Outflows from dwarf starbursts are ubiquitous: kinematics of z0.3 GRB-SN hosts resolved with FLAMES ††thanks: based on ESO proposal 092.D-0389, PI C. Thöne
Key Words.:stars: gamma-ray bursts, galaxies: kinematics and dynamics, galaxies: starbursts
The hosts of long duration gamma-ray bursts are predominantly starburst galaxies at subsolar metallicity. At redshifts z 1, this implies that most of them are low-mass dwarf galaxies similar to the populations of blue compact dwarfs and dwarf irregulars. What triggers the massive star-formation needed for producing a GRB progenitor is still largely unknown, as are the resolved gas properties and kinematics of these galaxies and their formation history. Here we present a sample of six spatially resolved GRB hosts at z 0.3 observed with 3D spectroscopy at high spectral resolution (R 8,000-13,000) using FLAMES/VLT. We analyze the resolved gas kinematics in the entire sample and the abundances in a subsample with high enough S/N. Only two of the six galaxies show a regular disk-like rotation field, two are dispersion dominated and the others have two narrow emission components associated with different parts of the galaxy but no regular rotation field, which might indicate recent mergers. All galaxies show evidence for broad components underlying the main emission peak with FWHM of 150–300 km s. This broad component is more metal rich than the narrow components, blueshifted in most cases and follows a different velocity structure. We find a weak correlation between the star-formation rate and the FHWM of the broad component, its flux compared to the narrow component and the maximum outflow velocity of the gas, but we not find any correlation with the star-formation density, metallicity or stellar mass. We hence associate this broad component with a metal rich outflow from star-forming regions in the host. The GRB is never found at the brightest or most extreme region of the host but is always associated with a star-forming region showing a clear wind component. Our study shows the great potential of 3D spectroscopy to study the star-formation processes and history in galaxies hosting extreme transients, the need for high S/N and the dangers using un- or only partially resolved data.
Long Gamma-ray bursts (LGRBs) progenitors have indubitably been identified as massive stars through their connection to Type Ic broad-line supernovae (SNe) coincidnt with the GRB (for a recent review see Cano et al., 2017b). Observations and theoretical models suggest that their progenitors are Wolf-Rayet (WR) stars at low metallicity, massive stars stripped of their H and He envelopes. At low metallicities, stellar winds are weaker and they retain enough angular momentum necessary for the GRB jet to form (Woosley & Heger, 2006). Massive star-formation in low metallicity gas happens, at low redshift, primarily in dwarf starburst galaxies. Unsurprisingly, dwarf galaxies are in fact the majority of low redshift GRB hosts. The average luminosity and stellar masses of GRB hosts are –19 mag or log M* 9.0 M at z0 but they rise to log M* 9.6 M and 10 M at redshifts of and (Perley et al., 2016; Palmerio et al., 2019). GRB hosts at low redshift are similar to the galaxy populations of blue compact dwarfs (BCDs, defined as M –18 mag and size kpc) and dwarf irregulars (dIrrs), a few have also been found to be dwarf spiral galaxies (e.g. GRB 980425 and GRB 060505). One of the very few exceptions is the host of the recently detected GRB 171205A, one of the closest GRBs ever detected and hosted by a large spiral with a mass of log M*10.1 (Perley & Taggart, 2017; Izzo et al., 2019) and the more distant face-on grand design spiral host of GRB 990705, which has a similar stellar mass (Hunt et al., 2014).
Since GRBs are distant and occur in small galaxies rarely monitored by high-angular resolution surveys, it has not been and will likely not be possible in the near future to image the progenitor of a GRB, hence we are only able to infer properties of the progenitor star from its environment. The average redshift of LGRBs is (Coward et al., 2013) where 1” corresponds to a size of 8 kpc, preventing spatially resolved observations until the advent of future, sensitive, high angular resolution spectral facilities such as JWST with resolutions 0.1”. To date only a handful of GRB hosts have been studied with IFU data: GRB 980425 (Christensen et al., 2008; Krühler et al., 2017), the SN-less long GRB 060505 (Thöne et al., 2008, 2014), GRB 100316D (Izzo et al., 2017), which contains one of the datasets presented in this paper, and GRB 111005A, another potential SN-less GRB (Tanga et al., 2018). A few more hosts have been studied with resolved long-slit spectroscopy such as the hosts of GRB 120422A (Schulze et al., 2014), the short GRB 130603A (de Ugarte Postigo et al., 2014), GRB 130925A (Schady et al., 2015), GRB 161219A (Cano et al., 2017a) as well as studies predating IFU data on the same hosts mentioned above (e.g. Thöne et al., 2008; Levesque et al., 2011)
GRB 980425 was located in a star-forming (SF) region next to a brighter SF region with properties more extreme than the GRB region. In observations at lower angular resolution those two regions could have been easily confused. The site of GRB 060505 was found to be metal poor and highly star-forming in the long-slit study, but IFU data show that the GRB region is not very different from other SF regions in the spiral arms at a similar galactocentric distance. The host of GRB 100316D is a textbook low-redshift GRB host, an irregular, low mass, low metallicity galaxy with several bright star-forming regions, in one of which the GRB was located. The large, highly inclined, dusty spiral host of GRB 111005A having a high metallicity and stellar population age looks different from other LGRB hosts. The fact that no SN was detected and the GRB seemingly being located in a dense region could imply that the putative host is actually a foreground galaxy, although the authors dispute this (Tanga et al., 2018; MichałowskI et al., 2018). Such a mis-association has recently been reported for the host of GRB 020819B, which in fact was a dusty starburst background galaxy at z 1.9 (Perley et al., 2017) and not the metal rich spiral at z 0.41 originally identified as the host (Jakobsson et al., 2005; Levesque et al., 2010c).
The origin of star-formation in dwarf starburst galaxies is still largely unknown. Their star-burst episodes seem to only last a few tens of Myr followed by Gyrs of quiescence or very low activity (Lee et al., 2009; McQuinn et al., 2010; Zhao et al., 2011) and the small potential well leads to only a few SN explosion are able to disrupt molecular clouds and interrupt SF in the galaxy. Some galaxies have been undergoing tidal interaction with a massive neighbor or are dwarf-dwarf mergers (see e.g. Bekki, 2008; van Zee et al., 1998), but many are seemingly isolated (see e.g. the SIGRID sample Nicholls et al. 2014, or galaxies in the LITTLE THINGS survey, Hunter et al. 2012). However, some post-merger systems might appear isolated until a detailed kinematical study, for which HI studies have proven particularly useful (see e.g. Ashley et al., 2013, 2017). Other triggering mechanisms have been suggested such as inflows of gas and clumps, either from clouds in the halo or material external to the galaxy (Elmegreen & Hunter, 2015; Verbeke et al., 2014) or stellar feedback and ram-stripping removing part of the gas that is later re-accreted (Ashley et al., 2017). It is likely that not one single mechanism can explain all starburst episodes in dwarf galaxies (see e.g. Koleva et al., 2014).
Studies of GRB host morphologies by Conselice et al. (2005) and Wainwright et al. (2007) suggested a considerable fraction of mergers and interacting systems, visually, however, a few LGRB hosts are clearly interacting systems (GRB 090323, Savaglio et al. 2012, GRB 090426, Thöne et al. 2011, GRB 120422A, Schulze et al. 2014 and GRB 080810, Wiseman et al. 2017). Other hosts might have experienced interactions on smaller scales leading to enhanced SF which could have given rise to the GRB progenitor (see e.g. Izzo et al., 2017; Arabsalmani et al., 2015), in some cases interacting neighbors have only been found after a thorough search (Arabsalmani et al., 2019, de Ugarte Postigo et al. in prep.). Another possibility to supply the gas needed for SF are inflows of metal poor gas, which has been suggested as a possibility for GRB hosts from resolved HI and radio continuum maps (Michałowski et al., 2012, 2015).
Kinematics even in the most compact dwarfs can be complex and hot gas/HI kinematics and stellar kinematics do not trace each other, supporting scenarios of interactions (Johnson et al., 2012; Koleva et al., 2014; Ashley et al., 2017). Star-bursting dwarf galaxies often show broad components in nebular emission lines which can reach velocities of up to a thousand km s (Izotov et al., 2007; Telles et al., 2014) and whose origins are not fully understood yet. One suggestion is that SN explosions form large bubbles that drive the gas outwards, alternatively, fast winds from WR stars could provide a similar kinematic input (Telles et al., 2014). Other suggestions include turbulent mixing layers on the surface of dense gas clouds (Westmoquette et al., 2007; James et al., 2009) or AGN activity, although the latter has been mostly ruled out by based on ionization mesurements (Izotov et al., 2007; James et al., 2009). While those broad components seem to be frequent, they might not be present in all starburst dwarfs (see e.g. James et al., 2010). The so-called green-pea (GP) galaxies (Cardamone et al., 2009), which are essentially extreme BCDs in an early stage of their starburst characterized by very strong [OIII] emission, have proven to show broad components that are likely associated with outflows from SN driven winds (Amorín et al., 2012).
Galactic winds are clearly linked to starburst activity and have been extensively studied in the local Universe, where they can be directly observed (e.g. in M82 observed in H), out to high redshifts, where they are studied via absorption lines (for reviews see Veilleux et al., 2005; Rupke, 2018). Emission and absorption lines study different parts of the winds but outside the local Universe, the combination of both has only been done in a few cases (Erb et al., 2012; Wood et al., 2015). Recently, the MEGAFLOW sample has started to study galactic winds detected in background QSO absorbers and their galaxy counterparts in emission with MUSE (Schroetter et al., 2016; Zabl et al., 2019). Galactic winds are now seen as a crucial factor in explaining and constraining the shape of the mass-metallicity relation (Mannucci et al., 2010; Chisholm et al., 2018) and enriching the intergalactic medium (IGM).
The place of GRB hosts among the variety of (dwarf) starbust galaxies is still a matter of debate and can best be addressed in spatially and spectrally resolved studies of these galaxies. Where does their star-formation come from? Are they mergers or are there different processes at work? Do we see evidence of outflows or winds from the host or specific starburst regions? What is the location of the GRB compared to the SF regions in the host? Are GRB hosts a particular sub-class of starbursts with specific properties and what are their closest analogues? The answer to these questions does not only provide insight to the properties of GRB hosts but also to the conditions, time and place of massive star-formation and its impact on the general evolution of starburst galaxies.
In this paper we present the first sample of long GRB hosts observed at high spatial and spectral resolution with FLAMES/VLT and which is complete up to GRB discovered until 2015 . This comprises the seven GRB hosts of GRB 020903, GRB 030329, GRB 031203, GRB 050826, GRB 060218, GRB 100316D and GRB 120422A. For GRB 120422A, we did not obtain data due to an incorrect pointing of the instrument, hence the sample is reduced to 6 galaxies. In Sect. 2 we present the observations, Sect. 3 gives an overview of the individual hosts as well as a the properties of the GRB itself. Sect. 4 details the kinematical analysis of the different hosts, Sect. 5 analyses resolved abundances in four of the hosts and finally, Sect. 6 discusses the results. Throughout the paper we use a flat lambda CDM cosmology as constrained by Planck with , and (Planck Collaboration et al., 2018).
2 Observations and data reduction
The GRB host sample was observed with FLAMES/GIRAFFE at the VLT between Nov. 2013 and March 2014 with a total observing time of 22 h. Observations were done in ARGUS mode, which is an IFU with two different spatial samplings available of which we used the 0.3” sampling providing a 6.6” 4.2” field-of-view (FOV, see Fig. 1). We used three different grisms, LR6, 7 and 8 with spectral resolutions of R 13,500, 8200 and 10,000 respectively, to cover at least the range from H to [SII] at the different redshifts of the hosts. The exact wavelength coverage for each spectrum is shown in Fig. 2. Since most observations were divided into several observing blocks observed at different dates, the seeing varies between the different datasets. The nominal seeing for the individual galaxies are 0.7” (GRB 020903, GRB 030329), 0.8” (GRB 031203 and GRB 100316D), 1.0” (GRB 050826) and 1.2” (GRB 060218). Data reduction of ARGUS data was done using the standard ESO pipeline (version 2.11) without the sky subtraction option. In order to verify the fiber-to-fiber wavelength calibration, we controlled the wavelength of two skylines in the ARGUS data cube. No flux calibration was done for the data cubes.
The sample was analyzed with dedicated software in IDL and IRAF, partially based on the tools presented in Flores et al. (2006); Yang et al. (2008). For the emission line maps we integrated the flux in the region around the emission line and subtracted the continuum from a line-free region around the emission line. In our sample, this approach is justified since the emission lines have shapes deviating more or less from a pure Gaussian, hence this method is able to catch the entire flux in the emission line. For the velocity maps we fit the lines with a simple Gaussian and plot the center and velocity dispersion , which implies that we plot the properties of the dominant line component. The dispersion is not corrected for the instrumental broadening, which is FWHM 20–30 km s, corresponding to a of 8–13 km s, depending on the resolution and grism used. All maps are interpolated to 10 times the resolution for better visualization. Multi-component fits to individual emission lines for kinematical analysis were done with ngaussfit in IRAF and PAN in IDL (Peak ANalysis, Dimeo, 2005; Westmoquette et al., 2007). Metallicities were obtained using the N2-parameter, which is based on the ratio of [NII]6585/H, and the calibration of this parameter from the CALIFA sample presented in Marino et al. (2013). Relative line fluxes for [NII] and H were derived from a single Gaussian fit to the lines, due to their proximity in wavelength, an absolute flux calibration is not needed to derive the ratio of the two lines.
3 Line detections and global properties of the individual hosts
In Fig. 2 we plot the integrated spectra of all host galaxies. While H is detected in all galaxies the detection of further lines depends on the S/N of the spectra. Some lines are only detected in the integrated spectra but not in individual spaxels. In the following section we give a brief overview of the literature on the different GRB hosts and the detected lines for each system. For informations on the individual hosts we made ample use of the GHostS database. 111www.grbhosts.org, a database partly funded by Spitzer/NASA grant RSA Agreement No. 1287913. We furthermore determined the sizes of the hosts using literature data or archival images to derive r and r, the radii containing 80% and 50% of the light, respectively (see Tab. 1).
References. Levesque et al. (2010b), Thöne et al. (2007), Levesque et al. (2010a), Starling et al. (2011), Svensson et al. (2010), Wiersema et al. (2007), Levesque et al. (2011), Izzo et al. (2017).
3.1 Grb 020903
GRB 020903 was a very soft X-ray flash (XRF) (Sakamoto et al., 2004) and belongs to the class of low-luminosity GRBs with an E of 1.110 ergs. The afterglow showed an initial rise during the first few hours (Bersier et al., 2006) but there is very little follow-up of the optical afterglow while the radio afterglow resembles other, regular, GRB afterglows in brightness and temporal behavior (Soderberg et al., 2004). Soderberg et al. (2005) and Bersier et al. (2006) detected a SN at late times with spectra similar to SN1998bw but 0.5–0.6 mag fainter. The host is in irregular galaxy at with either several SF regions or an interacting system as can be seen in HST images of the host (Fruchter et al., 2006b; Svensson et al., 2010). The galaxy is small with r of 1.43 kpc and has a stellar mass of log M* = 8.69 M and a SFR of 1.7 M yr (Svensson et al., 2010; Levesque et al., 2010a). Han et al. (2010) found a tentative detection of Wolf-Rayet features in this galaxy, indicating a large population of potential GRB progenitors. Levesque et al. (2010a) also found a high ionization parameter for this galaxy indicative of a hard radiation field, which is not surprising given the presence of WR stars.
We only detect H in the FLAMES data, albeit with high S/N, and weak lines of [S ii] in the integrated spectra. [N ii] is not detected, which is not surprising given the low metallicity reported in the literature. From the limit of [N ii] we get a limit on the metallicity of 12+log(O/H) 8.1, in agreement with a metallicity of 7.98 - 8.07 found by Levesque et al. (2010b) and also the limit derived by Bersier et al. (2006). The different SF regions clearly visible in the HST image of the host cannot be separated in the FLAMES data as the spatial resolution of our data is too low.
3.2 Grb 030329
GRB 030329 (), detected by the HETE-2 satellite (Vanderspek et al., 2003) had the first, immediately confirmed, associated broad-line Ic SN of a long GRB (Hjorth et al., 2003; Matheson et al., 2003). It is still the GRB with the largest follow-up dataset and its radio afterglow had been observed for over 10 years (Mesler & Pihlström, 2013) and is, in fact, still visible. With an isotropic energy release of 1.7410 erg (Kann et al., 2010) it is also one of the brightest GRBs at low redshift. The host has a low mass and is very compact (log M*7.74 M, r 1.03 kpc Svensson et al. 2010) and has a low metallicity of 12+log(O/H) 7.7 – 8.0 (Thöne et al., 2007; Levesque et al., 2010a; Starling et al., 2011). From the HST images it seems the host consist of only one major SF region (see Fig. 1). Östlin et al. (2008) determined a low age of 5 Myr for the stellar population at the GRB site (assuming an instantaneous burst of star formation) using high spatial resolution broad-band data from the HST to fit the SED of the star-forming region at the GRB site. Levesque et al. (2010a) detect the Balmer series down to H8 together with [NeIII] emission, pointing to a young stellar population. The detection of [OIII] 4363 Å which becomes increasingly faint above 12+log(O/H) 8.0 furthermore confirms a very low metallicity for this galaxy.
We only detect H in individual spaxels as well as the integrated spectrum. [N ii] is too weak at the metallicity of the host and the [S ii] doublet is contaminated by bright sky emission lines. He i 7065 would be in the observed spectral range but is not detected either. Thöne et al. (2007) presented high spectral resolution data of the afterglow including both absorption lines and emission lines with the first ones showing a complex kinematical structure. This host, together with the host of GRB 060218, is one of only two GRB hosts so far where high resolution data of both absorption and emission lines are available.
3.3 Grb 031203
GRB 031203, an INTEGRAL burst at had a very weak afterglow and a SN similar to the “template” GRB-SN 1998bw (Malesani et al., 2004). The host is a compact, but rather luminous host (log M*) and has been studied with extensive wavelength coverage. Prochaska et al. (2004) found a high SFR corrected for dust extinction of 11 Myr. The GRB was located next to but not directly at the galaxy center which is also the brightest region in the host. No individual SF regions have been identified, neither in imaging nor in the FLAMES data. Unfortunately, there are no HST images available of this galaxy. The galaxy seems to host a young stellar population, indicated by the lack of a 4000 Å bump and the detection of high excitation MIR lines of Ne ii+Ne iii and [S iii]+[S iv](Watson et al., 2011a), characteristic for young starbursts. Also tentative WR lines have been detected (Han et al., 2010). Strong IR emission points to significant extinction but a high dust temperature ( 70K) and the dust-to-mass ratio seems to be smaller than for other bright IR galaxies, suggesting different dust properties from local dwarf galaxies (Symeonidis et al., 2014). Michałowski et al. (2015) detected radio emission but offset towards the west of the galaxy and with a flat spectral slope suggesting a contribution from synchrotron self-absorption, indicating a very young stellar population. The spectral slope also rules out an AGN contribution, contrary to what was suggested by Levesque et al. (2010a).
Due to its brightness we detect several emission lines even in the individual spaxels: H, [N ii], the [S ii] doublet, [Ar iii] 7136 and He i 7065. The other two HeI transitions in the spectral range 6678 and 7221 Å are not detected, nor is the [O ii] 7230, 7330 doublet. All of these lines were also detected in X-shooter spectra of the host in Guseva et al. (2011) and Watson et al. (2011a). We will provide a resolved analysis of the different emission lines in Sect. 5.
3.4 Grb 050826
The host of GRB 050826 () is visually a compact but has the highest stellar mass of the sample with log M*. There is some discrepancy in the literature on the SFR inferred from different indicators: While Svensson et al. (2010) find a moderate SFR of 1.39 M/yr based on U-band photometry, Levesque et al. (2010b) derive a somewhat higher value of 2.9 M/yr from the H emission line. Levesque et al. (2010b) determined a supersolar metallicity for the host galaxy based on the R parameter, taking the upper branch of the solution. Mirabal et al. (2007) detected also [OII] from the host and report the only afterglow detection of this GRB. No SN component has been searched for in this object due to the late confirmation of the afterglow. The initially reported transient (which later turned out to the be the actual afterglow) was not coincident with the X-ray afterglow position (Halpern, 2005) and only on Feb. 12, 2006, the final identification of this optical transient with the GRB was reported (Halpern & Mirabal, 2006), several months too late for a SN follow-up.
We only detect H in the integrated spectrum of the galaxy. Given the stated high metallicity of the host by Levesque et al. (2010b), we should easily detect [N ii] even in individual spaxels. The integrated spectrum of the host shows an emission excess at low significance at the position of the [N ii] which is not enough to reliably determine the metallicity. Assuming the same line width for H and [N ii]6585, we derive a 3- limit of [N ii]/H = 0.25, which would result in a metallicity limit of 12+log(O/H)8.45, in clear disagreement with the value in Levesque et al. (2010b). To investigate this further, we analyzed the original longslit spectrum from Levesque et al. (2010b) (D. Perley, priv. comm.) which was taken using a 1” slit at an orientation of 343 degrees (almost N-S), hence covering most of the galaxy. In this spectrum we measure a ratio [N ii]/H of 0.2, in good agreement with our measurements in the FLAMES spectrum. The metallicity in Levesque et al. (2010b) was derived using the R parameter, which has been shown to overpredict the metallicity, and the authors use the upper branch of the two-valued metallicity calibrator which in this case is the wrong one. Since the flux ratio from both datasets are consistent, the metallicity value stated in Levesque et al. (2010b) is incorrect and so is the claim of the supersolar metallicity of this GRB host.
3.5 Grb 060218
This very low redshift burst at was an X-ray flash (XRF) but with a very long duration of 2000s. It had a peculiar afterglow with an additional thermal component in the X-rays and in the optical (Campana et al., 2006; Thöne et al., 2011) which was interpreted as a possible shock-breakout of the SN from the star (Campana et al., 2006). The host is still the least luminous (M = –15.9 mag, log M* = 7.4 M) and smallest one ever detected with a size of r0.55 kpc. It also has the lowest metallicity so far determined for a GRB host with 12+log(O/H) 7.6 or 0.1 solar (Wiersema et al., 2007; Kewley et al., 2007). Even HST imaging (Starling et al., 2011) does not resolve different HII regions, although Wiersema et al. (2007) find two kinematical components in emission (see Sect. 4.5).
We detect only H in the individual spaxels and marginally detect [S ii] in the integrated spectrum of the host. The limit on the ratio of log( [N ii]/H) –1.87 from the integrated spectrum implies a metallicity limit of 12+log(O/H) 7.88, consistent with the value determined in other longslit data.
3.6 Grb 100316d
GRB 100316D was another low redshift (), long duration, sub-luminous XRF (1500 s) with a soft prompt emission spectrum and a thermal component in X-rays (Starling et al., 2011). Due to some confusion of the optical counterpart with bright star-forming regions in the host in the first days after the GRB, there exists very little information on the actual afterglow. The host galaxy is a low-mass (log M 8.93 M), irregular, highly star-forming galaxy (Starling et al., 2011; Levesque et al., 2011) with several bright star-forming regions. Due to its close distance, however, it is extended enough for a detailed, resolved analysis (the angular size of the host is 12 arcsec). The metallicity of the galaxy is rather low with 12+log(O/H) 8.0–8.2 (Levesque et al., 2011; Izzo et al., 2017) and a total SFR of 1.2 M/yr (Izzo et al., 2017). The GRB was located at the edge of the brightest and most extreme star-forming region in terms of SFR and metallicity.
The FOV of FLAMES only covers about half of the galaxy due to an error in the observational setup, but does include the GRB region. In individual spaxels we detect H, [N ii] and the [S ii] doublet at high S/N. Other emission lines are outside the range covered by the grism. Izzo et al. (2017) (henceforth I1) used the FLAMES data presented here for the kinematic analysis of the host. Further MUSE data yield a large number of emission lines including those not previously detected in GRB hosts such as [N i] and [Fe ii]. The detection of HeI emission implies a very young stellar population and I17 derive an age of 5 Myr for the population at the GRB site.
For the analysis of the host kinematics in this section we use the H emission line. Depending on the S/N in the different spectra we either fit integrated regions in different parts of the galaxy, including the GRB sites and other regions of interest, or we use a multi Gaussian fit to show their contributions in different regions. The high spectral resolution of FLAMES allows us to identify several distinct kinematical components in all galaxies of the sample. In Fig. 3 we plot the velocity fields and maps for the 6 galaxies in our sample. In Figs. 4, 5, 6, 9, 10 and 12 we show the fits to the H line for the different galaxies and/or in different regions as described in the sections below.
|km s||km s||km s||M||M||km s|
|GRB||region||Component 1 (narrow)||Component 2 (narrow)||Component 3 (broad)||EW||F/F||V|
|km s||km s||km s||km s||km s||km s||Å|
|GRB 020903||main SF reg.||3.6||58.5||40.2||65.7||–10.9||142.5||–592||0.46||53|
|second SF reg.||–7.3||58.5||21.9||69.4||7.3||127.9||–352||0.50||49|
|GRB 100316D||Integ. 1||–15.6||41.9||–||–||–11.2||125.1||–251||0.59||67|
4.1 Grb 020903
At first sight the host of GRB 020903 seems to have a regular rotation curve with a of 60 km s. The galaxy is next to a system of two larger, interacting galaxies, however, as noted by Soderberg et al. (2004) they are at a redshift of compared to of the host galaxy (3900 km s) and hence not interacting with the host. The line width varies very little across the galaxy (see Fig. 3), only the HII region east of the GRB region has a slightly higher value. The clearly distinct HII regions visible in high resolution HST images are not reflected in the velocity field nor do they seem to be related to major deviations from a smooth rotation.
A closer look at the profiles of the emission lines reveals a slight asymmetry which we fit with a double narrow profile. The ratio of the two profiles varies across the galaxy and seems to be associated with different parts of the galaxy (see Fig. 4). The bluer component is related to the close complex of several SF regions in the North of the galaxy while the redder component is associated with the two SF regions in the South of the galaxy, one of which is the GRB region. However, a small contribution of the secondary component is seen in all parts of the galaxy in addition to the main component. The combination of these two components and their gradually varying relative strength across the galaxy give an impression of an ordered rotation field. In fact, the velocity field plotted in Fig. 3 seems to be varying rather linearly across the FOV which is not the velocity pattern expected for a rotating disk.
In addition to the two narrow components, the line shows some excess emission in the blue wing which we fit with a weak broad component (see Fig. 4). This component is most prominent in the part we call “main SF region” but less at the GRB site and the SF region next to the GRB site.
4.2 Grb 030329
The host of GRB 030329 has the smallest of all the sample (15 km s) and a basically uniform dispersion across the host. Considering its very low mass and size, this galaxy probably consists of a single massive star-forming region. The S/N in the individual spaxels is too low to distinguish different components and only H is detected with enough significance to analyze the kinematics of the global galaxy spectrum.
In Fig. 5 we fit H of the combined galaxy spectrum using both a single Gaussian and two additional components in the blue and red wing of the line at velocities of –74 and +70 km s and a FWHM of 23 and 100 km s respectively. There is a clear excess of emission in the blue wing which seems to be present also in individual spaxels. The profile is slightly asymmetric with more emission in the blue than in the red part of the Gaussian, however, the data have too low S/N to constrain this further.
Kinematics of the host had previously been analyzed in Thöne et al. (2007) using UVES/VLT spectra. In those spectra, the emission lines show only a single Gaussian component, however, the absorption lines of Mg i and M ii span over 200 km s in velocity blueshifted compared to the emission line. This has been taken as a strong indication for a starburst wind in this galaxy. The possible excess emission in the wings of H found in the FLAMES data would support this conclusion. Any faint component in the wings would have been missed in the previous UVES data since those were taken only a few days after the GRB when the afterglow continuum was still bright.
4.3 Grb 031203
The host of GRB 031203 is one of only three hosts in our sample that shows an ordered velocity field indicative of ordered rotation in the galaxy and a possible disk component. The total line-of-sight velocity difference across the galaxy, however, is only 53 km s, which would point to a slowly rotating disk, a high inclination of the disk or a low mass, the latter of which is in conflict with the observations. The line width is rather uniform across the galaxy including the center and the GRB location. However, there is a patch of low velocity width at the western end of the galaxy and a higher velocity region in the south, which we are going to investigate further in the following.
We extract integrated spectra of the GRB region as well as the high- and the low- region and fit their line profiles (see Fig. 6). In all regions we clearly detect two components, a narrow, main peak with a FWHM of 90–100 km s and a weaker, broad component with a FWHM of 230–270 km s, also detected in X-shooter (single slit) spectra of the host (Guseva et al., 2011; Watson et al., 2011a). The relative strength of the component varies across the galaxy and explains the low- and high- regions in the line width map, which is derived from a single Gaussian fit to H. The broad component is strongest in the high- region in the South (“broad region”) and only slightly lower in the GRB region and weakest in the low- region in the Western part of the galaxy (“narrow region”). Across all the galaxy, the broad component is blueshifted compared to the main emission component. In contrast to GRB 030329, there are no afterglow spectra with absorption lines available of this burst to study a possible outflow in absorption.
We also look for possible broad components in the other detected lines of [Ar iii], He i, [S ii] and [N ii] (in the following those are named “weak lines”). For this we made another integrated spectrum only comprising the very central region of the galaxy ( spaxels), where those weak lines are actually detected, in order not to be dominated by noise (see Sect.5). In Fig. 7 we plot the lines in velocity space compared to H, note that here the weak lines have been smoothed. All weak lines have somewhat irregular profile due to the low S/N and, except [S ii] 6717, are affected by atmospheric lines that can alter the shape of the lines. In order to check the line profile in the FLAMES spectra, we compare the X-shooter data of the host presented in Guseva et al. (2011); Watson et al. (2011a) to the shape of the lines to the ones in our extraction (see Fig.7). According to the finding chart in Guseva et al. (2011), the slit was oriented N-S with a width of 0.9–1.0 arcsec, hence covering a very similar region as our integrated spectra of the galaxy center.
Even in the higher S/N X-shooter data, it is difficult to determine a difference in the line profile between H and the weak lines since almost all of them are affected by atmospheric lines in the wings. However, all lines do show a broad wing that cannot be fit by a single Gaussian and some excess emission on both sides. Hence, while it is difficult to confirm the exact kinematics of the weak lines from our data, emission likely stretches from -150 to 150 km s and possibly even up to 300 km s in both red and blue.
Guseva et al. (2011) claim some small excess emission beyond the broad-narrow profile fit at –400 km s and +350 km s in all bright emission lines of H, H and [OIII]. In Fig. 8 we therefore plot a zoom of H in the three regions studied in Fig. 6 and the extraction of the central 5x4 spaxels used for the comparison with the X-shooter spectra in the last paragraph. We are not able to recover the excess emission at –400 km s but we do see emission redwards of the broad emission component at 200 – 300 km s in all regions, coincident with the excess emission in the X-shooter spectra (right panel). In some spectra this emission even looks like a relatively narrow extra component in the red wing. Surprisingly, this component has the highest relative strength compared to the rest of the line in the “narrow region” with the lowest contribution of the broad component. In this region, we also clearly see an extra “shoulder” in the blue wing of H at –150 to –200 km s, also apparent in the X-shooter spectra plotted in Fig. 8 (right top panel in the figure) but not in the fit shown in Guseva et al. (2011), probably due to smoothing of the spectra. The reason for not detecting it in the other regions might be that this extra component is blended with the broad component due to its larger FWHM in those regions.
4.4 Grb 050826
This GRB host has the highest redshift of the sample and, despite its comparatively high mass and luminosity, the S/N of even H is rather low. We also note that H is next to a bright atmospheric emission line, however this is in the very blue wing of the line not affecting most of the profile. Since H shows a double peak in part of the host (see the discussion below and Fig. 9), we smooth the cube in wavelength with a Gaussian kernel of 7 pixels before fitting the line with a single Gaussian to derive velocities and maps. The velocity field of the galaxy is very scattered with no clear rotation and the difference in velocity across the galaxy is only 43 km s. The width of the line is rather high with a of 73 km s owing to the asymmetric/double line profile. The line width also shows some trend from lower velocities around 50 km s on the N-W side, where the GRB is located, and going up to almost 100 km s on the opposite side of the galaxy. However, this has been derived from a smoothed line profile, hence we take a closer look at the detailed kinematics in the following by fitting line profiles in different regions.
In Fig. 9 we fit H to three different regions in the host and in the integrated spectrum of the entire galaxy. The H line is broad at the center of the galaxy, asymmetric in the South dominated by the bluer component and clearly double-peaked in the northern part at the position of the GRB, which also apparent in individual spaxels (hence it is not an artefact). This behavior explains the strange-looking velocity field, which is due to the shifting distribution of the double component across the host, much like in the host of GRB 020903.
We first fit the line with a double Gaussian component with a velocity difference of v60 km s and FWHM of 42–56 km s for both components. However, there is clear excess emission in both the blue and red wing in all three regions, for which we include a third, broad component with a FWHM of 180 km s. A fit to this excess emission with a single broad component might not be optimal, alternatively, two narrower components could be fitted in the blue and red wings as we did for the host of GRB 030329. The lack of S/N prevents a detailed analysis on the fit. The centroid of the different components basically does not change, only their relative contributions. The broad component is present everywhere at a similar relative strength compared to the double narrow component. Interestingly, in the combined spectrum of the host, the broad component is less evident and the spectra could be equally well fit with a double component with a slightly higher than the narrow components in the triple component fit, underlining the importance of spatially resolved spectroscopy (see also Sect. 6.5).
4.5 Grb 060218
The host of GRB 060218 has no regular velocity field and does not show any sign of rotation. The velocity width hardly varies, only at the S-W edge there might be a region with slightly higher width. The GRB lies in one of the regions with the lowest , just outside the brightest region of the galaxy. Since the galaxy is very compact and low-mass, the absence of a regular velocity field is not surprising.
Wiersema et al. (2007) presented a spectrum taken with the UVES high-resolution spectrograph at maximum light of the supernova, in which they find two kinematical components in NaD absorption separated by 24 km s. The same components can be clearly distinguished in the emission lines of [OIII], but H was out of the range in those spectra. Motivated by this finding, we tried to recover these two components in the H line of the global spectra of the galaxy (see Fig. 10). Indeed, the emission line does show a very small asymmetry. Fitting a single Gaussian does give a reasonable fit with only a small excess in the blue wing of H. We then fit a double component with a velocity difference of v24.8 km s and FWHM of 49 and 54 km s respectively. These values are similar to the ones found in Wiersema et al. (2007), who get a v of 21.6 km s and FWHM of 47 and 35 km s, respectively. This double component initially gives a worse fit, but adding a small component in the blue wing with a FWHM of 44 km s results in the best residuals.
The presence of two main emission components would have been probably missed in our FLAMES spectra without the information from the UVES spectra. The third component not seen by Wiersema et al. (2007) might have either been missed due to the higher continuum emission from the GRB-SN or its association with a part of the host not covered by the UVES spectra. While a single Gaussian profile has a similarly good fit as the triple profile, once fitting the double profile clearly evident in Wiersema et al. (2007) a third component is needed to account for the additional emission in the blue wing. Since our S/N is too low, we cannot extract spectra from different parts of the host to see whether this is associated with a specific region.
4.6 Grb 100316d
The spectra of GRB 100316D are the only ones with high enough S/N to make 2D maps of different emission components and their velocity fields. The spectra include three major SF regions in the host and most of the bright SF regions located in the northern part of the galaxy. Like many of the other hosts in this study, the H line shows a broad emission component in addition to the usual narrow component. We use the spectra of all regions with an H line flux of 80 counts and fit a double component to all of them using the PAN line fitting tool. In Fig. 11 we plot the resulting flux, velocity and of the two components across the galaxy.
The broad component is mostly present in the bright SF regions around and next to the region of the GRB. The narrow component shows a regular velocity field while the broad component is more chaotic. The velocity width of the narrow component is lowest at the center of the SF regions and higher on the edges while the width of the broad component is more erratic. However, this might be a simple effect of S/N, which is lower at the edge and outside the main SF regions, hence we might just not detect a clear broad component in those spaxels.
To investigate possible further components, we also plot the H profile of four integrated SF regions in the FOV (see Fig. 12). We do not see any additional component apart from the narrow-broad profile described before in the three brightest regions. Interestingly, the broad component in the individual spaxels and integrated HII regions is always blueshifted compared to the narrow component, however, in the profile of the spectrum integrated over all spaxels with H detected, the profile is skewed to the red and we fit a broad component which is redshifted compared to the main emission peak. This is only an artefact from the combination of different regions with different relative contributions and the velocity shift due to the rotation of the galaxy and has nothing to do with different components actually present in the individual HII regions.
5 Resolved abundances
5.1 Grb 031203
The host of 031203 is the only galaxy where we detect [Ar iii] 7136 and He i 7065 together with the more common lines of [N ii] and the [S ii] doublet. In Fig. 14 we plot the maps of all those lines together with the ratio of [S ii] /H, the metallicity derived from the N2 parameter as well as [Ar iii]/H and He i/H. The lines all show a very similar distribution in flux compared to H in being concentrated towards the bright central region. He i and [Ar iii] are detected only in a part of that region due to their lower S/N.
The metallicity is very uniform across the galaxy with values ranging between 12+log(O/H)7.9 and 8.4 with a median of 8.2. The GRB site and the South-East have the lowest metallicity while the center is marginally more metal rich. [S ii] /H follows a patter similar to [N ii]/H but with the lowest values more concentrated towards the center of the galaxy. High [S ii]/H ratios can gives some indications of shocked regions. We therefore plot in Fig. 15 the ratios of [N ii]/H and [S ii]/H for all spaxels where all the lines are detected with a S/N3. Almost all of the spaxels are in the region that can be considered to be ionized mainly by ionization from massive stars and not e.g. by shocks (Westmoquette et al., 2009a). Last, we also plot the Ar and He abundance versus H using the detected transitions of Ar and He i (see Fig.14). As mentioned above, both lines are only detected in the most central regions excluding any further detailed analysis of the distribution of those elements in the host.
Guseva et al. (2011) do not report a metallicity separately for the narrow and broad component, which they argue they are unable to provide due to the low S/N of the T sensitive [O III] 4363 line. Why they do not fit a double component to [N ii] 6585 is not mentioned in the paper. Although the [N ii] line is affected by atmospheric lines, we try to fit a double component similar to H both to the FLAMES data of the center of the galaxy and the X-shooter spectra. The result is plotted in Fig. 13. Despite the uncertainty of the fit, it is clear that the broad component relative to the narrow component is much stronger in [N ii] than in H. From the ratio of [N ii]/H we derive a metallicity of 12+log(O/H)8.05 for the narrow component and 8.18 for the broad component. The fitting of the contaminated [N ii] line makes a precise error definition complicated, but it is evident from the shape of the line that the broad component is much stronger in [N ii] compared to H and hence has a higher metallicity.
5.2 Grb 100316d
The entire galaxy was already studied in detail by I17 based on data taken with MUSE. They concluded that the GRB site is metal poor and next to the most extreme region in the host in terms of SF rate, metallicity and age. The spatial resolution of our data are slightly better than that of MUSE due to better seeing conditions (0.8” for the FLAMES data vs. 1.1” for MUSE), however, we can only make maps of the metallicity using the N2 parameter and [S ii]/H since those are the only lines detected in the FLAMES data. The metallicity follows the same pattern as in I17 with the most metal poor region (12+log(O/H)8.1) being the bright SF region next to the GRB site. In the FLAMES data we also see a metal poor region with the same metallicity to the South-West of the GRB region. I17 do not show a [SII]/H map, only a combined [S ii+N ii]/H map, which is mostly uniform in the part covered by the FLAMES data with higher values in a region they consider being affected by shocks. Our [S ii]/H map shows no large variations but has particularly low values in the bright SF region next to the GRB site.
Like in the host of GRB 031203 we search for indications of shock excitation in the [S ii]/H vs. [N ii]/H plot (see Fig.15). Most of the spaxels in the FLAMES dataset are not affected by shock excitation as is the case for GRB 031203. I17 found a region possibly affected by shocks in the Western part of the galaxy, outside the FOV of the FLAMES datacube. For comparison we plot the values found from an integrated spectrum of this possibly shocked region in Fig. 15 and it is indeed in a region that could be indicative of shocks, however only in [S ii]/H, while the value for [N ii]/H is in a region that could also be only excited by ionizing radiation.
For this host we also tried to determine the metallicity of the broad vs. the narrow component. Since the S/N of [N ii] 6585 is not high enough to perform a reasonable fit we use the stronger line of [S ii] 6717 instead. As we see in Fig. 16 the broad component is stronger compared to the narrow component than in H. The fluxes of both components are almost identical for [S ii] while the narrow component has a 2.5 times higher flux in the narrow component than in the broad component. Our ability to provide a concrete metallicity value is hindered by two problems: First, a metallicity based on the S2 parameter requires the measurement of both lines of the [SII] doublet, however, the 6732 line does not have enough S/N to confidently fit both components. The ratio between the doublet depends on the electron density of the ISM and varies between 1.5 and 0.5 for low and high densities, respectively. Second, the S2 metallicity has a weaker correlation with metallicity compared to the one from the N2 parameter. If we assume a low electron density for both components, supported by the fact that we measure a ratio of 1.5 for [S ii] 6717/6732 for both the peak flux ratio and the ratio of the total flux in the integrated region 3+4, we obtain metallicities of 8.3 and 8.7 for the narrow and broad component respectively, using the metallicity based on the S2 index from Yin et al. (2007).
6.1 GRB host classification and velocity fields
The galaxies in our sample have stellar masses between log M*7.5 and 9.9 M, absolute luminosities ranging from –15.9 to –20.3 mag and sizes of 0.6 to 12 kpc. All those presented here are either BCDs (broadly defined to have M –18 mag and sizes 1 kpc), or dIrr galaxies, although the transition between these two classes is rather unconstrained. Their rather large range in sizes and masses and could imply that conditions for the creation of a GRB progenitor could be met across the full spectrum of starburst (dwarf) galaxies and are not restricted to a very specific type of galaxy. BCDs show all kinds of velocity fields from ordered, disk-like rotation to highly disturbed velocity fields (see e.g. Östlin et al., 2001; Blasco-Herrera et al., 2013; Cairós et al., 2015; Cairós & González-Pérez, 2017). Very compact dwarfs such as e.g. the luminous, highly star-forming GPs (Lofthouse et al., 2017) often do not show a regular rotation field. dIrrs are considered to be on the low-mass end of disk galaxies and hence be similar in their velocity field to spiral galaxies with the only difference being the absence of a bulge component (e.g. Swaters et al., 2009).
Only two galaxies in our sample show a regular disk-like rotation curve while the two smallest hosts do not show any signs of rotation. In order to determine which hosts are dispersion dominated we apply the commonly used criterion of dispersion vs. rotation dominated galaxies: v/ 1 with v 0.5 , where is the difference between the minimum and maximum velocity of H, and the flux weighted average of the line width in all spaxels with S/N H3 (see values in Tab. 2). Using and v we can derive a dynamical mass for the hosts. There are several methods to derive this value, depending on whether the galaxy is rotation or dispersion dominated. For spherical, relaxed systems the dynamical mass derived from the virial theorem is . For rotating systems the mass can be derived using the rotational velocity (v=0.5, identical to v derived above) (Bellocchi et al., 2013). M technically comprises the mass within the half light diameter using the velocity at the half-light or effective radius while we use half the maximum velocity spread for v, however, at the size of our galaxies and resolution of our dataset we can assume that v is not too different from v. In Tab. 2 we list the values from both methods for all galaxies, however for any further calculations we use the values from M for the hosts of GRBs 030329, 060218, 020903 and 050826 and M for the rotation dominated systems of GRBs 100316D and 031203. Note that for none of the galaxies we do apply any correction for a possible inclination of the galaxy as we have no constraining information of the inclination of the BCGs and dIrrs.
The two smallest and most compact hosts, those of GRB 030329 and GRB 060218, clearly fall in the category of BCDs. Both galaxies show almost no indication for ordered rotation and are clearly dispersion dominated. Interestingly, the stellar masses derived are much lower than those expected from the broadening of the line, which could point to a more turbulent system. The broad component here likely does not play a large role in determining as its contribution is rather small (see F/F in Tab. 3).
The hosts of GRB 020903 and GRB 100316D are dIrr with many different SF regions, however with vastly different sizes: While the host of GRB 020903 is rather compact, the one of GRB 100316D has r=12 kpc and shows intense SF around the GRB site but rather low surface brightness in most of the rest of the galaxy. The host of GRB 100316D has the most pronounced disk-like rotation field of all our sample and indeed I17 fitted a disk rotation curve to the host and found that the galaxy is similar to other dIrr as being dominated by the disk rotation at small radii but by dark matter (DM) at larger radii. I17 also found that the galaxy might not be completely virialized and suggested a close encounter with another (unknown) neighbor as the cause of this. This could also explain the starburst going on in a small part of the galaxy (40 % of the SF is located in the 25% of the galaxy covered by FLAMES). The host of GRB 020903 does not show a regular rotation curve, but is largely dominated by two main emission components close in velocity space that shift in strength across the galaxy. The host has v/ 1 and hence is on the edge of being dispersion dominated, note, however, that it is difficult to determine v due to the double emission component.
Finally, the hosts of GRB 031203 and GRB 050826 are on the high mass end of low-redshift GRB hosts and the host of GRB 050826 has a stellar mass that is beyond what is usually defined as a “dwarf galaxy”. Watson et al. (2011b) classified the host of GRB 031203 as a BCD due to its hard radiation field, high SF rate, relatively low mass and absence of large amounts of dust and considered it to be an analogue of high redshift star-forming galaxies. The host does show a clear rotation curve but a very low maximum velocity of only 25 km s. The masses derived from both the dispersion and the rotational velocity match very well but are considerably lower than the stellar mass derived. An explanation for this discrepancy would be a high inclination of the disk, however, the fact that the broad wind component should be strongest if the disk is seen face on, would argue against this possibility. With the lack of high resolution imaging for this galaxy it is difficult to determine its exact morphology. For the host of GRB 050826, any further conclusion on the dynamics is difficult as there is no high-resolution imaging available. Formally, the host of GRB 050826 is dispersion dominated, but the two narrow components complicate the applicability of this criterion as in the case of the host of GRB 020903.The double component could be an indication for a merger and we do not have a clear velocity field, also M is much higher than M due to the “artificial” increase in width through the double component. However, even a double emission component is not necessarily related to a merger event, as can be seen in the host of GRB 100418A (de Ugarte Postigo et al., 2018) where the two emission components have very similar abundances, disfavoring a merger event.
Even if found in seeming isolation, many dwarf starbursts turn out to have complicated velocity fields and signs of past mergers. Pérez-Gallego et al. (2011) and Blasco-Herrera et al. (2013) studied several tens of local starbursts, many of them BCDs, and found complicated velocity fields and signs for past or ongoing interactions in 60% of cases. Neighbors may easily go undetected due to low luminosity or lack of data. Only a few GRB hosts have confirmed neighbors that could be candidates for interactions (Savaglio et al., 2012; Thöne et al., 2011; Schulze et al., 2014; Arabsalmani et al., 2019), however a more detailed study of this issue using the large FOV of the MUSE IFU spectrograph is currently being done (Schulze et al. in prep.), which might add several more candidate interacting systems. In particular HI observations are well suited to trace past interactions and disturbances in the gas as shown by Arabsalmani et al. (2019) for the host of GRB 980425. Ashley et al. (2017) found that almost all BCDs show an irregular HI velocity field with holes and parts showing different rotation axes. Unfortunately, none of the GRB hosts presented here have been observed in HI as their distances make it difficult to obtain any signal with current instrumentation.
6.2 GRB location
In all galaxies of our sample, the location of the GRB is in the vicinity but never exactly at the position of the region with the highest H flux and hence the highest SFR. Fruchter et al. (2006b) and subsequent studies by e.g. Kelly et al. (2008); Svensson et al. (2010) concluded from high spatial resolution HST imaging that GRBs are located in the brightest regions in their hosts and have a higher correlation with the brightest and bluest regions than any other SN type (which, however, does not mean that the GRB has to be exactly on the brightest pixels in the host). In Tab. 1 we list the distance of the GRB from the brightest pixel in the H map. This is somewhat different from other studies which were taking the brightest pixel (usually using HST imaging), but in different imaging filters. Blanchard et al. (2016) use whatever filter the galaxy was observed in while Lyman et al. (2017) use the IR filter F160W, centered at 15,400 Å. Some of the filters used include H at the corresponding redshift and are likely dominated by that line, but others probe only continuum emission.
The average distance using the only 6 galaxies in our sample is kpc and is dominated by the large physical offset in the host of GRB 050826. If we leave out this value we get an average distance from the brightest H spaxel of kpc. This value is in agreement with the distances reported in Bloom et al. (2002); Blanchard et al. (2016); Lyman et al. (2017), who found values of , and kpc in the respective works. In all the sample, the GRB location is clearly inside the 80% light radius of each galaxy (see Tab. 1). Normalizing this to the galaxy r gives an average of . Lyman et al. (2017) found an average normalized distance compared to r of only 0.3 (comprising hosts up to ), while a recent work comparing GRB hosts and those with BL-Ic SNe without GRB (Japelj et al., 2018) found a normalized distance d/r of 0.98 for the GRB host sample (which partly comprises our sample presented here).
Hammer et al. (2006) measured the distance of the GRB location from the brightest pixel in the nearest SF region for the hosts of GRB 980425 and GRB 020903, not the brightest pixel in the entire host, which in both cases is at a different location. GRB 980425 was 800 pc from a bright region showing WR features, however, high resolution HST imaging reveals that the GRB site itself is a smaller SF region in a complex of SF regions South-East from that WR region (Fynbo et al., 2000). For GRB 020903, Hammer et al. (2006) found a distance of 450 pc from the brightest pixel of the SF region that we named “GRB region” in this work, however, due to the high redshift, this region cannot be resolved into individual HII regions. The authors concluded that GRB progenitors might be runaway stars originating from the most extreme HII regions. However, while none of the GRB sites is associated with the very brightest region in their respective host, they seem to always be associated directly with some SF region and a bright star cluster. This might tell us that either the conditions for forming a GRB progenitor are more important than a highly luminous SF region hosting a large number of massive stars or that GRB progenitors do not necessarily have to come from the most massive stars.
Another interesting fact is that generally, the GRB location seems to be in a region of the galaxy with low line width (see Fig. 3). This is a common pattern in dwarf galaxies, where regions of larger dispersion lie usually in between HII regions (see e.g. Cairós & González-Pérez, 2017, and references therein). In the three hosts where we resolve a broad component in different regions (see discussion below), the region of the GRB does clearly show an underlying broad component but is usually not the region with the highest (relative) luminosity of the broad component. In the host of GRB 100316D, the region with the highest absolute and relative flux of the broad component is the SF region West of the GRB site. For the host of GRB 031203, the absolute flux of the broad component is in fact highest in the GRB region, but the relative contribution is larger in what we call the “broad region”. Note that in both cases, the size of the regions compared comprises the same number of spaxels. While the SF region of the GRB site itself does not seem to be associated with turbulent SF, the broad component is always very strong at the GRB site. In some cases, geometric effects might play a role in determining the actual observed strength of the broad component.
6.3 The broad emission component and evidences for a star-burst wind
For all hosts in our sample we find an underlying broad component or excess emission in the wings of H. The broad and narrow components seem to be kinematically detached from each other with the narrow component following (and determining) the general velocity field while the position of the broad component does not change significantly. In most of our hosts, the broad emission component is blueshifted compared to the narrow component, which provides further indication for an outflow. In GRB 100316D where we were able to make a 2D map of both components, the broad component is clearly concentrated in the most intense SF region in the galaxy. Furthermore, the broad component seems to be more metal rich than the narrow component, at least in the two hosts (GRB 031203 and GRB 100316D) where we were able to fit those components separately in both H and a metal transition. This make the association of the broad component with an inflow less likely as pristine gas from the IGM should have lower metallicity.
Three galaxies in our host sample have several emission components in addition to the broad component. The hosts of GRB 020903 as well as GRB 050826 show a clear double narrow component on top of a broader component. While for GRB 020903, those components can be associated with different parts in the host, the situation is less clear for GRB 050826. However, even compact looking galaxies can have a complicated velocity field simply due to the orientation of the galaxy as we showed e.g. for the host of GRB 100418A (de Ugarte Postigo et al., 2018). The hosts of GRB 030329 and GRB 060218, the smallest GRB hosts currently known, only show a weak broad component or excess emission. For GRB 030329 we fit a broad component, but the fit is not very good as the excess emission in the blue and red wing have a slightly different shape and are more pronounced in the blue wing. Also for the host of GRB 050826 two extra components in both wings instead of a broad component could be a better fit. Very similar looking excess emission in both wings of H has been observed in a sample of nearby extremely metal poor (XMP) dwarfs (Olmo-García et al., 2017) and has been explained with expanding shells that can be more or less symmetric and/or affected by differential dust extinction.
We observe the highest velocities for the broad component in the host of GRB 031203, a compact but more massive galaxy. The emission spans at least 300–400 km s with possible additional components stretching over a total of 700 km s. This clearly goes beyond a normal rotation field of a galaxy of this size and mass which should be of the order of 150 km s (see calculations in Sect. 6.1 above). Also here the broad component is blue-shifted and more metal rich, making the most clear outflow case in all our sample. Furthermore, this outflow seems to be asymmetric and/or connected to the main SF region in the galaxy as it is strongest in the central parts and weaker in the “tail” of the host to the West. Hence the outflow probably originates from the brightest parts of the galaxy and is either being blocked in some parts or intrinsically not spherically symmetric. The excess emission component in the red wing is curious and seems to be stronger in the Western part of the galaxy relative to the main emission component but is visible throughout the host. This might be an additional outflow component or shell or it could be connected to the radio emission from HI detected in Michałowski et al. (2015), which the authors interpret as infalling, metal poor gas.
GRB 060218 and GRB 030329 are the only hosts with absorption lines detected and resolved into different components as probed in the spectrum of the GRB-SN (Wiersema et al., 2007; Thöne et al., 2007). However, while the host of GRB 030329 shows a large spread in velocity of the absorbing gas in Mg I and Mg II, clearly superseding the velocities from pure rotation of this dwarf galaxy (Thöne et al., 2007), the host of GRB 060218 does not show anything similar. For GRB 030329 we concluded that the absorption components might be related to a galactic wind from the main and possibly only SF region in the host. The host of GRB 060218 is similar to the one of GRB 030329 in size and metallicity and the fact that both probably consist of a single large star-forming region. Galactic winds have been regularly detected also in NaD absorption where it is in fact the main line tracing those winds in the low redshift Universe due to its convenient wavelength (see e.g. Martin, 2006; Veilleux et al., 2005). It is possible that the outflow cones of winds have rather small opening angles, hence for GRB 030329 the wind would point towards us while for GRB 060218 we see the galaxy perpendicular to the wind, or at least outside of the wind cone. The fact that for the host of GRB 030329 we hardly see any velocity field while for the even smaller galaxy of GRB 060218 we do recognize some rotation curve could point to a different orientation of these two galaxies towards us, hence we would be seeing GRB 030329 face on while GRB 060218 is more inclined.
Broad components in emission have been observed in a number of starburst galaxies. The inner regions of M82, the possibly most famous starburst galaxy, show a very similar combination of a narrow and a broad profile, which can be traced out to large distances, and some regions at the base of the wind show double profiles associated to expanding shells (Westmoquette et al., 2009b, a). BCDs studied at high angular resolution frequently show broad components, e.g. Haro 14 (Cairós & González-Pérez, 2017), Haro 11 (Östlin et al., 2015), NGC 4449 (Kumari et al., 2017), UM448 (James et al., 2013), Mrk996 (James et al., 2009) or NGC 1569 (Westmoquette et al., 2007). Haro 11 (Östlin et al., 2015) shows a triple component in [S iii] of which one has a 90km s but which follows the velocity field of the stellar component and might be just a superposition of several narrow, unresolved emission lines and not the product of an outflow. Westmoquette et al. (2007) study the center of the starburst NGC 1569 and conclude the broad component originates from the interaction between a strong stellar wind and cold gas knots, producing a turbulent mixing layer on the surface which powers the observed starburst wind on global scales. Outflows in absorption have been detected for the vast majority of starburst galaxies with fractions ranging from 75% (Chisholm et al., 2015) to 90% for face-on galaxies (Heckman et al., 2015). An example of a study of winds in both absorption and emission is NGC 7552 (Wood et al., 2015), a face-on spiral galaxy. They detect blue-shifted broad emission components ( up to 300 km s) and blue-shifted absorption components up to 1000 km s, similar to what has been observed in GRB 030329 albeit with larger velocities.
At higher redshifts (Amorín et al., 2012) find broad components in in their entire sample of six GP galaxies at redshifts of in different emission lines. They associate these components with stellar winds or the expansion of supernova remnants and exclude possibilities such as turbulent mixing layers as those would not occur across the entire galaxy and are usually only observed in the Balmer lines but not in forbidden lines (as it is also the case in our FLAMES sample). Their components are somewhat broader with typical FWHM of 100–250 km s and full width zero intensities of 650 km s. IFU spectra of z0.2 GPs (Lofthouse et al., 2017) also show evidence for a broad component with FWHM 200 km s but at lower significance than the sample of Amorín et al. (2012). GPs are considered extremely young and very compact starbursts with a large number of WR stars, which could explain the larger kinetic energy observed in the outflowing gas. GPs also show outflows in UV rest frame absorption lines and many are Lyman Alpha Emitters with double-peaked Ly profiles, making them good analogues for high redshift galaxies responsible for the escape of Ly photons (see e.g. Henry et al., 2015; Yang et al., 2017). Broad-narrow line profiles have also been detected in massive star-forming galaxies at z2 in the SINS survey (see e.g. Genzel et al., 2011; Newman et al., 2012; Davies et al., 2019). There the usually blue-shifted broad profiles are associated with the brightest regions/clusters in their hosts and attributed to powerful winds in which the outflow rate can even supersede the SFR, hence quenching its own SF rather efficiently
Winds have also been made responsible for abundances differences in BDCs. In contrast to spiral galaxies, BCDs do not necessarily have a negative metallicity gradient. An interesting example is NGC 4449, where the central SF region is found to be more metal poor than SF regions in the outskirts of the galaxy, something which has also been found in other dwarf starbursts (Sánchez Almeida et al., 2015; Elmegreen et al., 2016). Kumari et al. (2017) propose as a possible explanation an outflow of metal rich gas, which acts stronger in the region with the highest SFR than in the outskirts, but also the inflow of metal-poor gas or the formation of the outer SF regions from pre-enriched gas during a merger event are viable explanations. James et al. (2009) found a large enhancement of N/O and nitrogen abundance in Mrk996 in the broad emission component while the narrow component shows a normal N/O ratio. The broad emission region was also shown to host a large number of WR stars. The opposite is the case in UM448 (James et al., 2013) where WR features (albeit rather faint) are associated with a region showing a lower fraction of the broad component but an enhancement in N/H an N/O, however they believe that WR stars alone cannot be responsible for the enhancement and suggest an inflow of metal-poor gas. Clearly, different galaxies are influenced by a varying interplay between in- and outflows, which are hard to study beyond the local Universe.
6.4 Correlations between components and galaxy properties
We search for possible correlations between the broad components and the properties of the host (stellar mass, luminosity, metallicity and SFR, see Tabs. 1, 2 and 3 and Fig. 18). As an additional value we also derive the maximum outflow velocity of the gas which is defined as V = abs(0.5 FWHM) with being the velocity difference between the broad and the narrow component (see Tab. 3). For the hosts of GRBs 020903, 050826 and 060218 we use between the bluest narrow and the broad component, for the host of GRB 030329 we use the blue shifted additional component. We only find correlations between the SFR and different measures of the broad component, namely with 1) the FWHM of the broad component and 2) the flux ratio between the broad and the narrow component(s) and 3) the maximum velocity of the gas V. The only outlier from correlation 3) is GRB 020903 where the position of the broad component relative to the bluer narrow component shifts between negative and positive velocities. The Pearson’s coefficient for all those correlations is good with values of 0.94, 0.96 and 0.94 for correlation 1), 2) and 3) respectively while the (non-) correlation between the FWHM and SFR or the stellar mass yield coefficients of 0.7.
Our results show that, generally, hosts with broader components, higher relative flux of the broad component, and higher V have also higher SFRs (see Fig. 18). In the host with the highest SFR, GRB 031203, the flux of the broad component is higher than the one in the narrow component by a factor of around 1.5, hence the contribution of the outflowing material is considerable compared to the emission from the bulk of the gas in the galaxy. We do not find a correlation between log M* and the width of the broad component (see Fig. 18) but we do find some correlation between the line with fitting a single Gaussian and log M* (values listed in Tab. 2 as ) with a Pearson’s coefficient of 0.88. The latter confirms the widely established relation “stellar mass Tully-Fisher” (sTF) M*- relation. The fact that the broad component does not correlate with the stellar mass but the narrow component does implies that the broad component is not related to the general velocity field of the gas in the galaxy. In fact, for GRB 100316D, where we can trace the velocity field of both components, only the narrow component is related to the rotation of the gas while the broad component has a more complicated velocity pattern.
Arabsalmani et al. (2018) find a correlation between GRB host metallicity and the width of both emission lines () and absorption lines (v) with the relation being tighter using emission lines. The velocities from absorption and emission of the same galaxy marginally agree after correcting for the intrinsic difference, being larger in absorption than in emission. However, their sample is small and the outliers are candidates for interacting systems or hosts with additional absorption from a satellite in the line-of-sight, but they also mention winds as a possible explanation. The emission line fits in their study are from a single Gaussian fit and thus rather reflect the velocity width of the narrow components in our sample than the one of the broad components, hence our results are not in disagreement. The velocity-metallicity relation is consistent with the one found for QSO-DLAs but GRB-DLAs tend to have higher v (Arabsalmani et al., 2015), however when deriving the metallicity of the host center for both samples using impact parameters and typical metallicity gradients there is an offset between the two samples and they speculate that the effect stems from sampling different sightlines. It is known that the velocity width in both samples are at least partially due to winds, which might imply a similar effect and strength of those winds in both samples to result in similar v-Z and M-Z relations.
Many studies have tried to establish scaling relations between winds traced by (mostly UV) absorption lines and SFR or stellar mass in star-forming galaxies. Most of these find a correlation between SFR, SFR or SFR/M* and outflow velocity (measured as or V) but with different slopes of the correlation (see e.g. Erb et al., 2012; Arribas et al., 2014; Chisholm et al., 2015; Heckman & Borthakur, 2016, and references therein). Heckman (2003) suggests that there is a minimum SFR density of 0.1 M/yr/kpc needed to launch an actual outflow from a rotating disk. Tanner et al. (2017) model a starburst wind and produce synthetic absorption lines to investigate possible relations between the wind speed and the SFR or SFR. They find that there is only a correlation for warm and cold gas below a certain threshold in SFR and SFR and that the velocities measured from that gas (e.g. via UV absorption lines) can be much lower than the actual wind velocity traced by the hot gas. This also explains the discrepancies in the slope of the correlation from different studies using absorption lines as they trace different SFRs above and below the threshold as well as different ions. Maseda et al. (2014) do not find a correlation between the stellar mass and (as measured from a single Gaussian fit) in a sample of extreme emission line galaxies (EELGs) at z1–2, however, their mass range is rather small. Those galaxies have a around 50 km s with some reaching up to 200 km s, values much higher than expected for their mass, but they do not detect a broad component as Amorín et al. (2012) for a sample of local GPs (which are nothing else than EELGs at a certain redshift). Maseda et al. (2014) conclude that the gas fraction must be much higher for those galaxies in order to be stable and part of the gas rapidly gets turned into stars before the starburst shuts down again. This process might be supported by winds and outflows of gas.
Surprisingly, most studies find a strong correlation with SFR but not necessarily the absolute SFR. GRB hosts also defy the criteria for a minimum SFR needed for outflows in other types of galaxies. Since the SFR density in contrast to the absolute SFR slightly depends on the inclination angle, an incorrect inclination could change the values slightly, but this does likely not have a significant impact on the result. Both SFR and SFR depend on the extinction applied, however, none of the hosts presented here have a large extinction, and also this would shift both values equally. Outflows from GRB hosts do seem to behave somewhat differently from those of other samples. If we compare the FLAMES sample to a large sample using MANGA data (Rodríguez del Pino et al., 2019), our galaxies have much higher relative strengths of the broad vs. narrow component but have lower outflow speeds V and FWHM of the broad component. Also, at the low masses of our sample galaxies, the detection rate of outflows should be rather low and not 100% (Rodríguez del Pino et al., 2019).
On the other hand, the low masses make it likely for the material to actually escape the galaxy. Using the dynamical mass of the galaxies derived in Sect. 6.1 (see also Tab. 2), we can derive escape velocities: taking a ratio of r/r of 10 and a radius of 3 kpc (Arribas et al., 2014). Escape velocities at low masses are rather small with log M, 9.0 and 10.0 having V of 20, 55 and 170 km s. For the hosts of GRB 030329 and GRB 031203 VV with the latter reaching a factor of V/V2.5, the highest in the sample. For GRBs 020903 and 100316D the ratio is 1, note, however that for GRB 020903 we used the v between the bluer narrow component and the broad component. The gas in the highest mass host, GRB 050826 does not reach escape velocity and, surprisingly, neither in the smallest host, GRB 060218. Rodríguez del Pino et al. (2019) and Arribas et al. (2014) concluded that beyond log M10.4, it is very rare for gas to reach escape velocities.
6.5 The perils of low spatial resolution
Our study also visualizes some of the complications of doing kinematic studies with poorly resolved and/or longslit data. In several galaxies of our sample, the global host spectra do not reveal the same amount of components or with much less significance than the integrated spectra of individual HII regions or parts of the galaxy. In the case of the host of GRB 100316D it even introduced an erroneous position of the broad component simply by superposition of different regions across the velocity field with different relative strength such that the integrated line resembles a different (and erroneous) profile. The main issue is the fact that the broad component changes very little in velocity while the narrow component follows the rotation field of the galaxy. Moiseev & Lozinskaya (2012) also analyzed the effect of decreasing resolution on IFU data of very nearby starbursts on the intensity and of different SF regions. As expected, peaks in flux get smoothed out and the range of extreme values decreases, however, the median value remains constant. Also, regions with high at the edge of HII regions blend together to larger structures of seemingly high .
One has to be especially careful with the orientation of longslit data since it makes a large difference whether the slit was oriented along or perpendicular (or any other angle) to the velocity field. In the first case, if the spectra along the slit are collapsed to a single 1D spectrum, broad components can get completely buried in the broadening of the narrow component due to the shift in velocity along the slit. If the orientation and velocity of the galaxy is unknown or unresolved, the interpretation is especially tricky and might lead to wrong conclusions on the presence/absence of different kinematic components. Another difficult case are hosts with double or multiple narrow components, possibly due to an ongoing interaction. Since the individual sub-components can also follow their own velocity fields, in the case that those double components are not separated very far in velocity, the blending together with the velocity change can mimic a velocity field of a normal rotating disk. In most cases, however, this should be relatively easy to disentangle as the pattern usually results somewhat different from the one of a rotating disk.
This paper presents the first spatially resolved data of low-redshift () long duration GRB hosts at high spectral resolution. It is likely also the highest redshift sample for which different kinematical components have be spatially resolved. We studied the kinematics of the H emission line as well as the abundances using other detected emission lines. Our main findings are the following:
Low redshift GRB hosts mostly belong to the class of BCDs and dIrrs with properties similar to the general population of these classes.
Only two out of six galaxies show a regular velocity field that would come from a rotating disk. The two most compact hosts are clearly dispersion dominated with very little rotation. The remaining two have a double narrow component which could point to systems that are in the process of merging.
The GRB location is very often close to but not on the brightest region in the host and not at the position of the highest line width. This might be an indication for kicks from a binary companion that has exploded as a SN previously or some factor related to the location of very massive star-formation inside a SF region.
All galaxies have underlying broad components with a FWHM of 200–300 km s and in most cases these are blueshifted compared to the main emission peak. We interpret this as the broad components being connected to outflows due to winds from young stars and/or possibly the first supernova explosions. For the lower mass hosts of our sample, the velocities of the wind (V) are higher than the escape velocity and hence the gas is actually capable of leaving the galaxy.
For GRB 030329, there are independent evidences for a metal-rich outflow from absorption lines, but GRB 060218, a very similar galaxy, does not show absorption lines at high velocities. Unfortunately, at low redshift we often lack absorption line spectra from the GRB itself, while at high- those absorption components stretching over several 100 km s are detected frequently.
The strength of this broad component changes across the galaxy but the largest contribution or width is not necessarily associated twith the region of highest SFR. On the other hand, the velocity does not follow the general rotation field of the galaxy but its centroid stays rather constant across the host.
The broad, possible wind components, are slightly more metal rich in the two hosts where we could perform a double component fit in H and another metal line. This is another possible indication that the broad component might be an outflow transporting enriched gas away from the galaxy via winds or SN explosions.
We find a correlation between the SFR of a host and 1) the width of the broad component, 2) the relative strength compared to the narrow component and 3) the maximum gas velocity V, but no correlation with the stellar mass or the SFR density of the host, the latter of which is usually correlated to the wind component in SF galaxies. In GRB hosts the broad component seems to be correlated with the total amount of current SF in the host and is different from the properties of the general ISM in the galaxy.
Detailed kinematics crucially depend on spectral and spatial resolution. Long-slit data or spatial integration over too large regions can result in erroneous line profiles, especially if the integration is done across the velocity field of the galaxy or in case of multiple emission line components close in velocity space.
Since this is the first attempt to resolve kinematics of GRB hosts, a larger sample with higher S/N and higher spectral resolution would be very much desired. It would also be interesting to compare our results of GRB hosts with those of other stellar explosions, e.g. SLSNe, which are an even more extreme version of GRB hosts with lower metallicities and higher SF rates (see e.g. Leloudas et al., 2015). Another unique feature of GRBs is that, at least in a narrow redshift range, we are able to probe both emission line kinematics from the same host galaxy, something which is very expensive to do e.g. for quasar intervening systems. The study of GRB host kinematics is still in its infancy but can give us important indications on the processes of SF in hosts of stellar explosions and allows for another aspect of comparison to field galaxies.
Acknowledgements.CT thanks D. A. Kann for a careful read and correction of the manuscript. CT, LI and AdUP acknowledge support from AYA2014-58381P and AYA2017-89384-P, CT and AdUP also from a Ramón y Cajal fellowships RyC-2012-09984 and RyC-2012-09975, LI from a Juan de la Cierva Integración fellowship IJCI-2016-30940. SDV acknowledges support from the French National Research Agency (ANR) under contract ANR-16-CE31-0003. LC is supported by YDUN grant DFF 4090-00079. Ground based observations were collected at the VLT under program 092.D-0389(A).
- Amorín et al. (2012) Amorín, R., Vílchez, J. M., Hägele, G. F., et al. 2012, ApJ, 754, L22
- Arabsalmani et al. (2018) Arabsalmani, M., Møller, P., Perley, D. A., et al. 2018, MNRAS, 473, 3312
- Arabsalmani et al. (2019) Arabsalmani, M., Roychowdhury, S., Starkenburg, T. K., et al. 2019, MNRAS[\eprint[arXiv]1903.00485]
- Arabsalmani et al. (2015) Arabsalmani, M., Roychowdhury, S., Zwaan, M. A., Kanekar, N., & Michałowski, M. J. 2015, MNRAS, 454, L51
- Arribas et al. (2014) Arribas, S., Colina, L., Bellocchi, E., Maiolino, R., & Villar-Martín, M. 2014, A&A, 568, A14
- Ashley et al. (2013) Ashley, T., Simpson, C. E., & Elmegreen, B. G. 2013, AJ, 146, 42
- Ashley et al. (2017) Ashley, T., Simpson, C. E., Elmegreen, B. G., Johnson, M., & Pokhrel, N. R. 2017, AJ, 153, 132
- Bekki (2008) Bekki, K. 2008, MNRAS, 388, L10
- Bellocchi et al. (2013) Bellocchi, E., Arribas, S., Colina, L., & Miralles-Caballero, D. 2013, A&A, 557, A59
- Bersier et al. (2006) Bersier, D., Fruchter, A. S., Strolger, L.-G., et al. 2006, ApJ, 643, 284
- Blanchard et al. (2016) Blanchard, P. K., Berger, E., & Fong, W.-f. 2016, ApJ, 817, 144
- Blasco-Herrera et al. (2013) Blasco-Herrera, J., Fathi, K., Östlin, G., Font, J., & Beckman, J. E. 2013, MNRAS, 435, 1958
- Bloom et al. (2002) Bloom, J. S., Kulkarni, S. R., & Djorgovski, S. G. 2002, AJ, 123, 1111
- Cairós et al. (2015) Cairós, L. M., Caon, N., & Weilbacher, P. M. 2015, A&A, 577, A21
- Cairós & González-Pérez (2017) Cairós, L. M. & González-Pérez, J. N. 2017, A&A, 608, A119
- Campana et al. (2006) Campana, S., Mangano, V., Blustin, A. J., et al. 2006, Nature, 442, 1008
- Cano et al. (2017a) Cano, Z., Izzo, L., de Ugarte Postigo, A., et al. 2017a, A&A, 605, A107
- Cano et al. (2017b) Cano, Z., Wang, S.-Q., Dai, Z.-G., & Wu, X.-F. 2017b, Advances in Astronomy, 2017, 8929054
- Cardamone et al. (2009) Cardamone, C., Schawinski, K., Sarzi, M., et al. 2009, MNRAS, 399, 1191
- Chisholm et al. (2018) Chisholm, J., Tremonti, C., & Leitherer, C. 2018, MNRAS[\eprint[arXiv]1808.10453]
- Chisholm et al. (2015) Chisholm, J., Tremonti, C. A., Leitherer, C., et al. 2015, ApJ, 811, 149
- Christensen et al. (2008) Christensen, L., Vreeswijk, P. M., Sollerman, J., et al. 2008, A&A, 490, 45
- Conselice et al. (2005) Conselice, C. J., Vreeswijk, P. M., Fruchter, A. S., et al. 2005, ApJ, 633, 29
- Coward et al. (2013) Coward, D. M., Howell, E. J., Branchesi, M., et al. 2013, MNRAS, 432, 2141
- Davies et al. (2019) Davies, R. L., Förster Schreiber, N. M., Übler, H., et al. 2019, ApJ, 873, 122
- de Ugarte Postigo et al. (2018) de Ugarte Postigo, A., Thöne, C. C., Bensch, K., et al. 2018, A&A, 620, A190
- de Ugarte Postigo et al. (2014) de Ugarte Postigo, A., Thöne, C. C., Rowlinson, A., et al. 2014, A&A, 563, A62
- Dimeo (2005) Dimeo, R. 2005, available at http://www.ncnr.nist.gov/staff/dimeo/panweb/pan.html
- Elmegreen & Hunter (2015) Elmegreen, B. G. & Hunter, D. A. 2015, ApJ, 805, 145
- Elmegreen et al. (2016) Elmegreen, D. M., Elmegreen, B. G., Sánchez Almeida, J., et al. 2016, ApJ, 825, 145
- Erb et al. (2012) Erb, D. K., Quider, A. M., Henry, A. L., & Martin, C. L. 2012, ApJ, 759, 26
- Flores et al. (2006) Flores, H., Hammer, F., Puech, M., Amram, P., & Balkowski, C. 2006, A&A, 455, 107
- Fruchter et al. (2006a) Fruchter, A. S., Levan, A. J., Strolger, L., et al. 2006a, Nature, 441, 463
- Fruchter et al. (2006b) Fruchter, A. S., Levan, A. J., Strolger, L., et al. 2006b, Nature, 441, 463
- Fynbo et al. (2000) Fynbo, J. U., Holland, S., Andersen, M. I., et al. 2000, ApJ, 542, L89
- Genzel et al. (2011) Genzel, R., Newman, S., Jones, T., et al. 2011, ApJ, 733, 101
- Guseva et al. (2011) Guseva, N. G., Izotov, Y. I., Fricke, K. J., & Henkel, C. 2011, A&A, 534, A84
- Halpern (2005) Halpern, J. P. 2005, GRB Coordinates Network, 3891
- Halpern & Mirabal (2006) Halpern, J. P. & Mirabal, N. 2006, GRB Coordinates Network, 5982
- Hammer et al. (2006) Hammer, F., Flores, H., Schaerer, D., et al. 2006, A&A, 454, 103
- Han et al. (2010) Han, X. H., Hammer, F., Liang, Y. C., et al. 2010, A&A, 514, A24
- Heckman (2003) Heckman, T. M. 2003, in Revista Mexicana de Astronomia y Astrofisica, vol. 27, Vol. 17, Revista Mexicana de Astronomia y Astrofisica Conference Series, ed. V. Avila-Reese, C. Firmani, C. S. Frenk, & C. Allen, 47–55
- Heckman et al. (2015) Heckman, T. M., Alexandroff, R. M., Borthakur, S., Overzier, R., & Leitherer, C. 2015, ApJ, 809, 147
- Heckman & Borthakur (2016) Heckman, T. M. & Borthakur, S. 2016, ApJ, 822, 9
- Henry et al. (2015) Henry, A., Scarlata, C., Martin, C. L., & Erb, D. 2015, ApJ, 809, 19
- Herenz et al. (2016) Herenz, E. C., Gruyters, P., Orlitova, I., et al. 2016, A&A, 587, A78
- Hjorth et al. (2003) Hjorth, J., Sollerman, J., Møller, P., et al. 2003, Nature, 423, 847
- Hunt et al. (2014) Hunt, L. K., Palazzi, E., Michałowski, M. J., et al. 2014, A&A, 565, A112
- Hunter et al. (2012) Hunter, D. A., Ficut-Vicas, D., Ashley, T., et al. 2012, AJ, 144, 134
- Izotov et al. (2007) Izotov, Y. I., Thuan, T. X., & Guseva, N. G. 2007, ApJ, 671, 1297
- Izzo et al. (2019) Izzo, L., de Ugarte Postigo, A., Maeda, K., et al. 2019, Nature, 565, 324
- Izzo et al. (2017) Izzo, L., Thöne, C. C., Schulze, S., et al. 2017, MNRAS, 472, 4480
- Jakobsson et al. (2005) Jakobsson, P., Frail, D. A., Fox, D. B., et al. 2005, ApJ, 629, 45
- James et al. (2010) James, B. L., Tsamis, Y. G., & Barlow, M. J. 2010, MNRAS, 401, 759
- James et al. (2013) James, B. L., Tsamis, Y. G., Barlow, M. J., Walsh, J. R., & Westmoquette, M. S. 2013, MNRAS, 428, 86
- James et al. (2009) James, B. L., Tsamis, Y. G., Barlow, M. J., et al. 2009, MNRAS, 398, 2
- Japelj et al. (2018) Japelj, J., Vergani, S. D., Salvaterra, R., et al. 2018, A&A, 617, A105
- Johnson et al. (2012) Johnson, M., Hunter, D. A., Oh, S.-H., et al. 2012, AJ, 144, 152
- Kann et al. (2010) Kann, D. A., Klose, S., Zhang, B., et al. 2010, ApJ, 720, 1513
- Kelly et al. (2008) Kelly, P. L., Kirshner, R. P., & Pahre, M. 2008, ApJ, 687, 1201
- Kewley et al. (2007) Kewley, L. J., Brown, W. R., Geller, M. J., Kenyon, S. J., & Kurtz, M. J. 2007, AJ, 133, 882
- Koleva et al. (2014) Koleva, M., De Rijcke, S., Zeilinger, W. W., et al. 2014, MNRAS, 441, 452
- Krühler et al. (2017) Krühler, T., Kuncarayakti, H., Schady, P., et al. 2017, A&A, 602, A85
- Kumari et al. (2017) Kumari, N., James, B. L., & Irwin, M. J. 2017, MNRAS, 470, 4618
- Lee et al. (2009) Lee, J. C., Kennicutt, Jr., R. C., Funes, S. J. J. G., Sakai, S., & Akiyama, S. 2009, ApJ, 692, 1305
- Leloudas et al. (2015) Leloudas, G., Schulze, S., Krühler, T., et al. 2015, MNRAS, 449, 917
- Levesque et al. (2010a) Levesque, E. M., Berger, E., Kewley, L. J., & Bagley, M. M. 2010a, AJ, 139, 694
- Levesque et al. (2011) Levesque, E. M., Berger, E., Soderberg, A. M., & Chornock, R. 2011, ApJ, 739, 23
- Levesque et al. (2010b) Levesque, E. M., Kewley, L. J., Berger, E., & Zahid, H. J. 2010b, AJ, 140, 1557
- Levesque et al. (2010c) Levesque, E. M., Kewley, L. J., Graham, J. F., & Fruchter, A. S. 2010c, ApJ, 712, L26
- Lofthouse et al. (2017) Lofthouse, E. K., Houghton, R. C. W., & Kaviraj, S. 2017, MNRAS, 471, 2311
- Lyman et al. (2017) Lyman, J. D., Levan, A. J., Tanvir, N. R., et al. 2017, MNRAS, 467, 1795
- Malesani et al. (2004) Malesani, D., Tagliaferri, G., Chincarini, G., et al. 2004, ApJ, 609, L5
- Mannucci et al. (2010) Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci, A. 2010, MNRAS, 408, 2115
- Marino et al. (2013) Marino, R. A., Rosales-Ortega, F. F., Sánchez, S. F., et al. 2013, A&A, 559, A114
- Martin (2006) Martin, C. L. 2006, ApJ, 647, 222
- Maseda et al. (2014) Maseda, M. V., van der Wel, A., Rix, H.-W., et al. 2014, ApJ, 791, 17
- Matheson et al. (2003) Matheson, T., Garnavich, P. M., Stanek, K. Z., et al. 2003, ApJ, 599, 394
- McQuinn et al. (2010) McQuinn, K. B. W., Skillman, E. D., Cannon, J. M., et al. 2010, ApJ, 724, 49
- Mesler & Pihlström (2013) Mesler, R. A. & Pihlström, Y. M. 2013, ApJ, 774, 77
- Michałowski et al. (2015) Michałowski, M. J., Gentile, G., Hjorth, J., et al. 2015, A&A, 582, A78
- Michałowski et al. (2012) Michałowski, M. J., Kamble, A., Hjorth, J., et al. 2012, ApJ, 755, 85
- MichałowskI et al. (2018) MichałowskI, M. J., Xu, D., Stevens, J., et al. 2018, A&A, 616, A169
- Mirabal et al. (2007) Mirabal, N., Halpern, J. P., & O’Brien, P. T. 2007, ApJ, 661, L127
- Moiseev & Lozinskaya (2012) Moiseev, A. V. & Lozinskaya, T. A. 2012, MNRAS, 423, 1831
- Newman et al. (2012) Newman, S. F., Shapiro Griffin, K., Genzel, R., et al. 2012, ApJ, 752, 111
- Nicholls et al. (2014) Nicholls, D. C., Dopita, M. A., Sutherland, R. S., et al. 2014, ApJ, 786, 155
- Olmo-García et al. (2017) Olmo-García, A., Sánchez Almeida, J., Muñoz-Tuñón, C., et al. 2017, ApJ, 834, 181
- Östlin et al. (2001) Östlin, G., Amram, P., Bergvall, N., et al. 2001, A&A, 374, 800
- Östlin et al. (2015) Östlin, G., Marquart, T., Cumming, R. J., et al. 2015, A&A, 583, A55
- Östlin et al. (2008) Östlin, G., Zackrisson, E., Sollerman, J., Mattila, S., & Hayes, M. 2008, MNRAS, 387, 1227
- Palmerio et al. (2019) Palmerio, J. T., Vergani, S. D., Salvaterra, R., et al. 2019, A&A, 623, A26
- Pérez-Gallego et al. (2011) Pérez-Gallego, J., Guzmán, R., Castillo-Morales, A., et al. 2011, MNRAS, 418, 2350
- Perley et al. (2017) Perley, D. A., Krühler, T., Schady, P., et al. 2017, MNRAS, 465, L89
- Perley & Taggart (2017) Perley, D. A. & Taggart, K. 2017, GRB Coordinates Network, Circular Service, No. 22194, #1 (2017), 22194
- Perley et al. (2016) Perley, D. A., Tanvir, N. R., Hjorth, J., et al. 2016, ApJ, 817, 8
- Planck Collaboration et al. (2018) Planck Collaboration, Aghanim, N., Akrami, Y., et al. 2018, ArXiv e-prints [\eprint[arXiv]1807.06209]
- Prochaska et al. (2004) Prochaska, J. X., Bloom, J. S., Chen, H.-W., et al. 2004, ApJ, 611, 200
- Rodríguez del Pino et al. (2019) Rodríguez del Pino, B., Arribas, S., Piqueras-López, J., Villar-Martín, M., & Colina, L. 2019, arXiv e-prints [\eprint[arXiv]1903.07432]
- Rupke (2018) Rupke, D. 2018, Galaxies, 6, 138
- Sakamoto et al. (2004) Sakamoto, T., Lamb, D. Q., Graziani, C., et al. 2004, ApJ, 602, 875
- Sánchez Almeida et al. (2015) Sánchez Almeida, J., Elmegreen, B. G., Muñoz-Tuñón, C., et al. 2015, ApJ, 810, L15
- Savaglio et al. (2012) Savaglio, S., Rau, A., Greiner, J., et al. 2012, MNRAS, 420, 627
- Schady et al. (2015) Schady, P., Krühler, T., Greiner, J., et al. 2015, A&A, 579, A126
- Schroetter et al. (2016) Schroetter, I., Bouché, N., Wendt, M., et al. 2016, ApJ, 833, 39
- Schulze et al. (2014) Schulze, S., Malesani, D., Cucchiara, A., et al. 2014, A&A, 566, A102
- Soderberg et al. (2004) Soderberg, A. M., Kulkarni, S. R., Berger, E., et al. 2004, ApJ, 606, 994
- Soderberg et al. (2005) Soderberg, A. M., Kulkarni, S. R., Fox, D. B., et al. 2005, ApJ, 627, 877
- Starling et al. (2011) Starling, R. L. C., Wiersema, K., Levan, A. J., et al. 2011, MNRAS, 411, 2792
- Svensson et al. (2010) Svensson, K. M., Levan, A. J., Tanvir, N. R., Fruchter, A. S., & Strolger, L.-G. 2010, MNRAS, 405, 57
- Swaters et al. (2009) Swaters, R. A., Sancisi, R., van Albada, T. S., & van der Hulst, J. M. 2009, A&A, 493, 871
- Symeonidis et al. (2014) Symeonidis, M., Oates, S. R., de Pasquale, M., et al. 2014, MNRAS, 443, L124
- Tanga et al. (2018) Tanga, M., Krühler, T., Schady, P., et al. 2018, A&A, 615, A136
- Tanner et al. (2017) Tanner, R., Cecil, G., & Heitsch, F. 2017, ApJ, 843, 137
- Telles et al. (2014) Telles, E., Thuan, T. X., Izotov, Y. I., & Carrasco, E. R. 2014, A&A, 561, A64
- Thöne et al. (2011) Thöne, C. C., Campana, S., Lazzati, D., et al. 2011, MNRAS, 414, 479
- Thöne et al. (2014) Thöne, C. C., Christensen, L., Prochaska, J. X., et al. 2014, MNRAS, 441, 2034
- Thöne et al. (2008) Thöne, C. C., Fynbo, J. P. U., Östlin, G., et al. 2008, ApJ, 676, 1151
- Thöne et al. (2007) Thöne, C. C., Greiner, J., Savaglio, S., & Jehin, E. 2007, ApJ, 671, 628
- van Zee et al. (1998) van Zee, L., Skillman, E. D., & Salzer, J. J. 1998, AJ, 116, 1186
- Vanderspek et al. (2003) Vanderspek, R., Crew, G., Doty, J., et al. 2003, GRB Coordinates Network, 1997
- Veilleux et al. (2005) Veilleux, S., Cecil, G., & Bland-Hawthorn, J. 2005, ARA&A, 43, 769
- Verbeke et al. (2014) Verbeke, R., De Rijcke, S., Koleva, M., et al. 2014, MNRAS, 442, 1830
- Wainwright et al. (2007) Wainwright, C., Berger, E., & Penprase, B. E. 2007, ApJ, 657, 367
- Watson et al. (2011a) Watson, D., French, J., Christensen, L., et al. 2011a, ApJ, 741, 58
- Watson et al. (2011b) Watson, D., French, J., Christensen, L., et al. 2011b, ApJ, 741, 58
- Westmoquette et al. (2007) Westmoquette, M. S., Exter, K. M., Smith, L. J., & Gallagher, J. S. 2007, MNRAS, 381, 894
- Westmoquette et al. (2009a) Westmoquette, M. S., Gallagher, J. S., Smith, L. J., et al. 2009a, ApJ, 706, 1571
- Westmoquette et al. (2009b) Westmoquette, M. S., Smith, L. J., Gallagher, III, J. S., et al. 2009b, ApJ, 696, 192
- Wiersema et al. (2007) Wiersema, K., Savaglio, S., Vreeswijk, P. M., et al. 2007, A&A, 464, 529
- Wiseman et al. (2017) Wiseman, P., Perley, D. A., Schady, P., et al. 2017, A&A, 607, A107
- Wood et al. (2015) Wood, C. M., Tremonti, C. A., Calzetti, D., et al. 2015, MNRAS, 452, 2712
- Woosley & Heger (2006) Woosley, S. E. & Heger, A. 2006, ApJ, 637, 914
- Yang et al. (2017) Yang, H., Malhotra, S., Gronke, M., et al. 2017, ApJ, 844, 171
- Yang et al. (2008) Yang, Y., Flores, H., Hammer, F., et al. 2008, A&A, 477, 789
- Yin et al. (2007) Yin, S. Y., Liang, Y. C., Hammer, F., et al. 2007, A&A, 462, 535
- Zabl et al. (2019) Zabl, J., Bouché, N. F., Schroetter, I., et al. 2019, MNRAS[\eprint[arXiv]1901.11416]
- Zhao et al. (2011) Zhao, Y., Gu, Q., & Gao, Y. 2011, AJ, 141, 68