EDP Sciences
Free Access
Issue
A&A
Volume 553, May 2013
Article Number A85
Number of page(s) 18
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/201220991
Published online 14 May 2013

© ESO, 2013

1. Introduction

Since the first results obtained by the Infrared Astronomical Satellite (IRAS; Soifer et al. 1984), there has been strong effort to study the physical processes that power the luminous (LIRGs; 1011L < LIR < 1012L) and ultraluminous (ULIRGs; 1012L < LIR < 1013L) infrared galaxy population (Sanders & Mirabel 1996; Lonsdale et al. 2006). The origin of the mid- and far-infrared emission (LIR[8–1000 μm]) that dominates their bolometric luminosity is established as mainly due to massive starbursts with a small AGN contribution for LIRGs and with an increasing contribution in ULIRGs (e.g. Goldader et al. 1995; Veilleux et al. 2009; Nardini et al. 2010; Alonso-Herrero et al. 2012 and references therein). The radiation that originates in the starburst and/or the active galactic nucleus is then reprocessed by a surrounding dust component, and then re-emitted at long wavelengths in the form of a huge infrared emission.

One of the main difficulties in understanding the underlying power source of the LIRGs and ULIRGs is the high opacity of their nuclear regions. Previous optical (García-Marín et al. 2009) and near-infrared studies in LIRGs and ULIRGs (Genzel et al. 1998; Scoville et al. 2000; Alonso-Herrero et al. 2006) reveal that the distribution of the dust in these object is not uniform and that, though the dust tends to concentrate in the inner kiloparsecs with average visual extinction of AV ~ 3–5 mag in LIRGs and even higher in ULIRGs, the global distribution shows a patchy structure on kiloparsec and sub-kiloparsec scales (Colina et al. 2000; García-Marín et al. 2006; Bedregal et al. 2009).

Besides the importance of knowing the 2D structure of the dust to understand the environment where the power source of the LIRGs and ULIRGs is embedded, dust plays a key role in the derivation of other physical and structural parameters of these objects, such as the derived star formation rate (García-Marín et al. 2009), the effective radius (Arribas et al. 2012), and as a consequence, the dynamical masses.

Understanding the distribution and effect of dust in star-forming galaxies is also important for correctly interpreting or comparing different tracers of star formation during the history of the Universe. This is in turn relevant when comparing local and high-z star-forming galaxy populations, which are often observed using different tracers and/or resolutions. The distribution of dust can in principle be studied in detail in local U/LIRGs with the advantage of the relatively high linear resolution and signal-to-noise ratio (S/N). These studies can, therefore, help us interpret observations of analogous high-z star-forming galaxies, for which such a level of resolution, and S/N is not attainable with current instruments.

The present work is part of a series presenting new H- and K-band Spectrograph for INtegral Field Observations in the Near Infrared (SINFONI, Eisenhauer et al. 2003) seeing-limited observations of a sample of local LIRGs and ULIRGs. Piqueras López et al. (2012, hereafter Paper I) presented the atlas of the sample, the data reduction, and a brief analysis and discussion of the morphology of the gas emission and kinematics. In this second paper in the series, we focus on the study of the 2D distribution of the dust derived using the Brγ/Brδ and Paα/Brγ ratios for LIRGs and ULIRGs, respectively, whereas in Piqueras López et al. (Paper III, in prep.), we will apply the results for the 2D dust structure to study both the overall star formation rate (SFR) and the kpc structure of the SFR surface density (ΣSFR) of the galaxies of the sample.

The paper is organized as follows. In Sects. 2 and 3 we briefly describe the sample, observations, and data reduction process, which are detailed in Paper I. The procedures for obtaining the emission and AV maps are described in Sect. 4, and the results and analysis of the AV maps and distributions are presented in Sect. 5. Finally, Sect. 6 includes a brief summary of the paper. Throughout this work we consider H0 = 70 km s-1 Mpc-1, ΩΛ = 0.70, ΩM = 0.30.

Table 1

SINFONI sample.

2. The sample

The SINFONI sample is a subsample of a larger set of local LIRGs and ULIRGs described in Arribas et al. (2008) that was observed with different IFS facilities. These facilities include optical and infrared IFS instruments in both hemispheres, such as INTEGRAL+WYFFOS (Arribas et al. 1998) at the 4.2 m William Herschel Telescope, VIsible MultiObject Spectrograph (VLT-VIMOS, LeFèvre et al. 2003), Potsdam MultiAperture Spectrophotometer (PMAS, Roth et al. 2005), and SINFONI. The large sample (~70 sources) covers the whole range of LIRGs and ULIRGs IR luminosities and the different morphological classes observed in these objects.

The sample used for the present study comprises a total of 17 objects, 10 LIRGs, and 7 ULIRGs covering the luminosity range log (LIR/L) = 11.10 − 12.43 (see Table 1). It was selected to be representative of the different morphological types of LIRGs and ULIRGs, from isolated galaxies to strongly interacting systems and mergers. The mean redshifts of the LIRG and ULIRG subsamples are zLIRGs = 0.014 and zULIRGs = 0.072 and the mean luminosities are log (LIR/L) = 11.33 and log (LIR/L) = 12.29, respectively. More details on the sample can be found in Paper I.

3. Observations and data reduction

The data were obtained in service mode between April 2006 and July 2008 using SINFONI on the VLT (periods 77B, 78B, and 81B). The sample was observed in the K band (1.95–2.45 μm) with a plate scale of 0125 × 0250 pixel-1 that results in a field of view (FoV) of 8″ × 8″ by a 2D 64 × 64 spaxel frame. The spectral resolution is R ~ 4000, and the full-width-at-half-maximum (FWHM) as measured from the OH sky line at 2.190 μm is ~6.0 Å with a dispersion of 2.45 Å per pixel. The observations have a typical resolution of ~0.63 arcsec (FWHM, seeing-limited) that corresponds, on average, to ~0.2 kpc and ~0.9 kpc for LIRGs and ULIRGs, respectively.

Owing to the limited FoV, the data sample typically ~ 3 × 3 kpc for the LIRGs and ~ 12 × 12 kpc for the ULIRGs subsample. Some of the more extended systems were then observed in different pointings to cover regions of interest like secondary nuclei or star-forming complexes. For a detailed description of the observations, pointings, and integration times, see Paper I.

The reduction process was performed using the ESO pipeline ESOREX (version 2.0.5) and our own IDL routines for the flux calibration. All the individual frames were corrected from dark subtraction, flat fielding, detector linearity, geometrical distortion, wavelength calibration and sky-subtraction. After this process, the cubes of those objects with several pointings were combined to build a final mosaic.

The maps of the Paα, Brγ, and Brδ lines were constructed by fitting a Gaussian profile on a spaxel-by-spaxel basis. We have developed our own routines, based on the IDL routine MPFIT (Markwardt 2009), to perform the fitting of the cubes in an automated fashion. As described in Paper I, the data were binned using the Voronoi method developed by Cappellari & Copin (2003) in order to maximise the S/N over the entire FoV. Each map was binned independently, since the S/N depends on the wavelength, as well as the spatial distribution of the emission, to achieve a minimum S/N on average in the whole FoV. This minimum S/N varies from object to object, and is typically between 15 and 25 for the brightest line (Brγ and Paα in LIRGs and ULIRGs, respectively), and between 8 and 10 for the weakest (Brδ and Brγ for LIRGs and ULIRGs, respectively). Further details on the flux calibration and the individual S/N thresholds used in the Voronoi binning can be found in the Paper I of these series.

thumbnail Fig. 3

Surface brightness distributions of the individual spaxels of LIRGs and ULIRGs. For clarity, we only plot a random distribution of the 20% and 50% of the points for LIRGs and ULIRGs, respectively. The colour code indicates the AV value of each spaxel, whereas the points with AV < 0 are outlined with a black contour line. The increase in the extinction towards low surface brightness values is mainly due to the high uncertainties of the flux measurements, in particular in ULIRGs, where the highest AV values correspond to the spaxels with the lower Brγ surface brightness.

Open with DEXTER

4. Data analysis

The 2D extinction / dust structure was derived using the Brγ/Brδ and Paα/Brγ line ratios for LIRGs and ULIRGs respectively. Although the Brδ line is detected in most of the ULIRGs, its S/N is not high enough to map the emission and, in most of the cases, it is not sufficient to perform an integrated analysis of the emission.

As mentioned before, the maps of the different lines were constructed by fitting a single Gaussian profile on a spaxel-by-spaxel basis (see Figs. 1 and 2). Based on the emission maps, we obtained the extinction in magnitudes (AV) following the procedure outlined in Bedregal et al. (2009). We compared the theoretical ratio between the two lines (Brγ/Brδ = 1.52 and Paα/Brγ = 12.07 at T = 10 000 K and ne = 104 cm-3, case B; Osterbrock 1989) with the measurements for each spaxel. The extinction in magnitudes could be expressed in the form (1)where Fλi,O and Fλi,T are the observed and theoretical fluxes for a line centred at λi. We made use of the extinction law described in Calzetti et al. (2000) to express Eq. (1) in terms of the visual extinction AV (ABrγ = 0.096 AV, ABrδ = 0.132 AV and APaα = 0.145 AV).

Since the individual values of AV are sensitive to the S/N of the weakest line (Brδ and Brγ for LIRGs and ULIRGs respectively), we have only considered those spaxels where the weakest line has been detected above an S/N threshold of four to obtain reliable AV. This effect is very significative in the case of the Brγ/Brδ ratio since the Brδ line lies close to the blue limit of the SINFONI K-band. As discussed in Paper I, this wavelength region is strongly affected by noise due to the sky emission, and the atmospheric transmission also decreases. This translates into a more complex local continuum determination, making the line fitting more uncertain. An excess in the continuum level estimation would decrease the line flux and, therefore, increase the extinction (see expression above). Although the Paα line also lies in this region of the spectra, this effect is not so relevant, given the strength of the line (more than ×10 the Brγ emission), and it is in the numerator of Eq. (1).

The 1σ uncertainties of the individual AV values vary typically from 10–20% in central regions with high S/N, up to 70–80% in external areas of low surface brightness (Σ ≲ 10-16 erg s-1 cm-2 kpc-2), with a median value of 30–35%. Owing to the larger individual errors in the low surface brightness areas, we observed an artificial increase in the AV measurements, especially in the ULIRGs. This systematic effect is observed in Fig. 3, where the highest extinction values are measured in those spaxels with the lower Brγ surface brightness. The AV uncertainties are obtained from the corresponding errors in the fluxes of the two lines, which as described in Paper I, are estimated using Monte Carlo simulations. The advantage of this kind of error estimation is that uncertainties are directly measured from the spectra, so they not only include the effect from photon noise but also take uncertainties due to an improper continuum determination or line fitting into account.

Table 2

Integrated properties and statistics of the AV distributions.

5. Results and discussion

5.1. Two-dimensional extinction structure in LIRGs and ULIRGs

Our 2D extinction maps cover areas from ~2 kpc to ~12 kpc for LIRGs and ULIRGs, respectively (see Figs. 1 and 2). The seeing-limited observations provide a linear resolution equivalent to a physical scale resolution of ~0.2 kpc and ~0.9 kpc for each subsample. On these scales, the extinction maps show that dust is not uniformly distributed, revealing a clumpy structure with almost transparent areas (AV ≲ 1 mag) and regions where the visual extinction is higher than ten magnitudes.

In LIRGs, the extinction maps show a very irregular and clumpy structure on scales of ~200–300 pc, already observed from near-IR continuum maps (Scoville et al. 2000). The higher AV values are usually associated with the nuclear regions of the objects, although obscured extranuclear regions are also common. The star-forming regions with high Brγ surface brightness, found along the dynamical structures like arms of rings, are typically low-extinction regions.

As shown in the radial profiles of NGC 3110 or IRASF 12115-4656 (Figs. 1b and e), the extinction increases inwards up to AV ~ 15–25 mag. This behaviour is also observed in ESO 320-G030 (Fig. 1d), where the emission is concentrated on a star-forming ring of ~500–600 pc radius. Although the nucleus might also be obscured, the radial profile of this object is not as steep as in NGC 3110 or IRASF 12115-4656 and indicates that the lack of emission could also be due to the intrinsic distribution of the star-forming regions around the ring.

The morphology of the AV maps in the ULIRG subsample suggest a patchy, non-uniform distribution of the dust, typically on physical scales of ≳1 kpc that correspond to our resolution limit. Owing to the higher linear resolution (i.e. kpc/spaxel) of the ULIRG subsample, the comparison with the LIRGs is not straightforward. As shown in Fig. 5, although our data samples similar areas of ~1–2 Reff, the dust structure is probed with significantly different spatial resolutions, owing to the factor ×5 in distance between both subsamples. This difference precludes our from resolving sub-kiloparsec structures in ULIRGs, such as the ones observed in the LIRG AV maps, limiting our physical resolution to ~1 kpc. As discussed in Sect. 5.4, these differences in the linear resolution between both subsamples not only shape the observed dust morphology in the more distant galaxies, but also might determine global measurements of the extinction, such as the median of the AV distributions. In Sect. 5.5 we discuss how this distance effect would have direct implications for the study of high-z galaxies.

5.2. AV distributions and radial profiles

Figures 1 and 2 show the AV distributions for each galaxy. Although it is clear that most of the spaxels with AV < 0 have no physical meaning individually, we have kept them in the distributions since they do have statistical relevance. If we remove them from the distributions, we introduce a bias toward the high AV values, displacing the mean and median of the distributions artificially. On the other hand, due to the S/N threshold adopted, we assure that most of those spaxels with AV < 0 are compatible with AV ~ 0 within the uncertainties.

thumbnail Fig. 4

Individual AV distributions, on a spaxel-by-spaxel basis, of the galaxies of the sample ordered by increasing LIR. The extremes of the distributions are the 5th and 95th percentiles (P5 and P95). The boxes illustrate the interquartile range, whereas the horizontal blue and red lines correspond to the median and the weighted mean of the distribution, respectively. The measurements of the nuclear extinction are plotted as black diamonds. The total AV distributions for LIRGs and ULIRGs are shown on the right-hand side of the plot.

Open with DEXTER

thumbnail Fig. 5

AV distributions and radial profiles of the LIRGs (top) and ULIRGs (bottom) subsamples on a spaxel-by-spaxel basis. The median AV values of each distribution are shown in the left panels. In the central and right panels, the radial profiles are plotted in terms of the radius in kpc (centre) and in units of the Hα effective radius (Reff, right panel) extracted from Arribas et al. (2012). The red line represents the weighted mean of AV and its error for different radial bins in steps of 1/30 of the total radial coverage. For the ULIRG subsample, we plotted the radial profile of the sample by considering each component of the systems separately (top) and the profile extracted by taking the brightest nucleus in the K-band continuum as the centre of the systems (bottom).

Open with DEXTER

The histograms show a wide variety of distributions, from narrow, peaked distributions, such as NGC 5135 or IC 4687 (Figs. 1f and 1h), concentrated towards low AV values, to wide distributions such as NGC 3110 or IRAS 14348-1447 (Figs. 1b and 2c) that extend up to ~30–35 mag. The median and weighted mean AV values, together with the 5th and 95th percentiles of the distributions, are listed in Table 2. Figure 4 also compares the individual distributions of each galaxy of the sample, ordered by increasing LIR. As shown in the figure, most of the individual values, within the interquartile ranges, are concentrated between AV ~ 1 and AV ~ 20 mag, and there is no clear evidence of any dependence with LIR.

In LIRGs, the visual extinction ranges between AV ~ 1 − 20 mag, whereas in ULIRGs, the AV values range between AV ~  2 − 15 mag. In LIRGs, these values of the visual extinction are very similar although slightly lower than previous results in the mid-infrared from Spitzer, AV ≲ 1 − 30 mag with a mean value of ~11 mag (Pereira-Santaella et al. 2010; Alonso-Herrero et al. 2012). On the other hand, ULIRGs show slightly higher values than previous measurements in the optical, from AV ≲ 0.2 mag to ~9 mag (García-Marín et al. 2009), and significantly lower than mid-infrared measurements based on the silicate absorption feature at 9.7 μm, from AV ~ 6 mag up to AV ≳ 40 mag (Imanishi et al. 2007).

These AV values have to be considered as lower limits of the dust extinction, since the theoretical values of the ratios are based on the assumption that the gas is optically thin. However, it is well known that the central regions of these objects are dusty environments and that the observed recombination lines might have been originated at different depths within the star-forming regions. That also means that we would infer different extinction values from different emission line ratios, since the lines are probing different optical depths.

To compare the AV distributions of LIRGs and ULIRGs, we combined all the available spaxels of each luminosity bin to obtain a typical AV distribution. These distributions are shown in Fig. 4 and, in more detail, in Fig. 5. The median values of AV for each subsample are similar, AVmed = 5.3 mag and AVmed = 6.5 mag, and the number of individual spaxels is ~16 000 for the LIRG distribution and ~3000 for the ULIRG one. The shape of the distributions are, however, slightly different. Whereas the LIRG distribution seems to have more than 50% of the points concentrated within the range AV ≃ 1–10 mag, the ULIRGs extend over a somewhat narrower range of AV ≃ 3–10 mag and tend to reach higher AV values on individual spaxels. On the other hand, the modes of the distributions are also different; whereas the LIRG distribution peaks at ~3–4 mag, the mode of the ULIRG distribution reaches up to ~7–8 mag. Although some of these differences might be intrinsic, we discuss in Sect. 5.4 that the physical sampling of the maps plays an important role in the study of the extinction, owing to the patchy structure of the dust.

We obtained the radial profiles of the extinction for every individual object of the sample and characteristic profiles for both LIRG and ULIRG subsamples. For this purpose, we adopted the same criterion as in Paper I to identify the central spaxel of each object with the brightest spaxel of the FoV in the K-band image (see Figs. 1 and 2). The only exception in NGC 3256, since the nucleus of this galaxy was not observed in the K-band (see Paper I for further details). For this object, we used the H-band continuum image of the nucleus to identify the central spaxel. For those systems with multiple nuclei (i.e. all the ULIRGs with the exception of IRAS 17208-0014), we also extracted the radial profiles of each component separately.

As shown in Figs. 1 and 2, the profiles typically sample the inner ~ 2 kpc for the LIRGs and ≲6 kpc for the ULIRGs, and most of them have an almost flat or negative slope. LIRGs show steeper negative slopes than ULIRGs, especially in the central 0.5–1 kpc. In LIRGs, these slopes are typically of ~  −2.4 mag kpc-1 on average, versus ~  −0.3 mag kpc-1 in ULIRGs. This could be explained by the different sampling scales of both subsamples, since we cannot resolve the innermost regions of the ULIRGs with the resolution achieved for the LIRG subset. In those pre-coalescent systems with multiple nuclei, we found no systematic differences between the radial profiles of both components. These profiles typically sample the innermost ~ 2–3 kpc of each component of the system, and show no clear radial dependence of the extinction on these scales.

We extracted the radial profiles of the total AV distributions of LIRGs and ULIRGs separately. Since each object has been observed with different sampling and physical scales, we also obtained the profiles in units of the effective radius Reff, using the values from Arribas et al. (2012) obtained from Hα maps. As shown in Fig. 5, the LIRG profile is almost flat beyond r ~ 1 kpc or ~0.5 Reff, with an average value of AV ~ 5.3 mag. Within the central kiloparsec, the extinction increases up to ~10 mag. The ULIRG subsample shows a very uniform profile, with a median value of AV ~ 6.0 mag and only local deviations due to the presence of double nuclei in some of the systems, or due to strong complexes of star formation at distances beyond 2–3 kpc radius (see Figs. 2a and b for some examples). The radial profile shows small differences when extracted for each component separately, with an almost flat slope over the inner 2–3 kpc radius (~Reff). The measurements beyond r ~ 6 kpc or r ~ 2 Reff are very uncertain, owing to the lack of available spaxels and to the low surface brightness of the Brγ emission in these regions. As mentioned before, the extinction in ULIRGs derived using Eq. (1) is highly affected by noise fluctuations of the Brγ line.

5.3. Nuclear and integrated AV measurements

We obtained integrated AV measurements for different regions of interest in each object (Table 2). The uncertainties of the parameters are obtained by a Monte Carlo method of N = 1000 simulations, and do not take the 1σ uncertainties of ~5% into account in the absolute flux calibration (see Piqueras López et al. 2012). We found that, in ~70% of the objects, its nucleus corresponds to a peak in the extinction, ranging from AV ~ 3 mag up to AV ~ 17 mag. These values of the nuclear extinction are higher than each median and mean AV values in ~ 57% of the objects. However, there are some galaxies, such as NGC 3110 or IC 4687 (Figs. 1b and h), where the nucleus is completely obscured, and no measurements of the extinction are available. That the nucleus of the objects coincides with a peak of the extinction agrees with other studies in LIRGs and ULIRGs (see Alonso-Herrero et al. 2006; García-Marín et al. 2009) based on Hα/Hβ and Paα/Brγ ratios, and with mid-infrared studies with Spitzer, based on the silicate absorption feature at 9.7 μm (Imanishi et al. 2007; Pereira-Santaella et al. 2010). These authors found that either the highest extinctions coincide with the nucleus of the objects or the nuclear regions are local maxima in the AV maps.

For those objects with a double nucleus, we also presented integrated measurements of the extinction for the secondary one. Although almost all of the interacting systems in our sample are ULIRGs, NGC 3256 presents a well known, highly obscured nucleus ~ 5 arcsec to the south of the main one. Our measurements of the extinction of the southern nucleus of this object shows that it is one of the most extinguished regions in our sample, with AV ≃ 12.2 mag, in good agreement with previous works (Kotilainen et al. 1996; Alonso-Herrero et al. 2006; Díaz-Santos et al. 2008; Rich et al. 2011).

5.4. Dust clumpiness and the effect of the linear resolution

thumbnail Fig. 6

Evolution of the computed median AV of the individual distributions of the LIRG subsample as a function of scale/distance. The AV values are normalised to the rest-framed distribution. The value that corresponds to the AV distribution of the whole LIRG subsample is plotted as black circles and a power-law fit to the data as a dashed black line. The mean distance of the ULIRG subsample (328 Mpc) is plotted as a vertical dotted line for reference.

Open with DEXTER

thumbnail Fig. 7

Different simulations of the LIRGs AV distribution for increasing distance (from top to bottom and left to right). The median of each distribution is highlighted with a vertical dashed line. The distance and the median AV of the simulated distribution is annotated in each panel. The top left panel corresponds to the average distance of the ULIRG subsample, and shows the observed AV distribution of the ULIRGs and its median, in yellow.

Open with DEXTER

As mentioned in Sect. 5.1, the AV maps reveal a patchy, clumpy structure of the dust at sub-kiloparsec and kiloparsec scales in LIRGs and ULIRGs, respectively. Given this non-uniform distribution, the measurements of the extinction at different distances might be affected by the physical scale of the observations. To probe how the pixel scale might bias the AV measurements, we have obtained the AV distribution of our LIRG sample simulating different scales, hence different distances. This distance effect because of the linear resolution might be even more relevant for high-z objects, where the structure is sampled on even larger scales of ~1–2 kpc.

To simulate the distribution at further distances, we degraded the individual Brγ and Brδ maps to different scales, and obtained maps of poorer spatial resolution. In this process, we only considered the same valid spaxels as in the original maps. Once the maps were degraded, we obtained the AV distributions of each individual object as described in Sect. 4, which were finally combined in one single distribution.

Since the FoV of our SINFONI observation is limited to ~8″ × 8″, we could only sample the innermost ~3 × 3 kpc of the LIRGs (typically ~Reff). If we translate these scales to a distance that is ten times larger, these ~3 kpc are equivalent to an angular distance of ~1′′, i.e. 1/8 of the FoV or ~8  spaxels. The lack of data from the external regions of the local objects keeps the extrapolation to larger distances from beeing straight-forward.

Figure 6 shows the relation of the median of the individual AV simulated distributions to the sampling scale/distance, for each galaxy in the LIRG subsample. Although each curve shows a different behaviour that depends on the particular gas and dust distribution, there is a general trend towards lower AV values as distance increases. This decrement of the median value of the visual extinction is also observed when we consider the distribution of LIRGs as a class, adding all the individual spaxels of the LIRG subsample in a single distribution, as shown in the figure. To parametrize this observed decrement of the median AV of the LIRG distribution with the sampling scale/distance, we fitted the data points to a simple power-law model, and found that the best-fit model corresponds to (2)where AV and S are the median values of the AV distribution and the physical scale at a distance D, respectively, and AV0 = 5.27 mag and S0 = 290 pc arcsec-1 are the median values of the rest-frame AV distribution and median physical scale at the mean distance of the LIRGs subset, D0 = 63.3 Mpc, respectively.

In Fig. 7, we show in detail the observed AV distribution of the LIRG subsample and the simulated distributions for increasing distances. The different panels reveal that not only the median value of the distribution decreases when the galaxies are sampled on a larger physical scale, but also that the shape of the distribution is slightly different, becoming narrower than the rest-framed distribution, and more compact. The mode of the simulated distribution also changes with respect to the ULIRG observed distribution, and becomes lower than the median. As shown in Fig. 6, the difference between the rest-frame extinction, AV0, and the simulated AV increases rapidly within the first D ~ 700 Mpc, and seems to reach an asymptotic value of AV/AV0 ~ 0.65 beyond that distance.

The non-uniform distribution of the dust, even on small scales, makes that, at a given resolution unit, we map both obscured regions and areas where the interstellar medium is more transparent. This average is biased towards the brightest regions, which are those were the emission is less absorbed. The result is that, on average, the dust distribution is smoothed, becomes narrower, and the AV values that we measure are biased towards the lowest values.

As mentioned in Sect. 5.2, we measured a median extinction of AV,LIRGs = 5.3 and AV,ULIRGs = 6.5 mag for our LIRG and ULIRG subsamples, respectively. Since the average distance of each subsample is ~ 63 Mpc and ~ 328 Mpc (see Paper I), this factor ~5 in distance and sampling scale (from ~0.2 kpc to ~0.9 kpc on average, respectively) translates to a decrease in the measured AV of ~1.2 mag, as shown in Fig. 7 (top right panel). If we assume that LIRGs and ULIRGs have similar structures and apply the same correction to the ULIRGs subsample, we find that the average AV in the ULIRGs, corrected from distance effects, reaches AV ~ 8.0 mag, i.e. ~2.8 mag higher than in LIRGs.

Besides this resolution/distance effect, the difference between the observed AV distribution in LIRGs and ULIRGs could also be interpreted in terms of intrinsic differences in the morphology between both classes. Although there seems to be a general trend for LIRGs as a class that suggest that the observed AV decreases with the distance/resolution, Fig. 6 also shows that the behaviour of each individual galaxy is very different, and depends strongly on the particular morphology of the dust distribution. These differences among the objects of the LIRGs subset suggest that the correction to the observed AV could not be accurate for individual galaxies, and would only reflect a general trend for LIRGs as a class.

5.5. Implication for extinction-corrected properties in high-z galaxies

There are different observational proofs that suggest that sub-millimetre galaxies (SMGs) could be the high-z analogous of local (U)LIRG. The sub-millimetre and radio fluxes of this population of galaxies indicate that their bolometric luminosities are comparable to local ULIRGs, whereas their mid-IR emission seems to be similar to the observed in local LIRGs (Kovács et al. 2006; Takata et al. 2006; Menéndez-Delmestre et al. 2009). Besides this, recent IFS-based observations of SMGs reveal that they present evidence of clumpy star formation on kiloparsec scales and similar star-formation rate surface densities (ΣSFR) to the local counterparts (Nesvadba et al. 2007; Harrison et al. 2012; Alaghband-Zadeh et al. 2012; Menéndez-Delmestre et al. 2013). It is well known that these high-z galaxies may be highly obscured and that a combination of intense dust-obscured star formation and dust-enshrouded AGN activity would be responsible for the high infrared luminosities of these objects (Blain et al. 2002). HST-NICMOS and ACS observations reflect structured dust obscuration in these objects (Swinbank et al. 2010). In particular, Takata et al. (2006) find that the internal extinction in these objects are similar to the extinction in local ULIRGs, and measured a median extinction of AV = 2.9 ± 0.5 mag in a sample of SMGs at z ~ 1.0–3.5 using the Hα/Hβ flux ratio.

Clumpy and dusty star-forming structures have also been identified at high redshifts in more common star forming galaxies (i.e. the so called main sequence star forming galaxies). These galaxies have (sub)kpc star-forming clumps mostly spread over galactocentric distances of few to several kpc (Förster Schreiber et al. 2011a,b; Genzel et al. 2011), and with internal nebular extinctions of 2 to 4 mag, assuming AV,Hα = AV,stellar/0.44 (Förster Schreiber et al. 2009; Wuyts et al. 2011).

In the previous subsection we showed that the distance/linear scale may play an important role in deriving the internal extinction properties of LIRGs and ULIRGs, by comparing both populations of galaxies locally, albeit with a difference of a factor ×5 in distance between both subsamples. According to the simulations, the median extinction measured in a given galaxy decreases when placed at increasing distances. This effect is due to the fixed angular resolution of the IFS data that translates into a larger physical scale per spatial resolution element (spaxel) as the distance to the galaxy increases, and is particularly important when the physical scales sampled by the spaxel are much larger than that of the intrinsic clumpy structure of the dust distribution and star-forming regions.

It is clear that if the intrinsic star-forming structure of high-z galaxies is in general similar to that of our local LIRGs (i.e. mostly disks) and ULIRGs (i.e. mostly interacting), the smearing effect mentioned above would have a direct impact in the derivation of their 2D internal extinction values, and of all relevant subsequent extinction corrected properties such as star formation surface densities, KS-laws, and overall star formation rates. This would certainly be the case not only with seeing-limited near-IR IFS, heavily undersampling the galaxies at redshifts of 1 to 3, with each spaxel corresponding to about 1.5–2 kpc, but even when using AO assisted IFS where the spaxel (50 to 100 mas) translates to about 0.4 to 0.8 kpc. Thus, if the results given by Eq. (2) (see also Fig. 6) are applied to high-z galaxies, the extinction values derived directly from the observed optical emission line ratios would require to be increased by a factor 1.4 on average. This correction would correspond to an additional increase in the Hα flux by a factor ~3 and hence, in the extinction-corrected SFR and ΣSFR.

Finally, it is worth noticing that this distance effect could also have a minor wavelength dependency. The method presented here to derive the AV values is based on specific emission line ratios in the near-infrared. Since these lines originate in regions of higher optical depths than the optical emission lines, and the stellar continuum measured at optical wavelengths, differences in the amount and/or distance dependency could appear. It would be worth exploring these effects with a suitable set of data, in particular if, as in many high-z studies, the AV corrections applied to the observed Hα flux is obtained indirectly using the standard Calzetti recipe AHα = 7.4 E(B − V), where E(B − V)gas = 0.44 E(B − V)stars (Calzetti et al. 2000, 2001).

6. Summary

In this paper, we presented a detailed 2D study of the extinction structure of a representative local sample of 10 LIRGs and 7 ULIRGs, based on VLT-SINFONI IFS K-band observations. We sample the central ~3 × 3 kpc for LIRGs and the ~12 × 12 kpc for ULIRGs, with average linear resolutions (FWHM) of ~0.2 kpc and ~0.9 kpc, respectively. The extinction maps are based on measurements of the Brγ/Brδ line ratio for LIRGs and of the Paα/Brγ line ratio for ULIRGs.

In agreement with previous studies, we found that the distribution of the dust in these galaxies presents a patchy structure on sub-kiloparsec scales, with regions almost transparent with AV ~ 0 to heavily obscured areas with AV values up to ~ 20 − 30 mag. In most of the objects in the sample (~70%), the nucleus of the galaxy coincides with the peak in the extinction maps, with values that range from AV ~ 3 mag up to AV ~ 17 mag.

We obtained the AV distribution of the individual galaxies on a spaxel-by-spaxel basis (see Fig. 4). The individual AV distributions show a wide range of values with most of them spread between AV ~ 1 and AV ~ 20 mag, with no clear evidence of any dependence with LIR. However, as a class (see Fig. 5), ULIRGs show AV values (median of 6.5 mag, mode of ~7–8 mag) higher than those for LIRGs (median of 5.3 mag, mode of ~3–4 mag). The AV distribution in LIRGs shows a mild decrease as a function of galactocentric distances of up to 1 kpc and a flattening at larger distances (2–3 kpc). No similar behaviour is detected in ULIRGs, most likely owing to the lower linear resolution of the observations.

To study the effect of the spatial sampling (i.e. physical scale per spaxel) in the derived extinction values at increasing galaxy distances, the individual Brγ and Brδ maps of our subsample of local LIRGs (at an average distance of 63 Mpc) have been artificially smeared. These simulations have shown that the spatial resolution plays an important role in shaping the AV distributions. The median value of the visual extinction measured on the LIRG subsample decreases as a function of the linear resolution/distance by a factor ~0.8 at the average distance (328 Mpc, 0.2 kpc/spaxel) of our ULIRG sample, and up to ~0.67 for distances above 800 Mpc (0.4 kpc/spaxel). This distance effect would have implications in the derivation of the intrinsic extinction, and subsequent properties, such as SFR, ΣSFR, and the KS-law, in high-z star-forming galaxies, even in AO-based spectroscopy. If local LIRGs are analogues of the main-sequence star-forming galaxies at cosmological distances, the extinction values (AV) derived from the observed emission lines in these high-z sources would need to be increased by a factor 1.4 on average.

Acknowledgments

We thank the anonymous referee for useful comments and suggestions that improved the final content of this paper. This work was supported by the Spanish Ministry of Science and Innovation (MICINN) under grants BES-2008-007516, ESP2007-65475-C02-01, and AYA2010-21161-C02-01. This paper made use of the plotting package jmaplot, developed by Jesús Maíz-Apellániz http://jmaiz.iaa.es/software/jmaplot/current/html/jmaplot_overview.html. Based on observations collected at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile, programmes 077.B-0151A, 078.B-0066A, and 081.B-0042A. This research made use of the NASA/IPAC Extragalactic Database (NED), which is operated by the Jet Propulsion Laboratory, California Institute of Technology, under contract with the National Aeronautics and Space Administration.

References

Online material

thumbnail Fig. 1a

NGC 2369. Top panels show the SINFONI K band continuum emission, the observed maps (not corrected from extinction) of the lines Brγ at 2.166 μm and Brδ at 1.945 μm, together with the AV map. The white contour englobes those spaxels above S/N = 4 considered to build the AV map and distribution. The brightest spaxel of the K band continuum is marked with a plus symbol (+), and has been used as reference to obtain the radial profile in the bottom right panel. The secondary nucleus, if present, is marked with a cross (×). Bottom left panel shows the AV distribution of all valid spaxels (blue histogram) and the distributions of the those spaxels above (red) and below (yellow) the median value of the Brδ S/N distribution. The relationship between the surface brightness of the lines and the AV is shown in the central panel, where the points with AV < 0 are outlined with a black contour. Finally, the radial distribution of the extinction is shown in the bottom right panel, where the blue line represents the mean value of AV for different radial bins and its error. The bins are obtained as the 1/30 of the total radial coverage of the map.

Open with DEXTER

thumbnail Fig. 1b

NGC 3110. Same as Fig. 1a but for NGC 3110.

Open with DEXTER

thumbnail Fig. 1c

NGC 3256. Same as Fig. 1a but for NGC 3256. Please note that the central spaxel lays outside the FoV since the nucleus was not observed in K-band. See text and Paper I for further details.

Open with DEXTER

thumbnail Fig. 1f

NGC 5135. Same as Fig. 1a but for NGC 5135.

Open with DEXTER

thumbnail Fig. 1h

IC 4687. Same as Fig. 1a but for IC 4687.

Open with DEXTER

thumbnail Fig. 1i

NGC 7130. Same as Fig. 1a but for NGC 7130.

Open with DEXTER

thumbnail Fig. 1j

IC 5179. Same as Fig. 1a but for IC 5179.

Open with DEXTER

thumbnail Fig. 2a

IRAS 06206-6315. Top panels show the SINFONI K band continuum emission, the observed maps (not corrected from extinction) of the lines Paα at 1.876 μm and Brγ at 2.166 μm, together with the AV map. The white contour englobes those spaxels above S/N = 4 considered to build the AV map and distribution. The brightest spaxel of the K band continuum is marked with a plus symbol (+), and has been used as reference to obtain the radial profile in the bottom right panel. The secondary nucleus, if present, is marked with a cross (×). Bottom left panel shows the AV distribution of all valid spaxels (blue histogram) and the distributions of the those spaxels above (red) and below (yellow) the median value of the Brγ S/N distribution. The relationship between the surface brightness of the lines and the AV is shown in the central panel, where the points with AV < 0 are outlined with a black contour. Finally, the radial distribution of the extinction is shown in the bottom right panel, where the blue line represents the mean value of AV for different radial bins and its error. The bins are obtained as the 1/30 of the total radial coverage of the map. For those objects with multiple components, the top inset shows the AV radial profile of the system taking the main nucleus as the centre, whereas the bottom subpanel shows the radial profile obtained by extracting the profiles of each component separately and plotting them in the same reference frame.

Open with DEXTER

All Tables

Table 1

SINFONI sample.

Table 2

Integrated properties and statistics of the AV distributions.

All Figures

thumbnail Fig. 3

Surface brightness distributions of the individual spaxels of LIRGs and ULIRGs. For clarity, we only plot a random distribution of the 20% and 50% of the points for LIRGs and ULIRGs, respectively. The colour code indicates the AV value of each spaxel, whereas the points with AV < 0 are outlined with a black contour line. The increase in the extinction towards low surface brightness values is mainly due to the high uncertainties of the flux measurements, in particular in ULIRGs, where the highest AV values correspond to the spaxels with the lower Brγ surface brightness.

Open with DEXTER
In the text
thumbnail Fig. 4

Individual AV distributions, on a spaxel-by-spaxel basis, of the galaxies of the sample ordered by increasing LIR. The extremes of the distributions are the 5th and 95th percentiles (P5 and P95). The boxes illustrate the interquartile range, whereas the horizontal blue and red lines correspond to the median and the weighted mean of the distribution, respectively. The measurements of the nuclear extinction are plotted as black diamonds. The total AV distributions for LIRGs and ULIRGs are shown on the right-hand side of the plot.

Open with DEXTER
In the text
thumbnail Fig. 5

AV distributions and radial profiles of the LIRGs (top) and ULIRGs (bottom) subsamples on a spaxel-by-spaxel basis. The median AV values of each distribution are shown in the left panels. In the central and right panels, the radial profiles are plotted in terms of the radius in kpc (centre) and in units of the Hα effective radius (Reff, right panel) extracted from Arribas et al. (2012). The red line represents the weighted mean of AV and its error for different radial bins in steps of 1/30 of the total radial coverage. For the ULIRG subsample, we plotted the radial profile of the sample by considering each component of the systems separately (top) and the profile extracted by taking the brightest nucleus in the K-band continuum as the centre of the systems (bottom).

Open with DEXTER
In the text
thumbnail Fig. 6

Evolution of the computed median AV of the individual distributions of the LIRG subsample as a function of scale/distance. The AV values are normalised to the rest-framed distribution. The value that corresponds to the AV distribution of the whole LIRG subsample is plotted as black circles and a power-law fit to the data as a dashed black line. The mean distance of the ULIRG subsample (328 Mpc) is plotted as a vertical dotted line for reference.

Open with DEXTER
In the text
thumbnail Fig. 7

Different simulations of the LIRGs AV distribution for increasing distance (from top to bottom and left to right). The median of each distribution is highlighted with a vertical dashed line. The distance and the median AV of the simulated distribution is annotated in each panel. The top left panel corresponds to the average distance of the ULIRG subsample, and shows the observed AV distribution of the ULIRGs and its median, in yellow.

Open with DEXTER
In the text
thumbnail Fig. 1a

NGC 2369. Top panels show the SINFONI K band continuum emission, the observed maps (not corrected from extinction) of the lines Brγ at 2.166 μm and Brδ at 1.945 μm, together with the AV map. The white contour englobes those spaxels above S/N = 4 considered to build the AV map and distribution. The brightest spaxel of the K band continuum is marked with a plus symbol (+), and has been used as reference to obtain the radial profile in the bottom right panel. The secondary nucleus, if present, is marked with a cross (×). Bottom left panel shows the AV distribution of all valid spaxels (blue histogram) and the distributions of the those spaxels above (red) and below (yellow) the median value of the Brδ S/N distribution. The relationship between the surface brightness of the lines and the AV is shown in the central panel, where the points with AV < 0 are outlined with a black contour. Finally, the radial distribution of the extinction is shown in the bottom right panel, where the blue line represents the mean value of AV for different radial bins and its error. The bins are obtained as the 1/30 of the total radial coverage of the map.

Open with DEXTER
In the text
thumbnail Fig. 1c

NGC 3256. Same as Fig. 1a but for NGC 3256. Please note that the central spaxel lays outside the FoV since the nucleus was not observed in K-band. See text and Paper I for further details.

Open with DEXTER
In the text
thumbnail Fig. 2a

IRAS 06206-6315. Top panels show the SINFONI K band continuum emission, the observed maps (not corrected from extinction) of the lines Paα at 1.876 μm and Brγ at 2.166 μm, together with the AV map. The white contour englobes those spaxels above S/N = 4 considered to build the AV map and distribution. The brightest spaxel of the K band continuum is marked with a plus symbol (+), and has been used as reference to obtain the radial profile in the bottom right panel. The secondary nucleus, if present, is marked with a cross (×). Bottom left panel shows the AV distribution of all valid spaxels (blue histogram) and the distributions of the those spaxels above (red) and below (yellow) the median value of the Brγ S/N distribution. The relationship between the surface brightness of the lines and the AV is shown in the central panel, where the points with AV < 0 are outlined with a black contour. Finally, the radial distribution of the extinction is shown in the bottom right panel, where the blue line represents the mean value of AV for different radial bins and its error. The bins are obtained as the 1/30 of the total radial coverage of the map. For those objects with multiple components, the top inset shows the AV radial profile of the system taking the main nucleus as the centre, whereas the bottom subpanel shows the radial profile obtained by extracting the profiles of each component separately and plotting them in the same reference frame.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.