Abundance determination of multiple star-forming regions in the HII galaxy SDSS J165712.75+321141.4
Guillermo F. Hagele, Ruben Garcia-Benito, Enrique Perez-Montero, Angeles I. Diaz, Monica V. Cardaci, Veronica Firpo, Elena Terlevich, Roberto Terlevich
aa r X i v : . [ a s t r o - ph . C O ] J a n Mon. Not. R. Astron. Soc. , 000–000 (0000) Printed 10 November 2018 (MN L A TEX style file v2.2)
Abundance determination of multiple star-forming regionsin the H ii galaxy SDSS J165712.75+321141.4 Guillermo F. H¨agele , ⋆ , Rub´en Garc´ıa-Benito , , Enrique P´erez-Montero ,´Angeles I. D´ıaz , M´onica V. Cardaci , , Ver´onica Firpo , Elena Terlevich † and Roberto Terlevich ‡ Departamento de F´ısica Te´orica, M´odulo 15, Universidad Aut´onoma de Madrid, 28049 Madrid, Spain Facultad de Cs. Astron´omicas y Geof´ısicas, Universidad Nacional de La Plata, Paseo del Bosque s/n, 1900 La Plata, Argentina Kavli Institute of Astronomy and Astrophysics, Peking University, 100871, Beijing, China Instituto de Astrof´ısica de Andaluc´ıa, CSIC, Apdo. 3004, 18080, Granada, Spain. Instituto Nacional de Astrof´ısica, ´Optica y Electr´onica, Tonantzintla, Apdo. Postal 51, 72000 Puebla, M´exico
10 November 2018
ABSTRACT
We analyze high signal-to-noise spectrophotometric observations acquired simul-taneously with TWIN, a double-arm spectrograph, from 3400 to 10400 ˚A of threestar-forming regions in the H ii galaxy SDSS J165712.75+321141.4. We have measuredfour line temperatures: T e ([O iii ]), T e ([S iii ]), T e ([O ii ]), and T e ([S ii ]), with high preci-sion, rms errors of order 2%, 5%, 6% and 6%, respectively, for the brightest region, andslightly worse for the other two. The temperature measurements allowed the directderivation of ionic abundances of oxygen, sulphur, nitrogen, neon and argon.We have computed CLOUDY tailor-made models which reproduce the O mea-sured thermal and ionic structures within the errors in the three knots, with deviationsof only 0.1 dex in the case of O + and S ionic abundances. In the case of the electrontemperature and the ionic abundances of S + /H + , we find major discrepancies whichcould be consequence of the presence of colder diffuse gas. The star formation historyderived using STARLIGHT shows a similar age distribution of the ionizing popula-tion among the three star-forming regions. This fact suggests a similar evolutionaryhistory which is probably related to a process of interaction with a companion galaxythat triggered the star formation in the different regions almost at the same time.The hardness of the radiation field mapped through the use of the softness parame-ter η is the same within the observational errors for all three regions, implying thatthe equivalent effective temperature of the radiation fields are very similar for all thestudied regions of the galaxy, in spite of some small differences in the ionization stateof different elements.Regarding the kinematics of the galaxy, the gas rotation curve shows a deviationfrom the circular motion probably due either to an interaction process or related toan expanding bubble or shell of ionized gas approaching us. A dynamical mass of2.5 × M ⊙ is derived from the rotation curve. Key words:
ISM: abundances - H ii regions - galaxies: abundances - galaxies: funda-mental parameters - galaxies: starburst - galaxies: stellar content. ⋆ CONICET, Argentina; e-mail: [email protected],[email protected] † Visiting astronomer at IoA, University of Cambridge, UK ‡ Research Affiliate, IoA, University of Cambridge, UK
Star formation is an ongoing process in the local universe,with observed rates of the order of 10 − M ⊙ yr − Mpc − (Madau et al. 1996). Most of the light and metals are pro-duced in the most massive among the newly formed stars.The most extreme regions forming massive stars are often c (cid:13) G. F. H¨agele et al. referred to as starbursts. In the local universe they accountfor about a quarter of all star formation (Heckman 1997),and this fraction may have been larger in the younger uni-verse. The origin of the term “starburst” (coined as “star-burst nuclei” by Weedman et al. 1981) dates back to theearly observations of dust-obscured star-forming regions inthe centres of nearby galaxies at the end of the seventiesand beginning of the eighties, but the basic concept extendsfurther back ( e.g. , Hodge 1969; Searle et al. 1973).The level of intensity of a starburst is highly variable.According to Terlevich (1997), in a starburst galaxy the en-ergy output of the starburst ( L SB ) is much larger than theone coming from the rest of the galaxy ( L G ), a galaxy with L SB ∼ L G is a galaxy with starbursts, and in a normalgalaxy L SB ≪ L G . This classification shows the variety ofenvironments of the bursts. It is clear that the visibility ofthe burst depends not only on its intensity but also on itsenvironment. Terlevich (1997) also proposed a division inphases of the starburst. The first one, the nebular phase, ischaracterized by the presence of strong emission lines fromgas photoionized by young massive stars, with an age of lessthan 10 Myr. The early continuum phase goes from 10 to 100Myr, when some Balmer lines appear in absorption and oth-ers in emission. Finally, the late continuum phase, is whenonly some weak emission lines appear in the spectrum. TheH ii galaxies are typical examples of the first phase.H ii galaxies are gas-rich dwarf galaxies experiencing aviolent star formation period which dominates the opticalspectrum of the host galaxy. They have one of the high-est intensity levels of star forming activity. In general, thesegalaxies have a central region which contains one or morestar forming knots, with a diameter of several hundred par-secs with high surface brightness, and a low luminosity un-derlying galaxy (M V > -17). The activity of the star for-mation episodes can not be sustained continuously for longperiods of time, since the central region can not have enoughgas to fuel these processes for longer than 10 years and tomatch the gas content and metallicity with theoretical con-siderations (Thuan et al. 2004).Spectroscopically, H ii galaxies are essentially identicalto the giant H ii regions found in nearby irregular and late-type galaxies. The correlation among structural parameters(H β luminosity, velocity dispersion, line widths) and be-tween these parameters and chemical composition (Terlevich& Melnick 1981) favours the interpretation of H ii galaxies asgiant H ii regions in distant dwarf irregular galaxies similarto the ones found nearby (Melnick et al. 1985).Other important characteristic of H ii galaxies is theirlow metallicity (Z ⊙ /50 Z Z ⊙ /3; Kunth & Sargent1983). The fact that H ii galaxies are metal-poor and veryblue objects seems to suggest that they are young. Nev-ertheless, there is evidence which indicates the presence ofpopulations older than the ones in the starburst. This isseen in the behaviour of the surface brightness profile whichis exponential in the external zones, or in the colour index,which turns redder in V-R and V-I (Telles & Terlevich 1997).IZw18 in particular, was considered as the best candidatefor a truly young galaxy. Early studies of the stellar popula-tion of IZw18 did not reveal any old population (Hunter &Thronson 1995). This contradicted some models which pre-dict that during a starburst, the heavy elements producedby the massive stars are ejected with high velocities into a hot phase, leaving the starburst region without immediatecontribution to the enrichment of the interstellar medium(Tenorio-Tagle 1996). In this scenario, the metals observednow would have their origin in a previous star formationevent, and an underlying old stellar population would beexpected. In fact, Garnett et al. (1997) attributed the highcarbon abundance that they found in HST spectroscopy ofIZw 18 as evidence for the presence of an old stellar popula-tion. In agreement with this result, using HST archive data(Aloisi et al. 1999) showed that stars older than 1 Gyr mustbe present in IZw 18. Moreover, studies of the resolved stellarpopulation in the near infrared with NICMOS ( ¨Ostlin 2000)found also that while the near infrared colour-magnitudediagram was dominated by stars 10-20 Myr old, the pres-ence of numerous AGB stars require an age of at least 10 years. Legrand et al. (2000) modelled the relative abundanceof metals in IZw 18 and concluded that, in addition to thepresent burst of star formation, a low star formation rate ex-tended over a long period of time was necessary to accountfor the observed values.In recent years, with the development of the IntegralField Unit (IFU) instruments to perform 3D spectroscopy,works that require a spatial coverage to study extendedgalactic or extra galactic star-forming regions have beenmainly focused on the use of this technique (Rela˜no et al.2010; Cair´os et al. 2010; Monreal-Ibero et al. 2010; Rosales-Ortega et al. 2010; P´erez-Gallego et al. 2010; Garc´ıa-Benitoet al. 2010; S´anchez et al. 2010; P´erez-Montero et al. 2011,see for example). However, medium or high dispersion slitspectroscopy are a better option for spectrophotometry, orwhen the object is very compact, or even extended butwith few star-forming knots. This is also the case whengood spatial and spectral resolution and simultaneous widespectral coverage are required (D´ıaz et al. 2007; Cumminget al. 2008; Firpo et al. 2010; H¨agele et al. 2006, 2007,2008; H¨agele 2008; H¨agele et al. 2009, 2010; P´erez-Monteroet al. 2009; L´opez-S´anchez & Esteban 2009, 2010a,b; L´opez-S´anchez 2010; Firpo et al. 2011, see for example).In this paper we present simultaneous blue and redlong-slit observations obtained with the double arm TWINspectrograph at the 3.5m telescope of Calar Alto of thethree brightest star-forming knots of the H ii galaxy SDSSJ165712.75+321141.4. This is part of a project to obtaina top quality spectrophotometric data base to determineionized gas parameters which are indispensable to criticallytest photoionization models and to explore discrepancies be-tween models and observations. In Section 2 we show thedetails of the observations and data reduction. Section 3presents the derived physical characteristics of the regions,including the electron temperature for four different species.Section 4 is devoted to the discussion of these results, and fi-nally the summary and conclusions are presented in Section5. c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 Table 1.
Right ascension, declination, redshift and SDSS photometric magnitudes of the observed knots obtained using the SDSS exploretools a . Object ID hereafter ID Knot RA Dec redshift u g r i z(spSpec SDSS)SDSS J165712.75+321141.4 SDSS J1657 A 16 h m . s
75 32 ◦ ′ . ′′
42 0.038 17.63 17.01 17.25 17.14 17.16(spSpec-52791-1176-591) B 16 h m . s
26 32 ◦ ′ . ′′
20 — 20.34 19.76 20.65 20.72 20.27C 16 h m . s
58 32 ◦ ′ . ′′
09 — 19.28 18.75 19.52 19.40 19.17 a http://cas.sdss.org/astro/en/tools/explore/obj.asp Using the implementation of the SDSS database in theINAOE Virtual Observatory superserver , we selected thebrightest nearby narrow emission line galaxies with verystrong lines and large equivalent widths of H α from thewhole SDSS data release available at the time of planningthe observations. These preliminary lists were then pro-cessed using BPT (Baldwin, Phillips & Terlevich 1981) di-agnostic diagrams to remove AGN-like objects. The finallist consisted of about 10500 bonafide bright H ii galaxies.They show spectral properties indicating a wide range ofgaseous abundances and ages of the underlying stellar pop-ulations (L´opez 2005). From this list, the final set was se-lected by further restricting the sample to the largest H α flux and highest signal-to-noise ratio objects (for a completedescription of the selection criteria see H¨agele et al. 2006,hereafter Paper I). Of the selected sample, we chose SDSSJ165712.75+321141.4 to be observed at the allocated time.For simplicity, we will call the galaxy SDSS J1657 in whatremains of the paper.Some general characteristics of the knots of SDSS J1657collected from the SDSS web page are listed in Table 1. Blue and red spectra were obtained simultaneously usingthe double beam Cassegrain Twin Spectrograph (TWIN)mounted on the 3.5m telescope of the Calar Alto Observa-tory at the Centro Astron´omico Hispano Alem´an (CAHA),Spain. These observations were part of a four night observingrun in 2006 June and they were acquired under excellent see-ing and photometric conditions (for details see; H¨agele et al.2008, hereinafter Paper II). The blue arm covers the wave-length range 3400-5700 ˚A (centred at λ c = 4550 ˚A), givinga spectral dispersion of 1.09 ˚A pixel − (R ≃ λ c = 8100 ˚A) with a spectral dispersion of 2.42 ˚A pixel − (R ≃ ∼ http://astro.inaoep.mx/en/observatories/virtual/ Figure 1.
False colour image of SDSS J1657 with the slit po-sition and the adopted knot names superimposed. This imagewas obtained using the SDSS explore tools. Circles and squaresrepresent the photometric and spectroscopic SDSS targets, re-spectively. The scale is 782 pc arcsec − , at the adopted distancefor SDSS J1657. [ See the electronic edition of the Journal for acolour version of this figure. ] Table 2.
CAHA instrumental configuration.Spectral range Disp. R a FWHM Spatial res.(˚A) (˚A px − ) ( ′′ px − )Blue 3400-5700 1.09 1420 0.56Red 5800-10400 2.42 1160 0.56 a RFWHM = λ /∆ λ FWHM nebular lines from [O ii ] λλ iii ] λλ ∼
60 for [O iii ] λ ∼
20 for [S ii ] λ Several bias and sky flat field frames were taken at the be-ginning and at the end of the night in both arms. In addi-tion, two lamp flat fields and one He-Ar calibration lampexposures were performed at each telescope position. The c (cid:13) , 000–000 G. F. H¨agele et al. -10 0 10distance (arcsecs)05e-161e-151.5e-15 F l ux ( e r g s - c m - ) H α differencecontinuum SDSS J1657
Knot CKnot AKnot B
Figure 2.
Spatial profile of the light distribution along theslit for the observed H α emission. The profiles correspond toline+continuum (dashed line), continuum (dashed-dotted line)and the difference between them (solid line), representing the pureemission from H α . images were processed and analysed with IRAF routinesin the usual manner. This procedure includes the removalof cosmic rays, bias subtraction, division by a normalisedflat field, and wavelength calibration. To finish, the spectraare corrected for atmospheric extinction and flux-calibrated.Four standard star observations were performed each nightat the same time for both arms, allowing a good spectropho-tometric calibration with an estimated rms error of about3%. Further details concerning each of these steps can befound in Paper II.Fig. 2 shows the spatial distribution of the H α flux andthe continuum along the slit for SDSS J1657. The emissionline profiles have been generated by collapsing 11 pixels ofthe spectra in the direction of the resolution at the centralposition of the emission lines in the rest frame, λ IRAF: the Image Reduction and Analysis Facility is distributedby the National Optical Astronomy Observatories, which is oper-ated by the Association of Universities for Research in Astronomy,Inc. (AURA) under cooperative agreement with the National Sci-ence Foundation (NSF).
The spectra of the three knots of SDSS J1657 (labelled fromA to C) with some of the relevant identified emission linesare shown in Fig. 3. The spectrum of each observed knot issplit into two panels. Knot A corresponds to the one anal-ysed in Paper II.The emission line fluxes were measured using the splot task in iraf following the procedure described in Paper I.Following P´erez-Montero & D´ıaz (2003), the statistical er-rors associated with the observed emission fluxes have beencalculated using the expression σ l = σ c N / [1 + EW/ ( N ∆)] / where σ l is the error in the observed line flux, σ c representsthe standard deviation in a box near the measured emissionline and stands for the error in the continuum placement, Nis the number of pixels used in the measurement of the lineflux, EW is the line equivalent width, and ∆ is the wave-length dispersion in ˚A per pixel (Gonz´alez-Delgado et al.1994). There are several emission lines affected by cosmeticfaults or charge transfer in the CCD, internal reflections inthe spectrograph, telluric emission lines or atmospheric ab-sorption lines. These cause the errors to increase, and, insome cases, they are impossible to quantify, in which casethey were ignored and excluded from any subsequent anal-ysis.Some observed lines (e.g., [Cl iii ] λλ σ l , as a conservative approach to include the uncertaintiesintroduced by the presence of the underlying stellar popu-lation.The absorption features of the underlying stellar pop-ulation may also affect the helium emission lines to someextent. However, these absorption lines are narrower thanthose of hydrogen (see, for example, Gonz´alez-Delgado et al.2005). Therefore it is difficult to set adequate pseudo-continua at both sides of the lines to measure their fluxes. c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 F l ux ( e r g c m - s - Å (cid:155) - ) [ O II] [ N e III] Η γ Η δ Η β [ O III] [ O III] [ O III] H e I [ S III] [ S II] , Η α [ S III] [ S III] [ O II] , SDSS J1657 - Knot A F l ux ( e r g c m - s - Å (cid:155) - ) [ O II] [ N e III] Η γ Η δ Η β [ O III] [ O III] [ O III] H e I [ S III] [ S II] , Η α [ S III] [ S III] [ O II] , SDSS J1657 - Knot B F l ux ( e r g c m - s - Å (cid:155) - ) [ O II] [ N e III] Η γ Η δ Η β [ O III] [ O III] [ O III] H e I [ S III] [ S II] , Η α [ S III] [ S III] [ O II] , SDSS J1657 - Knot C
Figure 3.
Blue and red spectra of Knots A, B and C of SDSS J1657 (upper, middle and lower panel, respectively) in the rest frame.The flux scales are the same in both spectral ranges.
We also applied the STARLIGHT code (Cid Fernan-des et al. 2005) to each region to separate the emissionspectra from the underlying stellar absorptions, but for thestrongest emission lines the difference between the measure-ments made after the subtraction of the STARLIGHT fitand the ones made using the pseudo-continuum is well be-low the observational errors. For a detailed discussion onthe differences in the emission line measurements see P´erez-Montero et al. (2010). The STARLIGHT project is supported by the Brazilian agen-cies CNPq, CAPES and FAPESP and by the France-BrazilCAPES/Cofecub program.
The reddening coefficients c (H β ) were calculatedfrom the measured Balmer decrements, F ( λ )/ F ( Hβ ). Weadopted the galactic extinction law of Miller & Mathews(1972) with R v =3.2. A least square fit of the measureddecrements to the theoretical ones, ( F ( λ )/ F ( Hβ )) , com-puted based on the data by Storey & Hummer (1995), wasperformed that provides the value of c (H β ). The theoreticalBalmer decrements depend on electron density and temper-ature. We used an iterative method to estimate them, takingas starting values those derived from the measured [S ii ] λλ iii ] λλ c (cid:13) , 000–000 G. F. H¨agele et al.
Table 3.
Relative reddening corrected line intensities [ F ( Hβ )= I ( Hβ )=10000] for the three star-forming knots.Knot A Knot B Knot C λ (˚A) f( λ ) EW I ( λ ) Error EW I ( λ ) Error EW I ( λ ) Error(˚A) (%) (˚A) (%) (˚A) (%)3727 [O ii ] a ±
230 1.2 107.8 13266 ±
420 3.2 112.2 14809 ±
282 1.93750 H12 0.266 1.9 232 ±
47 20.2 — — — — — —3770 H11 0.261 2.3 293 ±
40 13.8 9.7 617 ±
146 23.7 — — —3798 H10 0.254 4.1 500 ±
68 13.5 — — — 4.9 465 ±
141 30.43835 H9 0.246 7.1 780 ±
93 11.9 8.9 783 ±
180 23.0 7.5 761 ±
140 18.43868 [Ne iii ] 0.238 23.1 3262 ±
132 4.0 30.7 3663 ±
132 3.6 28.6 3533 ±
118 3.33889 He i +H8 0.233 14.0 1826 ±
95 5.2 22.3 1954 ±
208 10.7 26.2 2192 ±
267 12.23968 [Ne iii ]+H7 0.216 22.4 2456 ±
121 4.9 30.1 2705 ±
221 8.2 29.9 2699 ±
237 8.84026 [N ii ]+He i ±
16 10.4 4.3 327 ±
105 32.1 — — —4068 [S ii ] 0.195 1.4 198 ±
15 7.4 1.4 153 ±
36 23.7 1.2 130 ±
35 26.94102 H δ ±
65 2.7 30.2 2667 ±
163 6.1 29.2 2619 ±
173 6.64340 H γ ±
97 2.2 76.5 4813 ±
219 4.5 68.7 4712 ±
140 3.04363 [O iii ] 0.138 4.5 524 ±
24 4.6 8.3 846 ±
67 7.9 8.9 838 ±
59 7.04471 He i ±
33 7.4 4.7 405 ±
44 11.0 5.3 445 ±
41 9.34658 [Fe iii ] 0.053 1.0 107 ±
16 14.9 — — — — — —4686 He ii ±
14 11.0 4.3 313 ±
68 21.9 2.1 170 ±
24 14.24861 H β ±
79 0.8 153.3 10000 ±
178 1.8 167.2 10000 ±
128 1.34921 He i -0.014 0.8 75 ±
14 18.6 — — — — — —4959 [O iii ] -0.024 152.5 14333 ±
127 0.9 218.5 16118 ±
129 0.8 199.2 14940 ±
114 0.84986 [Fe iii ] b -0.030 1.4 135 ±
28 20.5 2.4 163 ±
57 35.1 3.2 215 ±
39 18.35007 [O iii ] -0.035 455.1 43082 ±
240 0.6 705.5 48653 ±
256 0.5 613.2 44727 ±
129 0.35015 He i -0.037 2.4 222 ±
23 10.1 3.2 205 ±
35 17.3 3.2 220 ±
28 12.65199 [N i ] -0.078 2.0 157 ±
26 16.4 — — — — — —5876 He i -0.209 18.9 1116 ±
44 3.9 23.4 991 ±
29 2.9 29.0 1149 ±
47 4.16300 [O i ] -0.276 8.1 438 ±
16 3.7 9.6 393 ±
26 6.7 8.6 366 ±
14 3.96312 [S iii ] -0.278 3.7 201 ± ±
10 5.8 3.5 148 ±
10 6.46364 [O i ] -0.285 2.8 152 ±
18 11.8 3.6 141 ±
20 14.4 3.0 124 ±
12 10.06548 [N ii ] -0.311 9.4 464 ±
23 5.0 5.5 216 ±
23 10.7 6.7 266 ±
14 5.46563 H α -0.313 571.3 27772 ±
153 0.5 772.5 28159 ±
105 0.4 730.5 27919 ±
133 0.56584 [N ii ] -0.316 28.8 1428 ±
47 3.3 16.2 632 ±
50 8.0 17.0 680 ±
29 4.36678 He i -0.329 6.7 315 ±
18 5.7 7.3 272 ±
11 4.1 7.9 296 ±
20 6.96717 [S ii ] -0.334 47.4 2207 ±
57 2.6 39.4 1489 ±
35 2.3 45.1 1699 ±
56 3.36731 [S ii ] -0.336 32.2 1598 ±
43 2.7 28.1 1060 ±
25 2.3 30.7 1154 ±
40 3.47065 He i -0.377 5.6 235 ±
10 4.4 7.5 226 ±
12 5.1 8.8 279 ±
26 9.37136 [Ar iii ] -0.385 16.4 717 ±
26 3.6 18.4 584 ±
19 3.3 17.6 581 ±
17 2.97281 He i c -0.402 0.9 41 ± ±
14 15.9 1.4 46 ±
13 28.67319 [O ii ] d -0.406 12.3 302 ±
17 5.6 4.8 165 ±
15 9.2 6.1 196 ±
18 9.27330 [O ii ] e -0.407 8.8 211 ±
14 6.5 7.2 251 ±
21 8.4 7.9 254 ±
26 10.17751 [Ar iii ] -0.451 4.6 177 ±
22 12.3 5.5 163 ±
13 8.0 5.8 170 ±
17 9.98665 P13 -0.531 7.7 144 ±
53 37.1 — — — 8.9 133 ±
38 28.38751 P12 -0.537 4.2 101 ±
29 28.2 — — — — — —8865 P11 -0.546 8.5 211 ±
34 16.3 17.9 277 ±
73 26.2 11.3 228 ±
46 20.29014 P10 -0.557 15.4 167 ±
36 21.5 — — — — — —9069 [S iii ] -0.561 59.2 1400 ±
99 7.1 68.2 984 ±
122 12.4 81.9 1577 ±
134 8.59229 P9 -0.572 16.9 263 ±
47 18.0 26.1 364 ±
129 35.6 — — —9532 [S iii ] -0.592 157.3 3674 ±
257 7.0 85.1 2700 ±
152 5.6 238.7 2915 ±
162 5.6I(H β )(erg sec − cm − ) 6.3 × − × − × − c(H β ) 0.05 ± ± ± a [O ii ] λλ b [Fe iii ] λλ c possibly blend with an unknown line; d [O ii ] λλ e [O ii ] λλ lying stellar population, only the strongest Balmer emissionlines (H α , H β , H γ and H δ ) were used.For the easiness of comparison, we have included in thefollowing sections the results presented in Paper II for knotA. Table 3 lists the reddening corrected emission lines foreach knot, together with the reddening constant and its er-ror taken as the uncertainties of the least square fit and thereddening corrected H β intensity. Column 1 shows the wave-length and the name of the measured lines. The adoptedreddening curve, f ( λ ), normalized to H β , is given in col-umn 2. The errors in the emission lines were obtained by propagating in quadrature the observational errors in theemission line fluxes and the reddening constant uncertain-ties. We have not taken into account errors in the theoreticalintensities since they are much lower than the observationalones. The physical conditions of the ionized gas, including electrontemperatures (T e ) and electron density (N e ≈ n([S ii ])), havebeen derived from the emission line data using the same c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 Table 4.
Electron densities and temperatures. Densities in cm − and temperatures in 10 K.n([S ii ]) T e ([O iii ]) T e ([O ii ]) T e ([S iii ]) T e ([S ii ])Knot A 30: 1.23 ± ± ± ± ± ± ± ± ± ± ± ± Table 5.
Ionic and total helium abundance.Knot A Knot B Knot CHe + /H + ( λ ± ± ± + /H + ( λ ± ± ± + /H + ( λ ± ± ± + /H + ( λ ± ± ± + /H + (Adop.) 0.087 ± ± ± /H + ( λ ± ± ± ± ± ± procedures as in Paper II, based on the five-level statisticalequilibrium atom approximation in the task temden , of thesoftware package IRAF (de Robertis et al. 1987; Shaw &Dufour 1995). As usual, we have taken as sources of errorthe uncertainties associated with the measurement of theemission-line fluxes and the reddening correction, and wehave propagated them through our calculations.For all three knots we have derived the elec-tron temperatures of [O ii ], [O iii ], [S ii ] and [S iii ]. The[O ii ] λλ iii ] electron temperatures, we have estimatedthese contributions to be less than 4 % in all cases and there-fore we have not corrected for this effect. The expression forthe correction of direct recombination, however, is valid onlyin the range of temperatures between 5000 and 10000 K.The temperatures found are slightly over that range. At anyrate, the relative contribution of recombination to collisionalintensities decreases rapidly with increasing temperature,therefore for the high T e values found in our objects thiscontribution is expected to be small.The derived electron densities and temperatures for thethree star-forming regions are given in Table 4 along withtheir corresponding errors. We have derived the ionic chemical abundances of the dif-ferent species using the strongest available emission linesdetected in the analyzed spectra and the task ionic of theSTSDAS package in IRAF, as described in Paper II.The total abundances have been calculated by takinginto account, when required, the unseen ionization stagesof each element, using the appropriate ionization correctionfactor (ICF) for each species, X/H = ICF(X + i ) X + i /H + asdetailed in what follows. We have used the well detected He i λλ ii λ iii ] as representative ofthe zone where the He emission arises since at any rate ra-tios of recombination lines are weakly sensitive to electrontemperature. We have used the equations given by Olive &Skillman to derive the He + /H + value, using the theoreticalemissivities scaled to H β from Benjamin et al. (1999) and theexpressions for the collisional correction factors from King-don & Ferland (1995). To calculate the abundance of twiceionized helium we have used equation (9) from Kunth &Sargent (1983). We have not made any corrections for flu-orescence since three of the used helium lines have a smalldependence with optical depth effects but the observed ob-jects have low densities. We have not corrected either forthe presence of an underlying stellar population. A sum-mary of the equations used to calculate these ionic abun-dances is given in Appendix B of Garc´ıa-Benito (2009). Thetotal abundance of He has been found by adding directlythe two ionic abundances, He/H = (He + +He )/H + . Theresults obtained for each line and the total He abundances,along with their corresponding errors are presented in Table5. Also given in the table is the adopted value for He + /H + as the average, weighted by the errors, of the abundancesderived from each He i emission line . The oxygen ionic abundance ratios, O + /H + and O /H + ,have been derived from the [O ii ] λλ iii ] λλ + and O , there-fore the approximation O/H = (O + +O )/H + is a valid one.S + /H + and S /H + , abundances have been derivedusing T e ([S ii ]) and T e ([S iii ]) and the fluxes of the[S ii ] λλ iii ] λλ may be expected for sul-phur depending on the nebular excitation. The total sulphurabundance has been calculated using an ICF for S + +S according to the formula by Barker (1980), which is basedon Stasi´nska (1978) photo-ionization models, with α = 2.5,which gives the best fit to the scarce observational data onS abundances (P´erez-Montero et al. 2006). Taking thisICF as a function of the ratio O /O instead of O + /O re-duces the propagated error for this quantity.The ionic abundance of nitrogen, N + /H + has been de-rived from the intensities of the [NII] λλ e ([N ii ]) ≈ T e ([O ii ]). The N/O abundance ratio hasbeen derived under the assumption that N/O = N + /O + andN/H was calculated as log(N/H) = log(N/O)+log(O/H).Neon is only visible in the spectra via the [Ne iii ] emis-sion line at λ has been derived using thisline. For this ion we have taken the electron tempera-ture of [O iii ], as representative of the high excitation zone(T e ([Ne iii ]) ≈ T e ([O iii ]); Peimbert & Costero 1969). Classi-cally, the total abundance of neon has been calculated as-suming that Ne/O ≈ Ne /O . Izotov et al. (2004) point c (cid:13) , 000–000 G. F. H¨agele et al.
Table 6.
Ionic chemical abundances derived from forbidden emis-sion lines, ICFs and total chemical abundances for elements heav-ier than helium. Knot A Knot B Knot C12 + log(O + /H + ) 7.37 ± ± ± /H + ) 7.87 ± ± ±
12 + log(O/H) ± ± ± + +O ) ∗
12 + log(O/H) ∗ ± ± ± + /H + ) 6.07 ± ± ± /H + ) 6.00 ± ± ± + + S ) 1.32 ± ± ±
12 + log(S/H) ± ± ± ± ± ± + + S ) ∗
12 + log(S/H) ∗ ± ± ± ∗ -1.65 ± ± ± + /H + ) 6.15 ± ± ±
12 + log(N/H) ± ± ± ± ± ± + ) ∗
12 + log(N/H) ∗ ± ± ± ∗ -1.34 ± ± ± /H + ) 7.22 ± ± ± ± ± ±
12 + log(Ne/H) ± ± ± ± ± ± ) ∗
12 + log(Ne/H) ∗ ± ± ± ∗ -0.70 ± ± ± /H + ) 5.49 ± ± ± ) 1.13 ± ± ±
12 + log(Ar/H) ± ± ± ± ± ± ) ∗
12 + log(Ar/H) ∗ ± ± ± ∗ -2.47 ± ± ± ∗ ICFs and total abundances from photoionization models (seetext). out that this assumption can lead to an overestimate ofNe/H in objects with low excitation, where the charge trans-fer between O and H becomes important. Thus, we haveused the expression of this ICF given by (P´erez-Monteroet al. 2007). It is interesting to note, however, that giventhe high excitation of the observed objects there is no sig-nificant difference between the total neon abundance derivedusing this ICF and those estimated using the classic approx-imation.The only available emission lines of argon in the opticalspectra of ionized regions correspond to Ar and Ar . Theabundance of Ar has been calculated from the measured[Ar iii ] λ e ([Ar iii ]) ≈ T e ([S iii ])(Garnett 1992). [Ar iv ] was not detected in the spectra.The total abundance of Ar was hence calculated using theICF(Ar ) derived from photo-ionization models by P´erez-Montero et al. (2007).The ionic abundances with respect to ionized hydrogenof the elements heavier than helium, ICFs, total abundancesand their corresponding errors are given in Table 6. Detailed tailor-made photoionization models were producedin order to ascertain the main properties of the ionizing stel-lar population and the ionized gas. The methodology is de-scribed in P´erez-Montero et al. (2010) who study the bright-est knots of the H ii galaxies described in Papers I and II, in-cluding knot A in SDSS J1657. Here we describe the modelsfor knots B and C, and compare them with the observationsand with the results obtained for knot A in P´erez-Monteroet al. (2010).We have resorted to the photoionization code CLOUDYv. 06.02c (Ferland et al. 1998), using the equivalent widthof H β , after removing the underlying stellar population ( i.e. the population younger than 10 Myr), the H α luminosityand the intensities of [O ii ] 3727 ˚A, [O iii ] 4363 and 5007 ˚A,[S ii ] 6717 and 6731 ˚A, and [S iii ] 9069 and 9532 ˚A relative toH β . We have used for the photoionization models the sameStarburst 99 stellar libraries as in the model fitting of thestellar population, described in § ⊙ ). We assummed a constant star formation his-tory which, according to P´erez-Montero et al., gives thebest agreement for the number of ionizing photons and theEW(H β ) corrected for underlying stellar population anddust absorption effects. A thick shell geometry and a con-stant density of 100 particles per cm have been set as inputconditions in all the models. To fit the observed properties,the distance to the ionizing source, the filling factor, thedust-to-gas ratio and the age of the stellar cluster were leftas free parameters.One of the most important parameters in the correctmodelling of ionized gas nebulae is the dust absorption fac-tor, f d , which gives the ratio between the number of ionizingphotons emitted by the stellar cluster and the number ofionizing photons absorbed by the gas (P´erez-Montero et al.2010). This factor must be taken into account in derivingproperties of the cluster from hydrogen Balmer recombina-tion lines. It has been obtained in the best model, after aniterative method to fit the observed relative emission-lineintensities and the corrected EW(H β ) and L(H α ). In Fig. 4we show the ratio between the intensities of the most repre-sentative observed and modelled emission lines for the threeknots. Data for knot A have been taken from P´erez-Monteroet al. (2010). As we can see, the agreement results excellentfor all involved [O iii ] lines, with a deviation smaller than5% in all three knots. In the case of the [O ii ] lines and[S iii ] 9069 ˚A it is better than 10%. The fitting of [S iii ] at6312 ˚A is a bit worse, with a 20% of disagreement in knotsA and C, and 30% in B. The largest discrepancy is found forthe [S ii ] lines, from 30% of disagreement up to 65% in thecase of 6717,6731 ˚A in knot B. In Table 7, we compare theobserved and modelled EW(H β ), corrected for the contribu-tion of the underlying stellar population. We also give thenumber of ionizing photons and other properties predictedby the individual models, such as the age of the ionizingcluster, filling factor, ionization parameter, dust-to-gas ra-tio and visual extinction. Regarding knot A, all quantitieshave been extracted from the model in P´erez-Montero et al.(2010). As we can see, the agreement between observed and c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 Wavelength (Å) I ob s ( λ ) / I m od ( λ ) [ O II] [ O III] [ S II] [ O III] [ S III] [ S II] [ O II] [ S III]
Knot AKnot BKnot C
Figure 4.
Ratio between observed and modelled intensities of the most representative emission lines for each one of the star-formigknots.
Table 7.
Observed and model-predicted properties of the threestudied regions Knot A Knot B Knot CAge (Myr) 7.9 5.1 4.6Abs. factor f d − ) Obser. 52.59 52.20 52.34Model 52.59 52.26 52.29EW(H β ) (˚A) Obser. a
132 153 167Model 127 158 164log Filling factor -2.52 -2.02 -2.22log U -2.84 -2.54 -2.62log Dust-to-gas ratio -1.96 -1.97 -1.81A V a Corrected for the underlying stellar population modelled values is excellent, both for the number of ionizingphotons and the EW(H β ).In Fig. 5 we show a comparison between the four mea-sured electron temperatures in each of the three knots andthe values predicted by the models. As we can see, the bestagreement is found for T e ([O iii ]). In T e ([S iii ]), a good agree-ment is found only for knot A. The model temperatures arehigher for T e ([S ii ]) and lower for T e ([O ii ]) than the derivedfrom the measured line intensities. In Fig. 6, we see the samecomparison for the four respective ionic abundances. As inthe case of electron temperatures, the agreement betweenO abundances derived from the observations and foundby the models is excellent, while in the case of O + and S ,only deviations not larger than 0.1 dex are found. The mostevident deviation is found for the values of S + /H + whichare higher in the direct measurements than in the model by0.3 dex in average for the three knots.We have calculated the total abundances of all the mea-sured ions, now using these models, taking into accountwhen required, the unseen ionization stages of each element,using the appropriate model predicted ICF for each species.The predicted ICFs for O, S, N, Ar and Ne and the to-tal abundances obtained are listed in Table 6. A discussionabout the differences between the ICFs calculated by thesemodels and those obtained from the most commonly usedformulae is found in Appendix A of P´erez-Montero et al.(2010). T e ([OII]) (derived) T e ([ O II]) ( m od e l s ) T e ([OIII]) (derived) T e ([ O III]) ( m od e l s ) T e ([SIII]) (derived) T e ([ S III]) ( m od e l s ) T e ([SII]) (derived) T e ([ S II]) ( m od e l s ) Figure 5.
Measured electron temperatures for the three knots,vs. values predicted by the photoionization models described inthe text. The symbols are: grey squares, knot A; red circles, knotB, and yellow triangles, knot C. Temperatures are in units of10 K. Four electron temperatures – T e ([O iii ]), T e ([O ii ]), T e ([S iii ])and T e ([S ii ])– have been estimated for the star-formingknots of SDSS J1657. The good quality of the data allows usto reach high precision, with rms errors of the order of 2%,5%, 6% and 6% in knot A for T e ([O iii ]), T e ([O ii ]), T e ([S iii ]),and T e ([S ii ]), respectively. For the faintest knots, B and C,the fractional errors are slightly higher, 3%, 8%, 6% and17%, respectively.The star-forming regions show temperatures within arelatively narrow range, between 12000 and 14800 K for c (cid:13) , 000–000 G. F. H¨agele et al. -4.8 -4.6 -4.4 -4.2 -4.0 log(O /H + ) (derived) -4.8-4.6-4.4-4.2-4.0 l og ( O + / H + ) ( m od e l s ) -5.0 -4.8 -4.6 -4.4 -4.2 log(O + /H + ) (derived) -5.0-4.8-4.6-4.4-4.2 l og ( O + / H + ) ( m od e l s ) -6.4 -6.2 -6.0 -5.8 -5.6 log(S /H + ) (derived) -6.4-6.2-6.0-5.8-5.6 l og ( S + / H + ) ( m od e l s ) -6.6 -6.4 -6.2 -6.0 -5.8 log(S + /H + ) (derived) -6.6-6.4-6.2-6.0-5.8 l og ( S + / H + ) ( m od e l s ) Figure 6.
Modelled vs. derived oxygen and sulphur ionic abun-dances (see description in the text). Knot A: grey squares; knotB: red circles; knot C: yellow triangles. T e ([O iii ]). It is worth remembering that the adopted se-lection criteria for SDSS J1657 was high H β flux and largeequivalent width of H α , which tend to render objects withabundances and electron temperatures close to the medianvalues shown by H ii galaxies. Although these criteria ap-plied to the main knot (the SDSS spectrum), we find similarelectron temperatures for all the regions. To our knowledge,there is no previously reported T e ([O iii ]) for this galaxy inthe literature. The estimated [O iii ] temperature for knot Bis very similar to that of knot C; both are higher than theT e ([O iii ]) for knot A by about 2000 K. At same the time,although differences in T e ([S iii ]) among the three knots aremuch larger, being this temperature 1900 K larger in knot Bthan in knot A, and 1600 K lower in knot C than in knot A,these deviations are still compatible within the errors withthe empirical relation found in Paper I between T e ([O iii ])and T e ([S iii ]) for a heterogeneous sample of Giant HII Re-gions and H ii galaxies. The abundances derived for the three knots using the directmethod show the characteristic low values found in strongline H ii galaxies (Terlevich et al. 1991; Hoyos & D´ıaz 2006).These values are in the range of 12+log(O/H) = 7.78 - 7.99,in very good agreement with what is found from the pho-toionization models discussed above, ranging between 7.80 -8.02. The data presented in this paper is of high quality andthe mean error values for the oxygen and neon abundancesare 0.05 dex, 0.12 for sulphur and 0.06 for argon. Knots Band C show a similar value of 12+log(O/H), while KnotA is almost 0.2 dex higher. This difference is greater thanthe estimated observational errors, and is similar (or evensmaller) to what is found in other works with spatial reso-lution of knots that belong to H ii galaxies or Blue CompactDwarf (BCD) galaxies (see e.g. Izotov et al. 1997; V´ılchez &Iglesias-P´aramo 1998; Kehrig et al. 2008; Cair´os et al. 2009; P´erez-Montero et al. 2009; Garc´ıa-Benito et al. 2010). How-ever, in general, these differences were attributed to the ob-servational uncertainties (pointing errors, seeing variations,etc.) or errors associated to the reddening correction and fluxcalibration, and the oxygen abundance variations were notassumed as statistically significant, concluding that thereis a possible common chemical evolution scenario in all ofthem. There are even greater differences when comparingthe estimated abundances of the individual knots with thosederived from the integrated spectra of the galaxies. For in-stance, Cair´os et al. (2009) found for the integrated spec-trum of Mrk 1418 a lower value of direct oxygen abundanceby about 0.35 dex (equivalent to a factor of 2.2) than forknots 1 and 2 of that galaxy. They pointed out that whilethis variation could reflect a real abundance difference indifferent scales (kpc-sized aperture for the integrated spec-trum and sizes of the order of 100 pc for individual H ii re-gions), it may also be due to relatively large measurementuncertainties for the weak [O iii ] auroral emission line. For-tunately, our data are not affected by pointing errors andseeing variations, and the other observational uncertaintieshave a second order effect, since TWIN is a double beamlongslit spectrograph that simultaneously acquire all the ob-served spectral range. Likewise, the errors associated withthe measurements of the weak auroral emission lines are rel-atively small, specially for [O iii ].The logarithmic N/O ratios found for SDSS J1657 us-ing the direct method are -1.23 ± ± ± ± ± ± ∼ ⊙ = -1.36 (Grevesse & Sauval 1998). On the otherhand, the logarithmic Ne/O ratio is remarkably constant,with a mean value of 0.75 (0.72 from the photoionizationmodels), despite the differences in oxygen abundance be-tween knot A and knots B and C. They are consistent withsolar, log(Ne/O) = -0.61 . The Ar/O ratios found (which arealmost the same using the direct method and the photoion- Oxygen from Allende-Prieto et al. (2001) and neon fromGrevesse & Sauval (1998). c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 -1.0 -0.5 0.0log (S + /S )-1.0-0.50.0 l og ( O + / O + ) HII galaxiesPaper I and IISDSS J1657 knotsSDSS J1657 knots (models)log η = −0.35 log η = 0 log η = 0.2 log η = −0.6
Mrk 709
Knot AKnot CKnot B -1.0 -0.5 0.0 0.5log ([SII] λλ6717,31/[
SIII] λλ9069,9532) -1.0-0.50.0 l og ([ O II] λλ , / [ Ο ΙΙΙ] λλ , ) HII galaxiesPaper I and IISDSS J1657 knotsSDSS J1657 knots (models)log η ’ = -0.4log η ’ = 0 log η ’ = -0.9 Mrk 709
Knot AKnot BKnot C
Figure 7.
Left panel: log(O + /O ) vs. log(S + /S ) using the direct method and the photoionization models (filled yellow diamondsand red triangles), the objects studied in Paper I and Paper II (blue circles) and H ii galaxies from the literature (open squares). Diagonalsin this diagram correspond to constant values of η . Right panel: log([O ii ]/[O iii ]) vs. log([S ii ]/[S iii ]), symbols as in left panel. Diagonalsin this diagram correspond to constant values of η ’. ization models) show a very similar value for Knot A andB, while Knot C has a ratio higher by 0.2 dex. The meanvalue is consistent with solar, log(Ar/O) = -2.29 , within theobservational errors.Finally, the derived helium abundances are the same forthe three knots within observational errors. The ionization structure of a nebula depends essentially onthe shape of the ionizing continuum and the nebular ge-ometry and can be traced by the ratio of successive stagesof ionization of the different elements. With our data it ispossible to use the O + /O and the S + /S to probe thenebular ionization structure. In fact, V´ılchez & Pagel (1988)showed that the quotient of these two quantities that theycalled “softness parameter” and denoted by η is intrinsi-cally related to the shape of the ionizing continuum anddepends on geometry only slightly. An insight into the ion-ization structure of the observed objects can be gained bymeans of the η diagram (see Paper I).In Fig. 7, left panel, we show the relation betweenlog(O + /O ) and log(S + /S ) derived using the directmethod and the photoionization models for the knots ofSDSS J1657 (filled yellow diamonds and red triangles, re-spectively), the objects studied in Paper I and Paper II (bluecircles) and H ii galaxies (open squares) from the literature(see description and references in Paper II). In this diagramdiagonal lines correspond to constant values of the η pa-rameter which can be taken as an indicator of the ionizingtemperature (V´ılchez & Pagel 1988). H ii galaxies occupy theregion with log η between -0.35 and 0.2, which correspondsto high values of the ionizing temperature. As noticed inPaper II, Knot A shows a very low logarithmic value of Oxygen from Allende-Prieto et al. (2001) and argon fromGrevesse & Sauval (1998). η , -0.6. Knots B and C present very similar values. Theseobjects, however, have the [O ii ] λλ iii ] lines measured in both theSDSS spectrum and ours is good, the [O ii ] lines measured onthe SDSS spectrum provide a T e ([O ii ]) = 1.23 ± + /O moving the cor-responding data point upwards in the left panel of Fig. 7.This would be consistent with the position of the objectin the right panel of the figure showing log([O ii ]/[O iii ]) vs.log([S ii ]/[S iii ]), which does not require explicit knowledgeof the line temperatures involved in the derivation of theionic ratios, and therefore does not depend on the methodto derive or estimate these temperatures. The right panelin Fig. 7 shows the purely observational counterpart of theleft panel. In this diagram diagonal lines represent constantvalues of log η ’ = log [([O ii ]/[O iii ])/([S ii ]/[S iii ])].Another possible explanation for the differences be-tween these two diagrams can be obtained by inspectingthe position of the models described above in relation to theobservational points (red triangles in the figures). Modelswith a thick shell geometry and a constant density predicthigher T e ([S ii ]) and, hence, lower S + /H + , which causes the η parameter to be higher than the values estimated fromthe measurements. In fact, the difference between the ionicabundances derived by the direct method and those pre-dicted by the models can be also seen in Fig. 6. On aver-age there is a difference of 0.23 dex for O + /H + , 0.07 dex forO /H + , 0.67 dex for S + /H + and 0.13 dex for S /H + . Thiseffect, already pointed out by P´erez-Montero et al. (2010),could be a consequence of an outer shell of cold diffuse ion-ization structure in these objects with an extra emission of[S ii ] which contributes to their integrated spectrum. Theagreement is better for the η ’ diagram, being knot B the c (cid:13) , 000–000 G. F. H¨agele et al. + l o g ( O / H ) O A B C 7.67.88.08.28.4 + l o g ( O / H ) N2 A B C 7.67.88.08.28.4 + l o g ( O / H ) O3N2
A B C + l o g ( O / H ) S A B C + l o g ( O / H ) S /O A B C + l o g ( O / H ) Ar /O A B C
Figure 8.
The oxygen abundances and their uncertainties for each observed knot of SDSS J1657 (blue solid line rectangles), as derivedusing different empirical calibrators. From left to right top panels: O , N2, and O3N2 and bottom ones: S , S /O , and Ar O . Thedash (black) and dotted (red) line rectangles represent the abundances and their uncertainties as derived from the direct method andthe photoionization models, respectively. most discrepant, which is consistent with the differences be-tween the observed and modelled lines (see Fig. 4), wherethe higher difference corresponds to the [S ii ] lines.In both diagrams, η and η ’, the three knots of SDSSJ1657 present a very similar ionization structure, showingalmost the same values within the observational errors. Thisimplies that the equivalent effective temperatures of the ion-ization radiation field are very similar for all the knots, al-though we find some small differences in the ionization stateof the different elements. The emission line spectra of the three star-forming knots inJ1657 are very similar, implying similar values for ioniza-tion parameter, ionization temperature, and chemical abun-dances. We derived the ionization parameters from the [O ii ]to [O iii ] lines ratio according to the expression given in D´ıazet al. (2000). The logarithmic ratio is similar in all the knotsranging from -2.47 for knot B to -2.65 for knot A.The different strong-line empirical methods for abun-dance derivations, which have been widely studied in theliterature, are based on directly calibrating the relative in-tensity of some bright emission lines against the abundanceof some relevant ions present in the nebula (see e.g. Garc´ıa-Lorenzo et al. 2008; Cair´os et al. 2009; Garc´ıa-Benito 2009;Garc´ıa-Benito et al. 2010, and references therein). For thecase of oxygen, we take the calibrations studied by P´erez-Montero & D´ıaz (2005), who obtain different uncertaintiesfor each parameter in a sample of ionized gaseous nebulae with accurate determinations of chemical abundances in thewhole range of metallicity.In Fig. 8, we show the total abundances as derivedfrom several strong-line empirical methods (with their cor-responding errors estimated taking into account the errorsof the line intensities and also the errors given by the cali-brations of the empirical parameters) and the oxygen abun-dances calculated from the electron temperatures measuredusing the direct method and those estimated from the pho-toionization models for each knot.Among the available strong-line empirical parameterswe studied the O parameter (also known as R and orig-inally defined by Pagel et al. (1979) and based on [O ii ] and[O iii ] strong emission lines). This parameter is character-ized by its double-valued relation with metallicity, with avery large dispersion in the turnover region. According tothe values measured, we used the McGaugh (1991) calibra-tion for the lower branch. For knots B and C, this calibratorfails to predict the value obtained with the direct method,overestimating the oxygen abundance, although the derivedvalues are very similar if we take into account the obser-vational errors and the large spread in the empirical O diagram (see Fig. 3 of P´erez-Montero & D´ıaz 2005).The N2 parameter (defined by Storchi-Bergmann et al.1994) is based on the strong emission lines of [N ii ]. It remainssingle-valued up to high metallicities in its relation to oxy-gen abundance, and it is almost independent of reddeningand flux calibrations. Nevertheless, it has the high disper-sion associated to the functional parameters of the nebula(ionization parameter and ionizing radiation temperature)and to N/O variations. We used the empirical calibrationof this parameter from Denicol´o et al. (2002) to derive theoxygen abundance in the three star-forming knots of this c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 galaxy. We can see in Fig. 8 that N2 behaves similarly toO in predicting the abundances.The parameter O3N2, defined by Alloin et al. (1979),depends on strong emission lines of [O iii ] and [N ii ]. We usedthe calibration due to Pettini & Pagel (2004) and, as we cansee in Fig. 8 it has a very similar behaviour to that of N2.The S parameter was defined by V´ılchez & Esteban(1996) and is based on the strong emission lines of [S ii ]and [S iii ]. The calibration by P´erez-Montero & D´ıaz (2005)yields comparable oxygen abundances for the three observedknots that are in turn in very good agreement with the abun-dances derived using the direct method for knots A and B,and slightly higher for knot C, but still consistent within theerrors.The ratio of the S and O parameters as S /O (D´ıaz & P´erez-Montero 2000) is a parameter that increasesmonotonically with the oxygen abundance up to the overso-lar regime and is very useful to study variations over wideranges of metallicity ( e.g. disks). We applied the calibrationfrom P´erez-Montero & D´ıaz (2005) and found a good con-cordance with the values determined by the direct method.The Ar O parameter, defined and calibrated byStasi´nska (2006) as the ratio of [Ar iii ] λ iii ] λ parameter and in better agreement with those deriveddirectly.Different strong line empirical metallicity calibratorsare commonly used to estimate the oxygen abundances inobjects for which direct derivation of electron temperaturesis not possible. However, as illustrated in Fig. 8, differentempirical calibrations give results that are not in completeagreement with direct measurements, and the goodness ofthe result even changes between knots when using the samecalibrator. P´erez-Montero et al. (2009) found a very simi-lar behaviour for the star-forming knots of IIZw71, exceptfor the Ar O parameter, which is a better estimator of theoxygen abundances for the knots in SDSS J1657. For twoknots in Mrk 1418, Cair´os et al. (2009) derived the oxy-gen abundances from the observed [O iii ] λ parameter). The study of the stellar content of our objects was car-ried out using the STARLIGHT code, which calculates thecombination of stellar libraries and the extinction law thatreproduces the spectral energy distribution, to derive theproperties of the stellar population in each of the knots.STARLIGHT fits an observed continuum spectral energydistribution using a combination of multiple simple stellarpopulations (SSPs; also known as instantaneous bursts) syn-thetic spectra using a χ minimization procedure. For con-sistency with the photoionization models used to model thegas of knot A in P´erez-Montero et al. (2010), we have used in Table 8.
Values of the extinction, total stellar mass, and fractionof the mass in stars younger than 10 Myr for each knot of SDSSJ1657 in the best-fit model using STARLIGHT.ID A(V) c(H β ) M ∗ M ion f(age <
10 Myr)(mag) (%)Knot A 0.18 0.08 280 1.6 0.58Knot B 0.00 0.00 100 0.2 0.21Knot C 0.00 0.00 140 0.4 0.28Note. Masses in 10 M ⊙ . A li g h t f r a c t i o n ( % ) B log(Age) C Figure 9.
Histogram of the distribution in visual light of themost probable stellar population models fitted by STARLIGHTfor knots A, B and C, as labelled. the fitting a synthetic stellar population obtained using Star-burst99 (Leitherer et al. 1999; V´azquez & Leitherer 2005)with the Geneva stellar evolutionary tracks for continuousstar formation with high mass loss (Meynet et al. 1994), theKroupa Initial Mass Function (IMF; Kroupa 2002) in twointervals (0.1-0.5 and 0.5-100 M ⊙ ) with different exponents(1.3 and 2.3, respectively), the theoretical wind model (Lei-therer et al. 1992), with the model atmospheres from Smithet al. (2002), and the stellar cluster metallicity being theclosest to the nebular one, Z = 0.004 (= 1/5 Z ⊙ , see PaperII and P´erez-Montero et al. 2010). The STARLIGHT codesolves simultaneously the ages and relative contributions ofthe different SSPs and the average reddening. The reddeninglaw from Cardelli et al. (1989) is used. Prior to the fittingprocedure, the spectra were shifted to the rest-frame, andre-sampled to a wavelength interval of 1 ˚A in the entire wave-length range between 3500 ˚A and 9000 ˚A by interpolationconserving flux, as required by the program. Bad pixels andemission lines were excluded from the final fits.In Fig. 9 we show the age distribution of the visual lightfraction for the individual knots. All of them present a veryyoung stellar population with ages around 10 Myr, responsi-ble for the ionization of the surrounding gas. Practically allthe mass of the knots seems to come from a very old popula-tion of about 8.3 Gyr. However, this result is puzzling given c (cid:13) , 000–000 G. F. H¨agele et al. that no absorption metal lines characteristic of old stellarpopulations, such as Mg ii λ ii H λ ii K λ ii Triplet λλ β )] estimated by the model. Noaperture correction factors have been taken into account forthe H α luminosities, due to the compact nature of the ob-jects. Indeed, the discrepancy factors between our H α lumi-nosities for Knot A and those measured in the SDSS catalogusing a 3 arcsec fiber is no larger than 1.3.We do not expect to find the same values of extinctionin the gas and the stellar population. In this galaxy, althoughthe extinction is larger in knot A than in B and C (the op-posite from the derived values using the Balmer decrement)the general result is consistent with low extinction.The mean ages and stellar mass fraction (with respectto the total stellar mass of the corresponding cluster) of theionizing stellar population fitting by STARLIGHT for eachknot is very similar, in very good agreement with the ion-ization structure derived in § Some properties of the emission knots can be obtained fromthe measured H α flux, such as H α luminosity, number ofionizing photons, mass of ionizing stars and mass of ionizedhydrogen (see e.g. D´ıaz et al. 2000). The observed H α fluxwas corrected in each knot for reddening using the valuesof the reddening constants, c(H β ), given in Table 3. We as-sume a distance to SDSS J1657 of 161.2 Mpc (Mould et al.2000). Besides, we have used the dust absorption factors, f d derived using photoionization models to correct these quan-tities. These factors are independent of reddening correctionbecause they affect above all the number of ionizing photonsemitted by the stellar cluster. As we can see in Table 7, thesefactors are larger in knots B and C than in knot A, consis-tent with the differences found in the reddening as derivedby measuring the Balmer decrement. These differences inreddening are not surprising attending to the different agesof the ionizing clusters found in the same photoionizationmodels (see § Table 9.
Derived properties of the observed knots using the ex-tinction and dust absorption corrected measured H α fluxes.ID F(H α ) L(H α ) Q(H ) M ion M(H ii ) SFRKnot A 28.4 8.3 6.3 3.6 2.1 0.697Knot B 15.2 4.8 3.3 1.5 3.5 0.375Knot C 23.0 7.0 5.3 2.2 5.3 0.564Note. Fluxes in 10 − erg s − cm − , luminosities in10 erg s − , ionizing photons in 10 photons s − , massesin 10 M ⊙ , and SFR in M ⊙ yr − . have not already broken the dust cocoon usually shroud-ing these very young star formation knots. The derived andcorrected values are given in Table 9.The derived values of the masses of the ionizing clusters( i.e. with an age younger than 10 Myr) can be comparedwith those provided by the STARLIGHT fit. For knot ASTARLIGHT gives a slightly lower value than the one wederive by a factor of 2, while for knot B and C this factoris larger, about 7 and 5, respectively. If we take the massesderived from the H α fluxes, which are, in principle, lowerlimits since a possible escape of photons is not taken intoaccount, and we use the calculated proportions of young tototal mass given in Table 8 we obtain total masses of about7 × M ⊙ for each knot.The weight of the underlying stellar population in thevisual light is larger in knot A than in knots B and C. Thecorrected EW(H β ) is 12% larger in knot A but less than 1%in the other two knots.The star formation rate (SFR) for each knot was de-rived from the H α luminosity using the expression given byKennicutt (1998), SFR = 7.9 × − × L(H α ). The derivedvalues are also given in Table 9, taking into account dustabsorption factors obtained from our models as discussed inprevious sections. The SFR obtained range from 0.375 M ⊙ yr − for knot B to 0.697 M ⊙ yr − for knot A. In Fig. 10 we show the longslit spectrum between 5030and 5220 ˚A (upper panel) and between 6690 and 7110 ˚A(middle panel). These spectral ranges contain H β and[O iii ] 4959,5007 ˚A, and H α , [N ii ] 6548,6584 ˚A, He i ii ] 6717,6731 ˚A emission lines, respectively. All theselines show a similar behaviour in the position-velocity (wave-length) plane, with the typical helical shape of the rotationcurve expected for a rotating disc. Knot B is redshifted whileKnot C is blueshifted with respect to the systemic velocityof knot A. We have analyzed the differential radial velocityfield along the slit using the H α and [O iii ] emission lines.The radial velocities derived for each pixel (0.56 arcsecs) aredisplayed in the lower panel of Fig. 10. We have also plot-ted the H α spatial profile. The velocities derived from bothemission lines are in good agreement, giving the same valuewithin the observational errors, which are about 20 km/s,taking into account the errors in the wavelength calibrationand a single Gaussian fitting. Larger differences ( ∼
30 km/s)between the derived velocities using the different emissionlines have been found around the secondary knots, B andC, but they are also within the observational errors. Fromthe data for Knot A, we have estimated a redshift of 0.0383 c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 -10 0 10distance (arcsecs)113501140011450115001155011600 v e l o c it y ( k m / s ) H α F l ux ( e r g s - c m - ) [OIII] 5007H α [OIII] 5007 KnotsH α KnotsKnot AKnot B Knot C
Figure 10.
Longslit spectrum between 5030 and 5220 ˚A (upperpanel) and between 6690 and 7110 ˚A (middle panel); the mainemission lines and the knots are labelled. The derived rotationcurve of SDSS J1657 is shown in the lower panel. Green circlesrepresent the radial velocities derived from H α and yellow dia-monds from [O iii ] 5007 ˚A for each pixel (0.56 arcsec) along theslit. Red squares and blue upward triangles represent the radialvelocities derived from H α and [O iii ] using the extracted spectrumfor each knot. The dashed line represents the spatial distributionof the H α flux to be compared with the rotation curve. (v r ≈ α and [O iii ]lines are in good agreement showing a similar behaviour,which leads us to suppose that this depression is a real de-viation from the circular motion. This could be due to thepresence of an expanding bubble or shell (or superposition ofmore than one) of ionized gas approaching us with respect tothe systemic velocity of the galaxy defined by the velocity ofthe brightest and most massive star-forming cluster, KnotA. In this case, we could only see the expanding materialwhich is moving in our direction, since we detect a system-atic variation of the radial velocity, but not a broadeningof the emission line with a systemic velocity in accordancewith that expected for the circular motion. Another pos-sibility would be that the ionized gas related to this weakemission area is affected by a possible interaction with a tailof emission gas located at the south of the Knots or with theredder galaxy observed to the north-east (see Fig. 1 and dis-cussion above). This interaction could be responsible for thedeviation from circular motion, as well as the fact that theionizing population of the different star-forming knots seemto have similar ages (see § r = 100 km/s given by the difference between KnotA and the furthest point where we can measure the emissionlines with good enough S/N, and considering an optical ra-dius of 14 arcsec for this point (equivalent to Ropt = 11 Kpcat the adopted distance of 161.2 Mpc for SDSS J1657, witha scale of 782 pc arcsec − ) we have estimated a dynamicalmass inside this radius of 2.5 × M ⊙ .Since the three star-forming knots clearly dominate theluminosity of the galaxy (see Fig. 1), we have derived theirblue magnitudes from the SDSS values (see Table 1). Weadopt these values as representative of the total light, al-though we know that they are (together) an upper limit tothe total magnitude of the galaxy since we are not takinginto account the light outside the 3 arcsec SDSS fiber (seeFig. 1). We have obtained the B magnitude from the Sloang and r photometric magnitudes using the transformationequation from Chonis & Gaskell (2008), B = g+0.327 × (g-r)+0.216, given BA=17.15, BB=20.75, and BC=19.67. Wehave obtained a total blue luminosity LB = 6.2 × L ⊙ ,which gives an upper estimation of 4 for M/LB, in agreementto what is found by Faber & Gallagher (1979) for irregulargalaxies. c (cid:13) , 000–000 G. F. H¨agele et al.
The star formation processes in H ii galaxies are known to oc-cur in low density environments. Spectrophotometric obser-vations of SDSS J1657 were carried out to study the physicalproperties of its individual bursts of star formation. In thethree individualized star-forming regions the electron den-sities were found to be well below the critical density forcollisional de-excitation.We extracted information on the three knots of SDSSJ1657, labelled A, B and C. For all of them we measuredfour line temperatures: T e ([O iii ]), T e ([S iii ]), T e ([O ii ]), andT e ([S ii ]), reaching high precision, with rms fractional errorsof the order of 2%, 5%, 6% and 6%, respectively, for Knot A,and slightly worse 3%, 8%, 6% and 17%, respectively, for thefainter knots B and C. The [O iii ] temperature of knot A wasfound to be about 2000 K lower than for the other two knots.For the [O ii ] temperature the estimated values for the threeknots are the same within the errors. The [S iii ] temperaturesshow a difference of 3500 K between the values derived forKnots B and C, and the estimated value for Knot A fallsin between. However, these large differences are compatiblewith the empirical relation between T e ([O iii ]) and T e ([S iii ])found in Paper I, within the observational errors and theempirical dispersion of that relation. Within observationalerrors, the [S ii ] temperature derived is similar for the threeknots.The temperature measurements allowed the directderivation of ionic abundances of oxygen, sulphur, nitrogen,neon and argon. The total abundances of these species are inthe same range of metallicities measured in H ii galaxies, be-tween 7.78 and 7.99, with estimated errors of about 0.05 dex.Knots B and C show similar abundances, while the value forKnot A is about 0.2 dex higher. This behaviour is mirroredby the N/O ratio. Knots A and B have similar Ar/O whileKnot C is the one with a different value. Within obseva-tional errors, S/O is almost the same for the three knots.The Ne/O ratio is remarkably similar for all three regions.We have studied the underlying and ionizing stellar pop-ulations for the three regions and modelled the propertiesof the emitting gas using a photoionization code. The es-timated electron temperatures and ionic abundances usingthe direct method are well reproduced by the models exceptfor those of S + /H + . This could be due to the presence ofdiffuse gas in these star-forming regions, which is not takeninto account in the models. Regarding the hardness of theradiation field, model predictions agree with observationswhen the softness parameter η which parametrizes the tem-perature of the ionizing radiation, is expressed in terms ofemission lines intensity ratios. However, there is a disagree-ment when ionic abundance ratios are used instead. This isprobably caused by the overestimate of the electron temper-ature of S + by the models and the corresponding underes-timate of the S + /S abundance ratio rather than by theexistence of an additional heating source in these knots.We have also estimated the total oxygen abundances bymeans of different strong-line empirical parameters. In allcases, the estimated abundances are consistent with thosederived by the direct method with the parameter involvingthe sulphur lines providing the closest fit. The rest of theparameters slightly overestimate the oxygen abundance.The star formation history for the three knots derived from the fitting of multiple simple stellar populations usingSTARLIGHT are remarkably similar, with an old popula-tion of about 8 Gyr presenting more than the 99% contri-bution to the mass fraction. This result is somewhat unex-pected and data on its surface brightness distribution wouldbe very valuable to explore further this finding. Since the agedistributions of the ionizing population among the differentknots of SDSS J1657 seem to be similar, this could indicatea common evolutionary stage of this population which isprobably related to a process of interaction with a compan-ion galaxy that triggered the star formation in the differ-ent knots at about the same time. Besides, this interactioncould be responsable for the deviation from the circular mo-tion shown by the rotation curve, which is basically linearin the region where we can measure the emission lines witha good enough S/N ratio. However, this deviation could bealso related to an expanding bubble or shell of ionized gasapproaching us with a velocity of 50 km/s with respect tothe predicted velocity from the rotation curve.The ionization structure mapped through the use of the η and η ′ diagrams derived from our observations shows verysimilar values within the errors for all the knots. This factimplies that the equivalent effective temperatures of the ion-ization radiation fields are very similar for all the studiedregions, in spite of some small differences in the ionizationstate of different elements.Finally, we have derived a dynamical mass of2.5 × M ⊙ from the rotation curve with a mean valueof 100 km/s at an radius of 11 Kpc. The total mass of theyoung clusters derived for the three star-forming knots us-ing the H α luminosities is 7.3 × M ⊙ , making up a smallfraction of the total dynamical mass, about 0.03 %. We haveestimated an upper limit of about 4 for the ratio M/LB.Data on the surface brightness distribution using broad andnarrow band images would be very valuable to explore fur-ther this result. ACKNOWLEDGMENTS c (cid:13) , 000–000 bundance of multiple knots in SDSS J1657 sity of Portsmouth, Princeton University, the United StatesNaval Observatory, and the University of Washington.This research has made use of the NASA/IPAC Ex-tragalactic Database (NED) which is operated by the JetPropulsion Laboratory, California Institute of Technology,under contract with the National Aeronautics and SpaceAdministration and of the SIMBAD database, operated atCDS, Strasbourg, France.Financial support for this work has been provided bythe Spanish Ministerio de Educaci´on y Ciencia (AYA2007-67965-C03-03 and 02). Partial support from the Comu-nidad de Madrid under grant S2009/ESP-1496 (ASTRO-MADRID) is acknowledged. EPM also thank to projectJunta de Andaluc´ıa TIC 114. ET and RT are grateful tothe Mexican Research Council (CONACYT) for support un-der grants CB2005-01-49847F and CB2008-01-103365. VFwould like to thank the hospitality of the AstrophysicsGroup of the UAM during the completion of this work.
REFERENCES
Allende-Prieto C., Lambert D. L., Asplund M., 2001, ApJ,556, L63Alloin D., Collin-Souffrin S., Joly M., Vigroux L., 1979,A&A, 78, 200Aloisi A., Tosi M., Greggio L., 1999, AJ, 118, 302Baldwin J. A., Phillips M. M., Terlevich R., 1981, PASP,93, 5Barker T., 1980, ApJ, 240, 99Benjamin R. A., Skillman E. D., Smits D. P., 1999, ApJ,514, 307Bravo-Alfaro H., Brinks E., Baker A. J., Walter F., KunthD., 2004, AJ, 127, 264Bravo-Alfaro H., Coziol R., Brinks E., 2006, Revista Mex-icana de Astronomia y Astrofisica, 42, 261Cair´os L. M., Caon N., Papaderos P., Kehrig C., WeilbacherP., Roth M. M., Zurita C., 2009, ApJ, 707, 1676Cair´os L. M., Caon N., Zurita C., Kehrig C., Roth M.,Weilbacher P., 2010, A&A, 520, A90+Cair´os L. M., Caon N., Zurita C., Kehrig C., WeilbacherP., Roth M., 2009, A&A, 507, 1291Cardelli J. A., Clayton G. C., Mathis J. S., 1989, ApJ, 345,245Chonis T. S., Gaskell C. M., 2008, AJ, 135, 264Cid Fernandes R., Mateus A., Sodr´e L., Stasi´nska G.,Gomes J. M., 2005, MNRAS, 358, 363Cumming R. J., Fathi K., ¨Ostlin G., Marquart T., M´arquezI., Masegosa J., Bergvall N., Amram P., 2008, A&A, 479,725de Robertis M. M., Dufour R. J., Hunt R. W., 1987, RoyalAstronomical Society of Canada, Journal, 81, 195Denicol´o G., Terlevich R., Terlevich E., 2002, MNRAS, 330,69D´ıaz A. I., 1988, MNRAS, 231, 57D´ıaz A. I., Castellanos M., Terlevich E., Luisa Garc´ıa-Vargas M., 2000, MNRAS, 318, 462D´ıaz A. I., P´erez-Montero E., 2000, MNRAS, 312, 130D´ıaz A. I., Terlevich E., Castellanos M., H¨agele G. F., 2007,MNRAS, 382, 251Faber S. M., Gallagher J. S., 1979, ARA&A, 17, 135 Ferland G. J., Korista K. T., Verner D. A., Ferguson J. W.,Kingdon J. B., Verner E. M., 1998, PASP, 110, 761Firpo V., Bosch G., H¨agele G. F., D´ıaz A. I., Morrell N.,2011, MNRAS, in preparationFirpo V., Bosch G., H¨agele G. F., Morrell N., 2010, MN-RAS, 406, 1094Garc´ıa-Benito R., 2009, PhD thesis, Universidad Au-tonoma de MadridGarc´ıa-Benito R., D´ıaz A., H¨agele G. F., P´erez-Montero E.,L´opez J., V´ılchez J. M., P´erez E., Terlevich E., TerlevichR., Rosa-Gonz´alez D., 2010, MNRAS, pp 1242–+Garc´ıa-Lorenzo B., Cair´os L. M., Caon N., Monreal-IberoA., Kehrig C., 2008, ApJ, 677, 201Garnett D. R., 1992, AJ, 103, 1330Garnett D. R., Skillman E. D., Dufour R. J., Shields G. A.,1997, ApJ, 481, 174Gonz´alez-Delgado R. M., Cervi˜no M., Martins L. P., Lei-therer C., Hauschildt P. H., 2005, MNRAS, 357, 945Gonz´alez-Delgado R. M., P´erez E., Tenorio-Tagle G.,V´ılchez J. M., Terlevich E., Terlevich R., Telles E.,Rodr´ıguez-Espinosa J. M., Mas-Hesse M., Garc´ıa-VargasM. L., D´ıaz A. I., Cepa J., Casta˜neda H., 1994, ApJ, 437,239Grevesse N., Sauval A. J., 1998, Space Science Reviews, 85,161H¨agele G. F., 2008, PhD thesis, Universidad Aut´onoma deMadridH¨agele G. F., D´ıaz A. I., Cardaci M. V., Terlevich E., Ter-levich R., 2007, MNRAS, 378, 163H¨agele G. F., D´ıaz A. I., Cardaci M. V., Terlevich E., Ter-levich R., 2009, MNRAS, 396, 2295H¨agele G. F., D´ıaz A. I., Cardaci M. V., Terlevich E., Ter-levich R., 2010, MNRAS, 402, 1005H¨agele G. F., D´ıaz A. I., Terlevich E., Terlevich R., P´erez-Montero E., Cardaci M. V., 2008, MNRAS, 383, 209 (Pa-per II)H¨agele G. F., P´erez-Montero E., D´ıaz A. I., Terlevich E.,Terlevich R., 2006, MNRAS, 372, 293 (Paper I)Heckman T. M., 1997, in Holt S. S., Mundy L. G., eds,American Institute of Physics Conference Series Vol. 393of American Institute of Physics Conference Series, Star-bursts and Cosmogony. p. 271Hodge P. W., 1969, ApJ, 156, 847Hoyos C., D´ıaz A. I., 2006, MNRAS, 365, 454Hunter D. A., Thronson Jr. H. A., 1995, ApJ, 452, 238Izotov Y. I., Lipovetsky V. A., Chaffee F. H., Foltz C. B.,Guseva N. G., Kniazev A. Y., 1997, ApJ, 476, 698Izotov Y. I., Stasi´nska G., Guseva N. G., Thuan T. X.,2004, A&A, 415, 87 (IT04)Johnson K. E., Indebetouw R., Watson C., KobulnickyH. A., 2004, AJ, 128, 610Kehrig C., V´ılchez J. M., S´anchez S. F., Telles E., P´erez-Montero E., Mart´ın-Gord´on D., 2008, A&A, 477, 813Kennicutt Jr. R. C., 1998, ARA&A, 36, 189Kingdon J., Ferland G. J., 1995, ApJ, 442, 714Kroupa P., 2002, Science, 295, 82Kunth D., Sargent W. L. W., 1983, ApJ, 273, 81Legrand F., Kunth D., Roy J., Mas-Hesse J. M., WalshJ. R., 2000, A&A, 355, 891Leitherer C., Robert C., Drissen L., 1992, ApJ, 401, 596Leitherer C., Schaerer D., Goldader J. D., DelgadoR. M. G., Robert C., Kune D. F., de Mello D. F., De- c (cid:13) , 000–000 G. F. H¨agele et al. vost D., Heckman T. M., 1999, ApJS, 123, 3L´opez J., 2005, MSc Thesis. INAOEL´opez-S´anchez A. R., 2010, A&A, 521, A63+L´opez-S´anchez A. R., Esteban C., 2009, A&A, 508, 615L´opez-S´anchez A. R., Esteban C., 2010a, A&A, 516, A104+L´opez-S´anchez A. R., Esteban C., 2010b, A&A, 517, A85+Madau P., Ferguson H. C., Dickinson M. E., Giavalisco M.,Steidel C. C., Fruchter A., 1996, MNRAS, 283, 1388McGaugh S. S., 1991, ApJ, 380, 140Melnick J., Terlevich R., Eggleton P. P., 1985, MNRAS,216, 255Meynet G., Maeder A., Schaller G., Schaerer D., Charbon-nel C., 1994, A&AS, 103, 97Miller J. S., Mathews W. G., 1972, ApJ, 172, 593Monreal-Ibero A., V´ılchez J. M., Walsh J. R., Mu˜noz-Tu˜n´on C., 2010, A&A, 517, A27+Mould J. R., Huchra J. P., Freedman W. L., Kennicutt Jr.R. C., Ferrarese L., Ford H. C., Gibson B. K., GrahamJ. A., Hughes S. M. G., Illingworth G. D., Kelson D. D.,Macri L. M., Madore B. F., Sakai S., Sebo K. M., Silber-mann N. A., Stetson P. B., 2000, ApJ, 529, 786Noeske K. G., Iglesias-P´aramo J., V´ılchez J. M., PapaderosP., Fricke K. J., 2001, A&A, 371, 806Olive K. A., Skillman E. D., 2001, New Astronomy, 6, 119Olive K. A., Skillman E. D., 2004, ApJ, 617, 29¨Ostlin G., 2000, ApJ, 535, L99¨Ostlin G., Cumming R. J., Amram P., Bergvall N., KunthD., M´arquez I., Masegosa J., Zackrisson E., 2004, A&A,419, L43Pagel B. E. J., Edmunds M. G., Blackwell D. E., ChunM. S., Smith G., 1979, MNRAS, 189, 95Peimbert M., Costero R., 1969, Boletin de los Observato-rios Tonantzintla y Tacubaya, 5, 3P´erez-Gallego J., Guzm´an R., Castillo-Morales A., Cas-tander F. J., Gallego J., Garland C. A., Gruel N., PisanoD. J., S´anchez S. F., Zamorano J., 2010, MNRAS, 402,1397P´erez-Montero E., D´ıaz A. I., 2003, MNRAS, 346, 105P´erez-Montero E., D´ıaz A. I., 2005, MNRAS, 361, 1063P´erez-Montero E., D´ıaz A. I., V´ılchez J. M., Kehrig C.,2006, A&A, 449, 193P´erez-Montero E., Garc´ıa-Benito R., D´ıaz A. I., P´erez E.,Kehrig C., 2009, A&A, 497, 53P´erez-Montero E., Garc´ıa-Benito R., H¨agele G. F., D´ıazA. I., 2010, MNRAS, 404, 2037P´erez-Montero E., H¨agele G. F., Contini T., D´ıaz A. I.,2007, MNRAS, 381, 125P´erez-Montero E., V´ılchez J. M., Cedr´es B., H¨agele G. F.,Moll´a M., Garc´ıa-Benito R., D´ıaz A. I., Kehrig C., Mart´ın-Gord´on D., 2011, A&A, in preparationPettini M., Pagel B. E. J., 2004, MNRAS, 348, L59Pustilnik S. A., Kniazev A. Y., Lipovetsky V. A., Ugryu-mov A. V., 2001, A&A, 373, 24Rela˜no M., Monreal-Ibero A., V´ılchez J. M., KennicuttR. C., 2010, MNRAS, 402, 1635Rosales-Ortega F. F., Kennicutt R. C., S´anchez S. F., D´ıazA. I., Pasquali A., Johnson B. D., Hao C. N., 2010, MN-RAS, 405, 735S´anchez S. F., Rosales-Ortega F. F., Kennicutt R. C., John-son B. D., Diaz A. I., Pasquali A., Hao C. N., 2010, MN-RAS, pp 1474–+ Searle L., Sargent W. L. W., Bagnuolo W. G., 1973, ApJ,179, 427Shaw R. A., Dufour R. J., 1995, PASP, 107, 896Smith L. J., Norris R. P. F., Crowther P. A., 2002, MNRAS,337, 1309Stasi´nska G., 1978, A&A, 66, 257Stasi´nska G., 2006, A&A, 454, L127Storchi-Bergmann T., Calzetti D., Kinney A. L., 1994,ApJ, 429, 572Storey P. J., Hummer D. G., 1995, MNRAS, 272, 41Taylor C., Brinks E., Skillman E. D., 1993, AJ, 105, 128Taylor C. L., Brinks E., Grashuis R. M., Skillman E. D.,1995, ApJS, 99, 427Taylor C. L., Thomas D. L., Brinks E., Skillman E. D.,1996, ApJS, 107, 143Telles E., Terlevich R., 1997, MNRAS, 286, 183Tenorio-Tagle G., 1996, AJ, 111, 1641Terlevich R., 1997, in Franco J., Terlevich R., Serrano A.,eds, Revista Mexicana de Astronomia y Astrofisica Con-ference Series Vol. 6 of Revista Mexicana de Astronomiay Astrofisica Conference Series, Starburst Galaxies – theDawn of a New ERA. pp 1–+Terlevich R., Melnick J., 1981, MNRAS, 195, 839Terlevich R., Melnick J., Masegosa J., Moles M., CopettiM. V. F., 1991, A&AS, 91, 285 (T91)Thuan T. X., Hibbard J. E., L´evrier F., 2004, AJ, 128, 617V´azquez G. A., Leitherer C., 2005, ApJ, 621, 695V´ılchez J. M., Esteban C., 1996, MNRAS, 280, 720V´ılchez J. M., Iglesias-P´aramo J., 1998, ApJ, 508, 248V´ılchez J. M., Pagel B. E. J., 1988, MNRAS, 231, 257Weedman D. W., Feldman F. R., Balzano V. A., RamseyL. W., Sramek R. A., Wuu C., 1981, ApJ, 248, 105 c (cid:13)000