The Lyman Alpha Reference Sample VII

A&A 587, A78 (2016) Astronomy DOI: 10.1051/0004-6361/201527373 & c ESO 2016 Astrophysics

The Lyman alpha reference sample VII. Spatially resolved Hα kinematics?,??

Edmund Christian Herenz1, Pieter Gruyters2, Ivana Orlitova3, Matthew Hayes4, Göran Östlin4, John M. Cannon5, Martin M. Roth1, Arjan Bik4, Stephen Pardy6, Héctor Otí-Floranes7, J. Miguel Mas-Hesse8, Angela Adamo4, Hakim Atek9, Florent Duval4,10, Lucia Guaita4,11, Daniel Kunth12, Peter Laursen13, Jens Melinder4, Johannes Puschnig4, Thøger E. Rivera-Thorsen4, Daniel Schaerer14,15, and Anne Verhamme14

1 Leibniz-Institut far Astrophysik Potsdam (AIP), An der Sternwarte 16, 14482 Potsdam, Germany e-mail: [email protected] 2 Lund Observatory, Box 43, 221 00 Lund, Sweden 3 Astronomical Institute, Academy of Sciences of the Czech Republic, Bocníˇ II 1401, 141 00 Prague, Czech Republic 4 Department of Astronomy, Oskar Klein Centre for Cosmoparticle Physics, Stockholm University, AlbaNova University Centre, 106 91 Stockholm, Sweden 5 Department of Physics & Astronomy, Macalester College, 1600 Grand Avenue, Saint Paul, MN 55105, USA 6 Department of Astronomy, University of Wisconsin, 475 North Charter Street, Madison, WI 53706, USA 7 CONACyT research fellow – Instituto de Radioastronomía y Astrofísica, UNAM, Campus Morelia, Michoacán, CP 58089, Mexico 8 Centro de Astrobiología (CSIC–INTA), Departamento de Astrofísica, PO Box 78, 28691 Villanueva de la Cañada, Spain 9 Laboratoire d’Astrophysique, École Polytechnique Fédérale de Lausanne (EPFL), Observatoire, 1290 Sauverny, Switzerland 10 Institute for Cosmic Ray Research, The University of Tokyo, 5-1-5 Kashiwanoha, Kashiwa, 277-8582 Chiba, Japan 11 INAF–Osservatorio Astronomico di Roma, via Frascati 33, 00040 Monteporzio, Italy 12 Institut d’Astrophysique de Paris, UMR 7095 CNRS & UPMC, 98bis Bd Arago, 75014 Paris, France 13 Dark Cosmology Centre, Niels Bohr Institute, University of Copenhagen, Juliane Maries Vej 30, 2100 Copenhagen, Denmark 14 Observatoire de Genève, Université de Genève, 51 Ch. des Maillettes, 1290 Versoix, Switzerland 15 CNRS, IRAP, 14 Avenue E. Belin, 31400 Toulouse, France

Received 15 September 2015 / Accepted 16 November 2015

ABSTRACT

We present integral field spectroscopic observations with the Potsdam Multi-Aperture Spectrophotometer of all 14 galaxies in the z ∼ 0.1 Lyman Alpha Reference Sample (LARS). We produce 2D line-of-sight velocity maps and velocity dispersion maps from the Balmer α (Hα) emission in our data cubes. These maps trace the spectral and spatial properties of the LARS galaxies’ intrinsic Lyα radiation field. We show our kinematic maps that are spatially registered onto the Hubble Space Telescope Hα and Lyman α (Lyα) images. We can conjecture a causal connection between spatially resolved Hα kinematics and Lyα photometry for individual galaxies, however, no general trend can be established for the whole sample. Furthermore, we compute the intrinsic velocity dispersion σ0, the shearing velocity vshear, and the vshear/σ0 ratio from our kinematic maps. In general LARS galaxies are characterised by high intrinsic −1 −1 velocity dispersions (54 km s median) and low shearing velocities (65 km s median). The vshear/σ0 values range from 0.5 to 3.2 with an average of 1.5. It is noteworthy that five galaxies of the sample are dispersion-dominated systems with vshear/σ0 < 1, and are thus kinematically similar to turbulent star-forming galaxies seen at high redshift. When linking our kinematical statistics to the global LARS Lyα properties, we find that dispersion-dominated systems show higher Lyα equivalent widths and higher Lyα escape fractions than systems with vshear/σ0 > 1. Our result indicates that turbulence in actively star-forming systems is causally connected to interstellar medium conditions that favour an escape of Lyα radiation. Key words. galaxies: ISM – galaxies: starburst – cosmology: observations – ultraviolet: galaxies – radiative transfer

1. Introduction equivalent width line provides enough contrast to be effectively singled out in specifically designed observational campaigns. As As already envisioned by Partridge & Peebles(1967), the hydro- of yet, mostly narrowband selection techniques have been em- gen Lyman α (Lyα, λLyα = 1215.67 Å) line has become a promi- ployed (e.g. Hu et al. 1998; Taniguchi et al. 2003; Malhotra nent target in successful systematic searches for galaxies in the & Rhoads 2004; Shimasaku et al. 2006; Tapken et al. 2006; early Universe. Redshifted into the optical, this narrow high Gronwall et al. 2007; Ouchi et al. 2008; Grove et al. 2009; ? Based on observations collected at the Centro Astronómico Shioya et al. 2009; Hayes et al. 2010; Ciardullo et al. 2012; Hispano Alemán (CAHA) at Calar Alto, operated jointly by the Max- Sandberg et al. 2015), but multi-object and integral-field spectro- Planck Institut für Astronomie and the Instituto de Astrofísica de scopic techniques are now also frequently used to deliver large Andalucía (CSIC). Lyman α emitter (LAE) samples (Cassata et al. 2011, 2015; ?? The reduced data cubes (FITS files) are only available at the CDS Adams et al. 2011; Mallery et al. 2012; Bacon et al. 2015). via anonymous ftp to cdsarc.u-strasbg.fr (130.79.128.5) or via Moreover, all galaxy redshift record holders in the last decade http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/587/A78

Article published by EDP Sciences A78, page 1 of 27 A&A 587, A78 (2016)

−1 were spectroscopically confirmed by virtue of their Lyα line (Iye (vexp . 100 km s ) or low neutral hydrogen column densities 19 −2 2011; Ono et al. 2012; Finkelstein et al. 2013; Oesch et al. 2015; NH  . 10 cm , while high expansion velocities (vexp ∼ 150 – Zitrin et al. 2015) and a bright LAE at z = 6.6 is even believed −1 300 km s ) with neutral hydrogen column densities of NH  & to contain a significant amount of stars made exclusively from 1020cm−2 produce the characteristic asymmetric profile with an primordial material (so called Pop-III stars; Sobral et al. 2015). extended red wing (Tapken et al. 2006, 2007; Verhamme et al. As important as Lyα radiation is in successfully unveiling 2008; Schaerer et al. 2011; Gronke & Dijkstra 2014). star formation processes in the early Universe, its correct inter- Further observational constraints on the kinematics of the pretation appears notoriously complicated. Resonant scatterings scattering medium can be obtained by measuring the offset in the interstellar and circum-galactic medium diffuse the intrin- from non-resonant, rest-frame optical emission lines (e.g. Hα sic Lyα radiation field in real and frequency space. These scat- or [O ]) to low-ionisation state metal absorption lines (e.g. terings increase the path length of Lyα photons within a galaxy O  λ1302 or Si  λ1304). While the emission lines provide sys- and consequently Lyα is more susceptible to being destroyed temic redshift, some of metal absorption lines in a low ionisa- by dust (see Dijkstra 2014, for a comprehensive review cover- tion state trace the kinematics of the cold neutral gas phase. ing the Lyα radiative transfer fundamentals). Consequently, a Both observables are challenging to obtain for high-z LAEs galaxy’s Lyα observables are influenced by a large number of and require long integration times on 8–10 m class telescopes its physical properties. Understanding these influences is crucial (e.g. Shapley et al. 2003; Tapken et al. 2004) or even additional to correctly interpret high-z LAE samples. In other words, we help from gravitational lenses (e.g. Schaerer & Verhamme 2008; have to answer the question: What differentiates an LAE from Christensen et al. 2012). As the continuum absorption often re- other star-forming galaxies that do not show Lyα in emission? mains undetected, just the offset between the rest-frame optical Regarding Lyα radiation transport in individual galaxies, a lines and the Lyα peaks are measured (e.g. McLinden et al. 2011; large body of theoretical work investigated analytically and nu- Guaita et al. 2013; Erb et al. 2014). Curiously, the observed Lyα merically how within simplified geometries certain parameters profiles also agree well with those predicted by the simple shell (e.g. density, temperature, dust content, kinematics, and clumpi- model, when the measured offsets (typically ∆v ∼ 200 km s−1) ness of the interstellar medium) affect the observed Lyα radi- are associated with shell expansion velocities in the simple shell ation field (e.g. Neufeld 1990; Ahn et al. 2003; Dijkstra et al. model (Verhamme et al. 2008; Hashimoto et al. 2013; Song et al. 2006; Verhamme et al. 2006; Laursen et al. 2013; Gronke & 2014; Hashimoto et al. 2015). Only a few profiles appear incom- Dijkstra 2014; Duval et al. 2014). Commonly a spherical shell patible with the expanding shells (Chonis et al. 2013). These model is adopted. In this model, a thin shell of expanding, con- profiles are characterised by extended wings or bumps in the tracting, or static neutral gas represents the medium responsible blue side of the profile (Martin et al. 2014; Henry et al. 2015). for scattering Lyα photons. As a result, from scattering in the However, given the aforementioned viewing angle dependencies shell complex Lyα line morphologies arise and the expansion in more complex scenarios, this overall success of the simple velocity and the neutral hydrogen column density of the shell shell model appears surprising and is therefore currently under are of pivotal importance in shaping the observable Lyα line. scrutiny (Gronke et al. 2015). Nevertheless, at least qualitatively By introducing deviations from pure spherical symmetry, Zheng the observations demonstrate, in concert with theoretical pre- & Wallace(2014) and Behrens et al.(2014) show that the ob- dictions, that LAEs predominantly have outflow kinematics and served Lyα properties also depend on the viewing angle under that such outflows promote the Lyα escape (see also Kunth et al. which a system is observed. The Behrens et al. result is also 1998; Mas-Hesse et al. 2003). found in large-scale cosmological simulations that were post- processed with Lyα radiative transport simulations (e.g. Laursen Kinematic information is moreover encoded in the rest- & Sommer-Larsen 2007; Laursen et al. 2009; Barnes et al. frame optical line emission alone. These emission lines trace 2011). Recently more realistic hydrodynamic simulations of iso- the ionised gas kinematics and, in particular, the hydrogen related galaxies have been paired with Lyα radiation transport sim- combination lines such as Hα relate directly to the spatial and ulations (Verhamme et al. 2012; Behrens & Braun 2014). These spectral properties of a galaxy’s intrinsic Lyα radiation field. studies again underline the viewing angle dependence of the Lyα Therefore spatially resolved spectroscopy of a galaxy’s Hα ra- observables. In particular, they show that disks observed face-on diation field constrains the initial conditions for the subsequent are expected to exhibit higher Lyα equivalent widths and Lyα Lyα radiative transfer through the interstellar, circum-galactic, escape fractions than if they were observed edge-on. More im- and intergalactic medium to the observer. At high z, however, portantly, these state-of-the-art simulations also emphasise the most LAEs are so compact that they cannot be spatially resolved importance of small-scale interstellar medium structure that was from the ground and, hence, all spatial information is lost in previously not included in simple models. For example Behrens the analysis of the integrated spectra. Resolving the intrinsic & Braun(2014) demonstrate how supernova-blown cavities are Lyα radiation of typical high-z LAEs spatially and spectrally able to produce favoured escape channels for Lyα photons. at such small scales would require integral field spectroscopy, Observationally, when Lyα is seen in emission, the spec- preferably with adaptive optics, in the near infrared with long tral line profiles can be typified by their distinctive shapes. In integration times. Although large samples of continuum-selected a large number of LAE spectra the Lyα line often appears asym- z ∼ 2−3 star-forming galaxies have already been observed with metric, with a relatively sharp drop on the blue and a more ex- this method (see Glazebrook 2013, for a comprehensive review), tended wing on the red side. A significant fraction of spectra little is known about the Lyα properties of the galaxies in those also shows characteristic double peaks, with the red peak of- samples. ten being stronger than the blue (e.g. Tapken et al. 2004, 2007; In this paper we present results obtained from our integral- Yamada et al. 2012; Hong et al. 2014; Henry et al. 2015; Yang field spectroscopic observations with the aim to relate spatially et al. 2015). Interestingly, a high percentage of double peaked and spectrally resolved intrinsic Lyα radiation field to its ob- and asymmetric Lyα profiles appear well explained by the spher- served Lyα properties. Therefore we targeted the Hα line in all ical symmetric scenarios mentioned above. Double-peaked pro- galaxies of the z ∼ 0.03–0.18 Lyman alpha reference sample files are successfully reproduced by slowly-expanding shells (LARS). The sample consists of 14 nearby laboratory galaxies,

A78, page 2 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics that have far-UV (FUV, λ ∼ 1500 Å) luminosities similar ESO338-IG04 (Hayes et al. 2005; Östlin et al. 2009; Sandberg to those of high-z star-forming galaxies. Moreover, to ensure et al. 2013) can be linked to outflows seen in the Hα radial ve- a strong intrinsic Lyα radiation field, galaxies with large Hα locity field. Moreover, the kinematic constraints provided by our equivalent widths were selected (EWHα ≥ 100 Å). The back- observations were already used for modelling the Lyα scattering bone of LARS is a substantial program with the Hubble Space in one LARS galaxy (Paper VI), and also here galactic scale out- Telescope (HST). In this program, each galaxy was observed flows were required to explain the galaxy’s Lyα radiation. With with a combination of ultraviolet long-pass filters, optical broad- the full data set presented here, we now study whether such theo- band filters, as well as Hα and Hβ narrowband filters. These im- retical expected effects are indeed common among Lyα-emitting ages were used to accurately reconstruct Lyα images of those galaxies. 14 galaxies. Moreover, UV spectroscopy with HSTs Cosmic The outline of this manuscript is as follows: in Sect.2 our Origins Spectrograph (COS) is available for the whole sample. PMAS observations of the LARS sample are detailed. The re- This paper is the seventh in a series presenting results of the duction of our PMAS data is explained in Sect.3. Ancillary LARS project. In Östlin et al.(2014; hereafter Paper I) we de- LARS data products used in this manuscript are described briefly tailed the sample selection, the observations with HST and the in Sect.4. The derivation and analysis of the H α velocity and process to reconstruct Lyα images from the HST data. In Hayes dispersion maps is presented in Sect.5. In Sect.6 the results et al.(2014; hereafter Paper II) we presented a detailed analy- are discussed and finally we summarise and conclude in Sect.7. sis of the imaging results. We found that six of the 14 galaxies Notes on individual objects are given in AppendixA. are indeed analogous to high-z LAEs, i.e. they would be selected by the conventional narrowband survey selection requirement1: EWLyα > 20 Å. The main result of Paper II is that a galaxy’s 2. Observations with PMAS morphology seen in Lyα is usually very different compared to its morphological appearance in Hα and the FUV; in particular, We observed all LARS galaxies with the Potsdam Multi- Lyα is often less centrally concentrated and so these galaxies Aperture Spectrophotometer (PMAS; Roth et al. 2005) at the are embedded in a faint low surface brightness Lyα halo. The Calar Alto 3.5 m telescope during four nights from March 12 to results in Paper II (see also Hayes et al. 2013) therefore pro- March 15, 2012 (PMAS run212), except for LARS 13, which vide clear observational evidence for resonant scattering of Lyα was observed on October 10 2011 (PMAS run197). We used photons in the neutral interstellar medium. Pardy et al.(2014; PMAS in the lens array configuration, where 256 fibers are cou- hereafter Paper III) subsequently presented 21 cm observations pled to a 16 × 16 lens array that contiguously samples the sky. tracing the neutral hydrogen content of the LARS galaxies and Depending on the extent of the targeted galaxy, we used either the results supported the complex coupling between Lyα radia- the standard magnification mode, which provides an 800 × 800 tive transfer and the properties of the neutral medium. In Guaita field of view (FoV) or the double magnification mode2, where et al.(2015; hereafter Paper IV), the morphology of the LARS the FoV is 1600 × 1600. The backwards-blazed R1200 grating galaxies was thoroughly re-examined. By artificially redshifting was mounted on the spectrograph. To ensure proper sampling the LARS imaging data-products, it was confirmed that morpho- of the line spread function the 4k × 4k CCD (Roth et al. 2010) logically LARS galaxies indeed resemble z ∼ 2−3 star forming was read out unbinned along the dispersion axis. This set-up de- galaxies. Rivera-Thorsen et al.(2015; hereafter Paper V) then livers a nominal resolving power from R ∼ 5000 to R ∼ 8000 presented high-resolution FUV COS spectroscopy of the full within the targeted wavelength ranges3. In Sect. 3.3 we show sample. Analysing the neutral interstellar medium kinematics as that while the nominal resolving powers are met on average, the traced by the low-ionisation state metal absorption lines, it was instrumental broadening varies at small amplitudes from fibre to shown that all galaxies with global Lyα escape fractions >5% fibre. appear to have outflowing winds. Finally, in Duval et al. (2016; On-target exposures were usually flanked by 400 s expo- hereafter Paper VI), a detailed radiative transfer study of one sures of empty sky near the target. These sky frames serve LARS galaxy was performed using all the observational con- as a reference for removing the telluric background emission. straints assembled within the LARS project, and including data Owing to an error in our observing schedule, no sky frames were that is presented in this paper. In particular it was shown that this taken for LARS 4, LARS 7, and LARS 9 (2012-03-12 pointing). galaxy’s spatial and spectral Lyα emission properties are consis- Fortunately, this did not render the observations unusable, since tent with scattering of Lyα photons by outflowing cool material there are enough blank-sky spectral pixels (so called spaxels) along the minor axis of the disk. within those on-target frames to provide us with a reference sky Now, in the first analysis of our LARS integral-field spec- (cf. Sect. 3.2). troscopic data presented here, we focus on a comparison of re- Observing blocks of one hour were usually flanked by con- sults obtainable from spatially resolved Hα kinematics to results tinuum and HgNe arc lamp exposures used for photometric from the LARS HST Lyα imaging and 21 cm H  observations. and wavelength calibration. We obtained several bias frames In a subsequent publication (Orlitova, in prep.) we will combine throughout each night when the detector was idle during target information from our 3D Hα spectroscopy with our COS UV acquisition. Spectrophotometric standard stars were observed spectra to constrain the parameters of outflowing winds. at the beginning and at the end of each astronomical night This paper is not the first in relating observed spatially re- (BD+75d325 & Feige 67 from Oke 1990, texp. = 600 s). Twilight solved Hα observations to a local galaxy’s Lyα radiation field. flat exposures were taken during dawn and dusk. Recently in a pioneering study exploiting MUSE (Bacon et al. 2014) science verification data, Bik et al.(2015) showed that 2 The naming of the mode refers to the instruments internal magnifi- an asymmetric Lyα halo around the main star-forming knot of cation of the telescopes focal plane; doubling the magnification of the focal plane doubles the extent of the FoV. 1 We adopt the convention established in high-z narrowband surveys 3 Values taken from the PMAS online grating tables, available at of designating galaxies with EWLyα ≥ 20 Å, LAEs, and galaxies with http://www.caha.es/pmas/PMAS_COOKBOOK/TABLES/pmas_ EWLyα < 20 Å, non-LAEs. gratings.html#4K_1200_1BW

A78, page 3 of 27 A&A 587, A78 (2016)

Table 1. Log of PMAS lens array observations of the LARS sample.

LARS Observing- texp. FoV λ-range Resolving power Seeing Observing- 2 00 ID date [s] [arcsec ] λstart − λend [Å] RFWHM FWHM [ ] conditions 1 2012-03-16 2 × 1800 s 16 × 16 5948−7773 5306 1.8 phot. 2 2012-03-15 2 × 1800 s 16 × 16 5752−7581 5205 1.1 non-phot. 3 2012-03-14 3 × 1800 s 16 × 16 5948−7773 5362 1.1 phot. 4 2012-03-13 3 × 1800 s 16 × 16 5752−7581 5883 1.0 phot. 5 2012-03-14 3 × 1800 s 8 × 8 5874−7700 5886 1.3 phot. 6 2012-03-13 2 × 1800 s 16 × 16 5752−7581 5865 1.0 phot. 7 2012-03-16 3 × 1800 s 16 × 16 5948−7773 3925 1.3 non-phot. 8 2012-03-15 3 × 1800 s 16 × 16 5948−7773 4415 0.9 non-phot. 9 2012-03-12 3 × 1800 s 16 × 16 5752−7581 5561 0.9 phot. 2012-03-14 3 × 1800 s 16 × 16 5948−7773 4772 0.9 non-phot. 10 2012-03-14 2 × 1800 s 8 × 8 5874−7700 5438 1.5 phot. 11 2012-03-15 3 × 1800 s 16 × 16 5948−7773 4593 1.0 non-phot. 12 2012-03-14 3 × 1800 s 8 × 8 6553−8363 6756 0.9 phot. 13 2011-10-02 4 × 1800 s 8 × 8 6792−8507 7766 1.2 phot. 2011-10-02 4 × 1800 s 8 × 8 6792−8507 7771 1.2 phot. 14 2012-03-13 2 × 1800 s 8 × 8 6553−8363 7718 0.9 phot.

Notes. See Paper I for coordinates and common names of the galaxies in the LARS sample. Galaxies LARS 09 and LARS 13 were covered with two pointings each, since their extent was larger than the used FoV.

Sky Emission, Absorption Bands & Hα in LARS Galaxies 5 3 7 2 8 9 4 1 13 12 10 14 In Table1 we provide a log of our observations. By chang- 11 ing the rotation of the grating, we adjusted the wavelength ranges 1.4 ] 2 −

covered by the detector, such that the galaxies Hα – [N ] com- c e

s 1.2 c plex is located near the centre of the CCD. We emphasise that r a 1 ±350 Å at the upper/lower end of the quoted spectral ranges are − 1.0 2 aﬀected by vignetting (a known “feature” of the PMAS detector; − m 0.8 c 1 see Roth et al. 2010). We also quote the average spectral reso- − s

g 0.6 α r lution of our ﬁnal data sets near H . The determination of this e [ 6

quantity is described in Sect. 3.3. The tabulated seeing values re- 1 0.4 − 0 fer always to the average full width at half maximum (FWHM) 1 / 0.2 λ of the guide star point-spread function (PSF) measured during f the exposures with the acquisition and guiding camera of PMAS. 00 6600 6800 7000 7200 7400 7600 This value is on average ≈0.2 higher than the DIMM measure- λ [ ] ments (see also Husemann et al. 2013). Fig. 1. Wavelengths of the redshifted Hα lines of the LARS galax- In Fig.1 we show the wavelengths of the redshifted H α line ies (vertical dashed lines) compared to the night sky emission spec- for all LARS galaxies overlaid on the typical night sky emis- trum at Calar Alto (Sánchez et al. 2007a). Grey regions indicate the sion spectrum at Calar Alto from Sánchez et al.(2007a). As can telluric absorption bands (O2 B-Band: 6887 Å−6919 Å, O2 A-Band: be seen for two galaxies (LARS 13 and LARS 14), the Hα line 7607 Å−7633 Å, and H2O a-Band: 7168 Å−7304 Å, see e.g. Cox 2000). signal is contaminated by telluric line emission and two other galaxies (LARS 9 and LARS 12) have their Hα line within an absorption band. This, however, has no effect on the presented and co-addition of exposures (see also Turner 2010). We now analysis. For LARS 13 and LARS 14, we could optimally sub- describe how we applied the tasks of p3d on our raw data for tract the interfering sky lines from the science exposures using each of these steps5: the separate sky frames. Moreover, as we use only spaxels with high signal-to-noise Hα lines in our analysis (see also Sect.5), – Bias subtraction: Master bias frames were created by the high frequency changes within the telluric absorption bands p3d_cmbias. These master bias frames were then subtracted do not alter the quantified features, i.e. width and peak position, from the corresponding science frames. Visual inspection of of the profiles. the bias subtracted frames showed no measurable offsets in unexposed regions between the four quadrants of the PMAS CCD, which attests to optimal bias removal. 3. The PMAS data reduction – Flat fielding: Each of the 256 fibers has its own wavelength 3.1. Basic reduction with p3d dependent throughput curve. To determine these, the task p3d_cflatf was applied on the twilight flats. The deter- 4 For reducing the raw data the p3d-package (Sandin et al. 2010, mined curves are then applied in the extraction step of the 2012) was utilised. This pipeline covers all basic steps needed science frame (see below) to normalise the extracted spectra. for reducing fiber-fed integral field spectroscopic data: bias subtraction, flat fielding, cosmic ray removal, tracing and extraction 5 As the amount of raw data from the observations was substantial, of the spectra, correction of differential atmospheric refraction, we greatly benefited from the scripting capabilites of p3d (Sandin et al. 2011). Our shell scripts and example parameter files can be found at 4 http://p3d.sourceforge.net https://github.com/Knusper/pmas_data_red

A78, page 4 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

– Cosmic ray removal: Cosmic ray hits on the CCD were re- moved with p3d_cr. Visual inspections guided by the pa- 14 80 rameter study of Husemann et al.(2012) lead us to apply the 76 12 L.A. Cosmic algorithm (van Dokkum 2001) with σlim = 5, 72 σ = 1, f = 15, a grow radius of 2 and a maximum of 10 frac lim 68 four iterations. 8 – Tracing and extraction: Spectra were extracted with the 64 modified Horne(1986) optimal extraction algorithm (Sandin 6 60 56 et al. 2010). The task p3d_ctrace was used to determine the 4 LArr v_FWHM [km/s] traces and the cross-dispersion profiles of the spectra on the 52 2 detector. Spectra were then extracted with p3d_cobjex, also 48 using the median recentring recommended in Sandin et al. 0 (2012). 0 2 4 6 8 10 12 14 – Wavelength calibration: We obtained dispersion solutions (i.e. mappings from pixel to wavelength space) from the Fig. 2. Representative resolving power map for the observation of LARS 1. The resolution is expressed as vFWHM of a 1D Gaussian fit. HgNe lamp frames using a sixth-order polynomial with Black spaxels at the positions (x, y) = (0, 14) and (8, 15) are dead fibres. p3d_cdmask . The wavelength sampling of the extracted science spectra is typically 0.46 Å px−1. We further improved the wavelength calibration of the science frames by ap- plying small shifts (typically 0.1−0.3 px) determined from this calculation the input variances were derived from squar- the strong 6300 Å and 6364 Å [O ] sky lines (not avail- ing the error cubes. Before co-addition we ensured by visual able within the targeted wavelength range for LARS 12, 13, inspection that there are no spatial offsets between individual and 14) exposures. – Sensitivity function and flux calibration: Using extracted and wavelength-calibrated standard star spectra, we created a sensitivity function utilising p3d_sensfunc. Absorption 3.3. Spectral resolving power determination bands in the standard star spectra, as well as telluric absorption bands (see Fig.1), were masked for the fit. Extinction To determine the intrinsic width of our observed Hα lines, curves were created using the empirical formula presented we need to correct for instrumental broadening by the spec- in Sánchez et al.(2007b) with the extinction in V-band as trograph’s line spread function (the spectrograph’s resolving measured by the Calar Alto extinction monitor6. We flux- power; Robertson 2013). It is known that for PMAS the resolv- calibrated all extracted and wavelength-calibrated science ing power varies from fibre to fibre and with wavelength (e.g. frames with the derived sensitivity function and extinction Sánchez et al. 2012). To determine the broadening at the wave- curve with p3d_fluxcal. length of the Hα line, we use the HgNe arc lamp exposures, which were originally used to wavelength calibrate the on-target The final data products from the p3d-pipeline are flux- and exposures. We first create wavelength-calibrated data cubes from wavelength-calibrated data cubes for all science exposures, as these arc frames. Next, we select two strong arc lamp lines that well as the corresponding error cubes. From here we now per- are nearest in wavelength to the corresponding galaxy’s Hα line. form the following reduction steps with our custom procedures Finally, we fit 1D Gaussians in each spaxel to each of those lines 7 written in python . (e.g. similar to Alonso-Herrero et al. 2009). Typically the separation from a galaxy’s Hα line to one of those arc lamp lines is 3.2. Sky subtraction and co-addition ∼50 Å. The difference between the FWHM of the fits to the different lines is typically ∼1 km s−1, hence we take the average of – Sky subtraction: All sky frames were reduced as science both at the position of Hα as the resolving power for each spaxel. frames, as described in the previous section. These extracted, We point out that the arc-lamp lines FWHM is always sampled wavelength- and flux-calibrated sky frames were then sub- by more than 2 pixels, therfore aliasing effects can be neglected tracted from the corresponding science frames and errors (Turner 2010). were propagated accordingly. Unfortunately, no separate sky This procedure provides us with resolving power maps. As exposures were taken for three targets (LARS 4, LARS 7, an example, we show a so-derived map for the observations of and LARS 9; see also Sect.2). For these targets we cre- LARS 1 in Fig.2. We note that the spatial gradient seen in Fig.2 ated a narrowband image of the Hα-[N ] region by sum- is not universal across our observations (see Sánchez et al. 2012, ming up the relevant layers in the datacube. In this image, for an explanation). We further note that the formal uncertainties we selected spaxels that do not contain significant amounts on the resolving power determination are negligible in compari- of flux. From these spaxels then an average sky spectrum was son to the uncertainties derived on the Hα profiles in Sect.5. created, which was then subtracted from all spaxels. As the In Table1 we give the mean resolving power at the position spectral resolution varies across the FoV (see also Sect. 3.3), this method produces some residuals at the position of the of Hα for each galaxy as RFWHM. Variations within the FoV typically have an amplitude of ∼30 km s−1. The average resolving sky lines, which however do not touch the Hα lines of the −1 affected targets. power of all our observations is R = 5764 or 52 km s , which is consistent with the values given in the PMAS online grating – Stacking: We co-added all individual flux-calibrated and sky- 8 subtracted data cubes using the variance-weighted mean. For manual .

6 http://www.caha.es/CAVEX/cavex.php 8 http://www.caha.es/pmas/PMAS_COOKBOOK/TABLES/pmas_ 7 http://www.python.org gratings.html#4K_1200_1BW

A78, page 5 of 27 A&A 587, A78 (2016)

3.4. Registration on astrometric grid of HST observations 1e4 1e3 1e3

] ] ] 6 1 LARS 1 1 2.5 LARS 2 1 LARS 3 − − − 0.8 5 To facilitate a comparison of our PMAS observations with the 2 2 2.0 2 − − − m m m c 0.6 c c 4 HST imaging results from Hayes et al.(2014), we have to reg- 1 1 1.5 1 − − − s s s 3 g g g ister our PMAS data cubes with respect to the LARS HST data r 0.4 r r e e 1.0 e 7 7 7

1 1 1 2 products. As explained in Östlin et al.(2014), all HST images are − − − 0 0 0

1 0.2 1 1 [ [ 0.5 [ 1

aligned with respect to each other and have a common pixel scale λ λ λ 00 −1 f 0.0 f 0.0 f 0 of 0.04 px . We use the continuum-subtracted Hα line image 6740 6745 6750 6755 6750 6755 6760 6765 6760 6765 6770 6775 as a reference. From this image we create contours that high- λ [ ] 1e3 λ [ ] 1e3 λ [ ] 1.4 ] ]

LARS 4 1 LARS 5 1 LARS 6 light the most prominent morphological features in Hα. We then − 6 − 1.2 2 2 produce a continuum-subtracted Hα map from the PMAS data − 5 − 1.0 m m c c cubes by subtracting a version of the data cube that is median- 1 4 1 0.8 − − s s

g 3 g 0.6 r r

ﬁltered in spectral direction. Finally, we visually match the con- e e 7 7 1 1 0.4

− 2 − tours from the reference image with the PMAS Hα map. This 0 0 1 1 [ 1 [ 0.2 λ λ

constrains the position of the PMAS FoV relative to the HST f f 0 0.0 imaging. We emphasise that we make full use of the FITS header 6775 6780 6785 6790 6780 6785 6790 6795 1e3 1e3 λ [ ] 1e3 λ [ ]

world-coordinate representation in our method, i.e. the headers ]

] ] 5 1 1 LARS 7 1 LARS 8 7 LARS 9

4 − of our ﬁnal data cubes are equipped with keyword value pairs − − 6

2 2 4 2 − − − to determine the position of each spaxel on the sky (Greisen & m 3 m 5 c c m

1 1 3 c

− − 4

Calabretta 2002; Calabretta & Greisen 2002). s s 1

g 2 g − r r 3 e e 2 s

For galaxies with a single PMAS pointing, we present the 7 7 1 1 g − −

r 2 0 0

1 e 1 1 [ ﬁnal result of our registration procedure in Fig.5 (or Figs.6 [ [ 1 1 λ λ λ f f and7, for the two galaxies with two PMAS pointings). For each 0 0 f 0 LARS galaxy, the Hα line intensity map extracted from our HST 6805 6810 6815 6820 6805 6810 6815 6820 6865 6870 6875 6880 1e3 λ [ ] 1e2 λ [ ] 1e3 λ [ ] 1.2 observations is shown in the top panel and can be compared to ] ] ] 1 LARS 10 1 6 LARS 11 1 LARS 12 − − − 1.4 the PMAS Hα signal-to-noise ratio (S/N) map in the panel be- 1.0 2 2 5 2 1.2 − − −

m 0.8 m m low. The S/N values for each spaxel are calculated by summa- c c 4 c 1.0 1 1 1 − − − tion of all ﬂux values within a narrow spectral window centred s 0.6 s 3 s 0.8 g g g r r r

e e e 0.6

7 0.4 7 2 7 on Hα and subsequent division by the square root of the sum of 1 1 1 − − − 0.4 0 0 0

1 1 1 1

[ 0.2 [ [ the variance values in that window. The width of the summation 0.2 λ λ λ

f f 0 f window is taken as twice the Hα lines FWHM. Our display of 0.0 0.0 the HST Hα images uses an asinh-scaling (Lupton et al. 2004) 6930 6935 6940 6945 7110 7115 7120 7125 7225 7230 7235 7240 1e2 λ [ ] 1e3 λ [ ] λ [ ] ] ] cut at 95% of the maximum value (see Sect. 4 and Fig. 3 of 1 LARS 13 1 7 LARS 14

− 7 − 6 Hayes et al. 2014, for absolute Hα intensities) and we scaled 2 6 2 − −

m 5 m 5 our PMAS Hα S/N maps logarithmically. The inferred ﬁnal po- c c 1 1

− 4 − 4 s s sition of the PMAS FoV is indicated with a white square (or two g g r 3 r

e e 3 7 7 1 2 1 squares for the galaxies with two pointings) within the Hα panel. − − 2 0 0 1 1 [ 1 [ 1

Also shown are the Hα contours used for visual matching. As λ λ

f 0 f 0 can be seen, most of the prominent morphological characteris- 7520 7525 7530 7535 7740 7745 7750 7755 tics present in the HST Hα maps are unambiguously identifiable λ [ ] λ [ ] in the lower resolution PMAS maps, exemplifying the robust- ness of our registration method. Fig. 3. Integrated continuum-subtracted Hα profiles of the LARS galaxies (black) compared to summed 1D Gaussian model profiles (grey) from which radial-velocity and velocity-dispersion maps were generated. 4. Ancillary LARS data products We compare our PMAS data to the HST imaging results of the LARS project presented in Hayes et al.(2013) and Paper II. spatial scale that is resolved within the VLA H  images is much Specifically, we use the continuum-subtracted Hα and Lyα im- larger than our PMAS observations. Moreover, the single-dish ages that were presented in Paper II. These images were pro- GBT spectra are sensitive to H  at the observed frequency range duced from our HST observation using the LARS extraction within 80 of the beam. Full details on data acquisition, reduction, software LaXs (Hayes et al. 2009). Details on the observational and analysis are presented in Paper III. strategy, reduction steps, and analysis performed to obtain these HST data products used in the present study are given in Paper II and Paper I. We also compare with the published results of the LARS 5. Analysis and results H  imaging and spectroscopy observations that were obtained 5.1. Hα velocity fields with the 100m Robert C. Byrd Green Bank Telescope (GBT) and the Karl G. Jansky Very Large Array (VLA). The GBT single- We condense the kinematical information traced by Hα in our dish spectra are present for all systems, except that the H  sig- data cubes into two-dimensional maps depicting velocity dis- nal could not be detected in the three LARS galaxies with the persion and line-of-sight radial velocity at each spaxel. Both largest distances (LARS 12, LARS 13, and LARS 14). The VLA quantities are derived from 1D Gaussian fits to the observed line interferometric imaging results are only available for LARS 2, shapes from the continuum-subtracted cube (see Sect. 3.4). We LARS 3, LARS 4, LARS 8, and LARS 9. We emphasise that calculate the radial velocity offset vLOS to the central wavelength with beam sizes from 5900 to 7200 (VLA D-configuration) the value given by the mean of all fits and the FWHM of the fitted

A78, page 6 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

Fig. 4. Uncertainties on derived velocity dispersions (left panel) and radial velocities (right panel) for Gaussian profile fits to Hα for all galaxies. All uncertainties for a specific galaxy have the same symbol according to the legend in the right panel.

line in velocity space vFWHM. To obtain higher S/N ratios in the the noise from our noise cubes and then fit the 1D Gaussian pro- galaxies’ outskirts we use the weighted Voronoi tessellation bin- file in exactly the same way to each of these 100 realisations. ning algorithm by Diehl & Statler(2006), which is a generali- The width of the distribution of all fit results characterised by sation of the Voronoi binning algorithm of Cappellari & Copin their standard deviation gives a robust measure for the uncer- (2003). tainty on the derived quantities (cf. Appendix B.4. in Davies We visually scrutinised the fits to the spectra and decided et al. 2011). The central moment of the distribution is the final that lines observed at a S/N of six were still reliably fit. Hence, value that is shown in our maps. In Fig.4 we indicate, for all per- six defines the minimum S/N in the Voronoi binning algorithm. formed fits, the error on the velocity dispersion ∆vFWHM and on Furthermore, we adopted a maximum bin size of 3 × 3 pixels. In radial velocity ∆vLOS as a function of the lines S/N. On average those calculations signal and noise are defined as in Sect. 3.4. In our uncertainties on vFWHM and vLOS follow the scaling laws ex- what follows, only results from the fits to bins that meet the min- pected for fitting Gaussian profiles to noisy emission-line spectra 1/2 −1 imum S/N requirement are considered. Most of the bins used are ∆vLOS ∝ ∆vFWHM ∝ vFWHM(S/N) (Landman et al. 1982; Lenz actually only a single spaxel, and very few bins in the outskirts & Ayres 1992). For reference, at our minimum S/N = 6 we have −1 −1 of the galaxies meeting the minimum S/N criterion do consist typical errors of ∆vFWHM ≈ 30 km s and ∆vLOS ≈ 10 km s , of 2 or 3 spaxels. In practice this means that we are typically and at the median (mean) S/N of our data – S/Nmedian = 22 −1 −1 sensitive to regions at Hα surface brightness higher than 5 × (S/Nmean = 50), we obtain ∆vFWHM ≈ 7 km s (≈3 km s ) and −1 −1 10−16 erg s−1 cm−2 arcsec−2. Using the Kennicutt(1998) conver- ∆vLOS ≈ 3 km s (≈1 km s ). sion and neglecting extinction, this translates to a detection limit In Fig.5 we show the resulting vFWHM and vLOS maps to- −2 −1 −2 of star-formation rate surface densities ∼3 × 10 M yr kpc , gether with the LaXs Lyα images (lower three panels). Maps and more than an order of magnitude deeper than probed by high red- images of the galaxies observed with two pointings are shown in −1 −2 shift integral field spectroscopy studies (∼1 M yr kpc , see Figs.6 and7 for LARS 9 and LARS 13, respectively. All vFWHM e.g. Law et al. 2009). maps have been corrected for instrumental broadening using the spectral resolving power maps derived in Sect. 3.3. In all v By visual inspection we also ensured that 1D Gaussian pro- LOS and v panels, we overlay contours derived from the Lyα files are a sufficient model of the observed Hα profiles seen in FWHM image to emphasise distinctive morphological Lyα properties. the PMAS LARS data cubes. We only have some spaxels where Individual maps are discussed in AppendixA. Using the a 1D Gaussian certainly fails to reproduce the complexity of the classification scheme introduced by Flores et al.(2006; see also observed spectral shape in two galaxies (LARS 9 and LARS 13). Sect. 3.3 in Glazebrook 2013) the LARS galaxies can be charac- Moreover, in LARS 14 weak broad wings are visible in the indi- terised by the qualitative appearance of their velocity fields: vidual Hα profiles. We have not attempted to model these special cases, but we describe their qualitative appearance in detail in the – Rotating disks: LARS 8 and LARS 11. Both galaxies show a Appendices A.9, A.13, and A.14, and acknowledge their exis- regular symmetric dipolar velocity field with a steep gradient tence in our interpretation in Sect.6. To demonstrate the overall in the central regions. This steep gradient is also the reason quality of our Hα velocity fields, we compare in Fig.3 the inte- for artificially broadened emission near the kinematical cen- grated Hα profiles of the LARS galaxies to the sum of the fitted tre. Morphologically both galaxies also bear resemblance to 1D Gaussians used to generate the vFWHM and vLOS maps. As can a disk and the kinematical axis is aligned with the morpho- be seen, the observed profile and that derived from the models logical axis. However, we point out in Appendix A.8 that are largely in agreement. The strongest deviation between data LARS 8 can also be classified as a shell galaxy, hinting at a and models is apparent in LARS 14, where the broad wings of recent merger event. the observed Hα profile are not reproduced at all by 1D Gaussian – Perturbed rotators: LARS 1, LARS 3, LARS 5, LARS 7, models (cf. Sect. 6.1 and Appendix A.14). and LARS 10. In those galaxies traces of orbital motion are To estimate formal uncertainties on vFWHM and vLOS, we use still noticeable, but the maps look either significantly per- a Monte Carlo technique: we perturb our spectra 100 times with turbed compared to a classical disk case (LARS 3, LARS 5,

A78, page 7 of 27 A&A 587, A78 (2016)

LARS 1 LARS 2 LARS 3 62°08'00"

53°27'25" 07'55" 43°55'55" 20" 07'50"

15" 50" 07'45"

HST 10" 07'40" Hα 45" 05" 07'35"

00" 40" 07'30" 13h28m44.5s 44.0s 43.5s 43.0s 9h07m04.0s 03.0s 02.0s 13h15m35.0s34.0s 33.0s 32.0s 31.0s 103 103 103 14 14 14

12 12 12

2 2 2 10 10 10 10 10 10 PMAS 8 8 8 Hα 6 6 6 101 101 101 4 4 4

2 2 2

0 0 0 0 0 0 10 10 10 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 90 90 45 80 80 40 70 70 35 60 60 30

50 ] ] 25 ]

2 50 2 2 − − − 40 40 20 m m m c c c

30 1 30 1 15 1 − − − HST s s s g g g

20 r r r

e 20 e 10 e 8 8 8

Lyα 1 1 1 − − − 0 0 0 1 1 1 10 [ 10 [ 5 [ α α α y y y L L L F F F

0 0 0 150 180 480 165 135 420 150 120 ] ] ]

1 1 360 1

135 − − −

s 105 s s 300 120 m m m PMAS k k k [ 90 [ [

M M 240 M

vFWHM 105 H H H

W 75 W W F F 180 F 90 v v v 60 75 120 45 60 60

60 32 160

45 24 120

30 16 80 ] ] ] 1 1 1

15 − 8 − 40 − s s s

0 m 0 m 0 m PMAS k k k [ [ [ S S S

v 15 O 8 O 40 O

LOS L L L v v v 30 16 80

45 24 120

60 32 160

Fig. 5. Comparison of LARS HST imaging results of the LARS sample to spatially resolved PMAS Hα spectroscopy. North is always up and east is always to the left. For each galaxy from top to bottom: the first panel shows the LaXs Hα line intensity map; tick labels indicate right ascension and declination and an asinh-scaling is used cut at 95% of the maximum value. The second panel shows a S/N map of the continuum-subtracted Hα signal observed with PMAS. Tick labels in the PMAS S/N map are in arc-seconds; the scaling is logarithmic from 1 to 103 and only spaxels with S/N > 1 are shown. The third panel shows the LARS Lyα images with a colour bar indicating the flux scale in cgs-units; scaling is the same as in the Hα map. The fourth panel shows resolution-corrected Hα vFWHM maps from our PMAS observations and the corresponding Hα vLOS maps are shown in the fifth panel. In the first and third panel, we indicate the position and extent of the PMAS field of view with a white box. Cyan contours in the HST Hα image are contours of constant surface brightness, adjusted to highlight the most prominent morphological features. Similarly, magenta contours in the HST Lyα images indicate the Lyα morphology; these contours are also shown in the fourth and fifth panel. To highlight the difference between Lyα and Hα, the Lyα and Hα panels contain as dashed yellow lines the contours from the Hα and Lyα panels, respectively.

A78, page 8 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

LARS 4 LARS 5 LARS 6

54°27'20" 57°26'45" 44°16'10"

15"

40" 05" 10" HST 05" 00" Hα 35"

00" 15'55"

13h07m27.0s 26.0s 25.0s 13h59m50.0s 49.5s 49.0s 48.5s 15h45m43.5s 43.0s 42.5s 103 103 103 14 7 14

12 6 12

2 2 2 10 10 5 10 10 10 PMAS 8 4 8 Hα 6 3 6 101 101 101 4 2 4

2 1 2

0 0 0 0 0 0 10 10 10 0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 0 2 4 6 8 10 12 14 13.5 22.5 120 20.0 12.0 105 10.5 17.5 90 15.0 9.0

] 75 ] 7.5 ]

12.5 2 2 2

− 60 − − 10.0 6.0 m m m c c c

1 45 1 4.5 1

7.5 − − − HST s s s g g g

r 30 r 3.0 r 5.0 e e e 8 8 8

Lyα 1 1 1 − − − 0 0 0 1 1 1 [ 15 [ [ 2.5 1.5 α α α y y y L L L F F F

0.0 0 0.0 165 88 225

150 200 80

135 175 72 ] ] ] 1 1 1

− 150 − −

120 s s s 64 m m m

PMAS k 125 k k 105 [ [ [ 56 M M M

vFWHM H 100 H H

90 W W W F F F

v v 48 v 75 75 50 40 60 25 32

100 40 60 75 30 40 50 20 ] ] ]

25 1 1 20 1

− 10 − − s s s

0 m 0 m 0 m PMAS k k k [ [ [ S S S

v 25 O 10 O O LOS L L 20 L v v v 20 50 40 30 75 60 40 100

Fig. 5. continued.

LARS 7, and LARS 10) and/or the observed velocity gradi- also Appendix A.3). From our imaging data alone LARS 7 ent is very weak (LARS 1, LARS 5 and LARS 10). Based and LARS 10 bear resemblance to shell galaxies (see also solely on morphology, we previously classiﬁed two of those Appendices A.7 and A.10). galaxies as dwarf edge-on disks (LARS 5 and LARS 7; – Complex kinematics: LARS 2, LARS 4, LARS 6, LARS 9, Paper II; Paper III). Moreover, LARS 3 is a member of LARS 12, LARS 13, and LARS 14. These seven galaxies a well known merger of two similar massive disks (see are characterised by more chaotic vLOS maps and they all

A78, page 9 of 27 A&A 587, A78 (2016)

LARS 7 LARS 8 LARS 10

7°35'10" 29°23'20"

29°23'20" 05" 15"

15" 10" HST 00"

Hα 10" 05" 34'55"

00" 05" 13h16m03.0s 02.5s 02.0s 12h50m13.0s 12.5s 12.0s 13h01m41.0s 40.5s 40.0s 39.5s 103 103 103 14 14 7

12 12 6

2 2 2 10 10 10 10 5 10 PMAS 8 8 4 Hα 6 6 3 101 101 101 4 4 2

2 2 1

0 0 0 0 0 0 10 10 10 0 2 4 6 8 10 12 14 0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 18 80 48 16 70 42 14 60 36 12 30 50 ] ] 10 ] 2 2 2

40 − 24 − 8 − m m m

c 18 c c 30 1 1 6 1 − − − HST s s s g 12 g g 20 r r 4 r e e e 8 8 8

Lyα 1 1 1 − − − 0 0 0 1 1 1 10 [ 6 [ 2 [ α α α y y y L L L F F F

0 0 0 140 200 220 130 180 120 200 160

] ] 110 ] 1 1 1

180 − 140 − − s s 100 s

m 120 m m

PMAS k k k 160 [ [ 90 [

M 100 M M vFWHM H H H 80 140 W W W F F F

v 80 v v 70 120 60 60 100 40 50

40 40 160

30 30 120

20 80 20 ] ] ]

1 1 10 1 10 − 40 − − s s s

0 m 0 m 0 m PMAS k k k [ [ [ S S S

v 10 O 40 O 10 O

LOS L L L v v v 20 80 20

120 30 30

40 160 40

Fig. 5. continued.

have also irregular morphologies. Four of the irregulars con- Notably, in all LARS galaxies, except for LARS 6, we ob- −1 sist of photometrically well-separated components that are serve high-velocity dispersions (vFWHM & 100 km s ); the also kinematically distinct (LARS 4, LARS 6, LARS13 most extreme case is LARS 3 where locally lines as broad as −1 and LARS 14), while the other three each have their own vFWHM ∼ 400 km s are found. The observed width of the individual peculiarities (LARS 2, LARS 9 and most spectac- Hα line seen in our vFWHM maps can be described as a succes- ularly LARS 12 − see also Appendices A.2, A.9, and A.12). sive convolution of the natural Hα line (7 km s−1 FHWM) with the thermal velocity distribution of the ionised gas (21.4 km s−1 These kinematic classes are also listed in Table2. FWHM at 104 K) and with non-thermal motions in the gas

A78, page 10 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics LARS 11 LARS 12 LARS 14 44°28'00" 54°28'48"

6°28'35" 45" 27'55" 30"

42"

HST 25" 27'50" Hα 39" 20"

36"

14h03m47.0s 46.5s 46.0s 45.5s 9h38m12.0s 11.5s 11.0s 9h25m59.5s 59.0s 58.5s 103 103 103 14 7 7

12 6 6

2 2 2 10 10 5 10 5 10 PMAS 8 4 4 Hα 6 3 3 101 101 101 4 2 2

2 1 1

0 0 0 0 0 0 10 10 10 0 2 4 6 8 10 12 14 0 1 2 3 4 5 6 7 0 1 2 3 4 5 6 7 54 28 27 24 48 24 21 42 20 18 36

] 15 ] 30 ] 16 2 2 2 − 12 − 24 − 12 m m m c c c

1 9 1 18 1 − − − HST 8 s s s g g g

r r 12 r

e 6 e e 8 8 8

Lyα 1 1 1 − − − 0 0 0

4 1 1 1 [ 3 [ 6 [ α α α y y y L L L F F F

0 0 0 250 320 280 225 280 200 240

] 240 ] ] 1 1 175 1 − − −

200 s s s 200 m m 150 m

PMAS k k k [ [ [ 160 M 160 M 125 M vFWHM H H H W W W F F F

120 v 120 v 100 v

80 80 75

50 40 40 200

120 45 150

30 100 80 ] ] ]

50 1 40 1 15 1 − − − s s s

0 m 0 m 0 m PMAS k k k [ [ [ S S S

v O O O LOS 50 L 40 L 15 L v v v

100 80 30

150 120 45

200

Fig. 5. continued.

(e.g. Jiménez-Vicente et al. 1999). Since all our dispersion mea- 5.2. Global kinematical properties: vshear, σ0, σtot surements are significantly higher than the thermally broadened and vshear/σ0 profile non-thermal motions dominate the observed linewidths. We now quantify the global kinematical properties of our Hα ve- Non-thermal motions could be centre-to-centre dispersions of locity fields that were qualitatively discussed in the previous the individual H  regions and turbulent motions of the ionised section. Therefore, we compute three non-parametric estima- gas. The latter appears to be the main driver for the observed tors that are commonly adopted in the literature for this purpose line widths, since the observed velocity dispersions are highly −1 4 (e.g. Law et al. 2009; Basu-Zych et al. 2009; Green et al. 2010; supersonic (i.e. &10 km s for H  at 10 K; see also Sect. 5.1 Gonçalves et al. 2010; Glazebrook 2013; Green et al. 2014; in Glazebrook 2013). Wisnioski et al. 2015): shearing velocity vshear, intrinsic velocity

A78, page 11 of 27 A&A 587, A78 (2016)

PMAS Hα 103 14

12 28°06'50" 2 10 10

45" 6 101 4 HST 2 40" 0 0 10 Hα 0 2 4 6 8 10 12 14 LARS 9 103 14

35" 12

2 10 10

8 30" 6 101 4

2 25" 0 0 10 0 2 4 6 8 10 12 14 8h23m54.5s 54.0s 53.5s 53.0s

240 27 320 24 180 280 21 120

18 240 ] ] 1 1 −

s 60 − 15 200 s m m k

[ 0 k 160 [ M

12 S H 60 O L W ] F v

2 120 v − 120

9 m 80 c

1 180 −

s 40

g 240

HST 6 r e

8 240

1 320

Lyα − 0 180

1 280 [

α 120 y

240 ] L ] 1 F 1 3 −

s 60 −

200 s m m k

[ 0 k 160 [ M S H 60 O L W F 120 v v 120 80 180 40 0 240 PMAS PMAS vFWHM vLOS

Fig. 6. Comparison of LARS HST imaging results to spatially resolved PMAS Hα spectroscopy for LARS 9. For detailed description of individual panels see caption of Fig.5. This galaxy was covered with 2 PMAS pointings. Hatched regions in the vLOS map indicate regions, where the Hα emission shows a more complex proﬁle that could not be described by a simple Gaussian (see also Appendix A.9).

dispersion σ0, and their ratio vshear/σ0. We describe the physical Here we take the median of the lower and upper ﬁfth percentile meaning of those parameters and our method to determine them of the distribution of values in the vLOS maps for vmin and vmax, in the following three subsections. Furthermore, we also calcu- respectively. This choice ensures that the calculation is robust late the integrated velocity dispersion σtot. This measure pro- against outliers while sampling the true extremes of the distri- vides a useful comparison for unresolved distant galaxies. Our bution. We conservatively estimate the uncertainty by propagat- results on vshear, σ0, σtot and vshear/σ0 are tabulated in Table2. ing the full width of the upper and lower ﬁfth percentile of the velocity map. We list our results for vshear in the second column of Table2. Our derived vshear values for the LARS sam- −1 −1 −1 5.2.1. Shearing velocity vshear ple range from 30 km s to 180 km s , with 65 km s being the median. This is less than half the typical maximum veloc- The shearing velocity vshear is a measure for the large-scale gas ity vmax of Hα rotation curves observed in spiral galaxies, e.g. −1 bulk motion along the line of sight. We calculate it via vmax = 145 km s is the median of 153 local spirals (Epinat et al. 2010). In contrast to our vshear measurement, vmax values are inclination corrected. In a sample with random inclinations 1 v π v ≈ . v v = (v − v ) . (1) on average the correction shear = 4 max 0 79 max is expected shear 2 max min A78, page 12 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

PMAS Hα 103 14 HST 12 2 Hα 10 10 8

6 101 4 13°09'18" 2

0 0 10 0 2 4 6 8 10 12 14 LARS 13 15" 103 14

12" 2 10 10

09" 6 101 1h50m27.6s 27.4s 27.2s 27.0s 4

0 0 10 0 2 4 6 8 10 12 14

HST 270 200 Lyα 240 150 210 100

8 ] ] 180 1

7 1 − 50 s −

6 s

150 m m k

5 [ 0 k [ M ] 120 S H

4 2 50 O L W − F v v

m 90

3 c 100 1

− 60 s 150 2 g r 30 e 8

1 200 −

0 270 200 1 1 [ α 240

y 150 L

F 210 100 ] ] 180 1 1 − 50 s − s

150 m m k

[ 0 k

0 [ M

120 S H 50 O L W F v

90 v 100 60 150 30 200 PMAS PMAS vFWHM vLOS

Fig. 7. Same as Fig.6, but for LARS 13. Hatched region in the vLOS map indicates the region, where the Hα emission shows a more complex proﬁle that could not be described by a simple Gaussian (see also Appendix A.13).

(e.g. Law et al. 2009), but even with this correction classical 5.2.2. Intrinsic velocity dispersion σ0 disks appear still incompatible with most of our measurements The intrinsic velocity dispersion σ0 (sometimes also called by a factor larger than two. We only find vshear values compati- ble with local rotators for the two galaxies that we classified as resolved velocity dispersion or local velocity dispersion, cf. Glazebrook 2013, Sect. 1.2.1) is in our case a measure for the rotating disks in Sect. 5.1. Nevertheless, high vshear values do not necessarily imply the presence of a disk. Large-scale bulk mo- typical random motions of the ionised gas within a galaxy. We tions at high amplitude also occur in close encounters of spatially calculate it by taking the flux weighted average of the observed and kinematically distinct companions (e.g. LARS 13). velocity dispersions in all spaxels, i.e. Sensitivity is an important factor in the determination of the P Hα observed v . An observation of less depth does not detect Fspaxelσspaxel shear σ · fainter regions at larger radii, which may have the highest veloci- 0 = P Hα (2) Fspaxel ties. Indeed, for most of our galaxies the spatial positions of vmin and vmax values are in those outer lower surface-brightness re- Hα gions. Hence observations of lower sensitivity are biased to Here Fspaxel is the Hα flux in each spaxel and σspaxel is the ve- lower vshear values. This sensitivity bias is strongly affecting locity dispersion measured in that spaxel; the sum runs over all high-z studies because these kinds of observations are often only spaxels with S/N ≥ 6 (Sect. 5.1). This kind of flux-weighted limited to the brightest regions (Law et al. 2009; Gonçalves et al. summation σ0 has been widely adopted in the literature to 2010). quantify the typical intrinsic velocity dispersion component

A78, page 13 of 27 A&A 587, A78 (2016)

Table 2. Global kinematic parameters from Hα for the LARS galaxies, calculated as described in Sect. 5.2.

Lyα ID Class σ0 σtot vshear vshear/σ0 EWLyα Lyα/Hα fesc. ξLyα [km s−1] [km s−1] [km s−1] [Å] 1 P 47.5 ± 0.1 60.5 56 ± 2 1.2 ± 0.1 33.0 1.36 0.119 3.37 2 C 38.6 ± 0.9 46.0 23 ± 3 0.6 ± 0.1 81.7 4.53 0.521 2.27 3 P 99.5 ± 3.7 130.9 138 ± 3 1.4 ± 0.2 16.3 0.16 0.003 0.77 4 C 44.1 ± 0.1 55.1 74 ± 9 1.7 ± 0.1 0.00 0.00 0.000 . . . 5 P 46.8 ± 0.3 54.7 37 ± 4 0.8 ± 0.1 35.9 2.16 0.174 2.61 6 C 27.2 ± 0.3 58.9 52 ± 7 1.9 ± 0.2 0.00 0.00 0.000 . . . 7 P 58.7 ± 0.3 67.4 31 ± 3 0.5 ± 0.1 40.9 1.94 0.100 3.37 8 R 49.0 ± 0.1 111.8 155 ± 3 3.2 ± 0.1 22.3 0.67 0.025 1.12 9 C 58.6 ± 0.1 112.7 182 ± 3 3.1 ± 0.1 3.31 0.13 0.007 >2.85 10 P 38.2 ± 1.0 49.2 36 ± 3 0.9 ± 0.2 8.90 0.47 0.026 2.08 11 R 69.3 ± 3.8 114.8 149 ± 4 2.1 ± 0.3 7.38 0.72 0.036 2.27 12 C 72.7 ± 1.0 81.3 96 ± 3 1.3 ± 0.1 8.49 0.48 0.009 3.48 13 C 69.2 ± 0.7 97.3 183 ± 4 2.5 ± 0.2 6.06 0.29 0.010 1.74 14 C 67.3 ± 1.3 72.9 40 ± 1 0.6 ± 0.1 39.4 2.24 0.163 3.62

Lyα Notes. For reference we also tabulate EWLyα, Lyα/Hα and fesc. from Paper II, as well as ξLyα from Hayes et al.(2013). Class refers to the three kinematic classes proposed by Flores et al.(2006): C = complex kinematics, P = perturbed rotators, and R = rotating disks.

(e.g. Östlin et al. 2001; Law et al. 2009; Gonçalves et al. 2010; are certain, that the adopted formalism in Eq. (2) gives a robust Green et al. 2010, 2014; Glazebrook 2013). estimate of the intrinsic velocity dispersion. The PSF smearing in the presence of strong velocity gradi- We list our derived values for σ0 in the third column of ents broadens the Hα line at the position of the gradients9. This Table2. In general, the ionised gas kinematics of the LARS kind of broadening could bias our calculation of the intrinsic galaxies are characterised by high intrinsic velocity dispersions velocity dispersion. At the distances of the LARS galaxies, our ranging from 40 km s−1 to 100 km s−1 with 54 km s−1 being typical seeing-disk PSF FWHM of 100 corresponds to scales of the median of the sample. Such high intrinsic dispersions are 0.6 kpc to 3 kpc. In most of our observations, this average PSF in contrast to H  velocity dispersions of ∼10−50 km s−1 typ- FWHM is comparable to the size of one PMAS spaxel (100 × 100) ically found in local spirals (e.g. Epinat et al. 2008a,b, 2010; and most of the LARS galaxies exhibit rather weak velocity gra- Erroz-Ferrer et al. 2015). In high-z, star-forming galaxies intrin- dients. The PSF smearing effect is only seen in galaxies with sic dispersions &50 km s−1 appear to be the norm (e.g. Puech strong velocity gradients (e.g. LARS 8; see also Appendix A.8), et al. 2006; Genzel et al. 2006; Law et al. 2009; Förster Schreiber especially when observed with PMAS’ 0.500 × 0.500 magnifica- et al. 2009; Erb et al. 2014; Wisnioski et al. 2015). In the local tion mode (e.g. LARS 12; see also Appendix A.12). In principle, z . 0.1 Universe, high intrinsic velocity dispersions are also the effect could be corrected by subtracting a PSF-smeared kine- found in the studies of blue compact galaxies (BCGs; Östlin matical model of the vLOS velocity field. Given the complexity et al. 1999, 2001), local Lyman break analogues (Basu-Zych of our observed velocity fields, however, this is not practicable. et al. 2009; Gonçalves et al. 2010) and (ultra-)luminous infrared And, moreover, at z ∼ 0.1 Green et al.(2014) find that for clas- galaxies (e.g. Monreal-Ibero et al. 2010). Finally, σ0 values of sical rotating disk kinematics the velocity dispersions corrected comparable amplitude are also common in a sample of 67 bright 40.6 42.6 by a kinematical model are, on average, negligible. Comparisons Hα emitters (10 ≤ LHα ≤ 10 )(Green et al. 2010, 2014). between adaptive-optics derived PSF-smearing unaffected σ0 Indeed, Green et al.(2010) and Green et al.(2014) show that star values to those derived from seeing-limited observations were formation rate (SFR) is positively correlated with intrinsic veloc- presented by Bassett et al.(2014). They find that while for one ity dispersion. We confirm this trend among our LARS galaxies galaxy σ0 remains unaffected by an increase in spatial resolu- and discuss its implications on Lyα observables in Sect. 6.3. −1 tion, for the other galaxy σ0 decreases by ∼10 km s . Bassett et al.(2014) attribute the ∼10 km s−1 discrepancy in one of their galaxies to a strong velocity gradient that is co-spatial with the 5.2.3. vshear/σ0 ratio strongest emission. A similar comparison is possible for three The ratio vshear/σ0 is a metric to quantify whether the gas kine- of our LARS galaxies – LARS 12, LARS 13 and LARS 14. matics are dominated by turbulent or ordered (in some cases or- For these galaxies adaptive optics IFS observations were pre- bital) motions. Our results for this quantity are listed in Table2. sented by Gonçalves et al.(2010) (see Appendices A.12–A.14). Given the above (Sects. 5.2.1 and 5.2.2) mentioned differences Gonçalves et al.(2010) derive σ = 67 km s−1, σ = 74 km s−1 0 0 to disks, it appears cogent that our vshear/σ0 ratios (median 1.4, and σ = 71 km s−1 for LARS 12, LARS 13, and LARS 0 mean 1.6) are much smaller than typical disk vshear/σ0 ratios 14, respectively. These values are in agreement with our mea- ∼4−8 (first to last quartile of the Epinat et al. sample). −1 −1 surements of σ0 = 73 km s , σ0 = 69 km s , and σ0 = −1 10 Objects with vshear/σ0 < 1 are commonly labelled 67 km s for those galaxies . From all these considerations we dispersion-dominated galaxies (Newman et al. 2013; Glazebrook 2013). According to this criterion, five galax- 9 This effect is similar to beam smearing in radio interferometric imag- ies of the 14 LARS sample are dispersion dominated: LARS 2, ing observations. LARS 5, LARS 7, LARS 10, and LARS 14; three perturbed ro- 10 For LARS 13, Gonçalves et al.(2010) do not sample the entire galaxy tators and two with complex kinematics. Dispersion-dominated in their observations; see Sect. A.13. galaxies are frequently found in kinematic studies of high-z,

A78, page 14 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics star-forming galaxies. Local samples with a similar percentage provides further evidence for the relevance of localised outflows (∼35%) of dispersion-dominated galaxies are the blue compact in shaping the observed Lyα morphology. galaxies studied by Östlin et al.(2001) or the Lyman break However, in LARS 3 the interpretation is already not as analogues studied by Gonçalves et al.(2010). In Sect. 6.3, we straightforwards. Here the broadest Hα lines do not occur co- show that dispersion-dominated systems are more likely to have spatial with the highest SFR density. Although also here the COS a significant fraction of Lyα photons escaping. spectrum shows a bulk outflow of cold neutral gas (Paper V), the locally enhanced dispersion values seen in the PMAS maps appear not to be related to this outflow. Instead, it appears more 5.2.4. Integrated velocity dispersion σ tot naturally that the high turbulence within the ionised gas is not For reference we also compute the integrated (spatially aver- primarily driven by star formation activity, but is rather the re- aged) velocity dispersion σtot. This measure provides a useful sult of the violent interaction with the north-western companion comparison for studies of unresolved distant galaxies where no (see e.g. Teyssier et al. 2010 for theoretical support). The inter- disentanglement between ordered and disordered, random mo- pretation is further complicated by the fact that the HST imaging tions are possible (e.g. McLinden et al. 2011; Guaita et al. 2013; data resolves the local Lyα knot on smaller scales than which we McLinden et al. 2014; Rhoads et al. 2014; Erb et al. 2014). are probing with our PMAS observations. We obtain σtot by calculating the square root of the second Nevertheless, there is another example among the LARS central moment of the integrated Hα profiles shown in Fig.3. galaxies for Lyα escape being related to an outflow: LARS 5 The results are given in the fourth column of Table2. Our in- (cf. paper VI). For this galaxy, we already presumed in Paper II tegrated velocity dispersions for the LARS sample range from the existence of a starburst-driven wind based solely on the fil- 46 km s−1 to 115 km s−1, with 70 km s−1 being the median of amentary structure seen in the HST Hα image. Our dispersion the sample. Given the high S/N of the integrated profiles the for- map now reveals that these filaments fill the interior of a biconi- −2 −1 mal uncertainties on this σtot are very small (∼10 km s ). As cal structure, with its base confined on the most prominent star- expected, when large-scale motions dominate in the integrated forming clump (see also Appendix A.5), which is fully in ac- spectrum (i.e. high vshear/σ0 ratios), σtot is significantly larger cordance with theoretical expectations for an evolved starburst- than σ0, but for dispersion-dominated systems the discrepancy driven superwind (e.g. Cooper et al. 2008). Also, here Lyα becomes less extreme. photons could preferentially escape the disk through the cavity blown by the wind. The strongest Lyα enhancement is found directly above and below the main star-forming clump and the 6. Discussion: influence of Hα kinematics low surface brightness Lyα halo is then produced by those pho- on a galaxy’s Lyα emission tons scattering on a bipolar shell-like structure of ambient neutral gas swept up by the superwind. Unfortunately, the brightest 6.1. Clues on Lyα escape mechanisms via spatially resolved hot spots in Lyα above and below the plane are at the base of Hα kinematics this outflowing cone and occur at scales that we cannot resolve Recent theoretical modelling of the Lyα radiative transport in our PMAS data. Similar but less spectacular examples of el- within realistic interstellar medium environments predicts that evated velocity dispersions above and below a disk can be seen small-scale interstellar medium physics are a decisive factor in LARS 7 and LARS 11, with both galaxies also embedded in in regulating the Lyα escape from galaxies (Verhamme et al. low surface brightness Lyα halos. 2012; Behrens & Braun 2014). In particular, Behrens & Braun Another possible example for the importance of outflow (2014) demonstrate how supernova blown outflow cavities be- kinematics in facilitating direct Lyα escape is the most luminous come favoured escape channels for Lyα radiation. These cavi- LAE in the sample: LARS 14 (see also Appendix A.14). Here ties naturally occur in zones of enhanced star formation activity, the observed Hα profiles are always characterised by an under- where the input of kinetic energy from supernovae and stellar lying fainter broad component (see Fig.3) that is believed to be winds into the surrounding interstellar medium is also expected directly related to outflowing material (e.g. Yang et al. 2015). to drive highly turbulent gas flows. Do we see supporting ev- We conclude that while in some individual cases a causal idence for this scenario in our IFS observations of the LARS connection between spatially resolved Hα kinematics and lo- galaxies? calised outflow scenarios might be conjectured, globally this is As described in Sect. 5.1, all our velocity fields are char- not a trend seen in our observations. Furthermore, Hα kinemat- acterised by broad profiles that are best explained by turbu- ics tell only one part of the whole story, and the suggested out- lent flows of ionised gas. The more violent these flows are, the flow scenarios should also be traceable by gas with a high degree more likely they carve holes or bubbles through the interstel- of ionisation. At least for one galaxy with Lyα imaging similar lar medium through which the ionised gas eventually flows out to the LARS galaxies, such a connection has been demonstrated: into the galaxy’s halo. Close inspection of Fig.5 reveals, that in ESO338-IG04. Recently, Bik et al.(2015) analysed observations LARS 1, LARS 3, and LARS 13 (Fig.7) zones of high Ly α sur- of this galaxy obtained with the MUSE integral field spectro- face brightness occur near or even co-spatial to zones where for graph (Bacon et al. 2014). They show that the Lyα fuzz seen these galaxies the maximum velocity dispersions are observed. around the main star-forming knot of this galaxy (Hayes et al. In both LARS 1 and LARS 13, these zones are also co-spatial 2005; Östlin et al. 2009) can be related to an outflow, which with the regions of highest Hα surface-brightness, hence, here can be traced both with Hα kinematics and by a high degree of the SFR density is highest. Moreover, in LARS 1 the Hα image ionisation. In contrast to our observations, Bik et al.(2015) see shows a diffuse fan-like structure emanating outwards from this the fast outflowing material in their vLOS map as significantly star-forming knot (Paper II; Paper I). Finally, the analysis of the redshifted Hα emission. No similar prominent effect is appar- COS spectra in Paper V shows blueshifted, low-ionisation state ent in our vLOS maps. Nevertheless, the vLOS fields of LARS 7 interstellar absorption lines in LARS 1 and LARS 13 indica- and LARS 12 display strongly localised redshifts in Hα that are tive of outflowing neutral gas. Hence, for those two galaxies, to- spatially coincident with filamentary Hα fingers, hence, here we gether with the high-resolution HST imaging data, our IFS data might also see a fast outflow pointed away from the observer. In

A78, page 15 of 27 A&A 587, A78 (2016)

that observational difficulties are the source of the σtot–W50 350 difference in this galaxy. In LARS 7, however, we suspect the difference to be genuine. In this galaxy, the Hα morphology is significantly puffed up and rounder compared to the disk-like continuum (see also Appendix A.7). Therefore, we suspect that 300 the smaller W50 measurement indicates that bulk of H  is in a

] kinematically more quiescent state then the ionised gas. This

1 Lyα − galaxy is one of the stronger LAEs in the sample ( fesc. = 0.14 s 250 and EWLyα = 40 Å). That considerable amounts of Lyα photons m

k escape from LARS 7 was noted as peculiar in Paper V, since [

0 the metal absorption lines indicated there was a large amount 5 200

W of neutral gas sitting at the systemic velocity of the galaxy. Our 01 07 observations now indicate that the intrinsic Lyα photons are pro- 02 08 duced in gas that is less quiescent than the scattering medium, 150 03 09 which therefore is more transparent for a significant fraction of 04 10 the intrinsic Lyα photons. 05 11 Spatially resolved VLA velocity fields are available for only 100 06 a subset of five LARS galaxies (LARS 2, LARS 3, LARS 4, LARS 8, and LARS 9). They represent rotations disturbed by 100 150 200 250 300 350 1 interactions with neighbours. In four of them, H  kinematical 2.355 σ [km s− ] × tot axes have similar orientations as those of the PMAS Hα velocity field. In the fifth object, the complex interacting system Fig. 8. Comparison of integrated H  linewidth W to FWHM of the Hα √ 50 LARS 9, the Hα and H  velocity fields show different charac- × integrated velocity dispersion 2 2 ln 2 σtot. The dashed line indicates teristics, but the complexities seen in the PMAS maps of this the one-to-one relation. For LARS 6, (encircled point) we plot as W50, a preliminary result based upon VLA C-configuration interferometry, galaxy are on scales far beyond the spatial resolving power of the VLA D configuration (see also Appendix A.9). However, as the beam of the GBT profile (used to derive W50 in Paper III) is too broad to separate LARS 6 from a neighbouring galaxy. some of the LARS galaxies have already been observed with the VLA in C and B configuration and the analysis is currently in progress. For the first time, these data will allow a comparison between Hα,H  and Lyα on meaningful physical scales. It is addition, in those galaxies, no co-spatial relation between Lyα evident that these comparisons will provide a critical benchmark emissivity and the suspected outflows can be established. for our understanding of Lyα radiative transfer in interstellar and circum-galactic environments. 6.2. Comparison of Hα to HI observations Radiative transfer of Lyα photons depends on the relative veloc- 6.3. Relations between kinematical properties and galaxy ity differences between scattering H  atoms and Lyα sources. If parameters the Lyα sources are out of resonance with the bulk of the neutral medium, they are less likely scattered, and in turn more likely to In Sect. 5.2 we quantified the global kinematical properties of escape the galaxy. It is exactly this interplay between the ionised the LARS galaxies using the non-parametric estimators vshear, and neutral ISM phases that is determining the whole Lyα ra- σ0 and vshear/σ0. Before linking these observables to global Lyα diative transfer. In Paper III we studied LARS galaxies in the properties of the LARS galaxies (cf. Sect. 6.4), we need to un- H  21 cm line, using single-dish GBT and VLA D-configuration derstand which galaxy parameters are encoded in them. observations. We now attempt a comparison between the neutral We find strong correlations between stellar mass M? and and ionised gas kinematics in the LARS galaxies, cognisant of vshear, as well as SFR and σ0 (M? and SFR from Paper II). the fact that the spatial scales probed by both instruments are Graphically we show these correlations in Fig.9 (left panel) much larger than our PMAS H  observations. and Fig. 10 (centre panel). The Spearman rank correlation co- Generally, our GBT H  spectra have low S/N, and for the efficients (e.g. Wall 1996) are ρs = 0.763 for the M?-vshear rela- three most distant LARS galaxies, LARS 12, LARS 13, and tion and ρs = 0.829 for the SFR-σ0 relation. This corresponds to LARS 14, we could not detect any significant signal at all. The likelihoods of the null hypothesis that no monotonic relation ex- profiles are mostly single or multiple peaked, but rarely show ists between the two parameters of p0 = 0.02% and p0 = 0.3% 11 a classical double-horn profile that would be expected for a flat (two-tailed test ) for the SFR-σ0 and M?-vshear relation, respec- rotation curve. Hence, qualitatively our observed GBT H  line tively. We discuss these two relations in more detail in Sect. 6.3.1 profiles are consistent with our PMAS results that most of the and Sect. 6.3.2, where we also introduce the comparison samples LARS galaxies are kinematically perturbed or sometimes even shown in Fig.9 and Fig. 10. strongly interacting systems. Besides this tight correlation between SFR and σ0, there is In Paper III we measured the width of the H  lines at also a weaker correlation between SFR and vshear in our sample 50% of the line peak from the GBT spectra. In Fig.8 we (ρs = 0.775, p0 = 0.1%). We show the data in Fig. 10 (left compare this quantity, W50, to the FWHM of the integrated panel). While not shown graphically, the DYNAMO disks would Hα velocity dispersion σtot (Sect. 5.2.4). Notably, two sys- scatter over the whole SFR-vshear plane, also filling the upper tems deviate significantly from the one-to-one relation: LARS 7 left corner in this diagram. Finally, in Fig. 10 (right panel) we and LARS 10. LARS 10 shows the lowest S/NH  spectrum and moreover our PMAS FoV does not capture two smaller 11 We consider correlations as significant when the likelihood of the star-forming clumps in the south-east. Therefore, we believe null hypothesis, p0, is smaller than 5%.

A78, page 16 of 27 110

100

80 ] E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics 1 − s 70

m 200 110 3.5 k

[ 60 01 08

0 02 09 σ 50 100 03 10 3.0 04 11 40 05 12 90 150 30 06 13 2.5 07 14 80 20 ] ] 1 1 0 20 40 60 − 80 100 120 s 0 corr. − 70 2.0 1 s SFR [M yr− ] σ /

Hα m r

100 m k

¯ a [ e k h [ r 60 s a

v 1.5 e 0 h σ s v 50 1.0 50 40

ρ = 0.763 ρ = 0.362 0.5 ρ = 0.723 s 30 s s p0 = 0.003 p0 = 0.2 p0 = 0.003 0 20 0.0 109 1010 1011 109 1010 1011 109 1010 1011 M [M ] M [M ] M [M ] ¯ ¯ ¯

Fig. 9. Global Hα kinematical parameters vshear (left panel), σ0 (middle panel), and vshear/σ0 (right panel) in comparison to stellar mass for LARS galaxies (from Paper II) and in comparison to literature values: DYNAMO z ∼ 0.1 [compact] perturbed rotators from Green et al.(2014) as [small] green circles; DYNAMO z ∼ 0.1 [compact] complex kinematics from Green et al.(2014) as [small] green squares; Local Lyman Break Analogues from Gonçalves et al.(2010) as blue circles; Keck /OSIRIS resolved z ∼ 2−3 star-forming galaxies from Law et al.(2009) as red circles; SINS z ∼ 2−3 galaxies from Förster Schreiber et al.(2009) as orange squares (only in the left panel). Uncertainties given where available. The LARS galaxies are represented by symbols according to the legend in the left panel. Spearman rank correlation coeﬃcients ρs and corresponding p0 values for the LARS galaxies are shown in each panel.

200 110 3.5

ρs = 0.775 100 ρs = 0.829 ρs = 0.507 3.0 p0 = 0.001 p0 = 0.0002 p0 = 0.06 90 150 2.5 80 ] ] 1 1 − s 0 − 70 2.0 s σ / m r

100 m k a [ e k h [ r 60 s a

v 1.5 e 0 h σ s v 50 1.0 50 40

0.5 30

0 20 0.0 100 101 102 100 101 102 100 101 102 corr. 1 corr. 1 corr. 1 SFRHα [M yr− ] SFRHα [M yr− ] SFRHα [M yr− ] ¯ ¯ ¯

Fig. 10. Global Hα kinematical parameters vshear (left panel), σ0 (middle panel), and vshear/σ0 (right panel) in comparison to SFR for LARS galaxies (from Paper II) and in comparison to literature values (same symbols as in Fig.9). Uncertainties given where available. Spearman rank correlation coeﬃcients ρs and corresponding p0 values for the LARS galaxies are shown in each panel.

find that vshear/σ0 is not significantly correlated with the SFR in 6.3.1. SFR-σ0 correlation LARS (ρs = 0.507, p0 = 6%). The highly significant correlation between σ0 and We also check in Fig.9 (centre panel) for a M?-σ0 correla- SFR has long been established for giant H  regions tion, but with ρs = 0.362 the null hypothesis that the variables (Terlevich & Melnick 1981). Recently, it has been shown are uncorrelated cannot be rejected (p0 = 20%). Similar low that it extends over a large dynamical range in star formation correlation-coefficients are found for the comparison samples in and mass, not only locally but also at high redshifts (Green Fig.9 (centre panel). Since there is likely a monotonic relation et al. 2010, 2014). The physical nature of the σ0-SFR relation between M?-vshear and M?-σ0 are uncorrelated, the monotonic might be that star formation feedback powers turbulence in the relation M?-vshear/σ0 (ρs = 0.723, p0 = 0.3%) – Fig.9 (right interstellar medium. On the other hand, the processes that lead panel) – is expected. to a high SFR might also be responsible for producing a high σ0. In particular, inflows of cold gas that feed the star formation From these results, we conclude that dispersion-dominated processes are expected to stir up the interstellar medium and galaxies in our sample are preferentially low-mass systems thus lead to turbulent flows (e.g. Wisnioski et al. 2015, and 10 with M? . 10 M . This result is also commonly found references therein). in samples of high-z, star-forming galaxies (Law et al. 2009; Graphically we compare, in Fig. 10 (centre panel), the Förster Schreiber et al. 2009; Newman et al. 2013). LARS SFR-σ0 points to the values from the DYNAMO

A78, page 17 of 27 A&A 587, A78 (2016) galaxies (Green et al. 2014) with complex kinematics and the and higher vshear measurements and that galaxies of higher mass DYNAMO perturbed rotators. To put this result in context with also show higher vshear values and higher vshear/σ0 ratios (Figs.9 high-z studies, in Fig. 10 we also show the SFR-σ0 points and 10). We now compare our vshear, σ0 and vshear/σ0-ratios to from the Keck/OSIRIS z ∼ 2−3 star-forming galaxies by the aperture integrated Lyα observables EWLyα, Lyα/Hα and Law et al.(2009), and the local Lyman break analogues by Lyα fesc from Paper II. We do this in form of a graphical 3 × 3 Gonçalves et al.(2010). An exhaustive compilation of SFR- σ0 matrix in Fig. 11. In each panel, we include the Spearman rank measurements is presented in Green et al.(2014), and the LARS correlation coefficient ρs and the likelihood p0 to reject the null SFR-σ0 points do not deviate from this relation. hypothesis. From the centre row in Fig. 11 it is obvious that none of the Lyα observables correlates with the averaged intrinsic velocity 6.3.2. M?-vshear correlation dispersion σ0. As σ0 correlates positively with SFR, this signi- Given the complexity of the ionised gas velocity fields seen fies that the observed Lyα emission is a bad SFR calibrator. In in the LARS galaxies, a tight relation between our measured, Fig. 11 it is also evident that galaxies with higher shearing veloc- −1 inclination-uncorrected vshear values with stellar mass appears ities (vshear & 50 km s ) have preferentially lower EWLyα, lower surprising. It indicates that, at least in a statistical sense, vshear Lyα/Hα, and lower f Lyα. Therefore, according to the above pre- is tracing systemic rotation in our systems and that the scatter in esc sented M?-vshear relation (Sect. 6.3.2) LAEs are preferentially our relation is dominated by the unknown inclination correction 10 found among the systems with M? . 10 M in LARS. This de- (see also Law et al. 2009). ficiency of strong LAEs among high-mass systems was already To put our data in context with other studies, we compare noted in Paper II. Here we see this trend from a kinematical per- M v our ?- shear data points in Fig.9 (left panel) to the Green et al. spective now. Also, a low vshear value, although a necessary con- (2014) DYNAMO sample and to the Gonçalves et al.(2010) lo- dition, seems not to be sufficient to have significant amounts Lyα cal Lyman break analogues. Our high-z comparison samples are photons escaping (e.g. LARS 6). Again, this shows that Lyα es- the Keck/OSIRIS z ∼ 2−3 star-forming galaxies Law et al. and cape from galaxies is a complex multi-parametric problem. the z ∼ 1−3 SINS sample by Förster Schreiber et al.(2009). For Although the LARS sample is small, the result that seven of the rotation-dominated and perturbed rotators in the DYNAMO eight non-LAEs have vshear/σ0 > 1 and four of six LAEs have sample, Green et al. tabulate rotation velocity at 2.2 disk scale v /σ < 1 signals that dispersion-dominated kinematics are lengths obtained from fitting model disks to their velocity fields. shear 0 an important ingredient in Lyα escape. Again, lower vshear/σ0 We convert these to an inclination-uncorrected value via multi- ratios are found in lower M objects in our sample and σ is plication with sin i. Again, for the DYNAMO sample we only ? 0 uncorrelated with M?. Therefore, the correlation between Lyα compare to their objects with complex kinematics or their per- observables and v /σ is a consequence of the correlation be- turbed rotators, as they are dominant in our sample. shear 0 tween Lyα observables and M?. Nevertheless, low vshear/σ0 ra- It is apparent from Fig.9 (left panel) that while our data tios also state that the ionised gas in those galaxies must be in a points line up well with the z ∼ 0.1 galaxies, a significant num- turbulent state. High SFR appears to be connected to an increase ber of high-z galaxies shows lower vshear values at given stellar turbulence, but it is currently not clear whether the processes that 10 11 mass in the range 10 M −10 M . The reason for this is that cause star formation or feedback from star formation are respon- high-z studies are not sensitive enough to reach the outer faint sible for the increased turbulence (Sect. 6.3.1). Regardless of isophotes, hence, they are biased towards lower v values (see shear the causal relationship between SFR and σ0, our results support Sect. 5.2.1). Indeed, Law et al.(2009) report a detection limit of that Lyα escape is being favoured in low-mass systems under- −1 −2 ∼1 M yr kpc , while Förster Schreiber et al.(2009) report going an intense star formation episode. Similarly Cowie et al. −1 −2 ≈0.03 M yr kpc as an average for their sample, which is (2010) found that in a sample of UV bright z ∼ 0.3 galaxies, comparable to our depth (Sect. 5.1). LAEs are primarily the young galaxies that have recently become strongly star forming. In the local Universe dispersion- dominated systems with vshear/σ0 < 1 are rare, but they be- 6.4. Relations between global kinematical properties come much more prevalent at higher redshifts (Wisnioski et al. and Lyα observables 2015). Coincidentally, the number density of LAEs rises towards higher redshifts (Wold et al. 2014). Therefore we speculate, that We now explore trends between global kinematical properties dispersion-dominated kinematics are indeed a necessary require- derived from the ionised gas in the LARS galaxies and their Lyα ment for a galaxy to have a significant amount of Lyα photons observables determined in Paper II, namely Lyα escape frac- escaping. Lyα tion fesc , ratio of Lyα to Hα flux Lyα/Hα, and Lyα equivalent width EW . For reference, we list these quantities here Lyα 6.5. Lyα extension and Hα kinematics again in Table2. We recall that Ly α/Hα is the observed flux ratio, while fesc is determined from the intrinsic luminosities, i.e. All of the LARS LAEs show a significantly more extended Lyα after correcting the fluxes for dust reddening. morphology compared to their appearance in Hα or UV con- Regarding the qualitative classification of the vLOS maps in tinuum. These large-scale Lyα haloes appear to completely en- Sect. 5.1 into rotating disks, perturbed rotators and objects show- compass the star-forming regions. LARS 1, LARS 2, LARS 5, ing complex kinematics there is no preference in any of the glob- LARS 7, LARS 12 and LARS 14 are the most obvious exam- ally integrated Lyα observables. Both the maximum and mini- ples of extended Lyα emission, but the phenomenon is visible mum of these observables occur in the complex kinematics class. in all our objects, even for those galaxies that show Lyα in Therefore, just the qualitative appearance of the velocity field absorption (Figs.5–7; cf. Hayes et al. 2013 and Paper II). At seems not to predict whether the galaxy is a LAE or not. high redshift the ubiquity of extended Lyα haloes around LAEs In the previous section, we showed that within the LARS was recently revealed by Wisotzki et al.(2015) on an individ- sample higher SFRs are found in galaxies that have higher σ0 ual object-by-object basis. However, in contrast to the LARS

A78, page 18 of 27 110 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

100 200 200 200 90 ρs = 0.618 ρs = 0.548 ρs = 0.561 80 − − − ]

1 p = 0.018 p = 0.043 p = 0.037 − 0 0 0 s 70 150 150 150 m k

[ 60 01 08 0

] 02 09] ] σ 50 1 03 1 10 1 − − − s 04 11s s 40 m 100 05 12m 100 m 100 k k k [ [ [ r r r

a 06 13a a

30 e e e h h h

s 07 14s s 20 v v v 0 20 40 60 80 100 120 corr. 1 50 SFRHα [M yr− ] 50 50 ¯

0 0 0 0 20 40 60 80 0 1 2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 110 110 Lyα/Hα 110 Lyα EWLyα [ ] fesc ρ =0.04 ρ =0.057 ρ = 0.084 100 s 100 s 100 s −

p0 =0.893 p0 =0.846 p0 = 0.776 90 90 90

80 80 80 ] ] ] 1 1 1

− 70 − 70 − 70 s s s m m m k k k

[ 60 [ 60 [ 60 0 0 0 σ σ σ

50 50 50

40 40 40

30 30 30

20 20 20 0 20 40 60 80 0 1 2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 Lyα/Hα Lyα 3.5 EWLyα [ ] 3.5 3.5 fesc ρ = 0.744 ρ = 0.669 ρ = 0.647 s − s − s − 3.0 3.0 3.0 p0 = 0.002 p0 = 0.009 p0 = 0.012

2.5 2.5 2.5 0 0 0 σ σ σ

/ 2.0 / 2.0 / 2.0 r r r a a a e e e h h h s s s v v v 1.5 1.5 1.5

1.0 1.0 1.0

0.5 0.5 0.5

0 20 40 60 80 0 1 2 3 4 5 0.0 0.1 0.2 0.3 0.4 0.5 Lyα/Hα Lyα EWLyα [ ] fesc

Fig. 11. Relations between global Lyα properties of the LARS galaxies to global kinematical parameters vshear, σ0 and vshear/σ0.

Lyα haloes, the high-z haloes appear to have ∼10× larger ex- of their circum-galactic neutral gas into forming stars and subse- tents at a given continuum radius (Wisotzki et al. 2015, their quently dust (Hayes et al. 2013). In such a scenario, the larger ex- Fig. 12). In order to quantify the spatial extent of observed rel- tent of high-z Lyα haloes could be related to the circum-galactic ative to intrinsic Lyα emission in Hayes et al.(2013) we de- gas-reservoirs being larger in the early universe. fined the “Relative Petrosian Extension” ξLyα as the ratio of the In Fig. 12 we show that there are no significant correlations Petrosian radii (Petrosian 1976) at η = 20% measured in Lyα between ξLyα and the global kinematical Hα parameters vshear, Lyα Hα and Hα: ξLyα = RP20 /RP20. For reference, we list the ξLyα val- σ0 and vshear/σ0. Therefore, we reason that the kinematics of the ues of the LARS galaxies here again in Table2. Based on an interstellar medium do not strongly influence the appearance of anti-correlation between ξLyα and UV slope we conjectured that the haloes and that the Lyα halo phenomenon is only related to smaller Lyα haloes occur in galaxies that have converted more the presence of a full gas reservoir. Further, 21 cm HI imaging

A78, page 19 of 27 A&A 587, A78 (2016)

200 disk in only two galaxies and in ﬁve a disturbed rotational signature is apparent. With respect to Lyα escape, we ﬁnd Lyα ρ = 0.295 no preference of high EWLyα or high fesc. values towards any s − 150 of those classes, but the minimum and maximum values both p0 = 0.352

] occur in objects showing complex kinematics in our sample. 1 −

s 2. A common feature in all LARS galaxies are high Hα velocity −1 m 100 k dispersions. With vFWHM & 100 km s our measurements [ r

a are in contrast to values typically seen in local spirals, but in e h s

v high-z star-forming galaxies such high values appear to be

50 the norm. 3. While we could not infer a direct relation between spatially resolved kinematics of the ionised gas and photometric Lyα properties for all LARS galaxies, in individual cases our 0 0 1 2 3 4 maps appear qualitatively consistent with outflow scenarios ξLyα 110 that promote Lyα escape from high-density regions. 4. Currently, a spatially resolved comparison of our H  veloc- 100 ity fields to our H  data is severely limited since the scales ρs =0.018 90 resolved by the radio observations are significantly larger. p0 =0.957 However, for one galaxy (LARS 7) the difference between 80

] globally integrated velocity dispersion of ionised and neutral 1 − s 70 gas oﬀers a viable explanation for the escape of signiﬁcant m

k amounts of Lyα photons. [ 60 0

σ 5. From our Hα velocity maps, we derive the non-parametric 50 statistics vshear, σ0 and vshear/σ0 to quantify the kinematics 40 of the LARS galaxies globally. Our vshear values range from −1 −1 30 km s to 180 km s and our σ0 values range from 30 −1 −1 40 km s to 100 km s . For our ratios vshear/σ0, we find a 20 median of 1.4, and five of the LARS galaxies are dispersion- 0 1 2 3 4 ξLyα dominated systems with vshear/σ0 < 1. 3.5 6. A positively correlated σ0 with SFR in the LARS galaxies is fully consistent with other IFS studies. We also find strong 3.0 ρ = 0.484 correlations between M? and vshear, M? and vshear/σ0, and s − SFR and vshear. In view of the M?-vshear correlation the SFR- 2.5 p0 = 0.111 vshear correlation implies that more massive galaxies have

0 2.0 higher overall star formation rates in our sample. σ

/ Lyα r

a 7. The Lyα properties EW , Lyα/Hα and f do not cor-

e Lyα esc. h s 1.5 v relate with σ0, but they correlate with vshear and vshear/σ0 (Fig. 11). We ﬁnd no correlation between the global kine- 1.0 matical statistics and the extent of the Lyα halo. 8. Four of six LARS LAEs are dispersion-dominated systems 0.5 with vshear/σ0 < 1 and 7 of 8 non-LAEs have vshear/σ0 > 1.

0.0 0 1 2 3 4 Observational studies of Lyα emission in local-Universe galax- ξLyα ies have so far focused on imaging and UV spectroscopy (for a recent review see Hayes 2015). For the first time, we pro- Fig. 12. Relations between ξLyα, the “relative Petrosian extension of vided empirical results from IFS observations for a sample of Lyα” as defined in Hayes et al.(2013), and global kinematical param- galaxies with known Lyα observables. In our pioneering study, eters vshear, σ0 and vshear/σ0 with symbols according to the legend in Fig.9. For galaxies with no Ly α emission (LARS 4 and LARS 6, cir- we focused on the spectral and spatial properties of the intrinsic Lyα radiation field as traced by Hα. We found a direct rela- cled symbols) ξLyα is defined as zero and for LARS 9 the measured ξLyα presents a lower limit (see also Hayes et al. 2013). Spearman rank tion between the global non-parametric kinematical statistics of Lyα correlation coefficients ρs and corresponding p0 values are calculated the ionised gas and the Lyα observables fesc. and EWLyα, and excluding LARS 4 and LARS 6. our main result is that dispersion-dominated systems favour Lyα escape. The prevalence of LAEs among dispersion-dominated galaxies could be simply a consequence of these systems be- of the LARS galaxies at high spatial resolution is needed to test ing the lower mass systems in our sample, which is a result al- this scenario. ready found in Paper II. However, the observed turbulence in actively star-forming systems should be related to ISM conditions that ease Lyα radiative transfer out of high-density envi- 7. Summary and conclusions ronments. In particular, if turbulence is a direct consequence of We obtained the following results from our integral field spec- star formation, then the energetic input from the star formation troscopic observations of the Hα line in the LARS galaxies: episode might also be powerful enough to drive cavities through the neutral medium into the lower density circum-galactic en- 1. Half of the LARS galaxies show complex Hα kinematics. vironments. Of course, the kinematics of the ionised gas offer There are kinematical properties consistent with a rotating only a limited view of the processes at play, and in future studies

A78, page 20 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics we will attempt to connect our kinematic measurements to spa- Green, A. W., Glazebrook, K., McGregor, P. J., et al. 2010, Nature, 467, 684 tial mappings of the ISMs ionisation state. The synergy between Green, A. W., Glazebrook, K., McGregor, P. J., et al. 2014, MNRAS, 437, 1070 IFS data, space-based UV imaging, and high-resolution H  ob- Greisen, E. W., & Calabretta, M. R. 2002, A&A, 395, 1061 Gronke, M., & Dijkstra, M. 2014, MNRAS, 444, 1095 servations of nearby star-forming galaxies will be vital to build Gronke, M., Bull, P., & Dijkstra, M. 2015, ApJ, 812, 123 a coherent observational picture of Lyα radiative transport in Gronwall, C., Ciardullo, R., Hickey, T., et al. 2007, ApJ, 667, 79 galaxies. Grove, L. F., Fynbo, J. P. U., Ledoux, C., et al. 2009, A&A, 497, 689 Guaita, L., Francke, H., Gawiser, E., et al. 2013, A&A, 551, A93 Guaita, L., Melinder, J., Hayes, M., et al. 2015, A&A, 576, A51 (Paper IV) Acknowledgements. E.C.H. especially thanks Sebastian Kamann and Bernd Hashimoto, T., Ouchi, M., Shimasaku, K., et al. 2013, ApJ, 765, 70 Husemann for teaching him how to operate the PMAS instrument. We thank Hashimoto, T., Verhamme, A., Ouchi, M., et al. 2015, ApJ, 812, 157 the support staﬀ at Calar Alto observatory for help with the visitor-mode Hayes, M. 2015, PASA, 32, 27 observations. All plots in this paper were created using matplotlib Hayes, M., Östlin, G., Mas-Hesse, J. M., et al. 2005, A&A, 438, 71 (Hunter 2007). Intensity related images use the cubhelix colour scheme by Hayes, M., Östlin, G., Mas-Hesse, J. M., & Kunth, D. 2009, AJ, 138, 911 Green(2011). This research made extensive use of the astropy pacakge Hayes, M., Östlin, G., Schaerer, D., et al. 2010, Nature, 464, 562 (Astropy Collaboration et al. 2013). M.H. acknowledges the support of the Hayes, M., Östlin, G., Schaerer, D., et al. 2013, ApJ, 765, L27 Swedish Research Council, Vetenskapsrådet, and the Swedish National Space Hayes, M., Östlin, G., Duval, F., et al. 2014, ApJ, 782, 6 (Paper II) Board (SNSB) and is Academy Fellow of the Knut and Alice Wallenberg Henry, A., Scarlata, C., Martin, C. L., & Erb, D. 2015, ApJ, 809, 19 Foundation. HOF is currently granted by a Cátedra CONACyT para Jóvenes Hong, S., Dey, A., & Prescott, M. K. M. 2014, PASP, 126, 1048 Investigadores. I.O. has been supported by the Czech Science Foundation grant Horne, K. 1986, PASP, 98, 609 GACR 14-20666P. D.K. is funded by the Centre National d’Études Spatiales Hu, E. M., Cowie, L. L., & McMahon, R. G. 1998, ApJ, 502, L99 (CNES). F.D. is grateful for ﬁnancial support from the Japan Society for the Hunter, J. D. 2007, Comput. Sci. Eng., 9, 90 Promotion of Science (JSPS) fund. P.L. acknowledges support from the ERC- Husemann, B., Kamann, S., Sandin, C., et al. 2012, A&A, 545, A137 StG grant EGGS-278202. Husemann, B., Jahnke, K., Sánchez, S. F., et al. 2013, A&A, 549, A87 Iye, M. 2011, Proc. Jap. Acad. Ser. B, 87, 575 Jaskot, A. E., & Oey, M. S. 2014, ApJ, 791, L19 Jiménez-Vicente, J., Battaner, E., Rozas, M., Castañeda, H., & Porcel, C. 1999, A&A, 342, 417 References Kennicutt, Jr., R. C. 1998, ARA&A, 36, 189 Kunth, D., Mas-Hesse, J. M., Terlevich, E., et al. 1998, A&A, 334, 11 Adams, J. J., Blanc, G. A., Hill, G. J., et al. 2011, ApJS, 192, 5 Landman, D. A., Roussel-Dupre, R., & Tanigawa, G. 1982, ApJ, 261, 732 Ahn, S.-H., Lee, H.-W., & Lee, H. M. 2003, MNRAS, 340, 863 Laursen, P., & Sommer-Larsen, J. 2007, ApJ, 657, L69 Alonso-Herrero, A., García-Marín, M., Monreal-Ibero, A., et al. 2009, A&A, Laursen, P., Sommer-Larsen, J., & Andersen, A. C. 2009, ApJ, 704, 1640 506, 1541 Laursen, P., Duval, F., & Östlin, G. 2013, ApJ, 766, 124 Astropy Collaboration, Robitaille, T. P., Tollerud, E. J., et al. 2013, A&A, 558, Law, D. R., Steidel, C. C., Erb, D. K., et al. 2009, ApJ, 697, 2057 A33 Lenz, D. D., & Ayres, T. R. 1992, PASP, 104, 1104 Bacon, R., Vernet, J., Borisiva, E., et al. 2014, The Messenger, 157, 13 Lupton, R., Blanton, M. R., Fekete, G., et al. 2004, PASP, 116, 133 Bacon, R., Brinchmann, J., Richard, J., et al. 2015, A&A, 575, A75 Malhotra, S., & Rhoads, J. E. 2004, ApJ, 617, L5 Barnes, L. A., Haehnelt, M. G., Tescari, E., & Viel, M. 2011, MNRAS, 416, Mallery, R. P., Mobasher, B., Capak, P., et al. 2012, ApJ, 760, 128 1723 Marino, A., Iodice, E., Tantalo, R., et al. 2009, A&A, 508, 1235 Bassett, R., Glazebrook, K., Fisher, D. B., et al. 2014, MNRAS, 442, 3206 Martin, D. C., Chang, D., Matuszewski, M., et al. 2014, ApJ, 786, 106 Basu-Zych, A. R., Gonçalves, T. S., Overzier, R., et al. 2009, ApJ, 699, L118 Mas-Hesse, J. M., Kunth, D., Tenorio-Tagle, G., et al. 2003, ApJ, 598, 858 Behrens, C., & Braun, H. 2014, A&A, 572, A74 McLinden, E. M., Finkelstein, S. L., Rhoads, J. E., et al. 2011, ApJ, 730, 136 Behrens, C., Dijkstra, M., & Niemeyer, J. C. 2014, A&A, 563, A77 McLinden, E. M., Rhoads, J. E., Malhotra, S., et al. 2014, MNRAS, 439, 446 Bettoni, D., Galletta, G., Rampazzo, R., et al. 2011, A&A, 534, A24 Monreal-Ibero, A., Arribas, S., Colina, L., et al. 2010, A&A, 517, A28 Bik, A., Östlin, G., Hayes, M., et al. 2015, A&A, 576, L13 Neufeld, D. A. 1990, ApJ, 350, 216 Calabretta, M. R., & Greisen, E. W. 2002, A&A, 395, 1077 Newman, S. F., Genzel, R., Förster Schreiber, N. M., et al. 2013, ApJ, 767, 104 Cannon, J. M., Skillman, E. D., Kunth, D., et al. 2004, ApJ, 608, 768 Oesch, P. A., van Dokkum, P. G., Illingworth, G. D., et al. 2015, ApJ, 804, L30 Cappellari, M., & Copin, Y. 2003, MNRAS, 342, 345 Oke, J. B. 1990, AJ, 99, 1621 Cardamone, C., Schawinski, K., Sarzi, M., et al. 2009, MNRAS, 399, 1191 Ono, Y., Ouchi, M., Mobasher, B., et al. 2012, ApJ, 744, 83 Cassata, P., Le Fèvre, O., Garilli, B., et al. 2011, A&A, 525, A143 Östlin, G., Amram, P., Masegosa, J., Bergvall, N., & Boulesteix, J. 1999, A&AS, Cassata, P., Tasca, L. A. M., Le Fèvre, O., et al. 2015, A&A, 573, A24 137, 419 Chonis, T. S., Blanc, G. A., Hill, G. J., et al. 2013, ApJ, 775, 99 Östlin, G., Amram, P., Bergvall, N., et al. 2001, A&A, 374, 800 Christensen, L., Laursen, P., Richard, J., et al. 2012, MNRAS, 427, 1973 Östlin, G., Hayes, M., Kunth, D., et al. 2009, AJ, 138, 923 Ciardullo, R., Gronwall, C., Wolf, C., et al. 2012, ApJ, 744, 110 Östlin, G., Hayes, M., Duval, F., et al. 2014, ApJ, 797, 11 (Paper I) Cooper, J. L., Bicknell, G. V., Sutherland, R. S., & Bland-Hawthorn, J. 2008, Ouchi, M., Shimasaku, K., Akiyama, M., et al. 2008, ApJS, 176, 301 ApJ, 674, 157 Pardy, S. A., Cannon, J. M., Östlin, G., et al. 2014, ApJ, 794, 101 (Paper III) Cowie, L. L., Barger, A. J., & Hu, E. M. 2010, ApJ, 711, 928 Partridge, R. B., & Peebles, P. J. E. 1967, ApJ, 147, 868 Cox, A. N. 2000, Sky Telesc., 100, 72 Petrosian, V. 1976, ApJ, 209, L1 Davies, R., Förster Schreiber, N. M., Cresci, G., et al. 2011, ApJ, 741, 69 Puech, M., Hammer, F., Flores, H., Östlin, G., & Marquart, T. 2006, A&A, 455, Diehl, S., & Statler, T. S. 2006, MNRAS, 368, 497 119 Dijkstra, M. 2014, PASA, 31, 40 Rauch, M., Haehnelt, M., Bunker, A., et al. 2008, ApJ, 681, 856 Dijkstra, M., Haiman, Z., & Spaans, M. 2006, ApJ, 649, 14 Rhoads, J. E., Malhotra, S., Richardson, M. L. A., et al. 2014, ApJ, 780, 20 Duval, F., Schaerer, D., Östlin, G., & Laursen, P. 2014, A&A, 562, A52 Rivera-Thorsen, T. E., Hayes, M., Östlin, G., et al. 2015, ApJ, 805, 14 (Paper V) Duval, F., Ostlin, G., Hayes, M., et al. 2016, A&A, 587, A77 (Paper VI) Robertson, J. G. 2013, PASA, 30, 48 Epinat, B., Amram, P., & Marcelin, M. 2008a, MNRAS, 390, 466 Roth, M. M., Kelz, A., Fechner, T., et al. 2005, PASP, 117, 620 Epinat, B., Amram, P., Marcelin, M., et al. 2008b, MNRAS, 388, 500 Roth, M. M., Fechner, T., Wolter, D., et al. 2010, in High Energy, Optical, and Epinat, B., Amram, P., Balkowski, C., & Marcelin, M. 2010, MNRAS, 401, 2113 Infrared Detectors for Astronomy IV, SPIE Conf. Ser., 7742, 774209 Erb, D. K., Steidel, C. C., Trainor, R. F., et al. 2014, ApJ, 795, 33 Sánchez, S. F., Aceituno, J., Thiele, U., Pérez-Ramírez, D., & Alves, J. 2007a, Erroz-Ferrer, S., Knapen, J. H., Leaman, R., et al. 2015, MNRAS, 451, 1004 PASP, 119, 1186 Finkelstein, S. L., Papovich, C., Dickinson, M., et al. 2013, Nature, 502, 524 Sánchez, S. F., Cardiel, N., Verheijen, M. A. W., et al. 2007b, A&A, 465, 207 Flores, H., Hammer, F., Puech, M., Amram, P., & Balkowski, C. 2006, A&A, Sánchez, S. F., Kennicutt, R. C., Gil de Paz, A., et al. 2012, A&A, 538, A8 455, 107 Sandberg, A., Östlin, G., Hayes, M., et al. 2013, A&A, 552, A95 Förster Schreiber, N. M., Genzel, R., Bouché, N., et al. 2009, ApJ, 706, 1364 Sandberg, A., Guaita, L., Östlin, G., Hayes, M., & Kiaeerad, F. 2015, A&A, 580, Genzel, R., Tacconi, L. J., Eisenhauer, F., et al. 2006, Nature, 442, 786 A91 Glazebrook, K. 2013, PASA, 30, 56 Sandin, C., Becker, T., Roth, M. M., et al. 2010, A&A, 515, A35 Gonçalves, T. S., Basu-Zych, A., Overzier, R., et al. 2010, ApJ, 724, 1373 Sandin, C., Weilbacher, P., Streicher, O., Walcher, C. J., & Roth, M. M. 2011, Green, D. A. 2011, Bull. Astron. Soc. India, 39, 289 The Messenger, 144, 13

A78, page 21 of 27 A&A 587, A78 (2016)

Sandin, C., Weilbacher, P., Tabataba-Vakili, F., Kamann, S., & Streicher, O. Turner, J. E. 2010, in Canary Islands Winter School of Astrophysics, Vol. XVII, 2012, in Software and Cyberinfrastructure for Astronomy II, SPIE, 8451, 3D Spectroscopy in Astronomy, eds. E. Mediavilla, S. Arribas, M. Roth, 84510 J. Cepa-Nogue, & F. Sanchez (Cambridge University Press), 87 Schaerer, D., & Verhamme, A. 2008, A&A, 480, 369 van Dokkum, P. G. 2001, PASP, 113, 1420 Schaerer, D., Hayes, M., Verhamme, A., & Teyssier, R. 2011, A&A, 531, A12 Verhamme, A., Schaerer, D., & Maselli, A. 2006, A&A, 460, 397 Shapley, A. E., Steidel, C. C., Pettini, M., & Adelberger, K. L. 2003, ApJ, 588, Verhamme, A., Schaerer, D., Atek, H., & Tapken, C. 2008, A&A, 491, 89 65 Verhamme, A., Dubois, Y., Blaizot, J., et al. 2012, A&A, 546, A111 Shimasaku, K., Kashikawa, N., Doi, M., et al. 2006, PASJ, 58, 313 Wall, J. V. 1996, QJRAS, 37, 519 Shioya, Y., Taniguchi, Y., Sasaki, S. S., et al. 2009, ApJ, 696, 546 Wisnioski, E., Förster Schreiber, N. M., Wuyts, S., et al. 2015, ApJ, 799, 209 Sobral, D., Matthee, J., Darvish, B., et al. 2015, ApJ, 808, 139 Wisotzki, L., Bacon, R., Blaizot, J., et al. 2015, A&A, in press, Song, M., Finkelstein, S. L., Gebhardt, K., et al. 2014, ApJ, 791, 3 DOI: 10.1051/0004-6361/201527384 Taniguchi, Y., Shioya, Y., Ajiki, M., et al. 2003, J. Kor. Astron. Soc., 36, 123 Wold, I. G. B., Barger, A. J., & Cowie, L. L. 2014, ApJ, 783, 119 Tapken, C., Appenzeller, I., Mehlert, D., Noll, S., & Richling, S. 2004, A&A, Yamada, T., Matsuda, Y., Kousai, K., et al. 2012, ApJ, 751, 29 416, L1 Yang, H., Malhotra, S., Gronke, M., et al. 2015, ApJ, submitted Tapken, C., Appenzeller, I., Gabasch, A., et al. 2006, A&A, 455, 145 [arXiv:1506.02885] Tapken, C., Appenzeller, I., Noll, S., et al. 2007, A&A, 467, 63 Zheng, Z., & Wallace, J. 2014, ApJ, 794, 116 Terlevich, R., & Melnick, J. 1981, MNRAS, 195, 839 Zitrin, A., Labbé, I., Belli, S., et al. 2015, ApJ, 810, L12 Teyssier, R., Chapon, D., & Bournaud, F. 2010, ApJ, 720, L149

A78, page 22 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

Appendix A: Notes on individual objects VLA H  reveals an extended tail towards the west. This tidal tail is much larger than the optical dimensions of the pair Arp In the following, we detail the observed Hα velocity fields for 238 (see also Cannon et al. 2004, for similar extended tidal H  each galaxy. We compare our maps to the photometric Lyα prop- structures around the Lyα emitting star bursts Tol 1924-416 and erties derived from the HST images from Paper II and H  obser- IRAS08339+6517). vations from Paper III (see Sect.4). In Hα the main kinematical axis runs from west to east and the radial velocity field appears symmetric, although slightly dis- A.1. LARS 1 (Mrk 259) turbed towards the north and the south. We measure vshear ≈ 140 km s−1, a value that is in the domain of typical maximum 41 −1 With LLyα = 8 × 10 erg s and EWLyα = 33 Å LARS 1 velocities of inclination corrected Hα rotation curves of spiral is a strong Lyα emitting galaxy. A detailed description of the galaxies (e.g. Epinat et al. 2008a, 2010; Erroz-Ferrer et al. 2015). −1 photometric properties of this galaxy was presented in Paper I. With vFWHM & 300 km s highest velocity dispersions are ob- If LARS 1 were at high-z, it would easily be selected in con- served in the western part. The largest Lyα surface brightness is ventional narrowband imaging surveys (Paper IV). The galaxy also observed in the western part of the nucleus. Lyα appears in shows a highly irregular morphology. Its main feature is UV absorption in the eastern region, where we also see a minimum −1 bright complex in the north-east that harbours the youngest in velocity dispersions with vFWHM ≈ 70 km s . stellar population. As already noted in Paper II a filamentary structure emanating from the north-eastern knot is seen in Hα. Towards the south-east, numerous smaller star-forming com- A.4. LARS 4 (SDSS J130728.45+542652.3) plexes are found that blend in with an older stellar population. Although LARS 4 is similar to LARS 1 in terms of dust content In contrast with the irregular appearance, the line-of-sight and star formation rate, globally this galaxy shows Lyα in ab- velocity field of LARS 1 is rather symmetric. The velocity field sorption. The galaxy can be photometrically decomposed into is consistent with a rotating galaxy. The kinematical centre is two main components: an elongated lower surface brightness close to centre of the PMAS FoV and the kinematical axis ap- structure in the west and a more puffed up and luminous com- pears to run from the north-west to the south-east. We measure panion in the east. The eastern structure is tilted at ≈40◦ with ± −1 vshear = 56 1 km s . The GBT single dish H  spectrum of this respect to the western one. source is reminiscent of a classic double-horn profile, also an Compared to the highly irregular morphology in UV and Hα, obvious sign of rotation, with a peak separation consistent with the radial velocity field appears rather regular, with a moderate our v measurement. shear amplitude of v ≈ 75 km s−1 along a well-defined axis from v ≈ −1 shear The highest velocity dispersions ( FWHM 150 km s ) are west to east. The kinematic centre appears to be right between found in the north-eastern region. Here LARS 1 also shines the two photometric components, indicating that the observed α α strong in Ly , but while a fraction of Ly appears to escape shearing is the velocity difference between the merging com- α directly towards us, a more extended Ly halo around this part ponents and not a consequence of rotation. The lack of a well- is indicative of resonant scatterings in the circum-galactic gas defined simple and organised disk is further supported by the (see also Paper I). In constrast, in the south-western part of absence of a double-horn profile in the GBT single-dish H  pro- the galaxy, where we observe the lowest velocity dispersions file. Nevertheless, higher sensitivity VLA imaging results show v ≈ −1 α ( FWHM 70 km s ), Ly photons do not escape along the line a coherent velocity field on larger scales with an amplitude com- of sight. parable to vshear (Paper III, Fig. 9). In the dispersion map, the two components are clearly dis- A.2. LARS 2 (Shoc 240) tinguishable, with the eastern component showing higher ve- −1 locity dispersions (vFWHM ≈ 100 km s ) than the western Anywhere in LARS 2 where Hα photons are produced, Lyα −1 (vFWHM ≈ 65 km s ). In both regions, however, Lyα is seen photons emerge along the line of sight. LARS 2 is the galaxy in absorption. The highest velocity dispersions are observed at Lyα with the highest global escape fraction of Lyα photons ( fesc. = the boundary region where the two components are separated 52.1%) and the highest Lyα/Hα ratio (Lyα/Hα = 4.53). photometrically, but here only a little Lyα is escaping. −1 With vshear = 23±2 km s , LARS 2 shows the smallest vshear in our sample. However, we consider this value as a lower limit since we missed in our pointing the southernmost star-forming A.5. LARS 5 (Mrk 1486) knot. Our observed velocity field appears to be disturbed, but we In the HST images LARS 5 appears as a small (i.e. projected could envision a kinematical axis orthogonal to the photometric diameter d ≈ 5 kpc) highly inclined edge-on disk. Within the ap- major axis, i.e. roughly from west to east. This idea is supported parent disk Lyα appears in absorption but the galaxy shows an by our VLA imaging of this source (see Fig. 7 in Paper III). −1 extended halo, which is azimuthally symmetric at low surface The H  velocity field also indicates vshear ≈ 30 km s , meaning brightness isophotes, while brighter isophotes resemble more that our incomplete measurement gives a sensible lower limit. the elongated Hα shape. Most prominent in the Hα image are Within the observed region, our velocity dispersion map is rather −1 filamentary finger-like structures extending below and above the uniform, with vVFWHM ≈ 100 km s . disk, reminiscent of an outflowing wind. Although the profile of our single-dish H  observations A.3. LARS 3 (Arp 238) shows multiple peaks close to the noise level that might cor- respond to the peaks of a faint double-horn profile (Paper III), LARS 3 is the south-eastern nucleus of the violently interact- our Hα kinematics appear incompatible with a typical disk sce- −1 ing pair of similar sized spiral galaxies Arp 238. With a global nario: Firstly, our measured vshear = 37 km s would indicate Lyα 41 −1 fesc = 0.1 and LLyα = 10 erg s , this dust-rich, luminous a rotation curve with a very low amplitude, which is untypical infrared galaxy is a relatively weak Lyα emitter. Imaging by even for a small disk (e.g. in the largest Hα disk sample of

A78, page 23 of 27 A&A 587, A78 (2016)

Epinat et al. 2010, for d ≤ 5 kpc the maximum rotation curve from north-east to south-west along the continuum major axis, −1 velocity is on average 95 km s ). Secondly, there is appar- but in the north-west and south-east the vLOS values do not follow ent asymmetry in the spatial distribution of the vLOS values, this weak gradient at all. Overall the galaxy shows high velocity −1 i.e. a large sector in the south-west is characterised by simi- dispersion in Hα (vFWHM & 160 km s ), with the lowest values −1 −1 lar line-of-sight velocities (vLOS ∼ +20 km s ), while only a occurring in north-east (vFWHM ≈ 120 km s ), where the bright- small sector in the north-east shows blueshifted Hα emission est fluxes in Lyα and the highest Lyα/Hα ratios are observed. −1 (vLOS ∼ −40 km s ). The most prominent feature in the velocity dispersion map A.8. LARS 8 (SDSS-J125013.50+073441.5) is a biconical zone of increasing velocity dispersions (from −1 −1 ∼120 km s to ∼220 km s ) with increasing distance from the LARS 8 appears as a face-on disk with the highest metalicity centre. The base of this zone coincides the brightest region in of the sample. The apparent disk also possesses a highly dust UV and Hα in the south-western sector. We emphasise that the obscured nucleus. Nevertheless, there is a high degree of irreg- 00 seeing PSF FWHM for this observation is 1.3 , thus ∼2× the ularity compared to classical disks and spiral arms are not well 00 00 extent of the 0.5 × 0.5 spaxels used on LARS 5. However, defined. As traces of shell-like structures are visible in the inner as the kinematical centre is not co-spatial with the base of the parts and in the outskirts of the galaxy (cf. Fig. 6 in Paper I), cones, and since the velocity gradient around the base of the LARS 8 could be classified as a shell system hinting at a recent cone is very weak we are confident that this feature is real and merger event (e.g. Bettoni et al. 2011). The Lyα light distribution not caused by PSF smearing effects. of LARS 8 does not resemble that seen in Hα. The brightest Lyα zone is in the north of the disk, while in the south Lyα appears only in absorption. At lower surface brightness isophotes a Lyα A.6. LARS 6 (KISSR 2019) halo becomes visible. With an equivalent width EWLyα = 20.3 Å −1 With .1 M yr conversion of gas into stars, LARS 6 shows this galaxy would be selected as LAE in conventional narrow- the lowest SFR in the sample (Paper II). The main star-forming band surveys. knot is in the north with a tail of much smaller and fewer lu- Our vLOS map seems disk-like. Assuming an infinitely thin minous knots extending to the south. In our seeing limited data disk the ellipticity of 0.2 (Paper IV) implies an inclination of ◦ −1 cubes we cannot disentangle these individual knots photometri- 40 , hence, our observed vshear = 155 km s translates to vmax = cally. Lyα is seen in absorption, even on the smallest scales. 241 km s−1. The GBT single-dish H  profile shows a broad, pos- Shearing is observed between the main component and the sibly double-peaked profile. Although these peaks are not signif- −1 tail, although the amplitude is moderate (vshear = 52 km s ). icantly separated from the overall H  signal, the implied velocity −1 The main component also shows higher velocity dispersions difference of ∼300 km s appears consistent with our vshear mea- −1 −1 (vVFWHM ≈ 70 km s ) than the tail (vVFWHM ≈ 50 km s ). surement. The orientation of the velocity field from VLA inter- Overall this galaxy has the lowest observed velocity dispersion ferometric H  observations is qualitatively consistent with our in the sample. PMAS Hα observations, although at the VLA D-configuration 00 Our GBT single-dish H  measurements are severely contam- beam size of 72 LARS 8 appears only marginally resolved. −1 inated by the nearby field spiral UGC 10028. However, newly Elevated velocity dispersions with vFWHM & 160 km s are obtained but as of yet unpublished VLA D-configuration images apparent to extend orthogonally to the kinematical axis along −1 allow a first-order separation from LARS 6 and this companion. the 0 km s iso vLOS contour. In the other regions, the vFWHM −1 From these data, coherent rotation is present in the H  gas asso- map is rather flat, typically with vFWHM ∼ 100 km s – a value ciated with LARS 6 at a level roughly consistent with the vshear that is also still commonly observed within the sample of 153 lo- estimate we obtain from PMAS. cal spirals by Epinat et al.(2010). Given the strong gradient in the velocity field, the observed elevated velocity dispersions are likely a result of PSF smearing effects. A.7. LARS 7 (IRAS F13136+2938)

In the continuum LARS 7 resembles a highly inclined disk com- A.9. LARS 9 (IRAS 08208+2816) parable to LARS 5, but in Hα an almost azimuthally symmetric structure emerges that is highly distended with respect to the This luminous infrared galaxy has highly irregular morphology. elongated continuum morphology. Moreover, two extended low Numerous star-forming knots are visible in Hα along two arms surface brightness red lobes are found at the opposite ends of the of large extend that are connected to a central, very bright Hα nu- inclined disk, reminiscent of a shell-like structure. This is sug- cleus. At the end of the southern tail, a foreground star appears gestive of a recent merger event (e.g. Bettoni et al. 2011). In Lyα in projection; contributions from this star have been masked images LARS 7 appears even more extended, with the bright out in the PMAS data. In LARS 9 Lyα is absorbed almost isophotes following the Hα shape and low surface brightness everywhere along the line of sight towards the star-forming re- isophotes resembling a more scaled up version of the apparent gions. Nevertheless, this galaxy is embedded in a faint extended disk. Lyα fuzz. Morphologically, this fuzz broadly traces a scaled up For similar reasons as outlined for LARS 5 in Sect. A.5, version of its optical and Hα shape. Having a globally integrated 41 −1 our Hα kinematics argue against a typical disk scenario: we ob- Lyα luminosity of LLyα ≈ 3×10 erg s and an equivalent width −1 serve a low shearing amplitude; with vshear = 31 km s even of EWLyα ≈ 8 Å, this galaxy would not be selected by its Lyα lower than in LARS 5. Moreover, the line of sight velocity emission at high-z in conventional narrowband surveys. maps appears disturbed compared to that expected for a clas- Since LARS 9 extends more than ∼0.50 on the sky, we sical disk. Our GBT single dish H  observations reveal a single needed to cover it with two PMAS 1600 × 1600 pointings (Fig.6). broad (92 km s−1) line. LARS 7 is the only object where the The line-of-sight velocity ﬁeld for this galaxy is very peculiar. GBT linewidth is incompatible to its integrated Hα linewidth From the north to the centre, a weak (∼100 km s−1) gradient from (cf. Sect. 6.2). A main kinematical axis can be envisioned to run blueshifted to systemic velocity is apparent. From the centre

A78, page 24 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics

LARS9A Voronoi bin 129 LARS9A Voronoi bin 133 LARS9A Voronoi bin 140

80 80 ] ] ]

−1 150 −1 −1 Å Å Å

−2 −2 −2 60 60 cm cm cm −1 −1 −1 100 40 erg s erg s 40 erg s −16 −16 −16

50 20 20 Flux [10 Flux [10 Flux [10

0 0 0 6860 6865 6870 6875 6880 6885 6860 6865 6870 6875 6880 6885 6860 6865 6870 6875 6880 6885 Wavelength [Å] Wavelength [Å] Wavelength [Å]

Fig. A.1. Representative LARS 9 Hα line profiles in blue hatched region of Fig.6 described by a fit using two Gaussian components. The red dashed lines show the individual components and the blue line shows the sum, while the black line shows the profile as observed. For all profiles shown the single component fit used to create the map shown in Fig.6 converged on the stronger red component.

LARS9A Voronoi bin 73 LARS9A Voronoi bin 65 LARS9A Voronoi bin 85 3 8 50 ] ] ] −1 6 −1 −1 2 Å Å Å

−2 −2 40 −2 cm cm cm

−1 4 −1 −1 30 1 erg s erg s erg s −16 −16 −16 2 20 0

Flux [10 Flux [10 10 Flux [10 0

0 −1 6860 6865 6870 6875 6880 6885 6860 6865 6870 6875 6880 6885 6860 6865 6870 6875 6880 6885 Wavelength [Å] Wavelength [Å] Wavelength [Å]

Fig. A.2. Representative LARS 9 Hα line profiles in red hatched region of Fig.6, similar to Fig. A.1. The single component fit used to create the map shown in Fig.6 converged on the stronger blue component in the centre panel, but for the profiles shown in the left and right panels the single component fit is artificially broadened.

LARS13B Voronoi bin 2 LARS13B Voronoi bin 4 LARS13B Voronoi bin 14 5 15 6

] 4 ] ] −1 −1 −1 Å Å Å

−2 −2 −2 10 3 4 cm cm cm −1 −1 −1

2 erg s erg s erg s 5 −16 −16 2 −16 1 Flux [10 Flux [10 Flux [10 0 0 0

7515 7520 7525 7530 7535 7515 7520 7525 7530 7535 7515 7520 7525 7530 7535 Wavelength [Å] Wavelength [Å] Wavelength [Å]

Fig. A.3. Representative LARS 13 Hα line profiles in blue hatched region of Fig.7, similar to Fig. A.1. The single component fit used to create the map shown in Fig.7 converged on the stronger blue component in the right panel, but for the profiles shown in the left and centre panels the single component fit is artificially broadened. this weak gradient continues towards the south. The southern distinct secondary component. This secondary component does tail then shows an opposite gradient from red- to blueshifts. In not follow the large-scale motions of the northern component, to this south- and south-western zone, a single Gaussian often is which the nucleus also belongs. We indicate, with red and blue not an optimal representation of the observed Hα lines. In this hatchings in Fig.6, the regions of this component that are red- zone the spectral profile is sometimes asymmetric with an ex- and blueshifted with respect to the systemic velocity (as given by tended red or blue wing, but often it also shows clearly double- the nucleus). As a natural outcome from this qualitative analysis peaked morphology (for examples see Fig. A.1). In some of the emission in south-eastern tail of LARS 9 is solely coming these regions, our fit tries to capture both lines, hence, the ob- from this secondary component. Also, a significant fraction of tained line-of-sight velocity is centred between the peaks and Lyα photons appear to escape towards the observer only at the the linewidth appears very broad (e.g. left and right panels of end of this tail. Fig. A.2). In some other spaxels, when the secondary peak is From our PMAS observation we conclude that LARS 9 very weak or when only an extended wing is seen, the fit con- is a closely interacting pair of galaxies in an advanced stage verges to the stronger line. The affected zones have hatched rect- of merging. The interaction scenario is also supported by our angles overlaid in Fig.6. These zones harbour a kinematical 21 cm observations. The integrated GBT spectrum shows a

A78, page 25 of 27 A&A 587, A78 (2016) broad single line, indicating that the bulk of the H  is not par- the isophotal contours approximately preserving the axis ratio of taking in an ordered flat rotation. LARS 9 was only marginally Hα. Within the disk Lyα occurs exclusively in absorption. resolved in our VLA H  maps (Paper III). However, as of yet Running along the disk from the north-west to the south- unpublished VLA C configuration observations (K. Fitzgibbon, east a strong gradient in the vLOS map is apparent. The ob- −1 in prep.) show extended H  towards the west, enclosing the served shearing amplitude vshear = 150 km s is consistent galaxy SDSS J082353.65+280622.2. For this galaxy, no spectro- with this gradient being caused by rotation (e.g. Epinat et al. scopic redshift is available, but the photometric redshift agrees 2008a, 2010; Erroz-Ferrer et al. 2015). Unfortunately the beam with being associated to LARS 9. These observations therefore of our GBT single-dish H  observations is likely contaminated strongly suggest that a third system is significantly involved in by other sources. Nevertheless, within the multiple peaks in the the interaction. Moreover, the VLA C-configuration observa- GBT spectrum a double-horn profile at a velocity separation tions of LARS 9 show a peak in the second-moment maps that consistent with our Hα vshear measurement is visible. Although trace the random motions of H  that coincides spatially with the the projected height of the disk is similar to the PSF FWHM, zones of multiple H  components in our PMAS maps. Hence, the dispersion map shows traces of higher velocity dispersions −1 at these locations (hatched regions in Fig.6) Ly α photons of above and below the disk (vVFWHM ≈ 200 km s ) than within −1 both distinct kinematical components are likely scattered by H  (vVFWHM ≈ 150 km s ), especially around the star-forming re- at resonance. gions in the south-east. However, the elevated dispersions near the kinematic centre are caused by PSF smearing of the strong gradient in the vLOS field. A.10. LARS 10 (Mrk 0061)

Similar to LARS 1 this galaxy appears morphologically to be A.12. LARS 12 (LEDA 27453) in an advanced merger state. It possesses a large UV bright core in the north-west and a spur of smaller, less luminous star- Because of its modest EWLyα = 13 Å, this UV bright merger forming regions towards the south-east. Several redder low sur- would not be selected as a LAE at high-z. At its brightest knot in face brightness structures are seen outside the main body of UV and Hα the galaxy shows Lyα only in absorption. Lyα only the galaxy, reminiscent of the appearance of shell galaxies (e.g. appears in emission at larger radii, where a number of fainter Marino et al. 2009; Bettoni et al. 2011). From the central star- star-forming regions can be appreciated. Moreover, LARS 12 is forming parts Lyα is only seen in absorption, while a faint halo embedded in a faint, low surface brightness Lyα halo. emerges at larger radii. When considering solely the central parts The non-regular Hα velocity field indicates complex kine- the galaxy would remain undetected in high-z LAE surveys since matics. In the north-western corner a strong ∼140 km s−1 gradi- EW 00 Lyα = 8 Å. However, when integrating over the lower surface ent over ∼1.5 (∼2.8 kpc at dLARS12 = 470.5 Mpc) is observed, brightness emission EWLyα rises to 31 Å and with its luminosity but the velocity field is essentially flat from there towards the 41 −1 of LLyα = 2 × 10 erg s the galaxy would be within the sen- south-west. The high velocity and velocity dispersion in the val- sitivity of the deepest contemporary LAE surveys (Rauch et al. ues seen in the south are not significant, as the low S/N of this 2008; Bacon et al. 2015). broad line in these binned spaxels lead to a 50% error on the de- The line-of-sight velocity field is symmetric and consistent termined vLOS and σ0 value. Because of PSF smearing effects the with rotation and contrasts the irregular continuum morphology. strong velocity gradient is responsible for the region of elevated The velocity gradient running from the south-east to the north- velocity dispersions in the north-western corner. −1 west is weak and we measure vshear = 36 km s . The seeing For LARS 12 adaptive optics Paschen α IFS observations for this observation was very substantial (1.500 FWHM, i.e 3× have been presented in Gonçalves et al.(2010); object ScUVLG the size of the 0.500 × 0.500 spaxels). Despite the weak gradient, 093813 in their nomenclature. Their line-of-sight velocity field this leads to non-negligible PSF smearing effects in the disper- is qualitatively fully consistent with ours, but their higher reso- sion map. Moreover, because of the short exposure time and the lution data allows them to pinpoint the highest redshifted values rather low Hα flux, this galaxy has the lowest S/N per spaxel of directly to the two filaments that extend towards the north-west. the whole sample; this manifests in a noisy vFWHM map. Both Since they are not affected by PSF smearing their dispersion map of these effects essentially make local disturbances in the ve- is not contaminated by high-velocity dispersions in the north- locity dispersion untraceable in our LARS 10 data set, but our west. Of course, our artifact in the south is also absent from their −1 −1 flux weighted global measurement of σ0 ≈ 40 km s is robust map. Nevertheless, their σ0 value of 62 km s is in good agree- −1 against these nuisances. ment with our measurement of 72 km s . The GBT integrated H  spectrum shows a broad single line profile (∼280 km s−1 FWHM), however, at low signal to noise. This appears hard to reconcile with the Hα results. If real, it A.13. LARS 13 (IRAS 01477+1254) might indicate that a large fraction of neutral gas in and around Based on its highly irregular morphology this starburst system is LARS 10 is kinematically in a different state than the gas around clearly interacting. As LARS 13 shows a narrow Lyα equivalent the galaxy’s star-forming regions. width (EWLyα = 6 Å), it would not be selected as a LAE at high redshift in conventional narrowband surveys. Overall, only 1% A.11. LARS 11 (SDSS J140347.22+062812.1) of Lyα photons are escaping and the resulting Lyα luminosity is 41 −1 LLyα = 7 × 10 erg s . This galaxy appears as a highly inclined edge-on disk in the The complex vLOS map of this galaxy is characterised by continuum and UV that is slightly thicker when seen in Hα. a gradient from blue- to redshifts along the east-western axis According to the Hα and UV emission, stronger star formation for the northern component, as well as an gradient from blue-to occurs in the south-eastern zone of the projected disk. In Lyα a redshifts from south to north for the eastern component. In the mildly extended halo above and below the plane is visible, with north-western zone where the longitudinally aligned structure

A78, page 26 of 27 E. Ch. Herenz et al.: LARS. Spatially resolved Hα kinematics overlaps with the latitudinally aligned one, our single component A.14. LARS 14 (SDSS J092600.41+442736.1) Gaussian is not an optimal representation of the observed Hα LARS 14 is the most luminous LAE in the sample (LLyα = line profiles. In this region, the profile is often asymmetric with 42 −1 an extended blue wing (e.g. right panel of Fig. A.3). Some spax- 4.2 × 10 erg s ). It is also classified as a green pea galaxy els show a double-peaked Hα profile (e.g. left and centre panels (Cardamone et al. 2009, see also their Fig. 7). Green pea galax- of Fig. A.3). The affected zone has a hatched rectangle over- ies show commonly strong Lyα emission and are also thought laid in Fig.7. In the region where two equally strong peaks are to be Lyman continuum leaking galaxies (Jaskot & Oey 2014; prominent in the spectrum, our fit tries to fit a broad line that Henry et al. 2015; Yang et al. 2015). Visible in the HST images encloses both lines; these spaxels can be seen in the east running is a small companion in the south and tidal tails extending to as a diagonal line of broad velocity dispersions from north-west the east and south (Cardamone et al. 2009). Our seeing limited to south-east. Spaxels east of this demarcation show an extended data does not provide enough resolution to disentangle this struc- blue wing. We interpret this to mean that we are seeing sepa- ture photometrically from the main component within the datacube. Kinematically, however, we observe that the southern re- rate components of ionised gas kinematics, where one compo- −1 nent belongs to the latitudinal elongated structure while the other gion shears at ≈60−70 km s with respect to the northern, with not much velocity substructure seen. In the main part we observe belongs to the longitudinal elongated structure. The spatial and −1 high velocity dispersions with vFWHM ≈ 175 km s , dropping to spectral proximity of both regions indicates that the galaxy is −1 still in an ongoing interaction with both progenitors not having 80−90 km s towards the companion. Our modelling of the Hα fully coalesced. profile with a simple 1D Gaussian did not capture the extended For LARS 13 adaptive optics Paschen α IFS observations broad wings seen in the Hα profiles. These broad wings likely have been presented in Gonçalves et al.(2010); object ScUVLG trace outflowing ionised gas, and are a feature regularly observed 015028 in their nomenclature. However, their small FoV allows in pea galaxies (Yang et al. 2015). them only to sample emission from the strongest star-forming For this galaxy, adaptive optics Paschen α IFS observa- region in the western part, and so they do not have any infor- tions have been presented in (Gonçalves et al. and 2009, mation on the longitudinally aligned eastern part. Consequently, object ScUVLG 92600 in their nomenclature). They report −1 slightly higher values, both for the shear and the dispersion. their observed vshear = 78 km s is lower than our value −1 Nevertheless, our derived v /σ = 0.6 ± 0.1 is consistent with (vshear = 173 km s ), since the north-eastern zone with highest shear 0 their value of 0.51. This ratio would be lowered by about a fac- vLOS values in our map is not present in their data. In the zone where they have signal, there is good qualitative and quantita- tor of two, if only the northern main component is considered, tive agreement between our maps and their maps. In particular, since its velocity offset relative to the companion dominates the −1 shear. The main component as such therefore presents the struc- they found high velocity dispersions vFWHM & 200 km s and a weak velocity gradient in the north-western star-forming knot. ture with the lowest vshear/σ0 of our whole sample. In this region of high velocity dispersions Lyα appears to escape directly towards the observer.

A78, page 27 of 27