprising not to find complex motions due to this SNR in the integrated velocity profile. A possible explanation is that N120 nebular complex contains also several H ii regions which are brighter than the SNR so that in the velocity profile integrated over the whole extension of the nebula, the SNR contribution is masked.

Large ΔV in the integrated velocity profiles are also detected in other regions not catalogued as SNRs. Such is the case for 23 regions. These regions are indicated with an asterisk in Table 1. Some of those regions are superbubbles or form part of larger structures like supershells. Like that the supergiant shell (SGS) LMC2 appears as a very prominent velocity feature, with several velocity components (Ambrocio-Cruz et al. 2004). These kinematic results are in agreement with previous kinematics by Points et al. (1999) obtained from optical echelle observations using the CTIO 4 m telescope.

Another example is the nebula N119 (DEML 132) located very close to the adopted kinematic centre of the LMC. It has a peculiar morphology (spiral-shaped) and its kinematics has been studied in Ambrocio-Cruz et al. (2008). This nebula shows a lower velocity than the one assumed to have if it rotated with the LMC disc suggesting that its peculiar velocity can be due to an inflow to the LMC centre near which N119 is located. We expected that the bright rims (moving at the systemic velocity of the nebula) would mask the complex motions of the central, and fainter regions. But our results show that Fabry–Perot velocity data cubes analysis is an effective way for identifying nebulae with such expansion motions.

Our objective was to study how much the rotation curves of farthest galaxies are affected by possible expansion motions of the nebulae taking into account in the construction of the rotation curve of the galaxy. Our results show that we need to be careful with nebulae whose velocity profiles are complex because of the superposition of several velocity components due to expansion motions. Complex velocity profiles could give wrong values of the systemic velocity of the expanding nebula if estimated by the ‘barycentric’ method. Instead, a careful analysis of the expansion should be done.

A special mention is for the 30 Dor complex, the brightest nebular complex in the LMC. The detailed analysis of the very complex velocity profiles in that region has been reported in Laval et al. (1995) and, more recently, in Torres-Flores et al. (2013).

The analysis of Laval et al. (1995) combined Hα observations presented here (spectral resolution ∼10 000) to Hα observations and [O iii] observations made with an higher resolution interferometer (R ∼ 31 000 at Hα 6563 Å and R = ∼24 000 at [O iii] 5007 Å). The observations of Torres-Flores et al. (2013) were done using the MEDUSA fibre mode of the Very Large Telescope (VLT) Fibre Large Array Multi Element Spectrograph (FLAMES)-Giraffe spectrometer, thus, attached to an 8.4 m telescope. They cover the spectral range from 6300 to 6691 Å, hence including the following emission lines: [O i] (6300 Å), [S iii] (6312 Å), [O i] (6363 Å), [N ii] (6548 Å), Hα, [N ii] (6584 Å) and He i (6678 Å), all these lines recorded simultaneously given the multiplexing advantage of the spectrograph. By far, the most intense is Hα and by this reason Torres-Flores et al. restrict their preliminary analysis to this line although the signal is so intense that they took care about saturation issues. The spectral resolution is R = 17 740, thus providing a sampling velocity resolution of 8 km s⁻¹ at Hα. The spatial sampling is done by two different data sets relevant to the nebular information while the most employed is a regular grid of 1.2 arcsec diameter optical fibers distributed in an array of 32 × 30 fibre positions separated by 20 arcsec between them; thus, the spatial resolution is 20 arcsec. This array gives a FoV of 10 arcmin enough to cover the central part of 30 Dor.

As we can see, both observations have the same spectral resolution. Our survey Fabry–Perot observations have a better spatial coverage than MEDUSA's both in FoV (38 arcmin versus 10 arcmin) as well as in spatial resolution (9 arcsec versus 20 arcsec). On the other hand, MEDUSA's observations cover simultaneously at least seven nebular emission lines that could serve to study nebular excitation conditions while our survey studies, by temporal scanning, two lines only. However, both types of observations give the same performances for kinematical studies because those studies rely in the Hα line.

That is why both studies report complex Hα velocity profiles in the central parts of 30 Dor as well as the existence of expanding ‘lobes’ sorting out from 30 Dor centre. The expansion velocities found with the Fabry–Perot data of our survey range from 195 to 340 km s⁻¹ for those lobes or shells. MEDUSA's observations succeeded in detecting six additional expanding structures. When we look at the detailed profiles in the central region of 30 Dor, in windows of 18 arcsec × 18 arcsec and 36 arcsec × 36 arcsec, we are also able to distinguish these six bubbles.

Finally, the SGS 2 appears as a very prominent velocity feature, with several velocity components (Ambrocio-Cruz et al. 2004). In addition, there are nebulae showing several velocity components; some of these nebulae have been studied in detail elsewhere in order to discern the origin of these complex motions.

5.3 The kinematical structure: about the L and D components

While observing the H i emission of the LMC Luks & Rohlfs (1992b) interpreted the double peaked H i velocity profiles as due to two different components of the LMC: a component spread over the whole LMC extension and that showed characteristics of galactic disc rotation (that they called the D component) and another component identified by its lower velocities called by these authors the L component. This latter component was found to be located in more restricted regions of the LMC extension and it was not clear whether this low-velocity H i component is associated with stars or if it is only a gaseous layer.

The D component follows a typical rotation curve while the L component seems to correspond to a H i layer located 50–500 pc above the LMC disc (whereas it was not clearly established whether this layer is parallel or inclined relative to the disc). The work by Kim et al. (1998), also using the neutral hydrogen emission, covered a diameter of about 7.3 kpc of the LMC revealing a large scale, very symmetrical H i disc with a complex rotation and showing some spiral arm features. Since Kim et al. (1998) observations have about 20 times better angular resolution than Luks & Rohlfs (1992b), they smooth their data in order to compare with the Luks & Rohlfs (1992b) large-scale LMC kinematical finding a good agreement with the rotation curve derived by Luks & Rohlfs (1992b) for their D component. However, Kim et al. (1998) observations were tuned to detect only the velocity channel corresponding to the maximum emission (the peak velocity) regardless of other possible components and, consequently, those more recent observations cannot throw light on the two velocity components already found by Luks & Rohlfs (1992b).

At other wavelengths, before the important discovery concerning the existence of two H i velocity components, it had been found by Hartwick & Cowley (1988) that the radial velocities of CH stars presented two kinematical components as well.

More recently, Subramaniam & Prabhu (2005) proposed a two-disc model for the central region of the LMC (up to a galactocentric radius of 3 deg). Their model consists in two rotating discs that intersect at the LMC kinematical centre and then diverge farther away due to the different orientations of their respective lines of nodes and the same inclination. Thus, an increasing separation is created between both discs. One of the discs, let us call it D1, has a Position Angle (PA) of the line of nodes of 130 deg and can be identified with a counter-rotating stellar disc deduced from the carbon star kinematical observed by van der Marel et al. (2002). While not said explicitly in Subramaniam & Prabhu (2005) work, disc D1 seems to correspond to the L component of Luks & Rohlfs (1992b). The other disc (let us call it D2) has a PA of the line of nodes of 170 deg and it could be identified with the D component found by Luks & Rohlfs (1992b). D2 is dominated by H i gas while it is made up of stars too. It seems that the model is rather insensitive to small changes in the systemic velocity of the discs and, consequently, the authors adopted a common systemic velocity of 260 km s⁻¹. Depending on the projection along the line of sight, there would be regions in the LMC that are seen projected in the same LMC direction but, because they belong to one or the other disc and consequently, they will show one or two velocity components, i.e. they could have two velocities or an unresolved average of the different velocities.

Our kinematical results allow us to determine whether the different H ii regions and superbubbles catalogued in this work (Table 1) belong to one or another velocity components, either from the H i gas of Luks & Rohlfs (1992b) or from the stars. The catalogued H ii regions and Super Bubbles (SBs) are directly related to young massive stars and thus, it is interesting to study in which disc massive stars are located: Are massive stars preferentially located in the Luks & Rohlfs (1992b) disc corresponding to H i D-component? Or, are massive stars distributed preferentially in the L component, that seems to correspond to the counter-rotating stellar disc of Subramaniam & Prabhu (2005) model.

To test these ideas, we have compared the values of our heliocentric velocity components of each of the catalogued H ii regions or SBs with the values of the different velocity components listed in table 1 of Rohlfs et al. (1984) according to their equatorial coordinates. We also checked those values with figs 1 and 13 of Luks & Rohlfs (1992b) work where the isovelocity contours are plotted for the D and L H i components, respectively. In that way, we have constructed Column 7 of Table 1.

The radial velocity histogram (Fig. 7) shows a bimodal distribution centred around 275 km s⁻¹ which is the systemic velocity of the LMC (Luks & Rohlfs 1992b; Kim et al. 1998). The coloured sub-histograms give the distributions of D (red) and L (blue) components of Luks & Rohlfs (1992b). As seen in this figure, the red sub-histogram shows two velocity peaks (a kind of horn profile with peaks at 255 and 300 kms⁻¹) characteristic of a rotating disc whereas the blue sub-histogram shows only the approaching side of a rotating disc. It is interesting to note that there are several H ii regions that are not catalogued by us as D or L but have velocities corresponding to the approaching side (velocity peaking at 260 kms⁻¹ in Fig. 7).

We note that 66 per cent of the catalogued regions belong to the D component, 24 per cent belong to the L component, 3 per cent have both components and 7 per cent cannot be associated with any component in H i. We can see that the regions with highest ΔV (>50 kms⁻¹) are those belonging mainly to the D component (11 regions). There are only six regions with ΔV >50 km s⁻¹ belonging to the L component.

We should stress that in the address of these objects that have both velocity components L and D there is a difference in velocity of approximately 40 km s⁻¹ but no significant physical difference is noted between them. On the other hand we can see that the regions with high SFR belong to both L and D components.

Thus, massive stars are associated mainly to the gaseous disc D2 of Subramaniam & Prabhu (2005) model. In both discs internal violent motions are present and so too are the highest SFRs.

5.4 The star-forming rate

From this SFR we have derived Σ_SFR (Table 1, Column 11) which is the SFR per unit area (M_⊙ yr⁻¹ kpc⁻²). It is obtained dividing SFR by the area calculated from the ionized gas emission radius. We can see that DEM 263 (N157) has the highest value of Σ_SFR (⁠|$38.39 \times 10^{-2} \,{\mathrm{M}_{\odot }}\,{\rm yr}^{-1}\,{\rm kpc}^{-2}$|⁠), which is not surprising since it is located in the centre of 30 Dor nebulosity. Our results show five more cases that call attention by a Σ_SFR higher the |$20 \times 10^{-2} \,{\mathrm{M}_{\odot}}\,{\rm yr}^{-1}\,{\rm kpc}^{-2}$|⁠: DEM 226 (N148), DEM 22 (N83), DEM 271 (N159), DEM 246 (N154) and DEM 10 (N79). There are 10 more cases with values of the order of tens |$\times 10^{-2} \,{\mathrm{M}_{\odot}}\ {\rm yr}^{-1}\,{\rm kpc}^{-2}$|⁠. It should be noted that there is no trend that these H ii regions belong to D or L components.

We can compare our estimated values for the SFR per unit area by comparing them to the ones obtained by Carlson et al. (2012) on nine LMC H ii regions. Those authors estimated the SFR by counting young stellar objects (YSOs) identified by means of several colour–magnitude diagrams using Spitzer/Infrared Array Camera on the Spitzer Space Telescope (IRAC), Multi-Band Imaging Photometer for Spitzer (MIPS) 24 μ, SIRIUS and 2MASS IR observations together with optical photometry from the Magellanic Cloud Photometric Survey (MCPS; Zaritsky, Harris & Thompson 1997).

We show Carlson et al. (2012) results and our results on the SFR per unit area (Σ_SFR) in Table 3. In that table we listed the Σ_SFR estimated by Carlson et al. (2012) assuming a SF time-scale of 1 Myr (Column 2) and 0.2 Myr (Column 3). We also listed the Σ_SFR obtained in this work, from the Hα luminosity as described before (Column 3) and the kinematical age of the H ii region (Column 4). Those last quantities were extracted from Table 1 of this paper for the nine regions studied by Carlson et al. (2012). This last quantity is listed in order to decide which time-scale of Carlson et al. (2012) calculations we should take for the comparison.

As one can see from the Table 3, our Σ_SFR values are intermediate between Carlson et al. (2012) estimates corresponding to time-scales of 1 and 0.2 Myr. It could possibly suggest that the real time-scale lies between 0.2 and 1 Myr. Indeed, except for N206, the kinematical ages determined for the H ii regions (Column 4 of Table 3) vary from 0.19 to 1.98 Myr, indicating that the time-scales selected by Carlson et al. (2012) are within that range.

Attempting to understand the star formation in this galaxy, we have plotted (Fig. 8) the SFR versus the kinematic age (Fig. 8A), the diameter (Fig. 8B), the ΔV (Fig. 8C) and the galactocentric distance (Fig. 8D). We can see that there is a correlation between the SFR and the Kinematical age (Fig. 8A); younger H ii regions have a minor SFR while oldest H ii regions have higher SFR. We can see that there is a point at the bottom right that comes out of the correlation, this point belongs to H ii region DEM 87; it is worth mentioning that this region belongs to the L component from Luks & Rohlfs (1992b). From Fig. 8B, we can see that there is a correlation between the SFR and the diameter of the H ii regions, we can see that larger H ii regions have a highest SFR than the smaller H ii regions. We can see that there are four points at the bottom of the correlation, these points belong to H ii regions DEM 286, 264, 257 and 310, it is worth mentioning that two of them belong to the D component, one to the L component and another to both components (L and D). In Fig. 8C, we can see that there is a faint correlation between ΔV and SFR; the H ii regions with a high dispersion show a slightly higher SFR. Finally form Fig. 8D, we can see that there is no correlation between the SFR and the galactocentric distance. The three points in the innermost part of the plot, belong to N119 H ii region, but we note that there are more regions towards the external parts of the LMC. However, plotting Σ_SFR versus the kinematical age, diameter, ΔV and galactocentric distance no correlation was found.

From our data we conclude that star formation in the LMC has proceeded until the present time at an average rate of roughly SFR = 0.11 M_⊙ yr⁻¹. By applying the reddening correction of A_v = 0.3 obtained by Imara & Blitz (2007), the SFR for the LMC becomes SFR = 0.14 M_⊙ yr⁻¹. Our SFR value is about half the value estimated from optical Hα + [N ii] images by Kennicutt et al. (1995). Since we have used its prescription published in Kennicutt (1998) it is possible that the previous value (from 1995) had been obtained with a different value for the constant of proportionality in the relation between SFR and L(Hα). Another possibility is that the SFR of Kennicutt et al. (1995) has been computed over a LMC area larger than the one of our observations. On the other hand, the calculated value in this work is only slightly lower than those obtained by Whitney et al. (2008) and by Lawton et al. (2010). These authors use IR data from Spitzer in order to evaluate the SFR from YSO direct counting and from the IR flux, respectively. Thus, our SFR value agrees quite well with those derived from IR observations (where reddening corrections are not important) and it is possibly slightly underestimated because of the incomplete coverage of our kinematical survey.

In the same way, we calculated the SFR of 30 Dor Nebula obtaining a value: SFR = 0.025 M_⊙ yr⁻¹, and applying the reddening correction of Av = 0.3, we obtain SFR = 0.033 M_⊙ yr⁻¹. This value is lower than the one derived by Doran et al. (2013) of SFR = 0.073 M_⊙ yr⁻¹. Those later authors derived their SFR value by counting the O stars present in 30 Dor. Our lower SFR could be explained if the reddening correction we apply to our SFR value is too low. Indeed, it is quite possible that this happens because 30 Dor is a place where large amounts of gas and dust are present thus producing a higher than the mean internal reddening.

Individually, for the H ii regions our values of Σ_SFR coincide in one order of magnitude with those obtained by Carlson et al. (2012).

5.5 The triggered star formation

Star formation triggered by SGSs in the LMC, as for example around LMC 4 and LMC 5, has been clearly put in evidence (e.g. Yamaguchi et al. 2001b; Book, Chu & Gruendl 2008). Recently, Dawson et al. (2013) redefined the sample of SGSs and shell complexes and their properties as well as their velocity range. We have then taken advantage of the Hα velocity information to look at the association of our H ii regions sample to the supergiant shells (noted SGS and following the Dawson et al. 2013 numbering). We consider a H ii region associated with a SGS if it is spatially located in the direction of the SGS and also if its velocity falls in the velocity range of the SGS. In such a way we find that 75 per cent of our 210 H ii regions sample are associated with SGS, suggesting that the massive star formation in the LMC is dominated by the SGS impact on the ISM. Only four regions of our sample located in the direction of SGS exhibit systemic velocity different by 10 km s⁻¹ from the SGS velocity range (DEM 79 in SGS 6, DEM 227 in SGS 4, DEM 209 in SGS 3 and DEM 316 in SGS 2). All these regions exhibit a large dv and/or a large ΔV suggesting regions with particular kinematic that could bias their systemic velocity determination. This can be illustrated with DEM 316 which is in fact a SNR with complex line profiles (Williams et al. 1997) and DEM 227 which, based on its associated molecular cloud velocity, is placed inside SGS 4 by Yamaguchi et al. (2001b).

To investigate the present star formation activity on the scale of H ii regions, we carried out a census of YSOs possibly triggered by H ii regions. We searched for class O/I and class II YSOs located in the direction of the H ii regions from the Spitzer catalogue (Spitzer-SAGE program; Meixner et al. 2006) following the [3.6]–[4.5] and [5.8]–[8.0] colour–colour criteria from Allen et al. (2004). To take into account the possible contaminating objects (e.g. asymptotic giant branch stars, background galaxies) with similar colours and the general star formation we evaluate the number of YSOs in several areas located between H ii regions. These reference areas give a mean number density of 1.08 × 10⁻⁴ pc⁻². Adopting 3σ rms above this value as limit for triggered star formation we find that 15 per cent of the catalogued H ii regions can have triggered star formation with a mean density of 6.1 × 10⁻⁴ pc⁻² above the reference level. In our Galaxy, Watson, Hanspal & Mengistu (2010) reported that 20 per cent of their sample (46 infrared bubbles) showed a significant number of associated YSOs, while Deharveng et al. (2010) and Thompson et al. (2012) estimated a triggered massive SFR of 20 per cent (from a sample of 102 bubbles) and 14–30 per cent (from a sample of 322 bubbles), respectively. Yamaguchi et al. (2001a) have shown that SGSs play an important role in triggering the formation of molecular clouds and clusters in the LMC. They found that the surface density of clusters (clusters younger than 30 Myr) is enhanced by a factor of 4 inside the SGSs as compared with the outer parts. They establish a mean number density of clusters in SGSs of 1.8 × 10⁻⁵ pc⁻².

To better understand the condition for the triggering of star formation we look at the behaviour of the YSO number density with the Hα luminosity and the kinematical age (Figs 6 and 9). We find (Fig. 9) that the highest YSOs number density is found for H ii regions with small kinematical ages (t_kin < 1 Myr), hence for small size H ii regions. This is in agreement with Vaidya, Chu & Gruendl (2009) who show that YSOs are found in small H ii regions in addition to be found either in dark clouds or on the tip of bright rimmed dust pillars and with Thompson et al. (2012) who underline that, in our Galaxy, bubbles that are associated with massive star formation tend to be smaller and thinner. Fig. 6 down confirms that H ii regions with possible triggered star formation have small size and Fig. 6 up shows that they correspond mainly to radiation-bounded H ii regions. Selecting all regions younger than 1 Myr the triggered SFR becomes 24 per cent in more agreement with the values found for our Galaxy. We can then conclude that the triggered star formation process seems to be more active in small (young) and radiation-bounded H ii regions.

6 CONCLUSION

We report the kinematical and photometric results of 210 H ii regions in the LMC. We find a bimodal distribution of the Hα luminosity of LMC H ii regions, fact suggested to be due to a transition between radiation-bounded and density-bounded H ii regions.

We derive the SFR for every nebula and we conclude that the star formation in the LMC has proceeded until the current time at an average rate of roughly 0.11–0.14 M_sun yr⁻¹ and 0.025–0.033 M_sun yr⁻¹ for 30 Dor nebula. It is remarkable to note that the regions with the highest value of SFR belong to both, L (one H ii region) and D (two H ii regions), components.

We find that 151 H ii regions (74 per cent of our sample) can be associated with a molecular cloud.

75 per cent of our 210 H ii regions sample are associated with SGS, suggesting that the massive star formation in the LMC is dominated by the SGS impact on the ISM.

We note that 66 per cent of the catalogue regions belong to the D component while 24 per cent belong to the L component and 3 per cent belong to both components, L and D. We can see that the regions with highest velocity dispersion (>50 km s⁻¹) are those that belong to the D component.

We find that most of the LMC SNRs included in the survey present velocity differences larger than those of typical H ii regions. Large ΔV in the integrated velocity profiles are also detected in other regions not catalogued as SNRs. Some of those regions are superbubbles or form part of larger structures like supershells.

The authors acknowledge financial support from CNRS (France) grant, IN108912 grant from DGAPA-UNAM, 82389-F grant from CONACyT and CONACyT ‘programa de retención 2010 exp 117100’ throughout the course of this work. We warmly thank J.C. Lambert at CeSAM/LAM for making available GLNEMO2 http://projets.lam.fr/projects/glnemo2 and his support for using it. MR acknowledges the financial support of grant CY-253085 from CONACyT. We are also sincerely indebted to the reviewer for his/her valuable comments and suggestions allowing us to improve the work.

REFERENCES