Ionized Gas in the XUV Disc of the NGC 1512/1510 System

MNRAS 450, 3381–3409 (2015) doi:10.1093/mnras/stv703

Ionized gas in the XUV disc of the NGC 1512/1510 system

A.´ R. Lopez-S´ anchez,´ 1,2‹ T. Westmeier,3 C. Esteban4,5 and B. S. Koribalski6 1Australian Astronomical Observatory, PO Box 915, North Ryde, NSW 1670, Australia 2Department of Physics and Astronomy, Macquarie University, NSW 2109, Australia 3International Centre for Radio Astronomy Research, The University of Western Australia, 35 Stirling Hwy, Crawley, WA 6009, Australia 4Instituto de Astrof´ısica de Canarias, E-38200 La Laguna, Tenerife, Spain 5Departamento de Astrof´ısica, Universidad de La Laguna, E-38205 La Laguna, Tenerife, Spain 6CSIRO Astronomy and Space Science, Australia Telescope National Facility, PO BOX 76, Epping, NSW 1710, Australia

Accepted 2015 March 27. Received 2015 March 26; in original form 2014 June 4 Downloaded from ABSTRACT We present deep, intermediate-resolution, optical spectroscopy of 136 genuine UV-bright regions located in both the inner and outer regions of NGC 1512. This galaxy is in close interaction with the blue compact dwarf galaxy NGC 1510 and possesses two prominent H I arms

where extended ultraviolet complexes are found. Our data were taken using 2dF/AAOmega at http://mnras.oxfordjournals.org/ the 3.9 m Anglo-Australian Telescope and are combined with the H I data from Local Volume H I Survey and Galaxy Evolution Explorer UV data. We detect ionized gas in 82 per cent of the complexes, many of them located between 1 and 6.6 R25. We found significant differences between regions along the Arm 1 – 8.25 12+log(O/H) 8.45 –, and knots located in the external debris of Arm 2, –8.40 12+log(O/H) 8.60–. Considering a radial and an azimuthal gradient following the H I arms, we confirm that Arm 2 has experienced an enhancement in star formation because of the interaction with NGC 1510 and flattened the radial metallicity

at large radii. Arm 1 appears to retain the original and poorly disturbed radial distribution. at Oxford Brookes University on June 4, 2015 We trace the kinematics of the system up to 78 kpc using the Hα emission, which matches well that provided by the H I. We estimate that the gas existing at large galactocentric radii had a metallicity of 12+log(O/H) ∼ 8.1 before the interaction started around 400 Myr ago. The metals within the H I gas are very likely not coming from the inner regions of NGC 1512 but probably from material accreted during minor mergers or outﬂow-enriched intergalactic medium gas during the life of the galaxy. Key words: galaxies: abundances – galaxies: dwarf – galaxies: evolution – galaxies: individual: NGC 1510 – galaxies: individual: NGC 1512 – galaxies: kinematics and dynamics.

& Jarrett 2012), being this material the fuel for present and future 1 INTRODUCTION star-forming events. The extended ultraviolet (XUV) emission dis- One of the most surprising discoveries obtained by the Galaxy covered by GALEX is located well beyond the Hα or B25 radius of Evolution Explorer (GALEX) satellite (Martin et al. 2005)wasthe galaxies, and it seems to exist in ∼20–30 per cent of the local disc ﬁnding of UV-bright complexes in the outskirts of nearby spiral galaxy population (Thilker et al. 2007; Zaritsky & Christlein 2007; galaxies (Gil de Paz et al. 2005, 2007a; Thilker et al. 2005). Diffuse Lemonias et al. 2011). XUV-discs have even been found around stellar tails and shells were already found surrounding nearby galax- E/S0 galaxies (Salim & Rich 2010; Thilker et al. 2010;Moffett ies using deep optical images (e.g. Malin & Carter 1983) but these et al. 2012). The origin of these UV-bright complexes seems to be faint features were thought to be mainly composed by old stars. H II young stellar clusters associated with a recent or still on-going star regions located up to 30 kpc from the main galaxy have been also formation activity (Gil de Paz et al. 2007b; Bresolin et al. 2009a; found recently (e.g. Ryan-Weber et al. 2004;Meureretal.2006). Bresolin, Kennicutt & Ryan-Weber 2012). XUV-discs should be Indeed, spiral galaxies typically possess a large H I disc that reaches embedded in larger H I envelopes – a 2X-H I disc as deﬁned by well beyond their optical size (e.g. Freeman et al. 1977;Walter Koribalski & Lopez-S´ anchez´ (2009) – which are providing the fuel et al. 2008; Westmeier, Braun & Koribalski 2011; For, Koribalski for their star formation activity (e.g. Koribalski & Lopez-S´ anchez´ 2009; Bigiel et al. 2010b;Werketal.2010a). The study of star formation processes in galaxy outskirts gives E-mail: [email protected] key clues about the formation and evolution of galaxies, not only

C 2015 The Authors Published by Oxford University Press on behalf of the Royal Astronomical Society 3382 A.´ R. Lopez-S´ anchez´ et al. because these regions may probe physical conditions similar to UV colours) and the most recent star formation event (from Hα those present of the early Universe, but also because they test the emission); (iii) what is the metallicity distribution throughout the inside–out scenario of galaxy disc formation. In the first case, the system, studying not only radial but azimuthal chemical gradients, analysis of XUV-complexes in the outer regions of galaxies com- and how this affects the metal redistribution in galaxies; (iv) what is plements the study of dwarf galaxies as building blocks of larger the nature of the XUV-complexes and search for tidal dwarf galaxy objects but also the nature of tidal dwarf galaxies (TDGs), as well (TDG) or external dwarf galaxies candidates; and (v) what are the as the viability of star formation in regions of low gas density. In relationships between the neutral gas, the stellar mass, the metallic- the second case, these UV-bright regions provide a powerful tool to ity, and the star formation in these regions and the comparison with investigate the mass and chemical enrichment of galaxies from their what is observed in dwarf galaxies. centre to their external regions via colours, gas-to-stars ratios, and Here, we present the results of our analysis of the XUV emission metallicity gradients. Following the inside–out scenario, the galac- in the impressive galaxy pair NGC 1512/1510. At the adopted dis- tic stellar disc grows gradually with time as it accretes gas from tance of 9.5 Mpc, 1 arcmin corresponds to 2.49 kpc, and 2 arcsec their gaseous, extended discs (e.g. White & Frenk 1991; Bouwens, (the diameter of a 2dF fibre) corresponds to 83 pc. This system Cayon & Silk 1997) producing a relatively quick self-chemical en- hosts hundreds of independent UV-bright regions associated with richment and an almost universal negative metallicity gradient once dense H I clouds in the outskirts, the blue compact dwarf galaxy this is normalized to the galaxy optical size (Zaritsky, Kennicutt & (BCDG) NGC 1510 (which is located at just 13.8 kpc from the Huchra 1994; Boissier & Prantzos 1999, 2000). Indeed, detailed centre of NGC 1512), a central star-forming ring and two TDG Downloaded from observations using integral-field spectroscopy (IFS) data from the candidates at very large distance (projected radius of 83 kpc) iden- CALIFA survey (Sanchez´ et al. 2012; Husemann et al. 2013)have tified by their H I emission. A review of the properties of the proven that a characteristic oxygen abundance gradient does exist NGC 1512/1510 system, as well as a very detailed UV and H I in galaxy discs (Rosales-Ortega et al. 2012;Sanchez´ et al. 2014). analysis, can be found in Koribalski & Lopez-S´ anchez´ (2009, here-

However, it has been observationally found that the outskirts after KLS09). Recently, Bresolin et al. (2012) obtained deep FORS2 http://mnras.oxfordjournals.org/ of galaxies do not follow the same pattern as their main disc. In – 8 m Antu unit of the European Southern Observatory Very Large particular, a flattening of the metallicity gradient in the external Telescope on Cerro Paranal, Chile – optical spectroscopy of 62 UV- regions of spiral galaxies has been found (Bresolin et al. 2009a; bright complexes within NGC 1512, including star-forming regions Kewley et al. 2010; Rupke, Kewley & Barnes 2010;Werketal. in the optical disc and in XUV knots of NGC 1512, although the 2010a,b; Bresolin, Kennicutt & Ryan-Weber 2012), and it is also majority (all but 16) were located within the effective radius, Re detected in our Milky Way (e.g. Vilchez & Esteban 1996; Esteban (4.26 arcmin = 11.76 kpc), of the galaxy. In six of the observed et al. 2013). It has been suggested that the flat metallicity gradient knots they were able to determine the oxygen and nitrogen abun- in the outer discs is a consequence of galaxy interactions, however dances following the direct Te method, yielding to average values there are still some non-interacting galaxies where this behaviour is of 12+log(O/H) ∼ 8.17 ± 0.09 and log(N/O) ∼−1.32 ± 0.12. also observed (Werk et al. 2011;Sanchez´ et al. 2014). In fact, for the at Oxford Brookes University on June 4, 2015 Milky Way (Esteban et al. 2013) suggest that a levelling out of the 2 OBSERVATIONS AND DATA REDUCTION star formation efficiency beyond the isophotal radius can explain the flattening of the abundance gradients in the external Galactic The observations were carried out at the 3.9 m AAT at Siding Spring disc. In any case, the finding of a flat metallicity gradient in the Observatory (NSW, Australia) between 2008 Nov 29 and Dec 2. We external regions of galaxies informs about the metal redistribution used the 2dF instrument (Lewis et al. 2002) in combination with the in these systems, an aspect which may also have consequences in AAOmega spectrograph (Saunders et al. 2004; Smith et al. 2004; the interpretation of the scatter in the observed mass–metallicity Sharp et al. 2006) to get our spectroscopic data. The 2dF instrument, relations (Lequeux et al. 1979; Tremonti et al. 2004; Kewley & which is installed at the prime focus of the AAT, consists of a robot Ellison 2008; Lara-Lopez´ et al. 2010; Lara-Lopez,´ Lopez-S´ anchez´ gantry which positions up to 400 optical fibres on a plate with a & Hopkins 2013c; Lara-Lopez´ et al. 2013b). field of 2◦ diameter projected on the sky. The accuracy of fibre Hence, the analysis of the ionized gas within XUV complexes positioning is 0.3 arcsec. 2dF also possesses a wide-field corrector, found at large galaxy radii allows us to investigate the nature, phys- an atmospheric dispersion compensator (ADC), and a tumbling ical conditions, chemical abundances, and kinematics of the inter- mechanism with two field plates which allows the next field to be stellar medium (ISM) in these regions. When combining with UV configured while the current field is being observed. Eight fibres and 21-cm H I observations, these data provide important constraints actually are guide fibre-bundles which are used to ensure accurate to the star formation activity and star formation history of galaxies, telescope positioning during the observation. The other 392 target the relationships between gas and stars, as well as the physics behind fibres from 2dF, each one with a 2 arcsec diameter projected on the them. We are therefore conducting such study of XUV discs found sky (Lewis et al. 2002), are fed to the AAOmega spectrograph. in nearby, H I-rich galaxies included in the Local Volume H I Survey In comparison to long-slit spectroscopy, multi-object fibre spec- (LVHIS) project (Koribalski 2008, Koribalski et al., in preparation) troscopy allows us to be more flexible in choosing the areas of using the multi-object fibre-positioner instrument Two-degree Field interest and to save observing time as all pointings can be done in (2dF) and the AAOmega spectrograph installed at the 3.9 m Anglo- one exposure. For the case of the galaxy pair NGC 1512/1510, we Australian Telescope (AAT, Siding Spring Observatory, NSW, Aus- needed to allocate fibres within a field of view of ∼1◦. tralia). The 2dF instrument is perfect for our analysis, as it provides The AAOmega spectrograph, which is stationed in the thermally a very flexible position of fibres within a wide field of view, allow- stable environment of one of the telescope’s Coude´ rooms, possesses ing the simultaneous observation of regions within both the XUV a dual beam system. Each arm of the AAOmega system is equipped and spiral discs. In particular, with these observations, we want to with a 2k × 4k E2V CCD detector and an AAO2 CCD controller. In explore (i) how many XUV complexes show ionized gas emission order to achieve the best compromise between sensitivity, spectral (i.e. on-going star formation activity); (ii) what is the correlation resolution and wavelength coverage, we used the 580V grating in between the ages of the dominant young stellar population (from the blue arm and the 2000R grating in the red arm. The 5700 Å

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3383 dichroic was used to split the light. The 580V grating, which has abundances) which will be later used for the discussion of the phys- a dispersion of 1 Å pixel−1, was centred at 4550 Å, and hence it ical and chemical properties of the NGC 1512/1510 system. gave a coverage of 3500–5600 Å. The 2000R grating, which has a dispersion of 0.23 Å pixel−1, was centred at 6530 Å, and hence it gave a coverage of 6300–6750 Å. With this setup, we can measure 3.1 Identification of the UV-bright regions all the critical diagnostic optical lines, [O II] λ3727, Hγ ,[OIII] Fig. 1 shows the deep NUV image of the NGC 1512/1510 galaxy λ4363, He II λ4686, Hβ,[OIII] λλ4959,5007, [N II] λ6548, Hα, pair with the identification of all the UV-rich regions observed using [N II] λ6583, He I λ6678 and [S II] λλ6717,6731, but we will also the 2dF/AAOmega instrument at the AAT. The UV-knots within the get the highest dispersion available in the Hα line (0.23 Å pixel−1), two TDG candidates (see fig. 8 in KLS09) were also observed but giving a spectral resolution of ∼35 km s−1. they do not appear in this figure. We have coloured each region We used the GALEX images to get accurate RA and Dec. posi- according to the results of its analysis, as explained below and tions of all the UV-bright star-forming regions observed by KLS09. it is compiled in Table 1. Each 1D spectrum of the 164 observed We then ran the 2dF CONFIGURE software, which implements a Simu- targets was carefully inspected by eye to identify the bright emission lated Annealing algorithm (Miszalski et al. 2006) to allocate optical lines such as [O II] λ3727, Hβ,[OIII] λ5007 and Hα. We identified fibres to the UV-rich star-forming regions. We also allocated 25 153 regions (93.3 per cent). The 11 regions for which we could not fibres for sky positions. We observed two different plate configura-

determine their nature are coloured in red in Fig. 1. From these 153 Downloaded from tions, including some objects to be observed in both plates to check regions, 17 were identified as background galaxies, as the bright the quality of our analysis. The 2dF CONFIGURE software allocated emission lines were identified at radial velocities larger than the 90 UV-selected knots in Plate 1 and 74 in Plate 2, being a total of expected radial velocity of the regions found in the system (between 164 regions. 28 objects (including NGC 1510 and the two TDG ∼750 and ∼1100 km s−1). Background galaxies are coloured in candidates) were included in both plates, hence we observed 136 × cyan in Fig. 1. We also identified a foreground star thanks to the

genuine targets. For each plate, we obtained 3 2400 s exposures, http://mnras.oxfordjournals.org/ presence of strong absorption lines (Ca II H, K, G band, Hγ ,Hβ, as well as the calibration frames (arcs and flat-fields). Hα) found at a radial velocity of −96.8 km s−1. This object has been The spectrophotometric stars GD 108 (Oke 1990) and HR 3454 coloured in blue in Fig. 1. 135 regions (82.3 per cent of the observed (Hamuy et al. 1994) were used to get the absolute flux calibration regions; 88.2 per cent of the identified regions) have been classified of the 1D spectra. At least two exposures of 300 s (for GD 108) as emission-line objects truly belonging to the NGC 1512/1510 or 10 s (for HR 3454) were taken immediately after the science system. Hence, our 2dF/AAOmega observations confirm that the frames were obtained. Calibration frames were also taken for the huge majority of the observed UV-rich star clusters has ionized gas spectrophotometric stars. emission. The raw data were processed using software developed at the Table 1 compiles the statistics of the results of our 2dF/AAOmega Australian Astronomical Observatory (AAO) called 2DFDR (Croom,

observations. Colours grey, pink, orange, yellow and green indicate at Oxford Brookes University on June 4, 2015 Saunders & Heald 2004; Sharp & Birchall 2010). The 2DFDR pro- the number of emission lines detected in the spectrum of each cessing applies the standard sequence of tasks for 1D spectral extrac- region. Almost half of the regions identified within the galaxy pair tion from 2D images. This includes bias subtraction, flat-fielding, (47.4 per cent) show the five emission lines needed([O II] λ3727, tramline fitting, wavelength calibration, and sky subtraction. A mas- Hβ,[OIII] λ5007, Hα,and[NII] λ6583) for computing the oxygen ter bias was created using the available bias frames, but for each abundance of the ionized gas using several methods. However, we plate configuration observed, the raw AAOmega frames were run are able to estimate the metallicity of the ionized gas in 120 regions through 2DFDR to provide the processed spectra. The sky subtrac- (88.9 per cent of the detected regions within the NGC 1512/1510 tion was performed using the 25 fibres allocated in each plate to system) as these are the objects for which we identify at least Hα sky positions. The continuum sky subtraction accuracy is typically and [N II] λ6583. 2–3 per cent of the sky level. 1 Because of the results we found in the further analysis of the We later used standard IRAF tasks to perform the absolute flux data, instead of identifying five areas within the NGC 1512/1510 calibration of the 2D spectra using the data obtained for the spec- system as KLS09 did, we have re-classified all UV-rich knots in trophotometric standard star. The correction for atmospheric ex- four zones: the star-forming ring in NGC 1512 (Ring, Zone 1), the tinction was achieved using an average curve for the continuous long XUV arm at the east (Arm 1, Zone 2), the arm between the atmospheric extinction at Siding Spring Observatory. We estimate long XUV arm and the ring of NGC 1512 (internal arm, Zone 3) that the flux-calibrated spectra are typically accurate to ∼5 per cent, and the knots found somewhat dispersed at the W and NW (external although this number may reach 10 per cent at the blue end of the debris of Arm 2, Zone 4). We then numbered the observed knots blue camera (wavelengths lower than 3800 Å) because of the de- within these four regions that we use to clearly identify each one crease of the sensitivity of the instrument in this spectral range. throughout our study. Table 2 compiles the coordinates of each one of the 164 targeted regions. 3 ANALYSIS AND RESULTS This section details the analysis of the ionized gas within the 3.2 Emission-line analysis UV-bright knots observed using our intermediate-resolution opti- The IRAF software was used to analyse the 1D spectra. Line cal spectroscopy and provides the results (Hα fluxes and equivalent fluxes and equivalent widths were measured by integrating all widths, nature of the ionization, reddening, oxygen and nitrogen the flux in the line between two given limits and over a local continuum estimated by eye. We want to emphasize that, opposite to an automatic process, the visual inspection of the spec- 1 IRAF (Image Reduction and Analysis Facility) is distributed by NOAO tra is needed to get a proper estimation of the adjacent contin- which is operated by AURA Inc., under cooperative agreement with NSF. uum and hence a reliable line flux estimation when emission lines

MNRAS 450, 3381–3409 (2015) 3384 A.´ R. Lopez-S´ anchez´ et al. Downloaded from http://mnras.oxfordjournals.org/ at Oxford Brookes University on June 4, 2015

Figure 1. Deep NUV image of the NGC 1512/1510 galaxy pair. The UV-bright regions observed using 2dF/AAOmega at the AAT that have been identified and analysed are indicated using a code of digits and colours. The first digit indicates the zone where the region is located. We distinguish four zones: Ring (1), Arm 1 (2), internal arm (3) and external debris of Arm 2 (4). The second digit is simply a sequential number. The colour indicates the emission lines detected in their spectrum (see Table 1). Two regions were observed in NGC 1510: one corresponds to its centre, while the other is at the west of the galaxy. The UV-knots within the two TDG candidates (see fig. 8 in KLS09) were also observed but they do not appear here. North is up and east to the left. are faint. The errors associated with the line flux measurements respect to Hβ and their associated errors for all the regions. The [N II] were estimated considering the noise in the adjacent continuum, λ6583/Hα ratio is also included as it is one of the magnitudes that the width of each emission line, and the photon noise at the line we will use to estimate the oxygen abundance Table 2 also includes profile. the observed Hα flux (uncorrected for extinction), the radial velocity Table 2 compiles the dereddened line intensity ratios and their at which the Hα line is observed, and the equivalent widths of the associated errors for all the regions. This table includes the line Hα,Hβ,andHγ lines. The error in the Hα flux was computed intensity ratio of [O II] λ3727, Hγ ,[OIII] λ5007, and Hα with considering the quadratic sum of the error in its flux measurement

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3385

Table 1. Statistical details of the results of our 2dF/AAOmega observations. Last column indicates the colour each region is identiﬁed in Fig. 1.

Fibre Per cent Per cent Colour in number total class Fig. 1

Observed 164 100.0 – – Non-identified 11 6.7 – red Identified 153 93.3 – – Background galaxies 17 10.4 11.1 cyan Foreground stars 1 0.6 0.7 blue Regions in NGC 1512/1510 135 82.3 88.2 – Only Hα detected 15 9.1 11.1 grey Hα and [N II] 30 18.3 22.2 yellow Hα,Hβ,and[NII] 6 3.7 4.5 pink Hα,Hβ,[NII], and [O III] 20 12.2 18.8 orange Hα,Hβ,[NII], [O III], and [O II] 64 39.0 47.4 green Downloaded from Hγ detected 51 31.1 37.8 – and the error in flux calibration, which we assumed to be 5 per cent. value agrees with the results found in XUV complexes surrounding Colons indicate errors of the order or larger than 40 per cent. other nearby spiral galaxies (Gil de Paz et al. 2007b). Fig. 2 shows some examples of the optical spectra obtained in our analysis of UV-bright regions within the NGC 1512/NGC 1510 http://mnras.oxfordjournals.org/ system. The two top panels plot the spectrum obtained for 3.4 First estimation of oxygen abundances NGC 1510 in each plate configuration. The fibre in Plate 1 was The preferred technique for determining the chemical abundances located in the central, brightest part of this BCDG, whereas the fi- of the ionized gas is via the Te method, which uses electron temper- bre in Plate 2 was located slightly towards the west. Fig. 2 also plots ature sensitive lines such as the [O III] λ4363 line (e.g. Peimbert & the spectrum of Zone 4 Region 17, for which we derive a relatively Costero 1969; Osterbrock & Ferland 2006). However, the auroral low oxygen abundance following the direct Te method (see below), lines are much fainter than the bright nebular lines and therefore are and the spectrum of a typical UV-bright region within the XUV disc rarely detected in galaxies, particularly at high metallicities. Hence, of the NGC 1512/1510 system. strong emission line (SEL) methods must be used to derive oxygen abundances in external galaxies. These empirical techniques at Oxford Brookes University on June 4, 2015 are widely used today. The most-common SEL calibrations involve using the N2 and O3N2 parameters, defined as 3.3 Nature of the ionized gas I([N II]λ6583) We checked the nature of the ionization of the UV-bright regions N2 = log , (1) I(Hα) using the so-called diagnostic diagrams, as first proposed by Bald- win, Phillips & Terlevich (1981) and Veilleux & Osterbrock (1987), I λ I λ which are now extendedly used to distinguish between pure star- ([N II] 6583) ([O III] 5007) O3N2 = log . (2) forming regions and AGN/low-ionization nuclear emission-line re- I(Hα) I(Hβ) λ β gion activity. Fig. 3 plots the typical [O III] 5007/H versus [N II] The huge advantage of using such ratios is that they consider emis- λ α 6583/H diagram for the 84 regions for which the four emission sion lines which are very close in wavelength, and hence both the lines are detected. We distinguish between objects observed using reddening correction and the absolute flux calibration can be ne- Plate 1 (blue stars) and those observed using Plate 2 (purple dia- glected. However, these SEL techniques do not take into account any monds) in order to quantify if any offset is observed between the ionization parameter, and hence accurate oxygen abundances (i.e. two data sets. We used the analytic relations given by Dopita et al. with uncertainties lower than ∼0.10 dex) cannot be derived. Fur- (2000) and Kewley et al. (2001), as well as the empirical relation thermore, precaution should always be taken when using any SEL provided by Kauffmann et al. (2003), to check the nature of the method. Reviews of the most-common empirical calibrations and excitation mechanism of the ionized gas within these regions. As their limitations can be found in Kewley & Ellison (2008), Lopez-´ we expected, all regions lie below the Kewley et al. (2001) theo- Sanchez´ & Esteban (2010b), and Lopez-S´ anchez´ et al. (2012). We retical line for starburst galaxies, which considers continuous star refer the reader to these studies for more details. formation. This clearly indicates that photoionization is the main Here, we first used the empirical calibrations provided by Pettini excitation mechanism of the gas and that there is no evidence for & Pagel (2004), which considers the N2 and O3N2 ratios, to get a significant contribution from shock excitation. Interestingly, the a first estimation of the oxygen abundance of the UV-rich regions loci of the UV-rich regions within this diagram agree well with the detected within the NGC 1512/1510 system. We used the linear predictions given by the models provided by Dopita et al. (2000), relation between the oxygen abundance and the N2 ratio, which considered extragalactic H II regions with instantaneous star + / = . + . , formation. 12 log(O H) 8 90 0 57N2 (3) Considering that a single O7V star has a Hα luminosity of for the 120 regions for which we detect both emission lines, and the L = 1.36 × 1037 erg s−1 (Schaerer & Vacca 1998), using the O7V linear relation between O/H and the O N ratio, extinction-corrected Hα flux we estimate that the number of mas- 3 2 sive ionizing stars in these regions typically is between 1 and 5. This 12 + log(O/H) = 8.73 − 0.32O3N2, (4)

MNRAS 450, 3381–3409 (2015) 3386 A.´ R. Lopez-S´ anchez´ et al.

for the 84 regions for which the four lines are detected. Table 3 0.1 0.2 compiles the oxygen abundance estimated from these SEL methods, ± ± – – – – –

H γ including the value of the N2 and O3N2 parameters for each region. − 6.6 − 10.1 0.1 0.2 2– 3.5 Correction for reddening ± ± ± – EW H β When at least Hα and Hβ are available, we used the method de- − 67.5: – − 12 scribed in Lopez-S´ anchez´ & Esteban (2009) to correct all line ﬂuxes − 32.2 − 17.5 for both reddening and underlying stellar absorption. This is an iter- 9 3 3 12

0.0 – ative procedure to derive simultaneously the reddening coefﬁcient, ± ± ± ± ± – c(Hβ), and the equivalent widths of the absorption in the hydrogen H α (Å) (Å) (Å) − 57.7: – − 12.7: – − 19.1: – − 78 − 56 lines, W , to correct the observed line intensities for both effects.

− 188 abs − 0.3 − 107 The method also assumes that Wabs is the same for all the Balmer lines and uses the relation given by Mazzarella & Boroson (1993) α to perform the absorption correction, ]/ H – – – – – ± 0.000 ± 0.003 II

⎡ ⎤ Downloaded from I λ W [N ( ) × 1 + abs 1 I(Hβ) W β c(Hβ) = log ⎣ H ⎦ , (5) F λ W f (λ) ( ) × + abs F (Hβ) 1 Wλ β 0.01 0.084 0.05 0.133 0.65 – – – – – – /H ± ± ± for each detected hydrogen Balmer line, where F(λ)andI(λ)are H α

the observed and the theoretical ﬂuxes (unaffected by reddening http://mnras.oxfordjournals.org/ or absorption), Wabs, Wλ,andWHβ are the equivalent widths of the underlying stellar absorption, the considered Balmer line and Hβ, 0.03 2.81 0.05 2.83 ]/ H β – – – – – 2.80: 0.070: – 2.86 – λ β ± ± respectively, and f( ) is the reddening curve normalized to H using III the Cardelli, Clayton & Mathis (1989) extinction law. [O However, we do not consider the ‘standard’ value of the theoretical Hα/Hβ ratio for Case B recombination, 2.86, as this number β 0.02 4.53 0.07 2.43 actually depends on both the electron temperature, Te, and the elec- – – – – /H – – – ± ± tron density, ne, of the gas. The same happens to other H I Balmer H γ ratios. Appendix discusses this issue and provides relationships between the electron temperature and the theoretical Hα/Hβ,Hγ /Hβ, at Oxford Brookes University on June 4, 2015 Balmer absorption lines. I 0.02 0.44 0.07 0.45 and Hδ/Hβ ratios, using the Storey & Hummer (1995) calculations ]/ H β ± ± −3 II and assuming ne = 100 cm . More importantly, Appendix A also

[O gives empirical relationships between the oxygen abundance and such theoretical Balmer ratios. We here used those equations and )

− 1 the oxygen abundance determined following the Pettini & Pagel rad ––

v (2004) calibrations to get the theoretical Hα/Hβ ratio (and the Hγ /Hβ ratio if available) for each individual region. The results α are listed in Table 2, as these will be the H I Balmer ratios obtained H 7 958.3 – 12 878.7 – 2200 1015.7 2.99 140 1031.2 4.76 4.3 960.6 – after applying the correction for reddening and underlying stellar ± ± ± ± ±

(1) (km s absorption. Equivalent widths are also compiled in Table 2. Flux 1 and other spectral properties of the regions. The full table is available online. β = The bottom panel of Fig. 4 compares the values derived for c(H ) )

β following the Hα/Hβ and Hγ /Hβ ratios. In order to quantify the Wabs which best matches both values, we looked for the linear ﬁt between both reddening coefﬁcient that was providing r ∼ 0 (i.e. there is not correlation between both variables). This analysis yielded an =

◦ average value of Wabs 0.85 Å, corresponding to a dispersion = β . It has not been corrected for extinction. d 0.071 dex. We note that the value of c(H ) derived from the − 43 23 58.97 43300 − 43 23 59.53 2810 − 43 39 41.36 107 − 43 39 41.36 37.2 − 43 39 28.06 4.9: 969.5 – − 43 39 28.06 3.7: 956.0 – − 43 19 54.91 5.0: 962.0 – − 43 19 54.91 – − 43 20 54.86 219

− 1 Hγ /Hβ ratio is systematically 0.05 dex higher than the reddening coefﬁcient derived using the Hα/Hβ ratio. This is very probably a erg s consequence of the much higher uncertainty in the measurement of − 17 RA Dec. γ

hms the H emission line. We then assumed an average value of Wabs = 0.85 Å and applied equation (5) to compute the reddening coefﬁcient c(Hβ) using the Hα/Hβ ratio. If Hγ is observed, we also applied equation (5) to derive c(Hβ)usingtheHγ /Hβ ratio. However, because of the low ﬂux in units of 10 signal-to-noise (S/N) ratio of the Hγ line in the majority of the α

Dereddened line intensity ratios with respect to I(H cases, the reddening coefﬁcient obtained using the Hα/Hβ ratio

(1) H (and not the average value between both reddening coefﬁcients) is considered in all cases for our analysis. Table 3 lists the reddening This object is a background galaxy.This The object number is given a here foreground is star. the The redshift number estimated given here using is emission the lines. radial velocity estimated using the H Table 2. Zone Region Plate0 Coordinates N1510 1 04 03 32.64 Notes. a b 0 N1510W 2 04 03 32.05 0 TDG1W 1 04 02 38.12 0 TDG1W 2 04 02 38.12 0 TDG1E 1 04 02 35.67 0 TDG1E 2 04 02 35.67 0 TDG2 1 04 01 18.09 0 TDG2 2 04 01 18.09 0 N1512 1 04 03 54.21 coefﬁcient derived using the Hα/Hβ ratio for 90 regions both lines

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3387 Downloaded from http://mnras.oxfordjournals.org/ at Oxford Brookes University on June 4, 2015

Figure 2. Example of the optical spectra obtained in our analysis of UV-bright regions within the NGC 1512/NGC 1510 system. We plot four regions: NGC 1510 (which has the highest S/N ratio), NGC 1510 W, and regions 4_17 and 2_2 (which has a medium-low S/N ratio and exempliﬁes the typical spectrum obtained for the observed UV-bright regions). The most important emission lines are labelled. The spectra have been corrected for radial velocity andsky emission, but not for extinction.

MNRAS 450, 3381–3409 (2015) 3388 A.´ R. Lopez-S´ anchez´ et al. 0.03 0.03 0.03 0.03 0.06 ± ± ± ± ± – log(N/O) 1.39 − 0.78 − 0.72 − − 1.34 − 0.64 0.07 0.07 0.07 0.06 0.08 0.15 – 0.15 – 0.13 – ± ± ± ± ± ± ± ± P ; KK04: Kobulnicky & a and 23 R log(O/H) 12 + , b )using 2001a PP04a PP04b P01 KK04 Adopted Downloaded from 0.017 8.65 8.80 8.68 9.12 8.74 0.029 8.66 8.74 8.71 9.10 8.72 0.015 8.40 8.33 8.10 8.51 8.36 0.004 8.29 8.18 8.22 8.44 8.23 0.083 8.68 8.70 8.72 9.09 8.71 2 ± ± ± ± ± O – 8.69 8.77 – – 8.77 – 8.69 – – – 8.69 – 8.24 – – – 8.24: – – 8.67 – – – 8.67 –––––– 2 N ; P01: Pilyugin ( 2 1.102 0.008 N − 0.105 − 0.060 − − − 1.101 3 http://mnras.oxfordjournals.org/ 0.037 0.050 2 0.017 0.12 0.004 ± ± N – – ± ± ± 3 O − 0.122: 2004 )usingO − 0.220 − 0.021

λ β 0.006 0.010 0.008 1.262 0.026 0.08 0.001 1.729 0.16 – 0.038 Figure 3. Comparison of the observed [O III] 5007/H and 0.13 – 2 ± ± ± ± ± ± ± ± II λ α –

[N ] 6583/H ﬂux ratios obtained for the regions where all four emis- N sion lines are available, with the diagnostic diagrams proposed by Dopita − 1.154: − 0.37 − 0.40 − 0.432 − 0.423 − 0.876 − 0.380 − 1.074 et al. (2000) – D00, black continuous line – Kewley et al. (2001) – K01, red − 0.370 dotted line – and the empirical relation provided by Kauffmann et al. (2003)

– Ka03, dashed green line. Blue stars indicate regions observed using Plate 1, at Oxford Brookes University on June 4, 2015 0.045 0.062 0.16 0.017 0.006 ; PP04b: Pettini & Pagel ( ± ± ± ± 2 – – whilst purple diamonds correspond to regions observed using Plate 2. – ± − 0.26 − 0.656 − 0.413 are detected, as well as c(Hβ)usingtheHγ /Hβ ratio for the 51 − 0.163 regions the three H I Balmer lines are measured. 0.017 0.037 0.14 0.013 0.007 0.310

The top panel of Fig. 4 compares the oxygen abundance and 2004 )usingN – – – Py ± ± ± ± the reddening coefficient derived for the regions these values are ± available. Although with a considerable scatter, it is evident that the reddening coefficient (i.e. the amount of dust and obscuration 0.410.07 0.36 0.181 0.12 0.279 0.16 0.408 in each region) increases with increasing metallicity. A linear fit to 0.06 0.671 23 – – – ± ± ± ± ± the data, R c(Hβ) = (−3.34 ± 0.41) + (0.449 ± 0.053)x, (6) 0.17 1.91 0.04 1.66 0.06 1.74 0.07 8.03 0.02 9.08 3 – – – ± ± ± ± ± where x = 12+log(O/H), is shown with a dotted black line in the R figure. The correlation coefficient of the fit is r = 0.6450. This behaviour was already observed by Lopez-S´ anchez´ & Esteban (2010b) 0.18 1.23 0.02 1.36 0.05 1.25 0.07 4.76 0.04 2.99 when analysing the physical and chemical properties of a sample of 2 ± ± ± ± ± strong starburst galaxies. A similar behaviour, but within the Balmer R decrement and the luminosity (which scales with the oxygen metallicity following the well-known luminosity–metallicity relation), /H β γ

was reported by Gunawardhana et al. (2013) when analysing the H ) properties of a large sample of star-forming galaxies using data of β c (H 0.02 – – 0.08 0.41 – 0.050.01 0.40 0.29 0.68 0.30 0.01 0.27 0.48 0.02 – – 0.01 0.39 3.27 the Galaxy And Mass Assembly (GAMA) survey (Driver et al. 2011; 0.01 0.55 6.10 /H β y . ± ± ± ± ± ± ± ± α ,

Hopkins et al. 2013). Interestingly, the position of the BCDG NGC H 23 1510 lies far from the observed relation. R

3.6 Recomputing oxygen abundances Reddening coefﬁcient, important emission-line ratios and parameters used by empirical calibrations and derived oxygen abundances and N/O ratio for our sample regions. The full table is available 2004 )using Empirical calibrations used and parameters involved: PP04a: Pettini & Pagel ( Once the reddening is known and in those regions where we measure a the [O II] λ3727 doublet, we can re-compute the oxygen abundances 6–– –161– – – – – 1310.44 1420.39 1520.24 1–– –111– – – – – 1220.25 0 TDG1W 1 0.09 ””20.03 ””20.27 Kewley ( Table 3. Zone Region Plate 0 N1510 1 0.39 Notes. using additional SEL methods. We use the R23 parameter, online.

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3389 Downloaded from

Figure 4. (Top panel) Comparison between the reddening coefﬁcient, Figure 5. Comparison of oxygen abundances derived following the Pettini c(Hβ) derived from the Hα/Hβ ratio (y-axis) and the oxygen abundance & Pagel (2004) and Pilyugin (2001a,b) methods. Top panel compares the estimated for each knot (x-axis). (Bottom panel) Comparison between the metallicity provided by the N2 parameter (x-axis) with the metallicity differ- c(Hβ) derived from the Hα/Hβ ratio (x-axis) and the difference between ence obtained between the O3N2 and the N2 calibrations (y-axis) following http://mnras.oxfordjournals.org/ Pettini & Pagel (2004). The red dashed vertical line at 12+log(O/H) = 8.65 the c(Hβ) derived from the Hγ /Hβ and the Hα/Hβ ratio. A Wabs= 0.85 Å was used in both panels. The dotted lines show a ﬁt to the data, excluding indicates the upper limit the N2 calibration is valid. Bottom panel compares NGC 1510. Blue stars indicate regions observed using Plate 1, whilst purple the average metallicity value provided by the Pettini & Pagel (2004) meth- diamonds correspond to regions observed using Plate 2. ods (x-axis) with the metallicity difference between the Pilyugin (2001a,b) method (which uses the R23 and P parameters) and the average value provided by the Pettini & Pagel (2004) method (y-axis). The red dashed vertical I λ + I λ line at 12+log(O/H) = 8.30 indicates the lower limit to the high-metallicity R = ([O III]) 4959 ([O III]) 5007 , 3 β (7) branch. Blue stars indicate regions observed using Plate 1, whilst purple H diamonds correspond to regions observed using Plate 2. I([O II])λλ3726, 3729

R = , (8) at Oxford Brookes University on June 4, 2015 2 Hβ abundance given by the low-metallicity branch, the adopted oxy- R = R + R , 23 3 2 (9) gen abundance, and the oxygen abundance computed following the Pettini & Pagel (2004) method we discussed in Section 3.4, and the P (Pilyugin 2000)ory (McGaugh 1991) parameters, respectively. Table 3 compiles the R23, P and y parameters used for the SEL R3 P = , calibrations which considers [O II] λ3727 for each region, as well as R (10) 23 the oxygen abundances derived from those data. Fig. 5 compares the R 1 y = log 3 = log , (11) oxygen abundances derived following the Pettini & Pagel (2004) −1 R2 P − 1 and Pilyugin (2001a,b) methods. The red dashed vertical line at 12+log(O/H) = 8.65 indicates the upper limit the N2 calibration is to take into account the hardness of the ionizing radiation. Follow- valid, as this parameter saturates at high metallicity (for an extended ing Lopez-S´ anchez´ et al. (2012), we use the empirical calibrations discussion and important consequences when analysing metallici- between R23 and P and the oxygen abundance for high- and low- ties of large galaxy samples, see Lara-Lopez´ et al. 2013c). We find metallicity H II regions provided by Pilyugin (2001a,b), respectively, a relatively good agreement between the three methods, with an as well as the Kobulnicky & Kewley (2004) method, which is based average scatter of only 0.049 dex (for the N2–O3N2 comparison) on photoionization models and uses the R23 and y parameters. and 0.060 dex (for the N2,O3N2–R23, P comparison), as it seen in We also have to consider that calibrations involving the R23 pa- Fig. 5. rameter are bivaluated; these are given for 12+log(O/H) 8.0 (low Hence, we adopt that the valid oxygen abundance for each region + metallicity) and 12 log(O/H) 8.3 (high metallicity). In the case of is the average value of the N2,O3N2,andR23, P techniques. We the UV-bright regions observed in the NGC 1512/1510 system, we note that we have not considered the value given by the N2 cali- + do not have any with 12 log(O/H) 8.0, but there are a few objects bration when the oxygen abundance estimated by the O3N2 method in the intermediate metallicity range, 8.0 12+log(O/H) 8.3. was higher than 12+log(O/H) = 8.65. The second last column in In these cases, we just considered a weighted oxygen abundance Table 3 lists the final adopted oxygen abundance determined for following each region. The error in the oxygen abundance provided by each c − x x − c calibration was estimated from the uncertainties associated with the x = x u PP04 + x PP04 l , med low c high c (12) flux ratios and assuming that each SEL technique has an uncertainty d d of ±0.10 dex. We then estimated the final uncertainty considering where x = 12+log(O/H), cl = 8.0, cu = 8.3, and cd = cu−cl = 0.3, the number of SEL calibrations averaged and the dispersion of their and the subindices high, low, med, and PP04 refer to the oxy- values. In the majority of the cases, the dispersion of the oxygen gen abundance given by the high-metallicity branch, the oxygen abundances provided by the empirical calibrations was inferior to

MNRAS 450, 3381–3409 (2015) 3390 A.´ R. Lopez-S´ anchez´ et al.

±0.10 dex. In many knots, it was even within ±0.05 dex, i.e. the dispersion is of the order or inferior to the typical uncertainties of the individual calibrations. The ﬁnal uncertainty associated with the oxygen abundance of each knot is also listed in the second last column of Table 3. For the best cases, the ﬁnal uncertainty is just ±0.05 dex. Colons indicate errors larger than 0.3 dex.

3.6.1 Comparison between results of regions observed twice Figs 3–9 distinguish between regions observed using plate 1 (blue stars) and plate 2 (purple diamonds). This has been done to investigate if any correlation depending on the plate exists. As we see, a plate dependence is not observed, and indeed the scatter in the ﬁgures is essentially the same considering each plate independently. Fig. 6 explicitly checks the validity of this argument comparing the

results obtained for the regions for which we have observations Downloaded from in both plates. The top panel in Fig. 6 plots the oxygen abundance difference between plates for each region following the N2 method (blue stars), O3N2 method (red diamonds), and the Pilyugin (2001a,b) method (purple triangles). In all cases but NGC 1510 (for which we are not observing the same region within the galaxy) the dispersion is always inferior to 0.06 dex, that is, lower than the http://mnras.oxfordjournals.org/ typical ±0.1 dex uncertainty associated with empirical calibrations. Top middle panel of Fig. 6 compares the oxygen abundance differences with the total absolute flux of the Hα line. Again, the scatter is 0.06 dex and no dependence in the flux calibration appears. Indeed, when we compare the absolute Hα flux computed for each region using different plates (bottom middle panel of Fig. 6), we find that the flux difference is better than 15 per cent in almost all the cases, being the average scatter 7.7 per cent. Note these numbers do

not include the uncertainty in the flux calibration, which we already at Oxford Brookes University on June 4, 2015 estimated to be ∼5 per cent. Finally, bottom panel of Fig. 6 shows that the agreement between the reddening coefficients determined for each region that has been observed twice is excellent, with an average scatter of 0.026 dex without considering uncertainties. In summary, Fig. 6 also shows that the reduction process, absolute flux calibration, and emission-line fluxes estimation has been done Figure 6. Comparison of properties of the same region derived using dif- consistently in both plates. Interestingly, the fact that the oxygen ferent plate configurations. Top panel compares the oxygen abundance dif- abundances derived for NGC 1510 using SEL techniques differ that ferences derived following the N2 (blue stars), O3N2 (red diamonds), and much exemplifies how critical the position on the fibre within the R23, P (Pilyugin 2001a,b, purple triangles) calibrations. Dotted horizontal ± galaxy can be in some cases, and hence detailed studies of such lines indicate a difference of 0.06 dex. Top middle panel plots the oxygen abundance differences with the absolute Hα flux in Plate 1. Bottom middle objects really need IFS observations, such those provided by the panel compares the absolute Hα flux computed for each region using dif- CALIFA (Sanchez´ et al. 2012; Husemann et al. 2013)orSAMI ferent plates. Bottom panel plots the reddening coefficient differences with (Croom et al. 2012;Bryantetal.2015) surveys. We will further the absolute Hα flux in Plate 1. In all panels, the position of NGC 1510 investigate the reason of this mismatch below when computing the is indicated with black squares, however in this case we are not observing oxygen abundances following the Te method. exactly the same region of the galaxy, as it is discussed in the text.

the conversion between abundance scales for the high-metallicity 3.6.2 Using the Kobulnicky & Kewley (2004) calibration branch provided by Lara-Lopez´ et al. (2013b), To compare the results provided by the Kobulnicky & Kewley KK04u − 0.1026 KK04uT , = , (13) (2004) method with the other empirical techniques, we have to keep e LL 1.0211 in mind that SEL methods based on photoionization models (such where KK04u refers to the oxygen abundance – in units of as those presented in Kobulnicky & Kewley 2004) systematically 12+log(O/H)– given by the high-branch calibration provided by overestimate by 0.2–0.4 dex or more the oxygen abundances derived Kobulnicky & Kewley (2004) and KK04uTe,LL is this oxygen abun- using empirical calibrations which are based on direct estimations dance shifted to the absolute abundance scale given by the Te of the electron temperature (such as those provided by Pilyugin method. The top panel of Fig. 7 compares the KK04uT ,LL results 2 e 2001a,b; Pettini & Pagel 2004). As a ﬁrst approach, we adopt with those provided by the Pilyugin (2001a,b) calibration. As we

2 However, we note that detailed tailor-made photoionization models of of-the-art photoionization models (e.g. Dors et al. 2011;Perez-Montero´ individual galaxies (e.g. Perez-Montero´ et al. 2010) and some recent state- 2014) matches well the oxygen abundances derived using the Te method.

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3391 Downloaded from

Figure 7. (Top panel) Comparison of the oxygen abundances determined Figure 8. Comparison of the N2O2 value with the adopted oxygen metallicity for our data. The vertical red dotted line at 12+log(O/H) = 8.3 indicates following the Kobulnicky & Kewley (2004) method (scaled to the Te method using Lara-Lopez´ et al. 2013b) with the results provided by the Pilyugin the low limit to the high-metallicity branch. The green dash–dotted line is (2001a,b) calibration using the R and P parameters. (Bottom panel) Com- the quadratic relationship derived by Kewley & Dopita (2002), manually http://mnras.oxfordjournals.org/ 23 − parison of the adopted oxygen abundances with the results provided by scaled 0.36 dex to match with the data in our high-metallicity end. The the high-metallicity Kobulnicky & Kewley (2004) calibration without ap- red dashed line is a quadratic fit to our high-metallicity data. Blue stars plying any shift. A dot–dashed green line indicates a linear fit to the data indicate regions observed using Plate 1, whilst purple diamonds correspond (equation 14). In both panels, dotted horizontal lines indicate a difference of to regions observed using Plate 2. ±0.1 dex, while the red dashed vertical line at 12+log(O/H) = 8.3 indicates the lower limit to the high-metallicity branch. Blue stars indicate regions ob- the results provided by the Kobulnicky & Kewley (2004) method served using Plate 1, whilst purple diamonds correspond to regions observed perfectly matches the oxygen abundances adopted using the Pettini using Plate 2. & Pagel (2004) and Pilyugin (2001a,b) techniques. see, even after performing the shift of the Kobulnicky & Kewley at Oxford Brookes University on June 4, 2015 3.6.3 Using the N2O2 parameter as a high-metallicity oxygen (2004) values to the Te scale using equation (13), a significant offset abundance estimator of 0.1–0.2 dex between both data sets still remains. Equation (13) is then not valid for our objects, which are physically very differ- To further check the goodness of our data, we use the N2O2 param- ent to those Lara-Lopez´ et al. (2013b) used to get their calibration. eter, which is defined as, Therefore, we again suggest caution when blindly applying an SEL I λ = ([N II]) 6584 . technique to objects with different physical properties to those used N2O2 log (15) I([O II])λλ3726, 3729 to derive the empirical calibration (see Stasinska´ 2010,foranex- tended discussion). Kewley & Dopita (2002) recommended to use the N2O2 ratio as an We then use our data to try to get an absolute scale conversion estimator of the oxygen abundance in the high-metallicity regime, as which is valid for this kind of objects. The bottom panel of Fig. 7 its behaviour is almost linear with the oxygen abundance. Fig. 8 plots compares the adopted oxygen abundance (x-axis) with the result the N2O2 value derived for each region (y-axis) with the adopted provided by the Kobulnicky & Kewley (2004) method without ap- oxygen metallicity (x-axis). A clear, almost linear relation appears. plying any shift (y-axis). A linear fit to the data considering only The green dash–dotted line is the quadratic relationship derived − those regions with 12+log(O/H) ≥ 8.3 (we only consider data in by Kewley & Dopita (2002), manually scaled by 0.36 dex to the high-metallicity branch) provides match with the data in our high-metallicity end. Even though, this relationship seems not to be suitable for our objects in the range 8.3 KK04u − 1.727 ≤ + ≤ KK04uT = , (14) 12 log(O/H) 8.5. We perform a quadratic fit to the data in the e 0.8475 high-metallicity branch, yielding being the correlation coefficient r = 0.9773 and the dispersion x = 8.7457 + 0.3570N O − 0.04229(N O )2, (16) d = 0.021 dex. For 12+log(O/H) = 8.8, the offset is provided by 2 2 2 2 equation (14) is 0.45 dex. We now use equation (14) to convert the being x = 12+log(O/H). The correlation coefficient of this fit is = = results of the Kobulnicky & Kewley (2004) calibration into the Te r 0.977 59 and the dispersion d 0.026. We again note that scale. As expected, now the agreement is excellent, with a scatter this relationship has not been used to derive the adopted oxygen of 0.025 dex. abundance of the UV-bright regions within the NGC 1512/1510 We remark that we do not use this calibration for getting the final system but just to quantify the validity of our analysis. oxygen abundance for each region, but just to justify that different SEL techniques are providing quantitatively the same results in our 3.7 The N/O ratio analysis. That is, neglecting the problem of the absolute oxygen abundance scale (which following equation (14) provides an offset We computed the nitrogen-to-oxygen (N/O) ratio for those regions = = λ λ of 0.46 dex for KK04uTe 8.30 but 0.39 dex for KK04uTe 8.75), where both the [O II] 3727 and [N II] 6583 lines are observed.

MNRAS 450, 3381–3409 (2015) 3392 A.´ R. Lopez-S´ anchez´ et al.

We then infer the Te([O III]) from the [O III] (λ4959+λ5507)/λ4363 ratio by making use of the IRAF nebular package. As we assumed a two-zone approximation to deﬁne the temperature structure of the nebula, we used Te([O III]) as representative of high-ionization potential ions. The electron temperature assumed for the low-ionization potential ions was derived from the linear relation between Te([O III]) and Te([O II]) provided by Garnett (1992). We followed the very same prescriptions and ionization correction factors, icf, indicated by Lopez-S´ anchez´ & Esteban (2009)to compute the ionic abundances, the total O, N, and Ne abundances, and the N/O and Ne/O ratios, for the ionized gas within these four regions. We cannot derive the total S and Ar abundances because we do not observe any emission line associated with S++ or Ar++. The IRAF nebular package (Shaw & Dufour 1995) was used to compute the ionic abundances from the intensity of collisionally ex- Downloaded from cited lines, assuming Te[O III] for computing the ionic abundances ++ ++ +3 of O ,Ne ,andAr and the Te[O II] value derived from the Garnett (1992) relation for O+,N+,andS+. We note that we used Figure 9. (Top panel) Comparison between the oxygen abundance (x-axis) theatomicdatasetforO+ and S+ ions recommended by Garc´ıa- and the N/O ratio (y-axis) for the knots analysed here. (Bottom panel) The Rojas et al. (2005). We considered the standard icf of Peimbert & N/O ratio (y-axis) is plotted against the O2N2 ratio (x-axis). The dashed red

Costero (1969) to derive the total N and Ne abundances. http://mnras.oxfordjournals.org/ line is a fit to the data. Blue stars indicate regions observed using Plate 1, whilst purple diamonds correspond to regions observed using Plate 2. The results for the chemical abundances derived using the Te method are compiled in Table 5. This table also lists the electron We assumed the relationship provided by Garnett (1992) between temperatures derived for each region following the direct method. Te[O III]andTe[O II] and used the emission-line ratios to compute Table 5 also includes the oxygen abundance and the N/O ra- the electron temperatures which best reproduce the adopted oxygen tio adopted for each region following the SEL methods. Interest- abundances following the direct method using the IRAF nebular ingly, in all cases except NGC 1510 W the values provided by both −3 package (Shaw & Dufour 1995). We assumed ne = 100 cm and the Te and SEL methods are almost identical. Following the Te + + used Te[O II] and the [N II] λ6583/Hβ ratio to obtain the N /H method, both regions of NGC 1510W have the same metallicity, ionic abundance and derive the N/O = (N+/H+)/(O+/H+)ratio 12+log(O/H) ∼ 8.25, however SEL methods are overestimating (Peimbert & Costero 1969). This ratio does not strongly depend on this value by ∼0.1 dex. This is very likely an effect of the ionization at Oxford Brookes University on June 4, 2015 the electron temperature, as both O+ and N+ ions are assumed to be degree of the gas, i.e. the fibre position used for getting the data of in the low-ionization region. Hence, the derived uncertainties of the NGC 1510W is located almost entirely in the low-ionization region N/O ratio are low (±0.03 dex in many cases) and only consider the within the BCDG. Hence, some caution is needed when computing errors in the line-flux ratios. The N/O ratio adopted for each region oxygen abundances in these kinds of objects using SEL methods is listed in last column of Table 3. even if IFS data are available (see extended discussion in Lopez-´ The bottom panel of Fig. 9 compares the derived N/O ratio with Sanchez´ et al. 2011). the value of the N2O2 parameter. As we should expect following KLS09 used the emission-line intensity data of NGC 1510 pro- Fig. 8, we observe a very tight relation between both values. A vided by Storchi-Bergmann, Kinney & Challis (1995), which in- linear fit to the data provides cluded a detection of the auroral [O III] λ4363 line, to derive 12+log(O/H) = 7.95 and log(N/O) =−1.2 following the T method. y =−0.6071 + 0.6721x, (17) e KLS09 claimed that the high N/O ratio in NGC 1510 may be re- with r = 0.9930, where x = N2O2 and y = N/O. lated to the presence of Wolf–Rayet (WR) stars in this BCDG. Indeed, our spectrum shows a detection of the nebular He II λ4686 line (see Table 4) which seems to be associated with WR stars (e.g. 3.8 Chemical abundances using the direct method Lopez-S´ anchez´ & Esteban 2010a, and references therein). However, There are four regions for which we detect the auroral [O III] λ4363 our deeper data show that the oxygen abundance, 12+log(O/H) = emission line: the two positions of the BCDG NGC 1510, region 8.24 ± 0.06, and the N/O ratio, log(N/O) =−1.34 ± 0.05, com- 17 of zone 4, and region 27 of zone 2. For these four cases, we puted for NGC 1510 are perfectly compatible to those usually found computed the oxygen abundances of the ionized gas following the in BCDGs. Te method, taking into account all lines detected in each spectrum. Table 4 lists the dereddened line intensity ratios with respect to I(Hβ) = 100, the equivalent widths of the brightest H I Balmer 4 PROPERTIES DERIVED FROM UV AND lines, and the c(Hβ)andW iteratively derived using all available abs RADIO DATA H I Balmer lines and the method explained in Section 3.5. The electron density of the ionized gas, ne, was computed via KLS09 provided a unique data base of UV and 21 cm properties for the [S II] λλ6716,6731 doublet by making use of the five-level pro- the regions of the NGC 1512/1510 system. We here compile and gram for the analysis of emission-line nebulae included in the IRAF update those values to later perform our multiwavelength analysis. nebular package (Shaw & Dufour 1995). The four regions were We note that we used the same aperture to derive all fluxes (UV,Hα, −3 found in the low-density limit, ne < 100 cm , and hence we adopt and H I) in each region considering maps having the same resolution −3 ne = 100 cm . and point spread function (PSF) matched.

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3393

Table 4. Dereddened line intensity ratios with respect to I(Hβ) = 100 for the regions for which we have a direct estimation of the electron temperature. We also compile the equivalent widths of the brightest H I Balmer lines, the reddening coefﬁcient, c(Hβ), and the equivalent widths of the absorption in the hydrogen lines, Wabs, used to correct the spectra for reddening, and the equivalent widths of the emission H I Balmer lines.

Line f(λ) NGC 1510 NGC 1510 W Zone 4 Region 17 Zone 2 Region 27

[O II] 3727 0.322 291 ± 19 452 ± 33 210 ± 17 251 ± 12 H11 3771 0.313 – – 6.5 ± 1.6 – H10 3798 0.307 – – 9.3 ± 2.2 – H9 3835 0.299 – – 10.5 ± 2.3 – [Ne III] 3869 0.291 44.2 ± 4.3 38.2 ± 7.6 62.5 ± 4.6 62 ± 18 H8 3889 0.286 21.7 ± 2.9 – 26.6 ± 2.6 – [Ne III] 3969+H7 0.267 20.8 ± 2.4 – 34.6 ± 3.5 39 ± 12 He I 4026 0.251 – – 2.85 ± 0.54 – [S II] 4069 0.239 3.2 ± 1.0 – – – Hδ 4101 0.230 26.1 ± 2.5 26.5 ± 4.2 26.1 ± 2.4 26 ± 8 Hγ 4340 0.157 45.4 ± 2.8 45.9 ± 4.0 48.1 ± 3.0 48 ± 12 Downloaded from [O III] 4363 0.150 4.64 ± 0.80 2.43 ± 0.54 5.97 ± 0.78 4.26 ± 0.64 He I 4471 0.116 4.70 ± 0.80 – 3.68 ± 0.76 – [Fe III] 4658 0.059 2.50 ± 0.38 – – – He II 4686 0.049 1.70 ± 0.54 – – – [Ar IV] 4711 0.043 0.49: – – – [Ar IV] 4740 0.034 0.56: – – – Hβ 4861 0.000 100.0 ± 3.3 100.0 ± 4.3 100.0 ± 3.9 100 ± 6 http://mnras.oxfordjournals.org/ [O III] 4959 −0.025 142.8 ± 6.7 84.1 ± 6.0 165 ± 9 156 ± 10 [O III] 5007 −0.037 452.6 ± 21 243 ± 14 494 ± 25 462 ± 20 [N II] 6548 −0.295 7.98 ± 0.51 12.7 ± 1.2 4.25 ± 0.47 8.4 ± 0.9 Hα 6563 −0.297 286 ± 9 288 ± 5 284 ± 14 286 ± 12 [N II] 6583 −0.300 23.7 ± 1.3 37.6 ± 2.6 13.5 ± 0.9 23.3 ± 2.0 He I 6678 −0.312 2.34 ± 22 1.76 ± 0.39 2.18 ± 0.28 – [S II] 6716 −0.318 12.71 ± 0.74 20.3 ± 2.2 8.89 ± 0.62 21.5 ± 1.7 [S II] 6731 −0.319 9.45 ± 0.57 21.1 ± 1.6 6.45 ± 0.49 15.2 ± 1.4 −W(Hα) [Å] 188 ± 978± 3 774 ± 38 756 ± 38 at Oxford Brookes University on June 4, 2015 −W(Hβ) [Å] 32.2 ± 0.2 17.5 ± 0.2 216 ± 2 114 ± 3 −W(Hγ ) [Å] 10.1 ± 0.1 6.6 ± 0.2 66 ± 238± 3 −W(Hδ)[Å] 4.6± 0.1 3.0 ± 0.3 28.7 ± 1.5 10.3 ± 2.1 c(Hβ)0.36± 0.04 0.20 ± 0.3 0.01 ± 0.02 0.03 ± 0.02 Wabs [Å] 0.3 ± 0.1 0.3 ± 0.1 0.5 ± 0.1 0.5 ± 0.2

4.1 UV luminosities and colours for ages older than ∼10 Myr. Therefore, we assume that the effect of dust in the FN colour is well reproduced by a screen of dust As explained in section 4.2 of KLS09,theGALEX FUV and NUV showing a Milky Way-like extinction law. images were used to derive mFUV and mNUV magnitudes and the FUV − NUV (FN) colour. They also estimated the FUV flux density, −1 −2 −1 4.2 H I and gas masses fFUV (in units of erg s cm Å ), that was used to derive the star formation rate (SFR) of each region via the calibration provided The H I masses of each UV-bright region were derived by KLS09 us- by Salim et al. (2007). We include these results for each analysed ing the low-resolution Australia Telescope Compact Array (ATCA) 2 region in Table 6. Column 7 compiles the area (in kpc ) of the region, data considering the same region determined by their UV emission. Column 8 shows the FN colour, Column 9 indicates the fFUV,and Considering that the vast majority of the hydrogen in low-mass ob- Column 10 lists the SFR derived from the FUV flux density. jects is atomic (e.g. Leroy et al. 2005), the total gas mass can be We note our FUV-based SFR are slightly different from those = M determined assuming Mgas 1.32 H I, where the factor 1.32 cor- estimated by KLS09, as we now use the reddening coefficient rects the H I mass for the presence of helium. Column 11 of Table 6 β estimated for each region (or that given by equation 6 if H compiles the total gas mass derived for each region. As already was not observed) to correct the FUV flux density for extinction, noted by KLS09, the gas masses derived in the central regions of − = β assuming E(B V) 0.692c(H )mag, while KLS09 used the NGC 1512 are very probably underestimated, as we lack of CO ob- Galactic value provided by Schlegel, Finkbeiner & Davis (1998), servations to account for the molecular gas (H ) component, which − = 2 E(B V) 0.011 mag. is expected to be important in these knots (see fig. 20 and section β We also note that the effect of the so-called IRX– relation (i.e. 4.4 in KLS09). the relation between the UV slope or the FN colour with the infrared excess in either entire galaxies or individual regions within galaxies; 4.3 Stellar and baryonic masses see Boquien et al. 2009, and references within) seems to be small for the analysed regions in the NGC 1512/1510 system. We estimate We have used the UV data to get a tentative estimation of the stel- ∼ that this represents a variation of only 0.05 mag in the FN colour lar mass, Mstars, within each UV-bright complex. For this, we have

MNRAS 450, 3381–3409 (2015) 3394 A.´ R. Lopez-S´ anchez´ et al.

Table 5. Physical conditions and chemical abundances of the ionized gas of the regions for which we have a direct estimation of the electron temperature The oxygen abundance and the N/O ratio adopted for each region following the SEL methods are also included for comparison.

NGC 1510 NGC 1510 W Zone 4 Region 17 Zone 2 Region 27

Te[O III] [K] 11700 ± 600 11500 ± 900 12300 ± 600 11300 ± 700 Te[O II] [K] 11200 ± 500 11050 ± 600 11600 ± 400 10900 ± 500 −3 ne [cm ] <100 <100 <100 <100 log(O++/O+)0.09± 0.11 −0.35 ± 0.14 0.28 ± 0.09 0.18 ± 0.11 12+log(O+/H+)7.88± 0.10 8.09 ± 0.11 7.66 ± 0.07 7.86 ± 0.09 12+log(O++/H+)7.97± 0.06 7.74 ± 0.09 7.94 ± 0.05 8.04 ± 0.07 12+log(O/H), Te 8.23 ± 0.08 8.25 ± 0.11 8.12 ± 0.06 8.26 ± 0.08 12+log(O/H), emp 8.23 ± 0.06 8.36 ± 0.07 8.12 ± 0.05 8.24 ± 0.07 12+log(N+/H+)6.54± 0.05 6.76 ± 0.06 6.24 ± 0.04 6.59 ± 0.06 12+log(N/H) 6.89 ± 0.08 6.92 ± 0.08 6.70 ± 0.07 6.99 ± 0.09 icf(N) 2.22 ± 0.34 1.45 ± 0.17 2.90 ± 0.41 2.50 ± 0.44 Downloaded from

log(N/O), Te −1.34 ± 0.06 −1.33 ± 0.09 −1.42 ± 0.07 −1.27 ± 0.07 log(N/O), emp −1.34 ± 0.03 1.39 ± 0.03 −1.44 ± 0.03 −1.29 ± 0.03 12+log(S+/H+)5.59± 0.05 5.97 ± 0.06 5.39 ± 0.04 5.84 ± 0.06 12+log(Ne++/H+)7.43± 0.11 7.39 ± 0.18 7.49 ± 0.08 7.61 ± 0.08

icf(Ne) 1.82 ± 0.61 3.2 ± 1.6 1.53 ± 0.40 1.67 ± 0.64 http://mnras.oxfordjournals.org/ 12+log(Ne/H) 7.69 ± 0.16 7.9 ± 0.2 7.68 ± 0.13 7.83 ± 0.16 log(Ne/O) −0.54 ± 0.11 −0.35 ± 0.19 −0.45 ± 0.09 −0.43 ± 0.09 12+log(Ar+3/H+)4.8:– – – neglected the contribution of the current star formation event as ob- Zone 1 = Ring. This area corresponds to the central regions served from the ionized gas. We then assumed that the FN colours of NGC 1512. As expected, the star-forming knots within this and the FUV luminosities are coming from a dominant stellar pop- area show the highest metallicities of the system, ranging between ulation created in a past star formation event. The age of this star 8.6 ≤ 12+log(O/H) ≤ 8.8. The average value is 12+log(O/H) = formation event can be determined using the FN colour, as KLS09 8.71. The northern regions of this ring show slightly higher oxygen at Oxford Brookes University on June 4, 2015 already did (see section 4.2 in their paper). We then used theoretical abundances than the southern regions. evolutionary synthesis models provided by the STARBURST99 code Zone 2 = Arm 1. This is the long spiral/tidally induced arm at (Leitherer et al. 1999;Vazquez´ & Leitherer 2005) to compute the the east of the system. Although it seems it starts at the SW of stellar mass that matches the observed monochromatic FUV lumi- the system (bridge between NGC 1512 and NGC 1510) with rela- nosity and FN colour of each region. Although the FUV luminosity tively high metallicities – 8.5 ≤ 12+log(O/H) ≤ 8.6 – it later has depends poorly on the metallicity (e.g. see discussion in section 3 an almost flat gradient, with relatively low oxygen abundances, of Bianchi et al. 2005), we have also considered it using the oxygen 8.4 ≤ 12+log(O/H) ≤ 8.3. However, some bright UV regions abundance determined for each region. The adopted stellar mass are showing ever lower values, reaching 12+log(O/H) ∼ 8.2 in is listed in Column 12 of Table 6. As these are just tentative esti- some knots. The average value is 12+log(O/H) = 8.39, although if mations, we do not provide errors for the stellar mass of the UV the innermost five regions are not considered the average value of regions or for any other property derived using this value. the metallicity of Arm 1 is 12+log(O/H) = 8.35. The total baryonic masses are estimated adding the gas and the Zone 3 = internal arm. This area corresponds to the inner part stellar masses and are listed in Column 12 of Table 6. Similarly, the of an original, long arm (Arm 2) which, starting at the NE of the gas-to-star mass ratio, Mgas/Mstars, has been also derived for each central ring, has been now partially destroyed because of the inter- region, and it is compiled in Column 13 of Table 6. action between NGC 1512 and NGC 1510 (KLS09). Because of its closeness to the centre of the system, we should expect intermediate to high metallicities in this area, and indeed that is what we find. Metallicities range between 8.65 and 8.50, the average value be- 5 DISCUSSION ing 12+log(O/H) = 8.55. Interestingly, the two regions found at the western end of this area (3_24 and 3_25, see Fig. 1) have lower abun- 5.1 An oxygen abundance map of the NGC 1512/150 system dances, 12+log(O/H) ∼ 8.40, than expected, as being located very Our study allows us to map the oxygen abundance of the NGC close to the inner areas of the system. Indeed, region 4_01, which 1512/1510 system. For this, we use the data of the 92 unique UV- is found at just 2 kpc N from region 3_25, has 12+log(O/H) = bright regions (see Table 6) for which we compute the oxygen 8.71. This fact is evidencing that the interaction between the two abundance to create the metallicity map we show in Fig. 10.We main galaxies has largely disturbed the distribution and properties have assigned different symbols to each of the four zones identified of the star-forming regions of the system. in the system (see Fig. 1). We also include the positions of the BCDG Zone 4 = external debris. this corresponds to the external part of NGC 1510 (star) and the TDG 1 (pentagon), but this actually lies the old Arm 2, and hence in the past it should have been connected far from the system, at 78 kpc (31.4 arcmin) from the centre of to Zone 3. Now the distribution of star-forming regions is not as NGC 1512. We analyse this map by zone. uniform as it was probably before, which we suspect was similar

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3395

) to what we now see in Arm 1. Furthermore, we note it is not easy − 2 /A) to associate some knots to be in Arm 1 or Arm 2. Hence, regions bar kpc 1_01 to 1_05 may actually be in the external debris zone, while some of the knots 4_19, 4_20, 4_26 and 4_27 may be originally

)(M associated with Arm 1. Knot 4_1 may actually belong to the Ring. /A) log( M − 2 KLS09 already discussed that the H I morphology and kinematics stars kpc of this zone is also much more disturbed than that observed in Arm 1. We now see the same behaviour in the metallicity distribution of

)(M its star-forming regions. Although the average value in this zone is − 2 /A) log( M + = + = 0.06 6.78 7.04 0.09 6.24 6.71 0.10 6.03 6.68 0.03 6.83 7.15 0.10 5.80 6.65 0.09 6.31 6.85 0.09 6.57 6.89 0.03 7.19 7.42 0.11 5.71 6.70 12 log(O/H) 8.44, we ﬁnd regions from 12 log(O/H) 8.71 gas kpc ± ± ± ± ± ± ± ± ± M to 12+log(O/H) = 8.12. Actually, three of the four most metal- poor star-forming regions within the NGC 1512/1510 system, 4_17,

)(M 4_32, and 4_38, are located here. Particularly interesting is the knot − 2 4_17, for which we derive 12+log(O/H) = 8.12 ± 0.05 consistently 0.04 6.53 0.06 6.57 0.03 6.87 0.04 6.69 0.07 6.58 0.05 6.70 0.05 6.61 0.12 7.03 0.09 6.65 kpc ± ± ± ± ± ± ± ± ± using both the direct and the SEL methods. − 1 yr 4.22 log(SFR/A) log( Downloaded from It is very interesting to note the chemical differences between − 2.10 − 1.96 − 1.49 − 1.68 − 2.07 − 2.18 − − 2.73 − 2.99 M Arm 1 and the external debris of Arm 2. We should expect that, originally, these two long arms shared similar properties, as we may gas stars M M assume that both structures were created from the same material at

) the outskirts of NGC 1512. We will further discuss this issue below. )( bar http://mnras.oxfordjournals.org/ 5.2 Radial metallicity gradient ) log ( M )(M

stars The left-hand panel of Fig. 11 compares the oxygen abundance of each of the analysed UV-rich regions with their distance to the centre of NGC 1512 in units of R25 (note that we are using projected ) log( M

)(M galactocentric distances in our analysis). Because of the complexity gas 0.09 6.08 6.55 1.95 0.10 5.71 6.36 3.41 0.03 7.41 7.73 1.08 0.06 6.68 6.94 0.81 0.10 5.44 6.29 6.07 0.09 5.97 6.50 2.46 0.03 7.59 7.82 0.684 0.09 6.35 6.67 1.09 0.11 5.13 6.11 8.67 ± ± ± ± ± ± ± ± ± of the system, it is not easy to see just a radial decrease of metallicity from the centre to the outskirts. Considering all knots in the ring and in the internal arm and all regions of the external debris with )(M at Oxford Brookes University on June 4, 2015

− 1 + > 0.04 6.37 0.06 6.25 0.03 7.44 0.04 6.59 0.05 6.36 0.12 7.43 0.05 6.39 0.07 6.22 0.09 6.06 12 log(O/H) 8.20, a linear ﬁt yields a metallicity gradient of yr ± ± ± ± ± ± ± ± ±

12 + log(O/H) = 8.703 − 0.158R25, (18) log(SFR) log( M − 2.26 − 2.28 − 0.92 − 1.78 − 2.53 − 3.82 − 2.96 − 2.42 − 3.58 with r = 0.7862. This fit is plotted in the left-hand panel of Fig. 11 9.3 7.2 5.7 110 13 6.1 2.5 4.2 3.4 with a red dashed line. Note that the zero-point of this fit matches ± ± ± ± ± ± ± ± ± FUV f the average oxygen abundance of the central ring. However, a linear fit to all UV-bright regions along Arm 1, neglecting knots 1–5 which have metallicities at least 0.2 dex higher than the rest, and including 0.05 187.3 0.08 83.1 0.03 3580 0.03 320 0.07 94.6 0.14 16.1 0.06 126.3 0.10 47.0 0.12 29.9 a the TDG 1, provides a metallicity gradient with a very flat slope, ± ± ± ± ± ± ± ± ± FN

12 + log(O/H) = 8.400 − 0.023R25, (19) ) (mag) (2) ( M 2 with r = 0.3294, which is plotted with a blue dot–dashed line in the figure. As was already said, note that although an almost flat metallicity gradient along Arm 1 exists, the dispersion is ∼0.2 dex. 0.03 0.690 0.21 0.06 0.481 0.21 0.03 3.748 0.23 0.03 0.797 0.31 0.03 0.450 0.26 Using the average metallicity gradient provided by equation (18), ± ± ± ± ± – 2.520 0.43 we prepared a residual metallicity map (left-hand panel of Fig. 12) − 0.78 − 0.64 − 1.34 − 0.78 − 0.72 which takes into account how much this value differs with respect to the adopted oxygen abundance for each region. Besides NGC 0.07 0.08 0.05 0.08 0.15 – 0.597 0.33 0.15 – 0.438 0.21 0.07 0.13 – 0.258 0.20 1510 and the three low-metallicity knots we already identified in . ± ± ± ± ± ± ± ±

− 1 the external debris, this map identiﬁes two more zones where knots + log(O/H) log(N/O) Area Å have oxygen abundances which are systematically lower than their − 2 surroundings by about 0.2–0.3 dex. These are the most southern cm )12 0.01 8.74 0.05 8.71 0.01 8.23 0.02 8.73 0.01 8.24: 0.08 8.77 0.01 8.72

β knots of the internal arm and some of the regions within Arm 1 − 1 ± ± ± ± ± ± ±

c (H located at the N of the system. erg s 11.76 kpc − 17 25 NUV. (1) (dex) (dex) (dex) (kpc R = 5.3 A metallicity gradient following the spiral H I arms − 25 R Properties of the UV-rich regions within the NGC 1512/1510 system. The full table is available online. Instead of using just a radial gradient, we considered a radial plus FUV (1) I = azimuthal gradient by deﬁning two spirals which follow the H arms and UV-richcomplexes observed in the NGC 1512/1510 system (see FN 1 5 0.28 0.24 1 4 0.31 0.39 1 7 0.32 0.35 1 6 0.28 – 8.69 (2) In units of 10 0 TDG1 6.27 0.06 Table 6. Zone Reg 0 N1510 1.49 0.39 Notes. a 1 3 0.36 0.44 11 1 2 0.51 0.41 0.25 – 8.67 Fig. 10). These spirals were constructed by eye using the equation

MNRAS 450, 3381–3409 (2015) 3396 A.´ R. Lopez-S´ anchez´ et al. Downloaded from http://mnras.oxfordjournals.org/ at Oxford Brookes University on June 4, 2015

Figure 10. Map of the oxygen abundance in the NGC 1512/1510 system, as it is provided via the analysis of the emission lines detected in the UV-rich regions. The four zones identiﬁed in the system (see Fig. 1) are shown using different symbols: Zone 1 = Ring (Diamonds), Zone 2 = Arm 1 (circles), Zone 3 = internal arm (triangles), Zone 4 = external debris (squares). The position of the BCDG NGC 1510 is shown with a star, while the TDG 1 (which actually lies far from the system, at 78 kpc from the centre of NGC 1512) is plotted using a pentagon. The scale included in the top-left corner indicates the lengthof 5 arcmin, which corresponds to 12.44 kpc at the distance of NGC 1512. The value for R25 in NGC 1512 is 11.76 kpc (4.26 arcmin). The two H I tidal/spiral arms are sketched following a blue (Arm 1) or green (Arm 2) dotted line.

Figure 11. (Left-hand panel) Oxygen abundance versus R25 for the UV-rich regions analysed in the NGC 1512/1510 system. The red dashed line in the left-hand panel indicates the linear fit given by equation (18), which considers al knots in ring and in the internal arm and all regions of the external debris with 12+log(O/H) > 8.20. The blue dot–dashed line in the left-hand panel is a linear fit to all the regions in Arm 1, neglecting knots 1–5 which have metallicities at least 0.2 dex higher than the rest, and the TDG 1 (which is not shown here). The data points used for getting this fit are marked with a blue circle. (Right-hand panel) N/O ratio versus R25 for the UV-rich regions analysed in the NGC 1512/1510 system. The red dashed line is a linear fit to the same regions used to get the fit provided in equation (18). In both diagrams, colour indicates the Mgas/Mstars ratio of each region. Symbols are the same than those in Fig. 10.

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3397 Downloaded from Figure 12. (Left-hand panel) Oxygen abundance residual map constructed using the average metallicity gradient given by equation (18). (Right-hand panel) N/O map of the system. In both panels, symbols are the same than those in Fig. 10. http://mnras.oxfordjournals.org/ at Oxford Brookes University on June 4, 2015

Figure 13. Oxygen abundances of the UV-rich regions following the spiral Arm 1 (bottom panel) and the spiral Arm 2 (top panel). Blue and red dotted lines mark the lower/higher oxygen abundances found in the outskirts of Arm 1. The dashed black line shows a schematic model combining a linear gradient (for −1 R < 10 kpc) with slope α =−0.045 dex kpc and zero-point z0 = 8.80 dex and a ﬂat gradient (for R > 10 kpc) with 12+log(O/H) = 8.35 constructed to guide the eye. Colour indicates the Mgas/Mstars ratio of each region. Symbols are the same than those in Fig. 10.

θ = θ 0 + 1/bln (R/a), R being the distance to the centre of the could be even lower than the 12+log(O/H) ∼ 8.20 we still see in spiral, using the observed knots as a guide. For each region, we some knots, as is observed in the outskirts of spiral galaxies such searched for the minimum distance between its actual position and as NGC 300 (Bresolin et al. 2009b), M 83 (Bresolin et al. 2009a), the spiral given by Arm 1 (all regions in Zone 2, which we already NGC 4625 (Goddard et al. 2011), and NGC 3621 (Bresolin et al. identified with this arm) or Arm 2 (for regions belonging to Zones 2012). 3 and 4). We are not including the regions in the central ring (Zone In any case, almost all regions associated with the original Arm 2 1), NGC 1510, and TDG 1 in this analysis. now have systematically higher abundances, 12+log(O/H) ∼ 8.4– The left-hand panel of Fig. 13 plots the oxygen abundance of 8.6, than those given by our schematic model (left-hand panel of the star-forming regions along these two spiral arms. Following this Fig. 13). The same is happening to the five internal regions within figure, it is quite evident that the two spiral arms have experienced Arm 1. This fact informs that the star formation activity and the different star-formation histories. Assuming that the regions with chemical enrichment have been much more important in regions R 10 kpc in Arm 1 are showing the original flat oxygen abun- associated with or close to Arm 2, at the NW of NGC 1512. KLS09 dance distribution (i.e. before the two main galaxies interacted) already showed that the interaction with NGC 1510 has modified with 12+log(O/H) ∼ 8.35, and assuming that the centre of NGC the H I morphology and kinematics of the neutral gas in this zone, 1512 has 12+log(O/H) ∼ 8.80, the slope for a linear gradient for and that this is the reason why a much more chaotic distribution R < 10 kpc is α =−0.045 dex kpc−1. Both the slope and the con- of star-forming regions is observed. We now confirm that both the stant oxygen value in the outer regions are similar to those reported enhancement of the metallicities and the large metallicity dispersion by Bresolin et al. (2012) for NGC 1512. found within the remaining Arm 2 and nearby regions are also a This schematic model for the oxygen abundance distribution is consequence of the interaction between NGC 1512 and NGC 1510. plotted with black dashed lines in the figure. As we discuss below, Interestingly, we also note a slight increase of the oxygen abun- we note that the original metallicities in the outskirts of NGC 1512 dance in the six most external regions of Arm 1. Although this

MNRAS 450, 3381–3409 (2015) 3398 A.´ R. Lopez-S´ anchez´ et al.

with r = 0.6401. The two areas we identified using the oxygen abundance residual map also appear evident in the N/O residual map created using this fit. Regions located at the N of Arm 1 have 12+log(O/H) ∼ 8.3 and log (N/O) ∼−1.4, while regions located in the external areas of the internal arm have 12+log(O/H) ∼ 8.4 and log (N/O) ∼−1.2. In both cases, the N/O ratio agrees with what is expected from their oxygen abundances. We note that, although the Pettini & Pagel (2004) calibrations use the [N II] lines to estimate the oxygen abundance, these relationships are purely based on observational values that consider only Te measurements for which O+/H+,O++/H+,andN+/H+ have been computed independently, and hence there is not any ad hoc N/O versus O/H dependence embedded in the calibrations. Nitrogen enrichment may be expected in objects which are experiencing a strong star formation event as consequence of the accretion of low-metallicity H I gas (Koppen¨ & Hensler 2005; Amor´ın, Perez-Montero´ & V´ılchez 2010;Perez-Montero´ et al. 2011;Molla´ Downloaded from & Terlevich 2012; Amor´ın et al. 2012). In this scenario, the oxygen Figure 14. Mgas/Mstars map for the UV-rich regions identified in Table 6. abundance released by the recently formed massive stars is lower Symbols are the same than those in Fig. 10. than that in the ISM before the starburst, while the nitrogen abundance remains the same until it is released by the intermediate and

low-mass stars hundreds of millions of years after the massive stars http://mnras.oxfordjournals.org/ increase is only ∼0.1 dex and the metallicities of these regions have die. The net effect is an increase of the N/O ratio, that we do not the highest errors in the system, we note that these regions also have observe in these areas. a relatively low M /M ratio when compared with nearby knots gas stars Hence, these regions with both lower O/H and N/O are actually (see Fig. 14). Has the gas been consumed recently here, used by less chemically evolved than their surroundings. In the case of several star formation events, and has the material therefore been the outer regions in the internal arm, they were probably originally enriched more than in other regions of the Arm 1? Interestingly, located at larger distances than they are today, and therefore they still the residual H I velocity field presented by KLS09 (see their Fig. 7, keep the relatively low metallicity of these outer regions. Something right-hand panel) shows deviations up to ∼30 km s−1 around these similar may have happened to the low-metallicity regions at the N of complexes, which typically have the older ages of the system ac- Arm 1; they are now closer to NGC 1512 than they were in the past. cordingtotheirFN colours. This suggests that the interaction has at Oxford Brookes University on June 4, 2015 Following the residual H I velocity field of NGC 1512 (KLS09), this indeed induced the star formation activity within these regions, and area shows little deviation from the rotating H I curve, and hence consequently enhanced their oxygen abundances and depleted their we may expect less disturbance from the gravitational interaction gas reservoir. between the two main galaxies. If this suggestion was true, it would Neglecting all regions in Arm 1, it is difficult to define the ra- explain that other regions along the Arm 1, which show higher dial metallicity distribution that was originally present in the sys- deviations from the rotating H I curve, have experienced more star tem. NGC 1512 is a perfect example that illustrates the fact that formation events than those areas, and therefore enhanced their galaxy interactions may flatten the metallicity gradients in galaxies metallicities from 12+log(O/H) ∼ 8.20 to 12+log(O/H) ∼ 8.40 (Kewley et al. 2010;Rupkeetal.2010;Werketal.2011). NGC 1512 and from log(N/O) ∼−1.5 to log(N/O) ∼−1.3, as it is observed nicely shows, at the same time, both the gradient flattening because now. of its interaction with NGC 1510 in one arm (Arm 2) and the assumed, original, poorly disturbed radial distribution (gradient+flat in the outskirts) on the other arm (Arm 1). 5.5 Dust in the NGC 1512/1510 system The intrinsic extinction map of the system, once the contribution of the extinction of the Milky Way has been subtracted, was computed 5.4 The N/O ratio using the values of the reddening coefficient derived for each region The N/O ratio provides important clues to the study of the star and applying formation history of galaxies (e.g. Mollaetal.´ 2006). While the E(B − V )instrinsic = 0.692c(Hβ) − E(B − V )MW, (21) nitrogen stellar yield mainly proceeds from low- and intermediate- − = mass stars, the oxygen yield comes from massive stars. The right- where E(B V)MW 0.011 mag (Schlegel et al. 1998). The E(B − hand panel of Fig. 12 shows the N/O ratio map of the system. The V)instrinsic map is shown in Fig. 15. As this figure shows, the external areas show log (N/O) values between −1.30 and −1.50, intrinsic extinction seems not to be very important in the system. < − < while the internal areas have −0.9 ≤ log (N/O) ≤−0.6. These The highest values, 0.25 E(B V)instrinsic 0.30, are found in values more or less correlate with the adopted oxygen abundances NGC 1510 and in the internal regions of NGC 1512. In the external < − of their knots (e.g. Izotov et al. 2006;Lopez-S´ anchez´ & Esteban star-forming regions, the colour excess is typically 0.05 E(B < 2010b; Moustakas et al. 2010). The right-hand panel of Fig. 11 V)instrinsic 0.10. These data suggest that the dust contribution in shows the N/O gradient in NGC 1512. Again, it is not clear that a the system is rather small, particularly in the external areas, which unique N/O gradient exists in the system. A fit to the data of the have extinctions similar to those found in dwarf, low-metallicity same regions used to get equation (18) yields galaxies (Lopez-S´ anchez´ 2010). Bresolin et al. (2012) found that some of the brightest regions along Arm 1 show c(Hβ)valuesup log(N/O) =−0.62 − 0.32R25, (20) to 0.5 dex, the mean value being c(Hβ) = 0.34 dex. These values

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3399

v = −1 using the H I velocity map, H I 865.7 km s . Note that despite the uncertainty in the absolute optical radial velocity, a relative velocity (i.e. the observed radial velocity of a region with respect to the observed radial velocity of the centre of NGC 1512) allows us to achieve resolutions of ∼0.1 Å ∼ 5kms−1. The right-hand panel of Fig. 16 shows the mean H I velocity field as determined by KLS09 using the same velocity range (both optical heliocentric). The comparison of both kinematics maps allows us to extract some interesting results. First, we note that the kinematics of the ionized gas within the ring show slightly blueshifted velocities with respect to the H I velocity at the east, while its western knots tend to show slightly redshifted velocities with. This may be a feature of the internal kinematics of the ring. Secondly, the mean dispersion in velocity for the regions located at the NW (corresponding to the internal knots of Arm 1 and the external debris of Arm 2), 9.8 km s−1, is slightly larger than that for regions located along Arm 1, which is 7.0 km s−1. This result Downloaded from agrees with the hypothesis that the NW zone of NGC 1512 has Figure 15. Intrinsic E(B − V) map in the NGC 1512/1510 system. Symbols been disrupted because of the interaction with NGC 1510. It is also are the same than those in Fig. 10. interesting to note some kinematic disturbances at the outer regions of Arm 1. We also note that the H I velocity measured in the TDG are translated into intrinsic colour excesses of 0.29 and 0.19 mag, α 1 matches well with its H velocity. http://mnras.oxfordjournals.org/ respectively, which are 0.1–0.2 mag higher than the values found in Finally, we measure a velocity difference of vr,Hα − our analysis. Although this difference may be a consequence of just v = −1 r,H I 136 km s for knot 3_20, which is located at the ex- not analysing exactly the same regions, we note we have had special ternal regions of the inner arm. We did not detect the [N II] λ6583 care deriving the reddening coefficient of our observed knots. emission line in this object, but a very broad Hα line is clearly observed. The velocity dispersion of the Hα line in knot 3_20 is σ = 3.9 Å, which means that it has a velocity dispersion of 178 5.6 Kinematics using the Hα emission km s−1. Typical dispersions in the Hα line profiles of the regions To determine the radial velocity of each Hα detection towards UV- observed in NGC 1512 are σ = 1.2–1.5 Å (55–70 km s−1). A careful bright stellar clusters in the galaxy pair NGC 1512/1510, we fit a inspection of its optical spectrum reveals that both Hβ and [O III] Gaussian to each Hα emission profile. λ5007 are barely detected at the same radial velocity than Hα.These at Oxford Brookes University on June 4, 2015 The resulting Hα velocity field is shown in left-hand panel of emission lines also show very high-velocity dispersions, 7.4 Å = Fig. 16. 456 km s−1and 7.8 Å = 467 km s−1, respectively. Examining the As we see, the kinematics resemble that of a rotating disc, as it region where [N II] λ6583 should be, we constrain that the highest was found in the H I analysis (KLS09). These authors also noticed oxygen abundance of this knot is 12+log(O/H) 8.1 using the that the outer disc showed a more disturbed rotating pattern than N2 parameter and the Pettini & Pagel (2004) calibration. That is a the inner disc. The radial velocity of the centre of NGC 1512 using difference of at least 0.4 dex with respect to the typical metallicities −1 the Hα line is vr = 879 ± 35 km s , which matches that measured found in this area. We suggest knot 3_20 to be an independent,

Figure 16. Velocity ﬁelds of the galaxy pair NGC 1512/1510 as determined from the Hα emission line (left; this paper) and the H I emission line (right; KLS09) for the targeted UV-bright stellar clusters. Symbols are the same than those in Fig. 10. Note the very different kinematics in knot 3_20 (marked with h m s ◦ an open black circle at coordinates RA = 04 04 03.61 and Dec. =−43 24 46. 26) when comparing the H I and Hα maps.

MNRAS 450, 3381–3409 (2015) 3400 A.´ R. Lopez-S´ anchez´ et al. dwarf galaxy (or, given to its high-velocity dispersion, its remnant), which was accreted into the NGC 1512 system in the past. The H I emission comes from the NGC 1512 disc while the Hα emission is from the dwarf.

5.7 The star formation in NGC 1512/NGC 1510 revisited The combination of the metallicities and other properties derived using our optical spectra with the H I andUVdataallowusto explore with more details the star formation properties and the interplay between stars and gas in the outer regions of the system. As mention in Section 4, we remind the reader that the Hα,UV,and H I fluxes were obtained using the same aperture for each region after matching resolution and PSF. We also note that the definition of the region was obtained using the UV images, and hence this may introduce a bias towards regions of high UV/Hα rations. However as the majority of the UV-bright regions show Hα emission, this Downloaded from effect is probably small. Figure 17. Comparison between the Hα-based SFR derived using our spectra (y-axis) with the FUV-based SFR derived using the GALEX image 5.7.1 The Hα-based SFR (x-axis). Symbols are the same than those in Fig. 10. α We have used the extinction-corrected H luminosity of each region SFR and FUV-SFR therefore somehow indicate the changes in the http://mnras.oxfordjournals.org/ to estimate the SFR. However, we have to consider that we are recent star formation history of the UV-rich clusters? Interestingly, probably not observing all the emission of the ionized gas, as the knots 4_17 and 4_32 show an order of magnitude higher Hα-SFR fibre size used to get the spectroscopic data is 2 arcsec. Meurer when compared to the FUV-SFR. This seems to indicate that, at et al. (2006) presented flux-calibrated Hα images of a large sample least in these cases, the knots are experiencing right now a very of H I-rich galaxies, including the galaxy pair NGC 1512/1510. As strong star formation event. discussed in that paper, the Hα flux has been corrected for extinction, [N II] contribution, and underlying Hα absorption. Hence, we have 5.7.2 Age(s) of the most-recent star-forming event(s) compared the Hα fluxes obtained using the image provided by Meurer et al. (2006) with those derived using our spectra for the 64 An estimation of the age of the most-recent star-forming event can regions for which we do have measurements from both data sets. be obtained from the Hα equivalent width, as it decreases with time at Oxford Brookes University on June 4, 2015 Although the dispersion is rather large, the Hα flux estimated using (e.g. Leitherer & Heckman 1995). The left-hand panel of Fig. 18 our spectroscopy is, on average, only ∼3.4, 2.2, and 1.4 times lower shows the Hα equivalent width map of the system. In contrast to than that using the Hα images for a Hα flux of 10−15,10−14,and the FN map (see fig. 13 in KLS09), we do not observe any kind of 10−13 erg s−1 cm−2, respectively. As the angular areas of the knots age gradient or pattern in W(Hα). In fact, the Hα equivalent width are typically 40–60 times more extended than the aperture size used appears randomly distributed throughout the system, with values for getting the spectroscopic data, this informs that the ionized gas ranging between log[−W(Hα)] ∼ 2.9 and 0.7. is very localized within each UV-rich cluster. Following a STARBURST99 (Leitherer et al. 1999;Vazquez´ & We then scaled (for objects included in both data sets) or cor- Leitherer 2005) model experiencing a starburst at Z = 0.008, the rected (using a fit to the data for those objects only observed using corresponding ages for log[−W(Hα)] ∼ 3.5, 2.5, 1.3, and 0.0 are spectroscopy) our Hα flux values for the missing gas emission to 1, 5, 10, and 25 Myr, respectively. Hence, assuming an instanta- derive the total Hα luminosity in each knot. We finally applied neous burst and following their W(Hα), all UV-rich regions ob- Kennicutt (1998) relationship to derive their SFR. Note that the served here have experienced their most recent star formation in the Kennicutt (1998) calibration considers an instantaneous burst, if last ∼13 Myr. In some cases, such as knot 4_17, this has happened using the Calzetti et al. (2007) calibration – which assumes a STAR- just ∼3 Myr ago. All these ages are much younger than those de- BURST99 model (Leitherer et al. 1999;Vazquez´ & Leitherer 2005)of rived using the FN colours. This behaviour was already observed constant SFR for 100 Myr at solar metallicity – the SFR values will by Boquien et al. (2009) when analysing independent star-forming only be two-thirds of those adopted here following the Kennicutt regions within nearby spiral galaxies. KLS09 used the FN colours (1998) calibration. of the UV-rich regions to conclude that their star formation typically Fig. 17 compares the Hα-based SFR derived using our spectra started between 100 and 400 Myr ago, being more recent at the east with the FUV-basedSFR derived using the GALEX image. Although (brightest part of Arm 1) and west (around NGC 1510) of NGC the dispersion is large, we see that, on average, the Hα-based SFR 1512 (see their Table 6). are ∼4 times lower than the FUV-based SFR. This result agrees with How can we explain the age difference? The right-hand panel the finding reported by some authors (Boselli et al. 2009; Lee et al. of Fig. 18 compares the FN colour with the Hα equivalent width 2009; Hunter, Elmegreen & Ludka 2010) in dwarf galaxies with for the UV-rich stellar clusters of NGC 1512. As already discussed, SFR 0.03 M yr−1 (log SFR −1.5): the Hα-flux underpredict both quantities provide an estimation of the age of the star for- the SFR relative to the FUV luminosity. mation activity, although their time-scales are different. We should However, it is important to remember that while Hα emission expect that very young bursts – objects with high log[−W(Hα)] – traces the most massive, ionizing stars (time-scales of ∼10 Myr), the show bluer FN colours than older bursts. This panel includes the UV emission probes star formation over time-scales of ∼100 Myr, predictions given by two STARBURST99 models (Leitherer et al. 1999; the lifetime of the massive OB stars. Do the differences between Hα- Vazquez´ & Leitherer 2005), both with Z = 0.008 but one assuming

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3401 Downloaded from

Figure 18. (Left) Map of the Hα equivalent widths derived from our optical spectroscopy data. (Right) Comparison between the Hα equivalent width (y-axis) and the FN colour (x-axis) for the regions analysed in this work. The predictions given by two STARBURST99 models (Leitherer et al. 1999;Vazquez´ & Leitherer 2005), both with Z = 0.008 but one assuming an instantaneous burst (purple continuous line) and the other considering a constant star formation (red dotted line) are plotted. Two linear combinations of these models, one considering a 50 per cent continuous plus 50 per cent instantaneous (blue dashed line) and a 90 per cent continuous plus 10 per cent instantaneous (green dot–dashed line) are also plotted. The symbol colour indicates the gas-star ratio in each region. http://mnras.oxfordjournals.org/ Symbols are the same than those in Fig. 10. an instantaneous burst (purple continuous line) and the other considering a constant star formation (red dotted line). These models alone can only explain the position of a few clusters in this ﬁgure. As discussed in Section 4.1, we note that we assumed a Milky Way- like extinction law to correct the FN colour for internal extinction. A different dust attenuation curve would lead to a small reddening

in the FN colour with AFUV or AHα. However, as this effect is neg- at Oxford Brookes University on June 4, 2015 ligible, its impact should be negligible too and will not modify this analysis. For explaining the position of the rest of the clusters we can assume two scenarios.

(i) The UV-bright stellar clusters have experienced two instantaneous star formation events, one that typically happened 100– 400 Myr ago (following the FN colours) and a very recent starburst that started between 3 and 13 Myr ago (according to the Hα equivalent width). Following this scenario, we can use W(Hα) to constrain the age of the most recent star formation event, obtain the FN colour corresponding to a stellar population with that very young age, and subtract it to the observed FN colour to estimate the age of the older Figure 19. Comparison between the surface density of the cold gas consid- star formation event. This scenario will provide even older ages ering only the atomic component (y-axis) and the oxygen abundance (x-axis) (20–100 Myr depending on the knot) for the very first ignition of for the UV-bright knots in NGC 1512/1510. Colours indicate the FUV − the star formation activity than that determined by KLS09. NUV colour. Symbols are the same than those in Fig. 10. (ii) The star formation activity of the UV-rich stellar clusters can be explained as a combination of on-going star formation plus a consumed more in those complexes with lower log[−W(Hα)]. This starburst event. Indeed, the right-hand panel of Fig. 18 also plots agrees with the finding by KLS09 that objects with redder FN two extra lines which have been constructed assuming a linear colours systematically have a lower amount of gas. combination of continuous star formation/starburst, with weights of 0.5/0.5 (blue dashed line) and 0.9/0.1 (green dot–dashed line), 5.7.3 Relationships between stars, gas, and metallicity respectively. The position of the majority of the objects in this diagram is then easily explained according to this scenario. KLS09 showed that the UV-rich clusters in the NGC 1512/1510 system do follow the Schmidt–Kennicutt scaling laws of star formation The symbol colour in the right-hand panel of Fig. 18 indicates (Schmidt 1959, 1963; Kennicutt 1998; Kennicutt et al. 2007). For the gas-to-star ratio, Mgas/Mstars, in each region. Following this plot, all regions but those located in Zone 1, the SFR per unit area corre- it also appears that regions which possess a more intense recent lates well with the surface density of the cold gas considering only star formation contribution tend to have a higher gas-to-star ratio the atomic component (see fig. 20 in KLS09). Fig. 19 compares than knots where the continuous star formation dominates (or with the surface density of atomic gas, Mgas/area, with the oxygen abun- older ages for the first instantaneous starburst event following the dance for regions analysed here. Besides the known fact of missing FN colour). Either way, this fact suggests that the gas has been the molecular component in knots in Zone 1 (diamonds), and the

MNRAS 450, 3381–3409 (2015) 3402 A.´ R. Lopez-S´ anchez´ et al. Downloaded from

Figure 21. Comparison of the observed oxygen abundance with those pre- Figure 20. Comparison between the gas-to-star ratio, Mgas/Mstars (y-axis) dicted by simple closed-box chemical evolution models. The model follow- and the oxygen abundance (x-axis). Colours indicate the FUV − NUV ing the theoretical yield of oxygen expected for stars with rotation following colour. Symbols are the same than those in Fig. 10. Meynet & Maeder (2002) models, yO = 0.0074, is plotted with a green continuous line. Closed-box models with yO = 0.3, 0.03, and 0.003 are also http://mnras.oxfordjournals.org/ relatively large dispersion of Mgas/area along Arm 1, we do not see shown with a yellow dot–dashed line, a red dotted line, and a purple dashed any dependence of the surface density of atomic gas on the metal- line, respectively. Symbols are the same than those in Fig. 10, the star rep- −2 licity. Typical values of Mgas/area are between 3 and 9 M pc , resents the position of the BCDG NGC 1510. The symbol colour indicates numbers that agree with the SFR/area found in the external regions the FUV-based SFR. of spiral galaxies (Kennicutt et al. 2007; Bigiel et al. 2008, 2010a,b). them have effective yields higher than the theoretical y = 0.0074 We also check if the gas-to-star ratio, Mgas/Mstars, has any de- O pendence with the oxygen abundance. Fig. 20 compares both prop- value. This is even true for the regions located in the internal ring, erties. Remembering that the position of all knots within Zone 1 for which we should expect even higher effective yields than those (diamonds) are lower limits because of the lack of molecular gas computed here, as we lack data on the molecular gas. Typically, the measurements, again we do not see any correlation, as the decrease effective yields found in XUV complexes are between 0.03 and 0.3, at Oxford Brookes University on June 4, 2015 i.e. between 1 and 2 orders of magnitude higher than the theoretical of the Mgas/Mstars ratio with increasing oxygen abundance observed in galaxies (Lopez-S´ anchez´ 2010; Lara-Lopez´ et al. 2013a). We do value. observe, however, that regions with a higher amount of gas system- Therefore, these UV-bright knots have an oxygen abundance atically have bluer UV colours, as KLS09 already pointed out. much higher than expected by the closed-box model, i.e. the H I More interestingly, we calculate the effective oxygen yield for gas that has been used for the star formation already had a large each UV-rich complex, y = Z /ln (1/μ), where μ is the gas fraction of metals. The mean effective yield found within the XUV O O = fraction with respect to the total baryonic (stars and gas) mass, regions belonging to Arm 1 is yO 0.133, with a dispersion of μ = M /M ,andZ is the oxygen mass fraction.3 Follow- 0.119, while the mean effective yield obtained for the XUV clusters gas bar O = ing a closed-box model (Schmidt 1963; Searle & Sargent 1972; in the external debris of Arm 2 is yO 0.049, with a dispersion of Edmunds 1990), a galaxy, which initially consists of gas with no 0.025. Again, and besides the dispersion, these numbers agree with stars and no metals, experiences instantaneous recycling through- the hypothesis that the material within the knots belonging to the out its life, and the products of stellar nucleosynthesis are neither destroyed Arm 2 have experienced a larger chemical enrichment diluted by infalling pristine gas nor lost via outﬂow of enriched gas. than those located along Arm 1. Therefore, the metallicity at a given time is only determined by the fraction of baryons that remain in gaseous form. Fig. 21 compares 5.7.4 Where do the metals of the outer regions come from? the observed oxygen abundances with those predicted by simple closed-box models. The theoretical yield of oxygen expected for Bresolin et al. (2009a), Werk et al. (2010b, 2011), and Bresolin stars with rotation following the Meynet & Maeder (2002) models et al. (2012) already reported that the outer discs of spiral galaxies is yO = 0.0074 (green continuous line). In galaxies, it is usually are overabundant for their large gas fraction. Similarly high values found that the oxygen abundances are lower than those predicted are reported by Torres-Flores et al. (2014) when analysing TDGs by the closed-box model with yO = 0.0074, their effective yields located in the long tidal tail of NGC 92. Indeed, Lopez-S´ anchez´ being yO = 0.003–0.005 (Lee et al. 2003, 2007; Tremonti et al. (2010) found that a TDG candidate within the compact group HCG 2004; van Zee & Haynes 2006; Lopez-Sanchez 2010). This means 31 (member F), showed a high effective yield, yO = 0.03, this that galaxies tend to lie above the green continuous line plotted in object being very different from the other galaxies analysed in his Fig. 21, as indeed is happening in NGC 1510 (star), which has an study. The author explains this behaviour as a consequence of the effective yield of yO = 0.003. But the opposite behaviour is ob- accretion of a large fraction of pre-enriched H I gas striped from the served in the UV-bright stellar clusters of NGC 1512; almost all of main galaxies after a ﬂy-by encounter (Lopez-S´ anchez,´ Esteban & Rodr´ıguez 2004a). Somehow, member F of HCG 31 resembles the 3 The oxygen mass fraction is derived as ZO = f ×(O/H), f = 11.81 being properties of the XUV complexes found around NGC 1512, as we the conversion factor from number to mass fraction. will discuss below.

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3403

Adapting equation (1) in Bresolin et al. (2012), NGC 1512 came from low-luminosity dwarf galaxies which have been slowly accreted and merged into the system. This may have O y × t SFR = O × , (22) happened long time ago (several Gyr); the gas was enriched in H f M H I the past as a consequence of these minor mergers and later stayed and knowing from the FN colours when the star formation started essentially untouched for a long period of time. The remnants of in each knot, t, we can estimate how much the UV-bright re- stellar component of such dwarf galaxies are now so diffuse they O are almost impossible to distinguish, unless really deep optical data gions have been enriched in oxygen, ( H ), since then. For this, μ ∼ −2 we also consider the effective yield of oxygen, yO, computed in – lim(V) 28.8 mag arcsec or higher – are obtained. Relics the previous subsection. Averaging all regions of Arm 1, we ﬁnd of old interactions have been found as diffuse stellar tidal streams [12+log(O/H)] = 0.21, with a dispersion of 0.10 dex, while the around normal disc galaxies (Mart´ınez-Delgado et al. 2009, 2010)or same computation averaging all regions of the destroyed Arm 2 even around dwarf galaxies (Mart´ınez-Delgado et al. 2012). These (Zones 3 and 4) yields [12+log(O/H)] = 0.32, the dispersion structures are actually expected as a consequence of the hierarchical being 0.25 dex. This analysis suggests that about 400 Myr ago, framework for galaxy formation following a cold dark matter before the interaction with NGC 1510 started, the metallicity in (CDM) cosmogony, as minor mergers (which should have been the outskirts of NGC 1512 was 12+log(O/H) ∼ 8.1 in either arm. very common in early times) do not destroy pre-existing stellar Therefore, the neutral gas located at the outer regions of NGC discs (Bullock & Johnston 2005; Johnston et al. 2008; Cooper et al. 1512 had experienced metal enrichment even before the processes 2010). Downloaded from we observe today started. Assuming that the central regions also Alternatively, the gas in the outskirts of NGC 1512 was accreted have been enriched by 0.2–0.3 dex in the last 400 Myr (this may from the intergalactic medium (IGM), that has been previously actually be an upper limit to the real enrichment experienced in enriched with metals by outﬂows that happened on other galaxies. those central clusters), they should have had oxygen abundances of This mechanism was also proposed by Bresolin et al. (2012)andis + ∼

12 log(O/H) 8.5–8.6, which means a difference of 0.4–0.5 dex reproduced by cosmic chemical evolution models (e.g. Kobayashi, http://mnras.oxfordjournals.org/ with respect to the outer regions. Springel & White 2007). Both the enrichment by satellites or the Given the distance of many of the XUV complexes, particularly accretion of outflow-enriched IGM gas require that a significant TDG 1, to the centre of NGC 1512, it seems quite unlikely that the fraction of the metals are cooled down and mixed with cold H I metal enrichment of the outer regions of this galaxy came from its before being available for further star formation, as the one observed inner regions. In order to explain the flattening of the metallicity today in the XUV disc of NGC 1512. gradients, Werk et al. (2011) discussed mechanisms to transport metals from the galaxy centres to their outskirts. Metal mixing in 5.8 About the TDG nature of the XUV complexes cold neutral gas probably happens, but it seems to be a very slow −1 process; assuming radial H I outflows of ∼10 km s (as those Interactions between galaxies may induce the development of tidal at Oxford Brookes University on June 4, 2015 observed in the extended H I disc of NGC 2915 by Elson et al. tails of material expelled from the parent galaxies into the IGM 2010), some few Gyr are needed to transport metals up to 25 kpc (e.g. Hibbard & van Gorkom 1996; Hibbard et al. 2001b). Objects from the galaxy centre. Werk et al. (2011) considered that metal with masses typical of dwarf galaxies are usually formed from the transport may be occurring predominantly in a hot gas component, debris of these tidal tails (Duc & Mirabel 1994, 1998; Hibbard et al. although both the drivers and the amount of mixing of metals in the 2001a; Knierman et al. 2003;Lopez-S´ anchez,´ Esteban & Rodr´ıguez hot-phase gas is unclear. 2004a,b; Hibbard et al. 2005; Neff et al. 2005). When one of these In any case, the amount of metals needed to enrich an almost objects appears to be self-gravitating, it is defined to be a TDG primordial gas to 12+log(O/H) ∼ 8.1 is large. Considering that (Braine et al. 2000; Weilbacher & Duc 2001; Bournaud et al. 2004; ∼75 per cent of the neutral gas is found in the outer part of the Bournaud & Duc 2006). Contrary to dwarf galaxies, these entities system (KLS09), this gas should have had ∼6.5 × 106 M in do not need a dark matter (DM) halo to rotate, as indeed TDGs form of oxygen to account for that oxygen abundance. Adding this are expected to have little to no DM. However, they are likely mass to that existing in the centre (which we computed assuming to contain a significant amount of old stars and they should have 12+log(O/H) ∼ 8.5, although this may have been larger), we de- chemical enrichment histories related to those of the parent galaxy rive an original oxygen abundance of 12+log(O/H) ∼ 8.85 (i.e. an (Duc et al. 2000; Weilbacher, Duc & Fritze-v. Alvensleben 2003), as increasing of ∼0.35 dex) in the centre of NGC 1512. Taking into tidal tails do not show evidence of abundance gradients. Hence, the account the mass–metallicity relation (Tremonti et al. 2004; Lara- definitive confirmation that a knot in the outer regions of a galaxy Lopez´ et al. 2013b), a galaxy with 12+log(O/H) ∼ 9.24 should is a TDG comes from both its kinematics (a self-gravitating object) and chemistry (metallicities similar to those found in the parent have log(Mstars) ∼ 11. This value is at least one order of magnitude higher than the stellar mass of NGC 1512 estimated using the op- galaxy). tical colours and only 1/3 of the total dynamical mass (see tables Can we define as TDG candidates the XUV clusters found in the 1and3inKLS09). Therefore, the metals (or at least an important outer regions of the NGC 1512/1510 system? Our data do not allow fraction) found in the outskirts do not seem to come from the inner us to investigate the intrinsic kinematics of the UV-bright regions in regions of NGC 1512. the NGC 1512/1510 system, but our metallicities can provide some Ruling out the hypothesis that the gas itself has experienced light to this issue. Recently, Sweet et al. (2014) identified strong 5 < + star formation activity for longer periods of time, we suggest that TDG candidates with oxygen abundances of 8.4 12 log(O/H) < the metals within the diffuse H I gas found in the outer regions of 8.7. The sample analysed by Weilbacher et al. (2003) also provided oxygen abundances typically higher than 8.3 dex in TDG 4 Note that both Tremonti et al. (2004) and Lara-Lopez´ et al. (2013b)usethe 5 absolute abundance scale given by photoionization models, which provide Note these values have been scaled here to the absolute Te-scale oxygen oxygen abundances that are at least 0.3–0.4 dex higher than those given by abundance, as Sweet et al. (2014) used the recent photoionization models the Temethod used here. by Dopita et al. (2013) to compute metallicities.

MNRAS 450, 3381–3409 (2015) 3404 A.´ R. Lopez-S´ anchez´ et al. candidates. Similarly, two objects found within the large M 81– higher than the observed one. Taking into account all available infor- M 82–NGC 3077 H I tidal stream also are TDG candidates, having mation about TDG 1 (recent star formation activity, relatively low 8.4 < 12+log(O/H) < 8.7 (Croxall et al. 2009). The majority of metallicity, gas-to-star ratio, and H I kinematics) and our previous the regions we have associated with Arm 2 in NGC 1512 are within discussion about the nature of the XUV complexes we consider that the typical metallicity range for TDGs. However, as we discussed this object is actually not a TDG but a star-forming clump within a before, our data suggest that the original metallicity of these re- denser H I cloud in the outskirts of the NGC 1512/1510 system, that gions was much lower than it is now, 12+log(O/H) ∼ 8.1, as we is, an extreme case of TSFC. Indeed, the H I cloud associated with see along Arm 1. Hence, this material has clearly experienced a this object has very likely a tidal origin, as seen in the impressive different chemical evolution than that observed in the central re- H I map of the system (fig. 11 in KLS09). gions of NGC 1512, with 12+log(O/H) = 8.7–8.8. Furthermore, Finally, what is the nature of regions 4_17, 4_32, and 4_38 at the according their star formation histories and their comparison with NW of Arm 2? These three clusters, particularly 4_17, show the low- the close-box model, all these regions seem to have been recently est metallicities of the system, 12+log(O/H) ∼ 8.10, despite their born from the large reservoir of neutral gas previously existing in surroundings having oxygen abundances of 12+log(O/H) ∼ 8.45. the external areas of NGC 1512, which was already enriched in We propose two scenarios: (i) because of unknown reasons, these metals that were not coming from its inner regions. Therefore, the clusters have experienced a quenching in their star formation, and material from which the external UV-rich complexes were formed therefore the recycling of the gas has not been as high as in other has been not expelled from the centre of the galaxy, and hence the nearby regions, or (ii) these are actually independent dwarf galax- Downloaded from entities created within this gas do not fulfil the characteristics to be ies which now are pulled into the internal areas of the system as a defined as TDG candidates. consequence of the strong gravitational forces induced by the inter- Note, however, that the H I structures we now see in the system action. Some spectroscopic properties (see Fig. 2 and Table 4), like and from which the XUV complexes are formed have been tidally the strength of the emission lines and their low reddening coefficient I α induced by the interaction between NGC 1512 and NGC 1510 and H Balmer absorptions, as well as the fact that the H -based http://mnras.oxfordjournals.org/ (KLS09). Therefore, besides their chemical history, these stellar SFR is almost an order of magnitude higher than the FUV-based clusters seem to have a tidal origin. We propose to define these SFR, agree with the hypothesis that knot 4_17 actually is a dwarf UV-bright, young, relatively low-metallicity, gas-rich, entities to be galaxy which has not been created from the debris of Arm 2. How- ‘tidally induced star-forming clusters’ (TSFCs) in the galaxy out- ever, its Hα kinematics (see Fig. 16) is undistinguishable from that skirts. The main quantitative observational properties of the TSFCs observed in nearby knots. The Hα kinematics of regions 4_32 and are that they are located in the outskirts of the galaxy (R/R25 > 1), 4_38 also agree with that found in their surroundings. Therefore, still possess SFR as traced by their FUV emission (SFRFUV > 0), we cannot confirm the hypothesis that these XUV complexes are they still possess a lot of gas (Mgas/Mstars > 1), and that their metal- independent dwarf galaxies. In any case, we remember that knot licities are at typically 0.2–0.3 dex lower than those found in the 3_20, which has both different kinematics and metallicities than its galaxy centre (as difference of TDG candidates, that should have nearby XUV regions, may actually be an independent dwarf galaxy. at Oxford Brookes University on June 4, 2015 the same metallicity that the inner regions of the galaxy). Hence, New, deeper optical data of all these clusters are needed to unveil many of the UV-rich complexes found along the XUV discs of spi- their real nature. ral galaxies are very likely to be made up by these entities. Knots E and F within the disturbed HCG 31 group (Lopez-S´ anchez´ et al. 2004a) may be another example of TSFCs, as they were formed 6 CONCLUSIONS from material in the long H I tail that the compact group possesses We have used the 2dF instrument in combination with the AAOmega (Williams, McMahon & van Gorkom 1991; Verdes-Montenegro spectrograph at the AAT to get deep, intermediate-resolution op- et al. 2005), although in these knots the metallicities are slightly tical spectroscopy of 136 genuine UV-bright regions located in + ∼ lower, 12 log(O/H) 8.05, than that observed in the UV-bright both the spiral and the XUV discs of the NGC 1512/1510 sys- regions of the NGC 1512 system. tem. In conjunction with the H I –provided by the LVHIS project What is the situation of the so-called TDG 1 (KLS09) located at a (Koribalski 2008; Koribalski et al., in preparation) – and UV data galactocentric distance of 78 kpc? Following the H I map presented we analyse the relationships between gas and stars, the chemical by KLS09, TDG 1 belongs to Arm 2, and hence it should be associ- enrichment of the ISM, and the properties of the stellar popula- ated with our Zone 4. The adopted metallicity of this star-forming tions and star formation processes within these objects. Our main + = knot is 12 log(O/H) 8.24, which quantitatively is the same oxy- conclusions are as follows. gen abundance as that observed throughout the long Arm 1. It is very interesting to note that, besides their low masses and distance (i) We confirm the detection of ionized gas in the huge majority from the centre of NGC 1512, the TDG candidate and the majority of the UV-rich regions in the system, as well as identified 17 back- of the regions studied in Arm 1 have these intermediate metallic- ground galaxies and 1 foreground star. We confirm that in all cases ities. This fact indicates that even in these far regions the gas has except the centre of NGC 1512 photoionization by massive stars is been processed several times in order to enrich the very metal-poor the main excitation mechanism of the gas. In the XUV complexes, and unprocessed material that presumably existed here in the past. only few (1–5) O7V stars are responsible for this. Interestingly, Fig. 21 shows that the data of TDG 1 agrees with a (ii) Using a comprehensive analysis of the emission lines of [O II] closed-box model with an effective yield of yO = 0.004, the ex- λ3727, Hγ ,Hβ,[OIII] λ5007, Hα,and[NII] λ6583, in conjunction pected for a dwarf galaxy. However, note that the H I envelope that with a careful study of the most-common SEL techniques, we have surrounds this object is really large (see fig. 8 in KLS09), but only computed the oxygen abundance and the N/O ratio for the majority the neutral gas within the UV-bright knot has been considered here of the detected UV-bright regions. We provide chemical abundance (once the images were matched in resolution and PSF, as described maps reaching a projected distance of 78 kpc (6.6R25)fromthe in Section 4), which is at least an order of magnitude lower than the centre of NGC 1512. We found significant differences between real H I reservoir. Hence, the effective yield of this object should be regions along the Arm 1 at the east, which have oxygen abundances

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3405

8.25 12+log(O/H) 8.45, and knots located in the external possess much more oxygen than expected from their gas mass debris of Arm 2, which typically have 8.40 12+log(O/H) 8.60. fraction. (iii) Thanks to the detection of the auroral [O III] λ4363 line, we (ix) Although the XUV complexes are very likely induced by have computed the chemical abundances of the ionized gas using the the interaction between NGC 1512 and NGC 1510, they cannot direct Te method in four regions. The results agree within the errors be strictly defined as TDGs, as they have not been formed from to those obtained using SEL methods, however precaution must be material stripped from the main galaxies, but from the diffuse gas taken with the ionization degree of the gas and/or the area integrated already existing at large galactocentric radii. We propose to define to get the spectrum when using SEL techniques. In the case of NGC these UV-bright, young, relatively low-metallicity, gas-rich entities 1510, we derive 12+log(O/H) = 8.24 ± 0.06 and a N/O ratio of to be TSFCs in the galaxy outskirts. log(N/O) =−1.34 ± 0.05, results that agree with the typical values (x) Our data suggest that the gas in the outer regions of NGC 1512 found in other BCDGs. We detect the nebular He II λ4686 emission already had a metallicity of 12+log(O/H) ∼ 8.1 about 400 Myr ago, line in the centre of NGC 1510. This line is attributed to the presence before the interaction with NGC 1510 started. The metals within of massive stars in the WR stage. Contrary to what was previously the diffuse H I gas are very likely not coming from the inner regions suggested, no local N enrichment is observed in this galaxy. of NGC 1512, but probably from material accreted during the life (iv) We have analysed the metallicity distribution of the sys- of the galaxy either by minor mergers or by outflow-enriched IGM tem. Considering just a radial gradient involving all complexes it gas. We found a probable remnant of an independent dwarf galaxy is difficult to see a radial decrease of metallicity from the centre (knot 3_20) and three knots (4_17, 4_32, and 4_38) which may also Downloaded from of NGC 1512 to the outer regions. However, assuming both a ra- be independent systems, all of them having 12+log(O/H) 8.1. dial and an azimuthal gradient following the spiral arms we are This hypothesis agrees with the finding around other nearby large able to clearly distinguish that Arm 1 has an almost flat oxygen galaxies such as M 83 or NGC 300, as well as constraints chemical abundance – 12+log(O/H) ∼ 8.35 – and flat N/O ratio – log(N/O) and dynamical models of galaxy evolution. Indeed, the extended, ∼−

1.4 – while regions located in the disrupted Arm 2 (inner arm metal enriched diffuse gas should be common in spiral galaxies http://mnras.oxfordjournals.org/ and external debris), show a large dispersion in oxygen abundances according to the CDM scenario. – 8.1 12+log(O/H) 8.6 – and N/O ratios. Arm 2 has experi- In summary, the knowledge of the metallicity distribution in the enced an enhancement in star formation because of the interaction outskirts of galaxies give key clues about their evolution and star with NGC 1510, flattening the radial metallicity at large radii, while formation histories. This seems to be true even in an S0 galaxy Arm 1 still shows the original and poorly disturbed radial distribu- like NGC 404 (Bresolin 2013). The analysis of the ionized gas in tion (gradient+flat in the outskirts). the outer regions of galaxies, which provides the information about (v) We have quantified the extinction within the UV-bright re- the chemical evolution, very nicely complements the studies of the gions using the Balmer decrement and the theoretical Hα/Hβ and neutral gas using the 21 cm H I emission, which constrains the Hγ /Hβ ratios expected for the oxygen abundances estimated in

dynamical evolution of the galaxies. When used together, these two at Oxford Brookes University on June 4, 2015 each knot. Underlying stellar absorptions with values W ∼0.85 Å abs analyses provide a very powerful tool to disentangle the nature and are needed to explain the Balmer decrements. The reddening coefﬁ- evolution of the galaxies we now observe in the Local Universe. cient correlates with the oxygen abundance, indicating that regions with higher metallicities have higher amount of dust. We provide an intrinsic extinction map of the system. ACKNOWLEDGEMENTS (vi) Using the Hα emission-line proﬁle, we are able to trace the kinematics of the system out to 78 kpc, which generally matches We thank Paul Dobbie for his help as AAO Support Astronomer well with that provided by the H I kinematics (velocity differences while conducting our observations at the AAT. We also thank the of less than 15 km s−1). We observe a hint of additional rotation anonymous referee for her/his very valuable comments and sugges- pattern in the Ring (Zone 1). Some small velocity discrepancies tions which have increased the quality of this paper. This research (differences between 20 and 40 km s−1) are found in some clusters made use of images provided by the Survey for Ionization in Neu- of the external debris and in particular areas of Arm 1. We locate a tral Gas Galaxies (Meurer et al. 2006) which is partially supported region, 3_20, for which its Hα radial velocity differs by 136 km s−1 by the National Aeronautics and Space Administration (NASA). with respect its H I radial velocity. The upper limit to the metallicity GALEX is a NASA Small Explorer, and we gratefully acknowledge of this knot is 12+log(O/H) = 8.1, which is 0.4 dex lower than that NASA’s support for construction, operation, and science analysis found in nearby complexes. We suggest that this region actually is for the GALEX mission, developed in cooperation with the Centre an independent dwarf galaxy or its remnant. National d’Etudes´ Spatiales of France and the Korean Ministry of (vii) Comparing the Hα-SFR and the FUV-SFR, we conclude Science and Technology. CEL thanks the funding provided by Span- that the ionized gas seems to be very localized at the centre of each ish Ministerio de Econom´ı a y Competitividad (MINECO) under UV-rich cluster. FUV-SFR are systematically ∼4 times higher than the project AYA2011-22614. This research has made extensive use Hα-SFR. We cannot explain the observed W(Hα)andFN colours of the NASA/IPAC Extragalactic Database (NED) which is oper- as just a single, instantaneous star formation event. A combination ated by the Jet Propulsion Laboratory, Caltech, under contract with of (i) two instantaneous events separated by 100–400 Myr or (ii) the National Aeronautics and Space Administration. continuous star formation plus a recent (less than 13 Myr) starburst, This research has made extensive use of the SAO/NASA Astro- are needed. The gas-to-star ratio decreases with both the FN colour physics Data System Bibliographic Services (ADS). and the W(Hα). (viii) We do not observe any correlation between the surface gas density or the gas-to-mass ratio with the oxygen abundance. REFERENCES However, a comparison with the closed-box model indicates that Amor´ın R. O., Perez-Montero´ E., V´ılchez J. M., 2010, ApJ, 715, L128 the XUV complexes have very high effective oxygen yields, as Amor´ın R., Perez-Montero´ E., V´ılchez J. M., Papaderos P., 2012, ApJ, 749, opposited to that found in dwarf galaxies, indicating that they 185

MNRAS 450, 3381–3409 (2015) 3406 A.´ R. Lopez-S´ anchez´ et al.

Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP, 93, 5 Vol. 240, Gas and Galaxy Evolution. Astron. Soc. Pac., San Francisco, Bianchi L. et al., 2005, ApJ, 619, L71 p. 657 Bigiel F., Leroy A., Walter F., Brinks E., de Blok W. J. G., Madore B., Hibbard J. E. et al., 2005, ApJ, 619, L87 Thornley M. D., 2008, AJ, 136, 2846 Hopkins A. M. et al., 2013, MNRAS, 430, 2047 Bigiel F., Leroy A., Walter F., Blitz L., Brinks E., de Blok W. J. G., Madore Hunter D. A., Elmegreen B. G., Ludka B. C., 2010, AJ, 139, 447 B., 2010a, AJ, 140, 1194 Husemann B. et al., 2013, A&A, 549, A87 Bigiel F., Leroy A., Seibert M., Walter F., Blitz L., Thilker D., Madore B., Izotov Y. I., Thuan T. X., 2007, ApJ, 665, 1115 2010b, ApJ, 720, L31 Izotov Y. I., Thuan T. X., Guseva N. G., 2005, ApJ, 632, 210 Boissier S., Prantzos N., 1999, MNRAS, 307, 857 Izotov Y. I., Stasinska´ G., Meynet G., Guseva N. G., Thuan T. X., 2006, Boissier S., Prantzos N., 2000, MNRAS, 312, 398 A&A, 448, 955 Boquien M. et al., 2009, ApJ, 706, 553 Johnston K. V., Bullock J. S., Sharma S., Font A., Robertson B. E., Leitner Boselli A., Boissier S., Cortese L., Buat V., Hughes T. M., Gavazzi G., 2009, S. N., 2008, ApJ, 689, 936 ApJ, 706, 1527 Kauffmann G. et al., 2003, MNRAS, 346, 1055 Bournaud F., Duc P.-A., 2006, A&A, 456, 481 Kennicutt R. C., Jr, 1998, ARA&A, 36, 189 Bournaud F., Duc P.-A., Amram P., Combes F., Gach J.-L., 2004, A&A, Kennicutt R. C., Jr, et al., 2007, ApJ, 671, 333 425, 813 Kewley L. J., Dopita M. A., 2002, ApJS, 142, 35 Bouwens R. J., Cayon L., Silk J., 1997, ApJ, 489, L21 Kewley L. J., Ellison S. L., 2008, ApJ, 681, 1183 Braine J., Lisenfeld U., Due P.-A., Leon S., 2000, Nature, 403, 867 Kewley L. J., Dopita M. A., Sutherland R. S., Heisler C. A., Trevena J., Downloaded from Bresolin F., 2013, ApJ, 772, L23 2001, ApJ, 556, 121 Bresolin F., Ryan-Weber E., Kennicutt R. C., Goddard Q., 2009a, ApJ, 695, Kewley L. J., Rupke D., Zahid H. J., Geller M. J., Barton E. J., 2010, ApJ, 580 721, L48 Bresolin F., Gieren W., Kudritzki R.-P., Pietrzynski´ G., Urbaneja M. A., Knierman K. A., Gallagher S. C., Charlton J. C., Hunsberger S. D., Whitmore Carraro G., 2009b, ApJ, 700, 309 B., Kundu A., Hibbard J. E., Zaritsky D., 2003, AJ, 126, 1227 Bresolin F., Kennicutt R. C., Ryan-Weber E., 2012, ApJ, 750, 122 Kobayashi C., Springel V., White S. D. M., 2007, MNRAS, 376, 1465 http://mnras.oxfordjournals.org/ Bryant J. J. et al., 2015, MNRAS, 447, 2857 Kobulnicky H. A., Kewley L. J., 2004, ApJ, 617, 240 Bullock J. S., Johnston K. V., 2005, ApJ, 635, 931 Koppen¨ J., Hensler G., 2005, A&A, 434, 531 Calzetti D. et al., 2007, ApJ, 666, 870 Koribalski B. S., 2008, in Koribalski B. S., Jerjen H., eds, Galaxies in the Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345, 245 Local Volume. Springer-Verlag, Netherlands, p. 41 Cooper A. P. et al., 2010, MNRAS, 406, 744 Koribalski B. S., Lopez-S´ anchez´ A.´ R., 2009, MNRAS, 400, 1749 Croom S., Saunders W., Heald R., 2004, AAO Newsl., 106, 12 Lara-Lopez´ M. A. et al., 2010, A&A, 521, L53 Croom S. M. et al., 2012, MNRAS, 421, 872 Lara-Lopez´ M. A. et al., 2013a, MNRAS, 433, L35 Croxall K. V., van Zee L., Lee H., Skillman E. D., Lee J. C., Cotˆ e´ S., Lara-Lopez´ M. A. et al., 2013b, MNRAS, 434, 451 Kennicutt R. C., Jr, Miller B. W., 2009, ApJ, 705, 723 Lara-Lopez´ M. A., Lopez-S´ anchez´ A.´ R., Hopkins A. M., 2013c, ApJ, 764, Dopita M. A., Kewley L. J., Heisler C. A., Sutherland R. S., 2000, ApJ, 542, 178 224 Lee H., McCall M. L., Richer M. G., 2003, AJ, 125, 2975 at Oxford Brookes University on June 4, 2015 Dopita M. A., Sutherland R. S., Nicholls D. C., Kewley L. J., Vogt F. P. A., Lee H., Zucker D. B., Grebel E. K., 2007, MNRAS, 376, 820 2013, ApJS, 208, 10 Lee J. C. et al., 2009, ApJ, 706, 599 Dors O. L., Jr, Krabbe A., Hagele¨ G. F., Perez-Montero´ E., 2011, MNRAS, Leitherer C., Heckman T. M., 1995, ApJS, 96, 9 415, 3616 Leitherer C. et al., 1999, ApJS, 123, 3 Driver S. P. et al., 2011, MNRAS, 413, 971 Lemonias J. J. et al., 2011, ApJ, 733, 74 Duc P.-A., Mirabel I. F., 1994, A&A, 289, 83 Lequeux J., Peimbert M., Rayo J. F., Serrano A., Torres-Peimbert S., 1979, Duc P.-A., Mirabel I. F., 1998, A&A, 333, 813 A&A, 80, 155 Duc P.-A., Brinks E., Springel V., Pichardo B., Weilbacher P., Mirabel I. F., Leroy A., Bolatto A. D., Simon J. D., Blitz L., 2005, ApJ, 625, 763 2000, AJ, 120, 1238 Lewis I. et al., 2002, MNRAS, 334, 673 Edmunds M. G., 1990, MNRAS, 246, 678 Lopez-S´ anchez´ A.´ R., 2010, A&A, 521, A63 Elson E. C., de Blok W. J. G., Kraan-Korteweg R. C., 2010, MNRAS, 404, Lopez-S´ anchez´ A.´ R., Esteban C., 2009, A&A, 508, 615 2061 Lopez-S´ anchez´ A.´ R., Esteban C., 2010a, A&A, 516, A104 Esteban C., Carigi L., Copetti M. V. F., Garc´ıa-Rojas J., Mesa-Delgado A., Lopez-S´ anchez´ A.´ R., Esteban C., 2010b, A&A, 517, A85 Castaneda˜ H. O., Pequignot´ D., 2013, MNRAS, 433, 382 Lopez-S´ anchez´ A.´ R., Esteban C., Rodr´ıguez M., 2004a, ApJS, 153, 243 For B.-Q., Koribalski B. S., Jarrett T. H., 2012, MNRAS, 425, 1934 Lopez-S´ anchez´ A.´ R., Esteban C., Rodr´ıguez M., 2004b, A&A, 428, 425 Freeman K. C., Karlsson B., Lynga G., Burrell J. F., van Woerden H., Goss Lopez-S´ anchez´ A.´ R., Mesa-Delgado A., Lopez-Mart´ ´ın L., Esteban C., W. M., Mebold U., 1977, A&A, 55, 445 2011, MNRAS, 411, 2076 Garc´ıa-Rojas J., Esteban C., Peimbert A., Peimbert M., Rodr´ıguez M., Ruiz Lopez-S´ anchez´ A.´ R., Dopita M. A., Kewley L. J., Zahid H. J., Nicholls M. T., 2005, MNRAS, 362, 301 D. C., Scharwachter¨ J., 2012, MNRAS, 426, 2630 Garnett D. R., 1992, AJ, 103, 1330 McGaugh S. S., 1991, ApJ, 380, 140 Gil de Paz A. et al., 2005, ApJ, 627, L29 Malin D. F., Carter D., 1983, ApJ, 274, 534 Gil de Paz A. et al., 2007a, ApJS, 173, 185 Martin D. C. et al., 2005, ApJ, 619, L1 Gil de Paz A. et al., 2007b, ApJ, 661, 115 Mart´ınez-Delgado D., Pohlen M., Gabany R. J., Majewski S. R., Penarrubia˜ Goddard Q. E., Bresolin F., Kennicutt R. C., Ryan-Weber E. V., Rosales- J., Palma C., 2009, ApJ, 692, 955 Ortega F. F., 2011, MNRAS, 412, 1246 Mart´ınez-Delgado D. et al., 2010, AJ, 140, 962 Gunawardhana M. L. P. et al., 2013, MNRAS, 433, 2764 Mart´ınez-Delgado D. et al., 2012, ApJ, 748, L24 Hamuy M., Suntzeff N. B., Heathcote S. R., Walker A. R., Gigoux P., Phillips Mazzarella J. M., Boroson T. A., 1993, ApJS, 85, 27 M. M., 1994, PASP, 106, 566 Meurer G. R. et al., 2006, ApJS, 165, 307 Hibbard J. E., van Gorkom J. H., 1996, AJ, 111, 655 Meynet G., Maeder A., 2002, A&A, 390, 561 Hibbard J. E., van der Hulst J. M., Barnes J. E., Rich R. M., 2001a, AJ, 122, Miszalski B., Shortridge K., Saunders W., Parker Q. A., Croom S. M., 2006, 2969 MNRAS, 371, 1537 Hibbard J. E., van Gorkom J. H., Rupen M. P., Schiminovich D., 2001b, Moffett A. J., Kannappan S. J., Baker A. J., Laine S., 2012, ApJ, 745, 34 in Hibbard J. E., Rupen M., van Gorkom J. H., eds, ASP Conf. Ser. Molla´ M., Terlevich R., 2012, MNRAS, 425, 1696

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3407

MollaM.,V´ ´ılchez J. M., Gavilan´ M., D´ıaz A. I., 2006, MNRAS, 372, 1069 Werk J. K., Putman M. E., Meurer G. R., Thilker D. A., Allen R. J., Bland- Moustakas J., Kennicutt R. C., Jr, Tremonti C. A., Dale D. A., Smith J.-D. Hawthorn J., Kravtsov A., Freeman K., 2010b, ApJ, 715, 656 T., Calzetti D., 2010, ApJS, 190, 233 Werk J. K., Putman M. E., Meurer G. R., Santiago-Figueroa N., 2011, ApJ, Neff S. G. et al., 2005, ApJ, 619, L91 735, 71 Oke J. B., 1990, AJ, 99, 1621 Westmeier T., Braun R., Koribalski B. S., 2011, MNRAS, 410, 2217 Osterbrock D. E., Ferland G. J., 2006, Astrophysics of gaseous nebulae and White S. D. M., Frenk C. S., 1991, ApJ, 379, 52 active galactic nuclei, 2nd. University Science Books, Sausalito, CA Williams B. A., McMahon P. M., van Gorkom J. H., 1991, AJ, 101, 1957 Peimbert M., Costero R., 1969, Bol. Obs. Tonantzintla Tacubaya, 5, 3 Zaritsky D., Christlein D., 2007, AJ, 134, 135 Perez-Montero´ E., Garc´ıa-Benito R., Hagele¨ G. F., D´ıaz A.´ I., 2010, Zaritsky D., Kennicutt R. C., Jr, Huchra J. P., 1994, ApJ, 420, 87 MNRAS, 404, 2037 Perez-Montero´ E. et al., 2011, A&A, 532, A141 Perez-Montero´ E., 2014, MNRAS, 441, 2663 APPENDIX A: RELATIONSHIPS BETWEEN Pettini M., Pagel B. E. J., 2004, MNRAS, 348, L59 THE THEORETICAL H I BALMER RATIOS Pilyugin L. S., 2000, A&A, 362, 325 FOR CASE B RECOMBINATION Pilyugin L. S., 2001a, A&A, 369, 594 AND OXYGEN ABUNDANCE Pilyugin L. S., 2001b, A&A, 374, 412 Rosales-Ortega F. F., Sanchez´ S. F., Iglesias-Paramo´ J., D´ıaz A. I., V´ılchez Although it is very common to assume that the theoretical Hα/Hβ J. M., Bland-Hawthorn J., Husemann B., Mast D., 2012, ApJ, 756, L31 ratio for case B recombination is 2.86, this value actually depends Downloaded from Rupke D. S. N., Kewley L. J., Barnes J. E., 2010, ApJ, 710, L156 on both the electron temperature and electron density of the gas Ryan-Weber E. V. et al., 2004, AJ, 127, 1431 (Osterbrock & Ferland 2006; Storey & Hummer 1995). In H II re- Salim S., Rich R. M., 2010, ApJ, 714, L290 gions within star-forming galaxies, the electron densities are typi- Salim S. et al., 2007, ApJS, 173, 267 cally found in the low-density limit, n = 100 cm−3, but electron Sanchez´ S. F. et al., 2012, A&A, 538, A8 e Sanchez´ S. F. et al., 2014, A&A, 563, A49 temperature depends on the effective temperature of the ionization http://mnras.oxfordjournals.org/ Saunders W. et al., 2004, in Moorwood A. F. M., Iye M., eds, Proc. SPIE source as well as on the metallicity of the gas, as the H II region Conf. Ser., Vol. 5492, Ground-based Instrumentation for Astronomy, cools down with increasing amount of metals. We note that the SPIE, Bellingham, p. 389 dependence between the theoretical Hα/Hβ ratio and the oxygen Schaerer D., Vacca W. D., 1998, ApJ, 497, 618 abundance is not very strong, but in order to get the best results and Schlegel D. J., Finkbeiner D. P., Davis M., 1998, ApJ, 500, 525 minimize the uncertainties when determining the reddening coef- Schmidt M., 1959, ApJ, 129, 243 ﬁcient of the ionized gas within star-forming galaxies it should be Schmidt M., 1963, ApJ, 137, 758 considered, specially at high metallicities where Te may be rather Searle L., Sargent W. L. W., 1972, ApJ, 173, 25 low. Strictly talking, the standard value of 2.86 is only valid for Sharp R., Birchall M. N., 2010, PASA, 27, 91 T = 10 000 K and n = 100 cm−3, which corresponds to an oxy- Sharp R. et al., 2006, in McLean I. S., Iye M., eds, Proc. SPIE Conf. Ser., e e Vol. 6269, Ground-based and Airborne Instrumentation for Astronomy, gen abundance of around 8.27. However, a cooler H II region (i.e. a at Oxford Brookes University on June 4, 2015 SPIE, Bellingham, p. 62690G high-metallicity H II region) will have a slightly higher theoretical β α β = ∼ Shaw R. A., Dufour R. J., 1995, PASP, 107, 896 Ha/H ratio, reaching H /H 3.04 for Te 5000 K. Similarly, Smith G. A. et al., 2004, in Moorwood A. F. M., Iye M., eds, Proc. SPIE hotter H II regions (i.e. low-metallicity H II regions) will have lower Conf. Ser., Vol. 5492, Ground-based Instrumentation for Astronomy. theoretical Ha/Hβ values (e.g. Hα/Hβ = 2.75 for T ∼ 20000 K). SPIE, Bellingham, p. 410 Here, we have used the data base of H II regions in star-forming Stasinska´ G., 2010, in Bruzual G. R., Charlot S., eds, Proc. IAU Symp. 262, galaxies presented by Lopez-S´ anchez´ & Esteban (2010b, Tables 2 Stellar Populations - Planning for the Next Decade. Cambridge Univ. and 3), for which a direct estimation of the electron tempera- Press, Cambridge, p. 93 ture is available, to get an empirical relationship between the Storchi-Bergmann T., Kinney A. L., Challis P., 1995, ApJS, 98, 103 Storey P. J., Hummer D. G., 1995, MNRAS, 272, 41 Sweet S. M., Drinkwater M. J., Meurer G., Bekki K., Dopita M. A., Kilborn Table A1. Theoretical H I Balmer ratios for Case B recombination at dif- = −3 V., Nicholls D. C., 2014, ApJ, 782, 35 ferent electron temperatures and ne 100 cm , computed from Storey & Thilker D. A. et al., 2005, ApJ, 619, L79 Hummer (1995), and approximate oxygen abundance for which the ionized Thilker D. A. et al., 2007, ApJS, 173, 538 gas has a particular electron temperature. Thilker D. A. et al., 2010, ApJ, 714, L171 Torres-Flores S., Scarano S., Mendes de Oliveira C., de Mello D. F., Amram Te Hα/Hβ Hγ /Hβ Hδ/Hβ Appr. 12+log(O/H) P., Plana H., 2014, MNRAS, 438, 1894 Tremonti C. A. et al., 2004, ApJ, 613, 898 5000 3.04 0.458 0.251 8.88 van Zee L., Haynes M. P., 2006, ApJ, 636, 214 6000 2.99 0.461 0.254 8.75 Vazquez´ G. A., Leitherer C., 2005, ApJ, 621, 695 7000 2.95 0.463 0.255 8.63 Veilleux S., Osterbrock D. E., 1987, ApJS, 63, 295 8000 2.91 0.465 0.257 8.50 Verdes-Montenegro L., Del Olmo A., Yun M. S., Perea J., 2005, A&A, 430, 9000 2.89 0.467 0.258 8.40 443 10 000 2.86 0.468 0.259 8.27 Vilchez J. M., Esteban C., 1996, MNRAS, 280, 720 11 000 2.84 0.470 0.260 8.15 Walter F., Brinks E., de Blok W. J. G., Bigiel F., Kennicutt R. C., Jr, Thornley 12 000 2.83 0.471 0.261 8.05 M. D., Leroy A., 2008, AJ, 136, 2563 13 000 2.81 0.472 0.261 7.88 Weilbacher P. M., Duc P.-A., 2001, in de Boer K. S., Dettmar R.-J., Klein U., 14 000 2.80 0.472 0.262 7.75 eds, Spectrophotometric Investigation of a Sample of Tidal Dwarf Galax- 15 000 2.79 0.473 0.262 7.62 ies, Dwarf Galaxies and their Environment. Shaker Verlag, Aachen, 16 000 2.78 0.473 0.263 7.50 p. 269 19 000 2.76 0.475 0.264 7.20 Weilbacher P. M., Duc P.-A., Fritze-v. Alvensleben U., 2003, A&A, 397, 20 000 2.75 0.475 0.264 7.10 545 Note. The bold font values indicate the standard values assumed in the Werk J. K. et al., 2010a, AJ, 139, 279 literature.

MNRAS 450, 3381–3409 (2015) 3408 A.´ R. Lopez-S´ anchez´ et al. Downloaded from http://mnras.oxfordjournals.org/ at Oxford Brookes University on June 4, 2015

Figure A1. Dependence of the theoretical H I Balmer ratios, Hα/Hβ (top), Hγ /Hβ (middle), and Hδ/Hβ (bottom) on the electron temperature, Te (left) and −3 the oxygen abundance (right), assuming Case B recombination and ne = 100 cm . A ﬁt using a quadratic (for Te) or third degree (for oxygen abundance) −3 polynomial is shown with a continuous line. The standard theoretical values for Te = 10 000 K and ne = 100 cm ,Hα/Hβ = 2.86, Hγ /Hβ = 0.468, and Hδ/Hβ = 0.259, are shown with dotted red lines.

theoretical Hα/Hβ ratio for case B recombination and oxygen between the logarithm of the Te (independent variable) and the abundance. Just for completeness, for the extreme low metallicity theoretical Hα/Hβ ratio (dependent variable), with the result regime, 12+log(O/H) ∼ 7.2–7.1, we have used the values given by Hα Izotov, Thuan & Guseva (2005) and Izotov & Thuan (2007). We = 10.35 − 3.254 log T + 0.3457(log T )2, (A1) Hβ e e considered the theoretical ratios of H I Balmer lines expected for case B recombination given by Storey & Hummer (1995) assuming being the correlation coefficient r = 0.9921, and the dispersion −3 ne = 100 cm .TableA1 compiles these values. d = 0.0024. Top-right panel of Fig. A1 shows the dependence of Top-left panel of Fig. A1 plots the dependence of the Hα/Hβ ratio Hα/Hβ ratio with the oxygen abundance. Similarly, we fitted a with the electron temperature. We fitted a quadratic polynomial degree 3 polynomial between O/H (independent variable) and the

MNRAS 450, 3381–3409 (2015) Ionized gas in the XUV disc of NGC 1512/1510 3409 theoretical Hα/Hβ ratio (dependent variable) to find SUPPORTING INFORMATION Hα =−24.767 + 10.882x − 1.442x2 + 0.0640x3, (A2) Additional Supporting Information may be found in the online ver- Hβ sion of this article: being x = 12+log(O/H). The correlation coefficient is r = 0.9479, β = and the dispersion is d = 0.001 16. Table 2. Dereddened line intensity ratios with respect to I(H ) 1 We have repeated this analysis for the case of the Hγ /Hβ and and other spectral properties of the regions. Hδ/Hβ ratios. The dependence of these ratios with both the electron Table 3. Reddening coefficient, important emission-line ratios and temperature and oxygen abundance is also shown in Fig. A1.The parameters used by empirical calibrations and derived oxygen abun- results are dances and N/O ratio for our sample regions. Table 6. Properties of the UV-rich regions within the NGC Hγ = . + . T − . T 2, 1512/1510 system. β 0 0254 0 1922 log e 0 0204(log e) (A3) H (http://mnras.oxfordjournals.org/lookup/suppl/doi:10.1093/mnras/ with r = 0.9907 and d = 0.000 29; stv703/-/DC1). Hδ =−0.071 32 + 0.1436 log T − 0.0153(log T )2, (A4) Hβ e e Please note: Oxford University Press are not responsible for the

content or functionality of any supporting materials supplied by Downloaded from = = with r 0.9903 and d 0.000 23; the authors. Any queries (other than missing material) should be Hγ directed to the corresponding author for the paper. = 1.7871 − 0.5248x + 0.0704x2 − 0.003 17x3, (A5) Hβ with r = 0.9477 and d = 0.000 29;

Hδ http://mnras.oxfordjournals.org/ = 1.5626 − 0.5110x + 0.0674x2 − 0.002 98x3, (A6) Hβ with r = 0.9982 and d = 0.000 23. This paper has been typeset from a TEX/LATEX ﬁle prepared by the author. at Oxford Brookes University on June 4, 2015

MNRAS 450, 3381–3409 (2015)