MAPPING THE CLUMPY STRUCTURES WITHIN SUBMILLIMETER GALAXIES USING LASER-GUIDE STAR ADAPTIVE OPTICS SPECTROSCOPY

We construct Hα intensity, velocity (v_Hα), and line-width (σ_Hα) maps for each target to help us understand the distribution of Hα properties across the SMGs in our sample (see Figures 1–6). Following a similar approach to Law et al. (2009) and Gonçalves et al. (2010), we convolve the individual image frames at each wavelength channel with a 2D Gaussian kernel (FWHM =1.5 pixels) in an effort to improve the S/N while remaining below the spatial resolution limit of our observations as determined by our point-spread function (PSF; FWHM ∼010–025), measured from the TT-star peak-up imaging. For SMM J123549.44 and SMM J163650.43, we also smooth the one-dimensional (1D) spectra at each spatial pixel with a 2 Å and 5 Å wide Gaussian kernel, respectively, to better distinguish the line emission from the underlying noise.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Imposing a set of line-search thresholds for the Hα line width (≲ 2500 km s⁻¹), S/N with respect to an off-source region (≳ 6), and velocity offset relative to the long-slit Hα redshift (≲ 1000 km s⁻¹) for our targets from Swinbank et al. (2004), we fit single Gaussian profiles to the Hα line emission along the dispersion direction at each spatial pixel. We use the collection of these Gaussian fits, in particular the central wavelength and the line width (corrected for instrumental width) at each spatial pixel, to construct an Hα velocity map, reflecting the relative spectral position of the line centroids, and an Hα line-width map characterizing the velocity dispersions of the Hα emission. Note that only the Gaussian fits complying with our imposed thresholds are used for the velocity and line-width maps, while other ones are disregarded (i.e., empty pixel in the resulting maps). The selection of these S/N, velocity, and line-width thresholds were not established a priori, but tuned carefully to each object, in order to eliminate spurious elements from the resulting maps. We produce Hα intensity maps by collapsing the final science data cube along the dispersion direction, stacking the science data cube's flux over a range Δv ≃ 1000–2000 km s⁻¹ centered at the redshifted Hα line. In Figures 2, 4, and 6, we show Hα velocity and line-width maps with overlaid Hα intensity contours for SMM J030227.73, SMM J123549.44, and SMM J163650.43, respectively. We oversample these maps by a factor of two to improve their visual appearance. We also show the 1D spectra integrated over several kiloparsec-sized Hα-bright regions, which are labeled in the maps shown in Figures 1, 3, and 5.

We flux calibrate the Hα line emission in our OSIRIS observations using as a reference the Hα line fluxes from long-slit spectroscopy by Swinbank et al. (2004), where slit loss correction has been performed based on K-band images. We set the galaxy-integrated long-slit Hα fluxes equal to the integrated OSIRIS flux resulting from collapse of the science cube along the dispersion direction (Δv ≃ 1000–2000 km s⁻¹) centered at the redshifted Hα line. By adopting this approach, we ensure that we are not artificially boosting the Hα emission within the brightest clump regions. In principle, peak-up imaging of the TT-star for each science target is also a viable reference to perform absolute flux calibrations. However, we note that when using the AO technique, a large flux uncertainty arises from the imperfect channeling of light into the AO-corrected PSF. Following detailed modeling of this issue by Law et al. (2007b), the total flux uncertainty for the reference TT-star flux corresponds to at least ∼30%. Furthermore, because flux calibration requires also taking into account the spatial offset between the science target and the TT-star to consider the true PSF at the science target location, the uncertainties in absolute flux calibration based on TT-star observations increase significantly.

In order to disentangle the diffuse Hα component (∼05–2'') from the compact one (FWHM ∼02–04), we rely on a target-specific PSF determined from peak-up imaging of the TT-star to subtract the compact region of high-S/N line emission from each map. This is particularly important considering the potential spread of emission beyond the AO-corrected PSF that may escape the immediate regions of these high-S/N regions of emission. In this way, we avoid potential contamination and ensure the best separation of the diffuse and compact Hα emission components. The PSF-subtracted images that we use to investigate diffuse emission are shown in Figures 1, 3, and 5 for SMM J030227.73, SMM J123549.44, and SMM J163650.43, respectively.

3.1.1. SMM J030227.73

3.1.2. SMM J123549.44

3.1.3. SMM J163650.43

Based on Kennicutt (1998), we derive a local SFR at each spatial pixel in our flux-calibrated Hα images for each SMG target. Taking into account the pixel scale of our observations (see Table 1), we construct the SFR surface density (Σ_SFR) maps shown in Figures 1, 3, and 5. The regions of broad-Hα emission, which we assume to be associated with AGN activity, have been subtracted following the approach described in Section 2.1 to cleanly eliminate the AGN contribution. The remaining narrow-line Hα emission shown in these figures, likely associated with star formation, extends out to scales >05–2'' (∼4–16 kpc). However, the bulk of this emission appears to be concentrated in multiple Hα unresolved clumps, limited in size by the PSF FWHM ∼010–025 (1–2 kpc at z ∼ 2).

We note that although the near-IR provides us with a less-obscured insight to these SMG clumps than the optical, the presence of significant obscuration remains an important caveat (Takata et al. 2006). In the absence of spatially resolved extinction maps, we make an estimated extinction correction based on the typical Balmer decrement found for SMGs: Takata et al. (2006) have shown that SMGs typically have observed Balmer decrements Hα/Hβ ∼ 5–20, corresponding to extinction levels in the range A_v ∼ 1–4 with a median 〈A_v〉 = 2.9 ± 0.5. Assuming a typical Balmer decrement ∼10, we apply the corresponding reddening correction of $e^{\tau _{{\rm Balmer}}} \sim 4$ to the measured Hα luminosities. Note that this factor is significantly larger than the typical attenuation found for other high-redshift populations, such as the optically selected LBGs, with attenuations of an average factor of 1.7 (Erb et al. 2006).

Prior to extinction correction, SMG clumps already have large SFR surface densities with peak values Σ_SFR ∼ 1–50 (see Figures 1, 3, and 5). After correcting for extinction in the manner we described, the peak SFR surface densities increase up to Σ_SFR ∼ 200 M_☉ yr⁻¹ kpc⁻² and the clump-wide SFRs reach values of up to 250 M_☉ yr⁻¹. In Figure 7, we show the total SFRs and SFR surface densities as a function of the line velocity dispersions based on the clump-wide spectral extractions shown in Figures 2, 4, and 6, with the extinction-corrected values shown by the endpoints of horizontal arrows. For comparison, we include both kiloparsec-scale and galaxy-wide measurements for optically and near-IR selected galaxies at z ∼ 2–3 (Lehnert et al. 2009; Förster Schreiber et al. 2009; Law et al. 2009) and for samples of low-redshift normal and starburst galaxies, compiled by Dib et al. (2006). We also compare the distribution in SFR surface densities in our SMG regions with that in low-redshift samples of normal galaxies and circumnuclear starbursts from Kennicutt (1998) and luminous infrared galaxies (LIRGs; L_{8–1000 μm} ∼ 10¹¹–10¹² L_☉) from Alonso-Herrero et al. (2006).

Zoom In Zoom Out Reset image size

The SMG clump-integrated SFRs extend over a large range of values (SFR_clump ∼ 1–60 M_☉ yr⁻¹) and are similar to the galaxy-wide values that have been attributed to LBGs (Law et al. 2009) and other high-redshift galaxies from the SINS sample (e.g., Förster Schreiber et al. 2009), particularly when extinction correction is taken into account (see left panel in Figure 7). It has been shown that a trend exists between Σ_SFR and Hα velocity dispersion in these high-redshift galaxy populations, where the most intense star-forming regions are also those with the highest velocity dispersions (Lehnert et al. 2009; Law et al. 2009; Förster Schreiber et al. 2009). On the right panel of Figure 7, we show that the SMG clumps in our sample occupy the high end of this trend. Although the SMG clumps in this sample present a wide range of extinction-corrected SFR and Σ_SFR values (SFR ∼3–250 M_☉ yr⁻¹, Σ_SFR ∼ 1–50 M_☉ yr⁻¹ kpc⁻²), the average values that characterize the star formation activity in these (〈SFR〉 ∼ 60 M_☉ yr⁻¹, 〈Σ_SFR〉 ∼ 10 M_☉ yr⁻¹ kpc⁻²) is similar in intensity—though higher by a factor of 2–5 in some cases—to that found in other high-redshift populations (e.g., 〈SFR〉_LBGs ∼ 25 M_☉ yr⁻¹, 〈Σ_SFR〉_LBGs ∼ 4 M_☉ yr⁻¹ kpc⁻²; Law et al. 2009). However, the Hα velocity dispersions are significantly different, likely associated to the different mass ranges that these different galaxy populations correspond to. Assuming that σ_Hα traces gas dynamics within virialized clouds and considering a typical clump size of 1 kpc, we determine clump dynamical masses in the range of M_clumps = σ²R/G ∼ 1–30 × 10⁹ M_☉. These clump masses are ∼1–2 orders of magnitude larger than the kiloparsec-scaled stellar clumps uncovered by Förster Schreiber et al. (2011) in the optically selected sample of z ∼ 2 SINS star-forming galaxies (M_clumps ∼ 0.1–8 × 10⁹ M_☉).

Lehnert et al. (2009) argue that the relationship between SFR surface density and the ionized gas velocity dispersion is a consequence of star formation self-regulation, where the mechanical energy of the starburst itself controls the local dynamics of the ionized gas. Following Lehnert et al. (2009), this relationship can be represented by a functional form involving the intensity of star formation with the velocities of the ionized gas to represent the energy injected into the interstellar medium surrounding the star-forming regions (see curve in Figure 7). Within this context, the high velocity dispersions observed in our SMG clumps are likely not the sole result of larger dynamical masses. They are likely a combination of differences in dynamical mass ranges and a reflection of the high pressures sustained by high local surface brightnesses, which in turn are likely due to the higher gas densities in SMGs (e.g., Harris et al. 2010).

Compared to galaxies at low redshift, the right panel of Figure 7 shows that SMG clumps have Σ_SFR values >3 orders of magnitude higher than those found in normal spirals (Kennicutt 1998; Dib et al. 2006), but which fall well within the range of starburst galaxies (Kennicutt 1998). If we correct for extinction the Σ_SFR in SMG clumps are shifted at the most within a factor of a few from the range occupied by the typical local circumnuclear starbursts, similar to the range found in low-redshift LIRGs (Alonso-Herrero et al. 2006).

In all three SMGs of our sample, we identify regions with distinct properties: regions with narrow-Hα emission where star formation is likely taking place and regions with broad-Hα emission, likely dominated by AGN activity (keeping in mind that outflows may also play a role; e.g., Shapiro et al. 2008, 2009; Newman et al. 2012). We find that the narrow-Hα emission is concentrated in multiple clumps unresolved on ∼1–2 kpc scales, asymmetrically distributed around the AGN. These clumps suggest either one of the following scenarios: (1) these Hα-bright regions correspond to distinct components in a disturbed system, as would be found in a merging system; or, (2) we are picking out the high surface-brightness peaks of Hα emission in an underlying lower surface-brightness plane of emission—perhaps even a disk, as has been found in a sub-sample of massive optically selected star-forming galaxies at z ∼ 2–3 (Förster Schreiber et al. 2006, 2011; Law et al. 2007b, 2009).

We find velocity offsets of ∼few × 100 km s⁻¹ between distinct galactic-scale regions (see Figures 2, 4, and 6). These velocity offsets could be explained by invoking a merger scenario (see also Engel et al. 2010), thus strengthening the conclusion derived from deep rest-frame optical HST-imaging that SMGs are disturbed systems, likely corresponding to mergers (Smail et al. 1998, 2004; Ivison et al. 2010; Swinbank et al. 2010). Within this context, the distinct components revealed by OSIRIS could potentially be associated with the remnants of the pre-merger galaxies that have not fully coalesced and still retain significant intrinsic velocities with respect to each other. This is the kinematic structure expected for the intermediate stages of a merging system (e.g., Cox et al. 2006; Hayward et al. 2011). We note that this is in agreement with Greve et al. (2005), where they discuss the likelihood that the double-peaked CO profiles observed in a sample of 18 SMGs correspond to either a disk or a merger. Taking into account estimates of gas masses and source sizes, they conclude that such high surface mass density would imply dynamical instabilities too powerful for a disk to survive. They thus conclude that the double-peaked nature of the CO lines traces distinct gas-rich components undergoing a merger or arise from a disk collapsing under gravitational instability.

On the other hand, the Hα-bright clumps of emission that OSIRIS picks out may correspond to distinct clumps within an underlying clumpy disk structure. This would be similar to the clumpy disks identified in face-on galaxies at somewhat lower redshifts (z ∼ 0.5–2.0; Elmegreen et al. 2004a, 2004b) and the more recently identified stellar clumps in optically and near-IR selected star-forming galaxies at z ∼ 2 by Shapiro et al. (2009), Förster Schreiber et al. (2011), Genzel et al. (2011), and Newman et al. (2012). Recent studies have indeed suggested the presence of disk structures in a large fraction of SMGs based on deep broadband imaging K-band data (e.g., Targett et al. 2011). Although these studies are unable to directly trace the kinematic structure of these objects, they may point to the possibility that SMGs may correspond to a diverse population presenting a varied range of kinematic structures.

Considering that the broad-Hα emission in our SMGs identifies a supermassive black hole likely at the dynamical center of system, and the large masses and extreme star-forming properties of the SMG clumps that are in turn asymmetrically distributed around the AGN, it is unlikely that these all reside in a regular potential well structure. The merger interpretation is likely the most accurate scenario for the SMGs in our sample, although the final test of whether an underlying disk structure is present will come from studies of the cold gas at the high spatial resolutions with ALMA.

We find compact unresolved regions (FWHM ∼1–3 kpc) of broad-line Hα emission that contribute a significant fraction (∼30%–90%) to the galaxy-wide Hα emission. However, fainter narrow-line Hα emission extends over large spatial scales of 4–16 kpc (see Figures 1, 3, and 5), with 1–2 kpc sized Hα-bright clumps indicating regions of particularly intense star formation (see line-width maps in Figures 2, 4, and 6). Spatially extended sizes for the diffuse Hα emission have also been identified in the seeing-limited IFU observations presented by Swinbank et al. (2006) and Alaghband-Zadeh et al. (2012), who report resolved nebular emission on scales from 4 to 11 kpc for a combined sample of 12 SMGs.

Evidence of spatially extended emission in SMGs has also been found at longer wavelengths. In high-resolution observations of a range of CO transitions (Engel et al. 2010; Ivison et al. 2011) and radio continuum (Biggs & Ivison 2008), linear sizes in the range of ∼1–16 kpc have been identified based on median angular-averaged FWHM sizes, with sample median sizes of ∼5–7 kpc. High-resolution radio continuum observations by Chapman et al. (2004) also unveil spatial extensions with a linear diameter out to ≳ 10 kpc, while a high-resolution far-IR study by Younger et al. (2010) reveal far-IR emission extending out to spatial scales in the range of ∼5–8 kpc.

The similarity in IR luminosities between local ultraluminous infrared galaxies (ULIRGs; L_{8–1000 μm} > 10¹² L_☉) and high-redshift SMGs, as well as the presence of AGN signatures in both populations have naturally motivated direct comparisons and discussions of their possible correspondence within the formation and evolution scenario of today's most massive spheroids. To a somewhat still controversial degree, the sizes of SMGs also play a role in this debate. High-resolution studies of the CO and nebular emission in SMGs have concluded that SMGs have compact sizes (Tacconi et al. 2006, 2008), although often with companions with an average separation of ∼8 ± 2 (Alaghband-Zadeh et al. 2012), while other studies have demonstrated the presence of more extended spatial extensions (Chapman et al. 2004; Tecza et al. 2004; Nesvadba et al. 2007; Biggs & Ivison 2008; Hailey-Dunsheath et al. 2010). The former argues for a similarity with local ULIRGs, which are found to be compact with sizes ∼1 kpc (in the mid-IR, Charmandaris et al. 2002; Díaz-Santos et al. 2010b; and in CO, Downes & Solomon 1998; Bryant & Scoville 1999); the latter argues for a sharp difference with local ULIRGs. However, an important role in this controversy has been played by the tenuous definition of what is compact with respect to what is extended, particularly in the case of a clumpy system, as we find in SMGs.

In light of our OSIRIS observations, together with these previous findings, it is clear that the SMG population displays a range in spatial extensions that are often significantly larger than those of local ULIRGs. On the other hand, SMGs share similar star formation intensities with LIRGs (see Figure 7), which are still within the range of extremely active environments, yet within a somewhat lower IR luminosity bin than ULIRGs. LIRGs have also been found to display rest-frame optical line emission over large spatial scales (∼3–7 kpc; Alonso-Herrero et al. 2006) as well as mid-IR extra-nuclear emission (Díaz-Santos et al. 2010a, 2010b), both indicative of extended star formation. We emphasize that evidence for extended star formation at high redshifts is not limited to SMGs. Some of the brightest and most massive LBGs do appear to have comparable Hα spatial extensions (R_Hα ∼ 2–7 kpc; Förster Schreiber et al. 2006).

Our results indicate that the SMG clumps in our sample have high surface densities of star formation activity, close to those found in local extreme environments, such as in circumnuclear starbursts and LIRGs. However, considering the much greater spatial extents found for these SMGs (∼8–16 kpc) in comparison to the 1 kpc sized nuclear starbursts (Kennicutt 1998), SMGs appear to be undergoing this intense activity on much larger spatial scales.