A kpc-scale resolved study of unobscured and obscured star-formation activity in normal galaxies at z = 1.5 and 2.2 from ALMA and HiZELS

Nine SHiZELS galaxies were observed with ALMA during Cycle 2, 3, 4 (projects 2012.1.00402.S, 2013.1.01188.S and 2015.1.00026.S; P.I.: E. Ibar) in Bands 6 or 7, depending on whether the galaxies were at $z=2.23$ or 1.47, respectively, corresponding to rest-frame $\sim$ 355$\mu$m. The observations were designed to detect continuum emission using four spectral windows (SPWs), each covering an effective bandwidth of 1.875 GHz at a spectral resolution of 15.6 MHz. Observations were taken under relatively good weather conditions with precipitable water vapour (P.W.V.) ranging from 0.2 mm to 3.9 mm, and using 29 to 45 antennas (the earlier the cycle, the smaller the number) with the longest baselines spanning 1.0–1.5 km. The phase, bandpass and flux calibrators for all observations used in this work are listed in Table 1.

Data reduction was carried out using the Common Astronomy Software Applications (casa) and using the ALMA pipeline up to the production of calibrated $uv$ data products. Three imaging approaches were performed using the task tclean, exploring a Briggs weighting (robust=0.5), natural weighting, and $uv$ tapering such that we created a synthesized beam of $\sim 1^{\prime\prime}$ (see Figs 8, 9). In all cases, the signal is interactively cleaned down to 2–3 $\sigma$ (r.m.s. $\sim$ 25$\mu$Jy beam^-1) in regions with significant emission at the source position. The astrometric accuracy of the ALMA Band-7 image is around 1.5 mas. Information about each target is presented in Table 2.

Sources SHiZELS-7, SHiZELS-9, SHiZELS-11 and SHiZELS-14 were detected by ALMA at a significance higher than 5$\sigma$ in at least one of the three imaging approaches, while SHiZELS-21, SHiZELS-8, SHiZELS-10, SHiZELS-2 and SHiZELS-3 remained undetected, regardless of the imaging approach.

Within the field of view (FoV) of SHiZELS-7 and SHiZELS-10, we identify two serendipitous detections. We denote them as SHiZELS7-ID2 and SHiZELS10-ID2 and their properties are listed in Appendix D.

One of the main challenges when analysing multi-wavelength high-resolution spatially-resolved observations is the global astrometric accuracy. In particular, IFU observations with a small FoV (as obtained with SINFONI, which has a FoV of about $3^{\prime\prime}\times 3^{\prime\prime}$) present a global astrometric uncertainty which is larger than the resolution elements of the observations, therefore they are difficult to anchor to other observations at high-resolution. For this reason, we perform the following astrometric corrections to analyse our data.

Firstly, we align the astrometry of the HiZELS narrow-band images by the public Gaia DR2 catalogue. The offset between the HiZELS catalogue generated by SExtractor and the Gaia catalogue is corrected to $\Delta{\rm R.A.}=0\aas@@fstack{\prime\prime}004\pm 0\aas@@fstack{\prime\prime}13^{\prime\prime}$ and $\Delta{\rm Dec.}=-0\aas@@fstack{\prime\prime}010\pm 0\aas@@fstack{\prime\prime}10$. We degrade the resolution of the H$\alpha$ moment-0 SINFONI image down to $0\aas@@fstack{\prime\prime}7$ using a Gaussian kernel to get a similar resolution to the HiZELS narrow-band images. Then we fit a Gaussian profile to this low-resolution image in order to obtain the position of the peak, identify the offset with respect to the narrow-band image, and use this information to correct the astrometry of the IFU image.

As mentioned before, we have also used the available HST F606W and F140W observations ¹¹1HST proposal ID: 14719, P.I.: P. Best. The HST FoV is about $2^{\prime}\times 2^{\prime}$, which is too small to apply a reliable astrometric correction using Gaia DR2. Therefore, we correct the global astrometry of the HST images using the HSC DR2 catalogue available in our fields (Aihara et al., 2019). The 5$\sigma$ limiting magnitude within $2^{\prime\prime}$ diameter apertures of the HSC $i$-band catalogue is 26.7 AB mag, which is similar to the HST images, and deep enough to have about 200 high-S/N sources for the alignment. The astrometric accuracy of HSC DR2 catalogue derives from the Gaia DR2 catalogue (Aihara et al., 2019). We match the HST and HSC catalogues for the offsets of R.A., Dec., and correct the astrometry of the HST image. After our astrometric correction, the catalogue match between HST and HSC catalogues is consistent with an offset of zero, with systematic errors of $\sim 0\aas@@fstack{\prime\prime}04$.

In Fig. 1, we show the true-colour image composite from HST F606W (in blue) and F140W (in red) images, including ALMA continuum contours in levels of $2\sigma$, $3\sigma$, $4\sigma$, $5\sigma$ for the four ALMA-detected galaxies. The galaxy centres revealed by HST more or less coincide with the ALMA imaging. The observations reveal a complex state for the interstellar medium in SHiZELS galaxies at $z\simeq$ 1.47 and 2.23. We show multi-wavelength postage stamps of the HST, H$\alpha$ (Gillman et al., 2019) and our ALMA images for each of the galaxies presented in this work in Figs 8 and 9.

SEDs of the SHiZELS sample were previously fitted using MAGPHYS (da Cunha et al., 2008) in Gillman et al. (2019). The ALMA detected fluxes of the ALMA-SHiZELS galaxies can help to constrain the FIR properties. Therefore, we re-fit the SEDs of the four ALMA detected galaxies including the FIR flux we obtained in this work, and list the main results in Table 2 and 3. For the ALMA non-detected galaxies, we list the dust mass given by MAGPHYS from fitting the optical-to-NIR SEDs. The stellar mass and dust mass show reasonable consistencies from the different approach.

We show the dust-to-stellar mass ratio ($\log_{10}(M_{\rm dust}/M_{*})$) in Fig. 2 and compare our results with the local galaxies from Dustpedia (Clark et al., 2018; Casasola et al., 2020). The only ULIRG in our sample, SHiZELS-14, has a larger $\log_{10}(M_{\rm dust}/M_{*})=-2.1\pm 0.2$ than the local galaxies, while this value is consistent with the typical value of $\log_{10}(M_{\rm dust}/M_{*})$ in SMGs at redshift about 2 (Calura et al., 2017). The rest targets of our sample all have consistent or 1$\sigma$ higher $\log_{10}(M_{\rm dust}/M_{*})$) values as the local galaxies, even for the targets with no ALMA detection.

To estimate the FIR luminosity (and SFR_FIR) using the ALMA observations (rest-frame continuum at $\sim$355 $\mu$m), we assume a FIR SED template based on previous stacking analyses for HiZELS galaxies at $z\sim 1.47$ (Ibar et al., 2013) and 2.23 (Thomson et al., 2017). Thanks to the rich multi-wavelength coverage, the stellar masses of the HiZELS galaxies can be estimated reasonably well (see Sobral et al., 2014; Gillman et al., 2019). For the purpose of this work, for the $z\sim 1.47$ targets, we adopt the stacked SEDs derived by Ibar et al. (2013) in the stellar mass bins of $9.9<\log(M_{*}/M_{\odot})<10.3$ and $10.3<\log(M_{*}/M_{\odot})<11.8$. For galaxies at $z\sim 2.23$, we consider the FIR template presented in Thomson et al. (2017).

We fit each FIR template SED following the modified-blackbody fitting method described by Beelen et al. (2006), assuming a fixed power-law index for the dust emissivity, $\beta=1.8$. The fitting results show a dust temperature of about 25 $\pm$ 1 K for $z\simeq 1.47$ targets and 32 $\pm$ 2 K for the $z\simeq 2.23$ targets, similar to previous measurements of luminous IR galaxies at high redshift (Hwang et al., 2010; Oteo et al., 2017; Liang et al., 2019). To estimate FIR luminosities, we normalise the assumed SED to the observed ALMA flux densities. We derive dust masses following the method presented in Beelen et al. (2006), finding values in the range of $10^{7.1}-10^{8.9}$ M_⊙. From Table 3 we can see that the dust masses derived from SED fitting and the MAGPHYS are consistent. For the galaxies with no ALMA detection, we use global upper limits using $5\times$ the r.m.s. of the tapered image.

Uncertainties for the FIR properties come from the SED fitting, FIR flux measured from ALMA results, and template assumption. We expect dust-temperature uncertainties for galaxy templates for $M_{*}>10^{10}$ M_⊙ galaxies of about 5 K (see the Fig. 5 in Ibar et al., 2013), which leads to a systematic uncertainty in $\log(L_{\rm FIR}/$L${}_{\odot})$ of about 0.3 dex. We run a Monte-Carlo simulation to sample the dust temperature, assuming a Gaussian distribution centred at $T_{\rm dust}$ from the FIR SED templates, and a scatter $\sigma_{T_{\rm dust}}=5$ K. We also sample the ALMA flux density (or flux density limit) from a Gaussian distribution centred at the measured flux density with the observed scatter. Then we derive $L_{\rm FIR}$ and $M_{\rm dust}$, and their r.m.s. scatter, such that the scatter of $L_{\rm FIR}$ includes the uncertainty in $T_{\rm dust}$ and ALMA flux density. Since we only have one detected FIR band for most sources, we cannot sensibly adopt a more complex model to understand the FIR properties.

The obscured SFR, as derived from the observed FIR emission, can be estimated by Kennicutt et al. (2012):

assuming a Chabrier initial mass function (IMF; Chabrier, 2003).

On the other hand, to correct the flux observed from the narrow-band H$\alpha$ imaging for dust extinction, we assume a parametrisation of the H$\alpha$ extinction as a function of stellar mass following Garn & Best (2010):

where $X=\log_{10}(M_{*}/10^{10}{\rm M}_{\odot})$. This correlation between H$\alpha$ extinction and stellar mass has also been confirmed by Sobral et al. (2012), Koyama et al. (2019) and Qin et al. (2019). Using the stellar masses presented by in Table 2, we derive the extinction-corrected H$\alpha$ luminosity ($L^{\rm corr}_{\rm H\alpha}$) to obtain the SFR:

assuming the same Chabrier IMF.

A comparison between these two SFR estimates is shown in Fig. 3.

Previous studies of the SFR from FIR and H$\alpha$ show that SMGs at $z\sim 2$ have much larger SFR_FIR than SFR${}_{\rm H\alpha}^{\rm corr}$ (Swinbank et al., 2004; Chen et al., 2020).

Due to the sensitive H$\alpha$ selection of our HiZELS parent sample, our sample comprises more typical ‘main sequence’ galaxies, and is not limited to the most dusty starbursts. Only one of the ALMA-SHiZELS galaxies has FIR-derived SFR above 100 M${}_{\odot}\,\rm yr^{-1}$. It is therefore not surprising that the SFRs derived from H$\alpha$ and FIR emission are more similar for our sample than for literature SMGs.

We consider the gas-phase oxygen abundance as a proxy for metallicity, and estimate it from the [N ii]/H$\alpha$ emission line ratio (Pettini & Pagel, 2004; Curti et al., 2017) from the near-IR IFU spectroscopy (see Table 3). The SHiZELS population has mainly Solar to sub-Solar values (Swinbank et al. 2012b; Molina et al. 2017). We show the mass-metallicity relation of our ALMA-HiZELS sample in Fig. 4 and identify the ALMA-detected and non-detected targets. We show the mass-metallicity relation at redshift $\sim 2$ from Erb et al. (2006) as a comparison. In our observations, we find that the lower the metal content or stellar mass, the weaker the ALMA continuum emission. Indeed, all of the ALMA-detected galaxies have [N ii]/H$\alpha\gtrapprox 0.2$ or $M_{*}>10^{10.5}{\rm M}_{\odot}$, corresponding to 12+log(O/H) $>8.5$. The typical [N ii]/H$\alpha$ emission line ratio for galaxies at $z\sim 2$ ranges from 0.03 to 0.3 (Strom et al., 2017), thus our ALMA-detected sample is indeed biased towards massive and high-metallicity galaxies. As expected, dust continuum is more easily detected for a galaxy with higher stellar mass or metallicity.

The ALMA detected targets have stellar mass values about one order of magnitude higher than the ALMA non-detected targets. Since the SFR of the ALMA-SHiZELS sample is in the range of 10 to 100 M_⊙ yr^-1 (within one order of magnitude; Gillman et al., 2019), the ALMA detected targets in this work mainly have low specific SFR compared to the rest of the SHiZELS sample.

Using the metallicity measurements, we can also derive the gas-to-dust mass ratio ($\delta_{\rm GDR}$) following Magdis et al. (2012): $\log\delta_{\rm GDR}=(10.54\pm 1.0)-(0.99\pm 0.12)\times[12+\log(\rm O/H)]$. We estimate ISM mass using the ALMA-derived 355 $\mu$m flux, following Scoville et al. (2016) (Eq 16). Our derived dust masses are presented in Table 3. As shown there, these estimates are in good agreement with the values derived from SED fitting.

To estimate the dust radius, we fit the $uv$-real visibility of the ALMA detected targets by casa task uvmodelfit, and show the results in Fig. 5. We chose the $disc$ model in uvmodelfit to fit the $uv$ profile. The circularised radius $R^{\rm uvfit}_{\rm dust}=\sqrt{R_{\rm maj}\,R_{\rm min}}$ of our targets are shown in Table 2. Since the FIR morphology is not regular, and thus the disk fitting results may not represent the flux distribution, we also measure the half-light radius ($R^{\rm half-light}_{\rm dust}$) of the ALMA images from the surface brightness distribution in the image plane and show the results in Table 2. The FIR surface brightness profiles in real plane are also fitted by Sérsic function and can be found in Section B. The different approaches adopted for the measurements of the dust radii all show that the radii are 4-5 $\pm$ 0.5 kpc, larger than the typical size of the high-redshift SMGs (Ikarashi et al., 2015; Simpson et al., 2015; Fujimoto et al., 2017). Dust emission from the four ALMA-detected galaxies is found to extend up to $\sim 2^{\prime\prime}$ ($\sim 16$ kpc, Fig. 10).

Figs 1 and 8 and 9 show the HST, SINFONI and ALMA images of our SHiZELS sample. For the galaxies at redshift about 1.47 and 2.23, the HST F606W filter corresponds to rest-frame 2400 Å and 1800 Å, therefore these images are tracing mainly the detectable UV emission, i.e. the star formation that is not obscured by the dust. On the other hand, the observed HST F140W image reveals the rest-frame $r-$ or $g-$band morphology (free of strong line emission), which is more sensitive to stellar mass. Thus Figs 1, 8 and 9 show proxies for the stars (F140W), UV-traced star formation (F606W), H$\alpha$-traced star formation (SINFONI), and dust mass distributions (ALMA). In this section, we describe properties of our ALMA-detected targets individually²²2Most of our targets have two versions of the H$\alpha$ maps (Swinbank et al., 2012a; Gillman et al., 2019). The main difference in the H$\alpha$ morphology is caused by the different S/N criteria to create the H$\alpha$ maps. Gillman et al. (2019) built H$\alpha$ maps with high S/N spectra while the results in Swinbank et al. (2012a) use all the available spectra, which would then trace more extended H$\alpha$ emission..

We caution that morphological information derived from the rest-frame UV/optical can be potentially very misleading in sources with substantial dust extinction and on-going star formation and in particular it is difficult to trace the true stellar mass (see Lang et al., 2019).

• •

  SHiZELS-7: The H$\alpha$ kinematics reveal the presence of an extended disc-like rotating structure. Two H$\alpha$ star-forming clumps, separated by 4 kpc, are identified by Swinbank et al. (2012a), although the fainter clump is not seen in Gillman et al. (2019) because of the different S/N criteria. The FIR continuum is spread over a diffuse structure. The two H$\alpha$ clumps are not likely to be affected by significant dust obscuration, so there might be no massive dust and gas clouds associated with the H$\alpha$ features. The HST F140W and F606W images show that the least obscured regions in this galaxy show a compact stellar core and a extended UV morphology.

• •

  SHiZELS-9: shows an extended rotation-dominated structure in H$\alpha$ with three bright clumps (separated by $\sim$ 3 kpc, Swinbank et al., 2012a). The FIR continuum shows a V-like structure in the central 3 kpc, linking a bright stellar core with a fainter one. A third stellar core shows both UV and FIR emission.

• •

  SHiZELS-11: This source was classified as a disk galaxy with a compact H$\alpha$ structure (Swinbank et al., 2012a); although a newer analysis shows a marginally extended morphology (Gillman et al., 2019). SHiZELS-11 has an active galactic nucleus (AGN) identified in the X-rays by XMM-Newton and in the radio by the Very Large Array (VLA — Simpson et al., 2006; Ueda et al., 2008). SHiZELS-11 is also detected by Chandra in X-UDS survey (Kocevski et al., 2018), with X-ray luminosity is $L_{X}=10^{43.3}$ erg $\rm s^{-1}$. The velocity dispersion of H$\alpha$ is about 90 km s^-1 (Gillman et al., 2019), much less than the typical velocity dispersion of type-I AGN, implying that the AGN in SHiZELS-11 must be obscured. Rotation is not clear from the H$\alpha$ dynamics, and the position of the maximum velocity dispersion is offset from the brightest H$\alpha$ pixel (Gillman et al., 2019). The HST F140W image shows a smaller source to the south, which may suggest a close merger or dust lane. The source has an apparently high metallicity and a steep metallicity gradient (Swinbank et al., 2012a), which might be due to contamination from an AGN. Both UV and FIR are bright in the galaxy centre, while the rest-frame optical is mainly coming from a clumpy structure surrounding the centre. The ALMA continuum shows compact and extended emission, which does not overlap with the H$\alpha$.

• •

  SHiZELS-14: This galaxy has been identified as a merger with three H$\alpha$ clumps separated by $\sim$ 5 kpc (Swinbank et al., 2012b). The H$\alpha$ morphology in Gillman et al. (2019) is less clumpy, but still extends to a half-light radius of about 7 kpc (Sersic model fitting results). The H$\alpha$ kinematics show a velocity-dispersion-dominated system. The galaxy has the highest metallicity in the SHiZELS sample of Swinbank et al. (2012a). The full extent of the H$\alpha$ emission is not traced by the rest-frame UV imaging. The ALMA continuum emission shows one dominant compact component at the centre, and more extended emission following a similar orientation as the F140W morphology. The FIR emission spreads up to $2^{\prime\prime}$, i.e. $\sim$16 kpc at $z=2.23$. This target is comprehensively studied by Cochrane et al. (in preparation), including new Jansky VLA data.