Oxygen Abundance Measurements of SHIELD Galaxies
Nathalie C. Haurberg, John J. Salzer, John M. Cannon, Melissa V. Marshall
aa r X i v : . [ a s t r o - ph . GA ] J a n Oxygen Abundance Measurements of SHIELD Galaxies
Nathalie C. Haurberg , , , John J. Salzer , , John M. Cannon , , & Melissa V. Marshall ABSTRACT
We have derived oxygen abundances for 8 galaxies from the Survey for H I in ex-tremely low-mass dwarfs (SHIELD). The SHIELD survey is an ongoing study of verylow-mass galaxies, with M HI between 10 . and 10 . M ⊙ , that were detected by theArecibo Legacy Fast ALFA (ALFALFA) survey. H α images from the WIYN 3.5m tele-scope show that these 8 SHIELD galaxies each possess one or two active star-formingregions which were targeted with long-slit spectral observations using the Mayall 4mtelescope at KPNO. We obtained a direct measurement of the electron temperature bydetection of the weak [O III ] λ II regions. Oxygen abundances forthe other H II regions were estimated using a strong-line method. When the SHIELDgalaxies are plotted on a B -band luminosity-metallicity diagram they appear to sug-gest a slightly shallower slope to the relationship than normally seen. However, thatoffset is systematically reduced when the near-infrared luminosity is used instead. Thisindicates a different mass-to-light ratio for the galaxies in this sample and we suggestthis may be indicative of differing star-formation histories in the lowest luminosity andsurface brightness dwarf irregulars. Subject headings: galaxies: abundances — galaxies: dwarf — galaxies: evolution —galaxies: star formation
1. INTRODUCTION
Chemical abundance studies of the interstellar medium in galaxies allow a glimpse into the star-formation and chemical enrichment history of these galaxies. The chemical evolution of local-universe galaxies with very low-metallicity gas is particularly important as these galaxies may beobjects of cosmological significance. These metal-poor systems represent potential local-universeanalogs to galaxies in the early universe and can provide a better understanding of the processes ofstar-formation and gas-enrichment in the early universe. Additionally, in order for very metal-poor Department of Astronomy, Indiana University, 727 E. Third St., Bloomington, IN 47405; [email protected], [email protected] Physics Department, Knox College, 2 E. South St., Galesburg, IL 61401; [email protected] Department of Physics & Astronomy, Macalester College, 1600 Grand Avenue, Saint Paul, MN 55105; [email protected] Visiting Astronomer, Kitt Peak National Observatory, National Optical Astronomy Observatory, which is oper-ated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with theNational Science Foundation. ≤ II regions (i.e., systems that have recently undergone a very strong starburst) and are relativelyinsensitive to isolated low-mass, relatively quiescent star-forming galaxies.A promising alternative method for finding XMD galaxies has been to study very low-luminositydwarfs since there is an historically well-established correlation between metallicity and luminos-ity among low-redshift dwarf galaxies (Lequeux et al. 1979; Skillman et al. 1989a; Pilyugin 2001;Melbourne & Salzer 2002; Tremonti et al. 2004; Salzer et al. 2005; Lee et al. 2006; van Zee & Haynes2006). This method has already led to the discovery of several XMD galaxies (Skillman et al.1989a,b; van Zee 2000; Pustilnik et al. 2005, 2011; Berg et al. 2012; Skillman et al. 2013). Despitethe fact that there should be large numbers of low-luminosity galaxies according to the galaxyluminosity function, searching for these systems is very difficult as they tend to be very low surfacebrightness and difficult to detect optically. However, blind H I surveys do not suffer optical selec-tion biases and thus can uncover low-mass H I sources that potentially have low-surface brightnessoptical counterparts that may be missed in optical studies.It is particularly important to identify and study low surface brightness XMD sources becauseour current understanding of XMD galaxies is derived almost exclusively from blue compact dwarfgalaxies (BCDs). However, normal low-luminosity star-forming galaxies should outnumber BCDsby at least an order of magnitude and thus should be more representative of the XMD population.The Arecibo Dual Beam Survey (ADBS; Rosenberg & Schneider 2000) detected many low-masssources with low-surface brightness dwarf irregular optical counterparts. Twelve of these ADBSgalaxies were selected for spectroscopic follow-up by Haurberg et al. (2013) to determine theirchemical abundances. While none of these galaxies were found to be XMD galaxies, the samplewas generally low-metallicity and showed that H I - surveys can efficiently uncover metal-poor andpotential XMD galaxies. The more recent Arecibo Legacy Fast ALFA (ALFALFA) blind H I -survey (Giovanelli et al. 2005; Haynes et al. 2011) has uncovered hundreds of low-mass H I sources,many with very low-surface brightness optical counterparts. The ALFALFA catalog provides a 3 –promising target list of potential XMD galaxies.The use of H I catalogs to search for low-luminosity XMD systems has already proven successfulwith the recent discovery of the Leo P dwarf. Leo P was selected as a candidate XMD fromthe ALFALFA catalog due to its low derived H I mass (Giovanelli et al. 2013) and very-low stellarmass inferred from optical follow-up imaging (Rhode et al. 2013). Optical spectroscopy revealed anextremely low oxygen abundance of 12 + log(O/H) = 7.17 ± I mass galaxies selectedfrom the ALFALFA catalog as part of The Survey for H I in Extremely Low-Mass Dwarfs (SHIELD).SHIELD is a comprehensive multi-wavelength project targeting 12 of the lowest-H I mass galaxiesdiscovered in the ALFALFA survey which have a clear optical counterpart (Cannon et al. 2011).The optical and H I properties of these galaxies (N. Haurberg et al. in preparation) make themprime candidates as XMDs.In Section 2 we describe the sample selection in more detail and include the derived photometricand H I properties. The spectroscopic observations are described in Section 3 and the reductionand measurement processes are outlined in Section 4. Section 5 describes how we determinedabundances for the H II regions and the abundance results. The analysis and discussion of theresults are in Section 6, and our conclusions are in Section 8.
2. SAMPLE SELECTION
The ALFALFA survey includes hundreds low-H I -mass objects with M HI < M ⊙ and additionallyhas provided the first robust sample of galaxies at the very low-mass end of the H I mass functionwith M HI < M ⊙ (Martin et al. 2010). Twelve of these galaxies, with 10 . M ⊙ < M HI < . M ⊙ , were selected for the SHIELD project (Cannon et al. 2011) from a preliminary ALFALFAcatalog. The galaxies selected for SHIELD were those with the lowest H I mass that had apparentoptical counterparts in Sloan Digitized Sky Survey (SDSS) images. These galaxies represent thelowest mass potentially star-forming galaxies that have formed a significant number of stars. Thus,following the traditional luminosity-metallicity trend (or mass-metallicity trend), the galaxies inthis sample should be some of the most metal-poor star-forming galaxies in the local universe.The SHIELD project encompasses a complex multi-wavelength data set including detailed H I gas mapping from multi-configuration Expanded Very Large Array observations (Cannon et al.2011), broadband BVR and narrowband H α imaging from the WIYN 3.5m telescope at Kitt PeakNational Observatory (KPNO; N. Haurberg et al. in preparation), 3.6 and 4.5 µ m imaging withthe InfraRed Array Camera (IRAC) on the Spitzer Space Telescope , Hubble Space Telescope Imaging with the Wide Field Camera 3 using the F606W ( ∼ R ) and F814W ( ∼ I ) filters, andlong-slit optical spectra with the Richey-Chretien Focus Spectrograph on the Mayall 4m at KPNO 4 –(this work).Since these galaxies were selected from a blind H I survey, the sample does not suffer any bias towardhigh-surface brightness objects as is often present in optical catalogs. The only optical qualificationfor galaxies in the SHIELD sample is that there be some evidence of a stellar counterpart associatedwith the H I source in SDSS images. Low H I mass sources in the ALFLFA catalog with no apparentoptical counterpart have been studied by Adams et al. (2013) and are also of great interest as thesemay represent sources with extremely low surface brightness counterparts or starless dark matterhalos.Thumbnail images showing the WIYN R -band and H α images for the 12 SHIELD galaxies areshown in Figure 1. The H II regions identified from the displayed H α images were targeted forthe long-slit spectroscopic follow-up that is the subject of this paper. Figure 1 also shows theapproximate slit locations overlaid on each galaxy. A comprehensive set of general and photometricquantities that describe the SHIELD galaxies is compiled in Table 1. Column 1 lists the galaxies inthe SHIELD sample with the coordinates for each source listed in Columns 2 and 3. The distance,calculated using HST optical imaging to determine the tip of the red giant branch (McQuinn et al.2014), is given in Column 4 and is used as the assumed distance for the derivation of all distance-dependent quantities. The absolute B -band magnitude, from the WIYN 3.5m (N. Haurberg etal. in preparation) images is listed in Column 5; the listed uncertainty includes the error in boththe distance and photometric error. The B − V color and B -band central surface brightness notcorrected for inclination effects ( µ ,B ) are listed in Columns 6 and 7, also from N. Haurberg et al (inpreparation). Columns 8 and 9 contain the 4.5 µ m and 3.6 µ m flux (in mJy) from the Spitzer SpaceTelescope observations, and Column 10 lists the stellar mass estimate derived from the Spitzerdata (Cannon, Marshall, et al. in preparation). The stellar masses presented in this table werecalculated using the method of Eskew et al. (2012). The H I masses given in Column 10 are fromCannon et al. (2011), but have been adjusted for the updated distances. Column 11 lists the star-formation rate (SFR) calculated from the H α images (N. Haurberg et al. in preparation) using thestandard prescription of Kennicutt (1998) to convert from H α luminosity to SFR.The optical images from the WIYN 3.5m telescope confirmed the optical counterparts of the The National Radio Astronomy Observatory is a facility of the National Science Foundation operated undercooperative agreement by Associated Universities, Inc. The WIYN Observatory is a joint facility of the University of Wisconsin-Madison, Indiana University, YaleUniversity, and the National Optical Astronomy Observatory. This work is based in part on observations made with the Spitzer Space Telescope, which is operated by the JetPropulsion Laboratory, California Institute of Technology under a contract with NASA. Support for this work wasprovided by NASA. Based on observations made with the NASA/ESA Hubble Space Telescope, obtained at the Space TelescopeScience Institute, which is operated by the Association of Universities for Research in Astronomy, Inc., under NASAcontract NAS 5-26555. These observations are associated with program 12658. I sources as low surface brightness dIrrs with blue colors (Table 1). The H α imag-ing revealed that the star formation in these galaxies was generally confined to 1 or 2 knots ofstar formation which were targeted with long-slit spectroscopic follow-up observations presentedin this work. Two of the SHIELD galaxies, AGC 749241 and AGC 748778 were not included inthis spectroscopic follow-up as the H α imaging revealed no star-forming nebulae. AGC 749241 hadno detected H α emission associated with it and AGC 748778 displayed only very weak, diffuse H α emission (2.6 σ above the sky) with no discrete H II regions (see Figure 1). Additionally, H II regions were detected in AGC 174585 and AGC 174605 but they were too faint to provide usefulspectra so they are not included in the following analysis. Hence, we were able to derive usefulabundance estimates for 8 of the 12 SHIELD galaxies.
3. OBSERVATIONS3.1. OPTICAL IMAGING
Optical imaging observations were performed using the Mini-Mosiac Imager on the WIYN 3.5mtelescope at Kitt Peak National Observatory (KPNO) over four nights: 2010 October 7-8 and29-30 in March 2011. The two Fall nights (Oct. 2010) had fantastic seeing ( ≈ . ′′ ) and weredone under photometric conditions, while the two nights in the Spring (Mar. 2011) had degradedseeing ( ≈ . ′′ ) and were not photometric. Short-exposure post-calibration observations for theSpring targets were performed in April of that 2011. The broadband Johnson B , V , R , and W036narrowband H α filters were used. The nominal field of view was 9.6 ′ × ′ across 2 chips and 4amplifiers.Exposure times for the broadband images in the fall were 900, 720, and 600 s for B , V , and R bands,respectively. The narrowband imaging was taken in sets of two long narrowband H α exposures (900s) with a short (180 s) R-broadband image between them. Since the conditions were less ideal inthe Spring, longer exposure times were used. The Spring broadband data were taken with 1200,720, and 900 s exposures (for B , V , and R , respectively) and 1200 s narrowband H α exposureswith a 240 s R -band exposure in between.The broadband observations times should be sufficiently long to have detected any diffuse emissionfrom the galaxies. These optical broadband images were analyzed with both isophotal fittingtechniques as well as large aperture photometry as discussed in detail in N. Haurberg et al. inpreparation. While we can not rule out that there may be an extraordinarily low surface brightnesscomponent that was not detected in our images, comparison of with the subsequent HST imaging(?????) shows now indication of such, thus we remain confident our images were sufficient atdetecting all of the optical emission in these galaxies. 6 –Fig. 1.— Thumbnail images of the SHIELD galaxies, oriented so that North points up and Eastpoints to the left. On the left of each column are short exposure R -band images and on the rightare continuum subtracted H α images from the WIYN 3.5m (N. Haurberg et al. in preparation).Each thumbnail is approximately 50 ′′ × ′′ . The approximate position of the slit is indicated forthe 8 SHIELD galaxies for which we obtained usable spectra. Those objects with no slit imagesuperimposed either had no visible H α emission or did not produce usable spectra. Table 1:: SHIELD Galaxy ProperitesGalaxy R.A. Dec Distance M B B − V µ ,B f3.6 f4.5 log( M ⋆ ) log( M HI ) log(SFR) [2000.0] [2000.0] [Mpc] [mag] [mag] [mag ′′− ] [mJy] [mJy] [ M ⊙ ] [ M ⊙ ] [ M ⊙ yr − ]AGC 748778 00:06:34.3 15:30:39 6.46 ± ± ± up. lim. AGC 112521 01:41:07.6 27:19:24 6.58 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± · · · · · · · · · ± ± ± ± ± ± ± ± · · · · · · · · · up. lim.1 McQuinn et al. (2014) derived from HST images, tip of the red giant branch; N. Haurberg et al (in preparation) derived from WIYN 3.5m observations; Cannon, Marshall, et al. in prep.; derived from Spitzer Space Telescope Images; Cannon et al. (2011), derived from radio observations and adjusted fornew distance estimate
The spectral observations presented in this work were carried out using the Mayall 4m telescope atKitt Peak National Observatory with the Richey-Chretien Focus Spectrograph and T2KA imagerover the course of 5 nights: 2012 April 17 – 19 and 2012 October 15 – 16. The KPC-10A grating(316 lines mm − ) and WG-345 blocking filter were used. The grating is blazed at 4000 ˚A giving adispersion of 2.78 ˚A pixel − and total coverage from 2850 – 8550 ˚A on the CCD. All spectra weretaken with a slit-width of 1.5 ′′ and the slit extended 324 ′′ along the spatial direction.Target sources were too dim to be seen with the acquisition cameras, so nearby bright stars wereused for blind-offsets. Offsets were checked by using multiple stars in the field, and were successfulin aligning sources within the slit. The slit was positioned as nearly along the parallactic angleas possible to avoid the effects of differential refraction through the atmosphere. Since the star-forming regions in these galaxies are relatively isolated, usually only the target source fell in theslit, but in two cases an additional emission nebula ended up in the slit as well.We observed several spectrophotometric standard stars throughout each night, which were used tocalibrate the flux scale for our spectra. The standards were selected from the lists of Oke & Gunn(1983) and Massey et al. (1988). Each night we additionally took images of HeNeAr spectral lampsfor wavelength calibration, zero-length exposure bias images, internal quartz lamps for flat fieldcalibration, and twilight sky images to correct for slit-width variations.
4. SPECTRAL REDUCTION AND MEASUREMENT
Spectral images were processed through the standard spectral reduction routines in iraf . Thebias level was determined from the overscan region in each image and the mean bias image wasused to account for any two-dimensional (2D) structure in the bias. Median combined quartz lampflats were used to account for pixel-to-pixel variations and twilight sky flats were used to createan illumination function correcting for variations in the slit width. The science images were thenprocessed through the lacos spec cosmic ray rejection routine of van Dokkum (2001) and theresultant “cleaned” images were examined carefully by eye in comparison with the original imageto ensure that emission line pixels were not rejected as cosmic rays.H II region spectra were extracted using the apall package in iraf . Since the diffuse backgroundemission was typically very weak, local sky regions that closely bracketed the emission line sourcewere selected for night-sky subtraction. The wavelength scale was determined for each night usingthe solution derived from HeNeAr lamp spectra. The spectrophotometric standard stars were thenused to create a sensitivity function that was applied to all the spectra. After the spectra were iraf is distributed by the National Optical Astronomy Observatories, which are operated by the Association ofUniversities for Research in Astronomy, Inc., under cooperative agreement with the National Science Foundation. α , H β , the [O II ] doublet, and the strong lines from the[O III ] triplet. The weak, temperature sensitive, [O
III ] λ II regions, AGC 110482 b and AGC 182595. The other extracted spectra were good quality but didnot contain an unambiguous detection of the [O III ] λ II regions from 8 galaxies. Both of the H II regions with discernible [O III ] λ splot routine in iraf . Thisroutine allows interactive setting of the continuum level at each line. To decide on the correctlocation for the continuum we used an average of the local continuum on either side of the emissionline and assumed a linear fit for for the continuum under the emission line. We generally assumedthat the continuum remained at a constant level under the emission line except for those cases wherethe lines were very broad or the slope of the continuum was steep enough to change significantlythrough the line.Since we are measuring emission from regions of relatively recent star-formation it is expected thatthe permitted Balmer lines are affected by underlying absorption from hot main-sequence stars. Theunderlying absorption was accounted for by first assuming that the equivalent width of absorptionis the same for all the Balmer lines and then using multiple Balmer emission-line ratios (H α /H β ,H γ /H β , and H δ /H β ) to simultaneously estimate the amount of underlying absorption and c(H β )(i.e. the absorption coefficient at H β ). We were able to estimate the proper absorption correctionby determining where the c(H β ) coefficients calculated from each Balmer ratio converged. Thiscorrection was applied to the hydrogen lines so that the relative Balmer intensities could be usedto accurately determine the reddening for each H II region. In cases where only one Balmer ratiowas available, an average absorption correction of 1.0 ˚A was used since most H II regions showedsome absorption that we felt needed to be accounted for. The reddening was determined usingthe reddening law of Cardelli et al. (1989) and using the temperature derived from the [O III ] lineswhen possible. If the [O
III ] λ β )for each line we measured are tabulated for each H II region in Table 2, along with the as-sumed equivalent width of underlying absorption (W abs ), c(H β ), and the assumed T e and n e .The electron temperature ( T e ) and electron density (n e ) of the nebula were estimated using the[O III ] λ / [O III ] λ λ , II ] λ λ , III ] λ II ] λ λ , − was assumed. This density is consistent with what is commonly observed for H II regions.The errors given on the line fluxes were calculated by fully propagating the derived errors from therelevant quantities (e.g., rms of the continuum, rms scatter in flux calibration, etc.).
5. ABUNDANCE DETERMINATION
The emission line ratio of [N II ] λ / H α and [O III ] λ / H β can be used to plot each H II -regionon a traditional diagnostic diagram (e.g., Baldwin et al. 1981; Veilleux & Osterbrock 1987) shownin Figure 3, which provides a rough diagnostic for excitation and temperature (thus metallicity)of emission nebulae. SHIELD galaxies are plotted on Figure 3 as solid black points; those withcircles around them are the two H II regions which displayed a measurable [O III ] λ II regions in low H I mass dIrr galaxies from the Arecibo Dual BeamSurvey (ADBS) which is a similarly selected, but more luminous, sample (Haurberg et al. 2013),and the grey points are galaxies from the KPNO International Spectroscopic Survey (KISS) (e.g.,Salzer et al. 2000; Melbourne & Salzer 2002; Salzer et al. 2005). The dashed line differentiatesactive galactic nuclei (AGN) from star-forming nebulae (Kauffmann et al. 2003) and the solid linerepresents the locus of high excitation star-forming nebulae from the models of Dopita & Evans(1986), which increase smoothly in metallicity from the upper left to the lower right.It is clear that H II regions from the SHIELD galaxies lie below and to the left of the majority ofthe KISS sample and the theoretical high-excitation star-forming model curve. This is consistentwith the manner in which these galaxies were selected and indicates that the observed H II regionsin the SHIELD galaxies are generally low-excitation systems. Emission-line surveys like KISS aremuch less likely to include these types of low-excitation systems because the key emission lines usedfor selection in those surveys need to be strong, high equivalent-width lines in order to be detected.The relatively low-excitation of these systems indicates that either the initial mass function (IMF)was not fully populated or the most massive stars formed in these regions have already evolved offthe main sequence. Both possibilities are consistent with the expectation for very low-luminositygalaxies as either would lead to reduced luminosity and surface brightness in comparison to themore well-studied low-metallicity BCDs.Three of the H II regions that we observed (AGC 110482 b , AGC 112521, AGC 111977) had suchweak [N II ] λ II ] λ / H α .Thus, these three H II regions have been plotted with arrows indicating the measured [N II ] λ / H α is an upper limit. The lack of distinguishable [N II ] λ II regions from the SHIELD galaxies alloccupy the region of the diagram expected for low-metallicity star-forming systems. However, theH II regions spread over a significant portion of the low-metallicity region while the derived abun-dances in Table 3 cover a quite narrow range. This is consistent with the finding of Haurberg et al. 11 –Fig. 2.— Four examples of spectra from H II regions in the SHIELD galaxies. The first two displaya measurable [O III ] λ -2.5 -2 -1.5 -1 -0.5 0-0.500.51 Fig. 3.— Emission-line diagnostic diagram with each H II region plotted separately. The blackpoints are from the SHIELD galaxies in this work, those with outer circles are H II regions for whichwe obtained a measurement of the temperature sensitive [O III ] λ Table 2:: Line Measurement ResultsLine Identification AGC 112521 AGC 110482 a AGC 110482 b AGC 111946 AGC1 11977 AGC 111164(˚A) F ion /H β F ion /H β F ion /H β F ion /H β F ion /H β F ion /H β [O II ] λ ± ± ± ± ± ± III ] λ · · · ± ± · · · ± ± I +H ζ · · · · · · · · · · · · · · · ± III ] λ ǫ · · · · · · ± · · · · · · ± δ ± ± ± ± · · · ± γ ± ± ± ± · · · ± III ] λ · · · · · · ± · · · · · · · · · H β ± ± ± ± ± ± III ] λ ± ± ± ± ± ± III ] λ ± ± ± ± ± ± I λ ± · · · ± · · · · · · · · · H α ± ± ± ± ± ± II ] λ ± ± · · · ± ± ± II ] λ ± ± ± ± · · · ± II ] λ ± ± ± ± · · · ± III ] λ · · · · · · ± · · · · · · · · · T e ∗ ∗ ∗ ∗ ∗ n e
389 48 146 100 ∗∗ ∗∗ abs (˚A) 5.0 2.5 1.0 0.0 1.0 2.5c(H β ) 0.069 0.277 0.496 0.532 0.375 0.171Continued on next page – Continued from previous pageLine Identification AGC 182595 AGC 731457 AGC 749237 a AGC 749237 b AGC 749237 c (˚A) F ion /H β F ion /H β F ion /H β F ion /H β F ion /H β [O II ] λ ± ± ± ± ± III ] λ ± · · · · · · ± ± I +H ζ · · · · · · · · · · · · · · · [Ne III ] λ ǫ · · · · · · · · · · · · · · · H δ ± ± · · · ± · · · H γ ± ± ± ± · · · [O III ] λ ± · · · · · · · · · · · · H β ± ± ± ± ± III ] λ ± ± ± ± ± III ] λ ± ± ± ± ± I λ · · · · · · · · · · · · · · · H α ± ± ± ± ± II ] λ ± ± ± ± ± II ] λ ± ± ± ± ± II ] λ ± ± ± ± ± III ] λ ± · · · · · · · · · · · · T e ∗ ∗ ∗ ∗ n e
127 119 489 41 100 ∗∗ W abs (˚A) 2.0 5.0 0.0 1.0 1.0 av c(H β ) 0.191 0.286 0.044 0.051 0.074** indicates a default n e of 100 cm − was assumed. T e values marked with ∗ were notcalculated directly, but were assumed to be 10,000 K. av next to the value for W abs indicates that the Balmer line data was not sufficient to derive a value for W abs so anaverage absorption correction of 1.0 ˚A was used instead. 15 –(2013) that there is significant scatter in the position of similar metallicity H II regions on thediagnostic diagram.In order to accurately determine the chemical abundances for emission line nebulae, it is preferableto have a direct measurement of the electron temperature T e . In the wavelength range coveredby our observations, the best way to calculate T e is from measuring the [O III ] λ III ] λ III ] λ λ , +2 ion. The relative strengths of the lines thus depend on the population ofthe different energy levels, which is strongly dependent on the temperature of the electrons (and toa lesser extent, the electron density). [O III ] λ II regions, although the detection in AGC182595 is noisy and thus fairly uncertain. In order to calculate the nebular abundances for theseH II regions we used the ELSA (Emission Line Spectrum Analyzer) program which is described indetail in Johnson et al. (2006).ELSA uses a five-level atom routine with ionization correction factors and a two-region ioniza-tion model to calculate abundances. The five-level atom calculations are based on the work ofHenry et al. (1989) but include multiple updates and improvements (see Johnson et al. 2006). The[O III ] λ III ] λ λ , II ] λ λ , II ] λ II ] λλ II ] λ T e,O +2 and following the method ofPagel et al. (1992). The ionization correction factors are determined from lines with similar ion-ization potentials and then used to account for the unseen ionization states. Many uncertainties,including those in line fluxes, reddening correction, plasma diagnostics, and ionic abundances arecarried through the ELSA program and propagated properly into the final uncertainties in theabundances.In cases where T e is not directly measurable, strong-line calibrations must be used. These methodsrely on ratios of strong emission lines to estimate an electron temperature and thus metallicity.Specifically, most strong-line abundance methods are calibrated to a specific oxygen abundance;we will use oxygen abundance to be synonymous with metallicity for the remainder of this paper.While these methods are very useful, some caution must be employed as they are reliant on boththeoretical models and empirical calibration to arrive at an abundance.One of the most robust strong-line methods for determining oxygen abundance is that of McGaugh(1991; henceforth the McGaugh method). The McGaugh method uses several key abundanceratios: R : log(([O III ] λ λ , II ]] λ β ) O : log([O III ] λ λ , II ] λ [N II ]/[O II ] : log([N II ] λ II ] λ II region is placed on a grid of model emission nebula using the R and O parameters as shown in Figure 4. Since the models are degenerate (i.e. there are two possibleoxygen abundances that can produce the same values for R and O ) the value of [N II ]/[O II ] isused to break the degeneracy between high- and low- metallicity “branches.” H II regions are placedon the low-metallicity branch if [N II ]/[O II ] < − . II ]/[O II ] > − .
9. If − . < [N II ]/[O II ] < − . II regions are deemed to be in the “turn-aroundregion” where precise determination of the abundance can be tricky. However, all the H II regionsin this work had very low [N II ] λ II ] λ parameter) to determineoxygen abundance, and also accounts for different hardness of the ionizing spectra by incorporatingthe O parameter which greatly improves its accuracy. The precise value of the oxygen abundancederived from the McGaugh method depends not only on the values of R and O but also onthe specifics of which models were used to create the grids. We have used the models described inMcGaugh (1991) where the highest mass star produced from the IMF is 60 M ⊙ . The exact value ofthe oxygen abundance was determined by using the position of each H II region and interpolatingbetween the model points to estimate abundance more precisely.A final table of abundances as well as key line ratios is presented in Table 3. Column 1 lists thegalaxy and specific H II -region where applicable. Columns 2 − e method when available. Column 7 lists the error-weighted mean oxygenabundance from the individual H II regions in AGC 110482 and AGC 749237. The abundancelisted in bold is the assumed abundance for each galaxy. We do not give explicit errors on theMcGaugh abundances, but will assume an error of 0.10 dex.The abundances derived via the McGaugh and T e methods are not in perfect agreement with eachother which is consistent with what has been reported by other authors. The inconsistency is oftenconsidered an effect of intrinsic errors in the strong-line calibration methods since they rely onindirect determinations of T e . While the T e method provides a direct measurement of the electrontemperature it is also subject to some level of uncertainty and bias as it has been shown to favorhigher temperature regions in the case that there are temperature fluctuations in the nebula beingobserved (Peimbert 1967). Thus, it can be difficult to determine which method more accuratelyyields the “correct” abundance.Some authors (e.g., Kewley & Ellison 2008; Zahid et al. 2012) have suggested that empirical calibra- 17 –Fig. 4.— Theoretical grids for estimating abundance from McGaugh (1991) plotted with H II regions from the SHIELD galaxies. The roughly horizontal lines indicate different ionization pa-rameters and the roughly vertical lines represent lines of constant oxygen abundance (expressed aslog(O/H)+12). A line of constant oxygen abundance is displayed for each 0.1 dex increase fromlog(O/H)+12 = 7.4 to 9.3. All the objects in this study appear to be located on the “far-side,”or low-metallicity branch (log(O/H)+12 < II regions where the [O III ] λ Table 3:: Derived SHIELD Galaxy AbundancesGalaxy log(R ) log(O ) log([NII]/[OII]) log(O/H)+12 log(O/H)+12 log(O/H)+12[McGaugh] [ T e ] [w. mean]AGC 112521 0.482 0.440 < -1.732 · · · · · · AGC 110482 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.81 ± a · · · · · · b < -1.685 7.98 ± · · · AGC 111946 0.629 -0.450 -1.142 · · · · · ·
AGC 111977 0.792 0.333 < -1.048 · · · · · · AGC 111164 0.654 0.338 -1.359 · · · · · ·
AGC 182595 0.874 0.455 -1.482 7.99 ± · · · AGC 731457 0.784 -0.193 -1.413 · · · · · ·
AGC 749237 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ± a · · · · · · b · · · · · · c · · · · · · Calculated abundances for SHIELD galaxies. When more than one HII region was present an error weightedmean, listed in Column 7, was calculated. The abundances listed in bold are the abundances we assumed foreach galaxy. The < indicates the [NII] line was not measurable so the [NII]/[OII] in this table is only an upperlimit.
19 –tions should be made to correct for systematic differences between various abundance determinationmethods. With only two T e abundances, we are unable to gauge whether we see a systematic offsetin our data. However, if we use the sample of low-luminosity Local Volume Legacy (LVL) galaxiesanalyzed in Berg et al. (2012), which we will use as a comparison set in Section 6, we can determinea reasonable empirical correction and use it to reconcile the two abundance methods. We have as-sumed the T e abundances listed in Berg et al. (2012) then calculated McGaugh abundances fromthe fluxes given in that paper, using the same method and models used for our galaxies. When thetwo different abundance calculation methods are compared we can see that there does appear tobe a systematic offset.A plot of abundance difference (calculated as the McGaugh abundance minus the T e abundance)vs. T e abundance for the Berg et al. (2012) sample is shown in Figure 5. The red diamonds areH II regions from the the LVL galaxies in Berg et al. (2012) and the black points are the H II regions with T e abundances from the SHIELD galaxies. Since it is unclear whether this differenceis a constant offset or if it varies with metallicity we have only used H II regions within the rangeof metallicities that is consistent with our sample (7.4 ≤ log(O/H)+12 ≤ .
076 dex, indicating that on average theMcGaugh abundances were 0 .
076 dex higher than those calculated using the T e method This offsetwill later be used as a “correction” when comparing our sample to others from the literature.“Corrected” average abundances for the SHIELD galaxies are compiled in Table 4. When more thanone H II region was measured a weighted mean of the multiple H II region abundances was used.Notice that the abundances for AGC 112521 and AGC 111164 qualify them both as XMD galaxieswhich is consistent with them being the two lowest luminosity galaxies for which we obtained usablespectra.
6. ANALYSIS
There is a well-established relationship between the luminosity and metallicity of galaxies, wherelower luminosity corresponds to lower metallicity, which has been shown to extend down to low-luminosity dIrrs (e.g., Lequeux et al. 1979; Skillman et al. 1989a; Pilyugin 2001; Melbourne & Salzer2002; Tremonti et al. 2004; Salzer et al. 2005b; Lee et al. 2006; van Zee & Haynes 2006; Berg et al.2012; Haurberg et al. 2013; Skillman et al. 2013). The underlying cause of this relationship is notfully understood, but it is believed to trace a more fundamental mass-metallicity (M-Z) relation-ship, where luminosity is representative of stellar mass. Studies of the luminosity-metallicity (L-Z)relationship that use near-infrared (NIR) luminosities (which is better correlated with stellar mass)show decreased scatter (e.g., Lee et al. 2004; Salzer et al. 2005b; Lee et al. 2006), implying the M-Zrelationship is the more fundamental physical relationship. The origin of the relationship, however,is still unclear but is a key to understanding the star-formation histories and chemical evolution of 20 –Fig. 5.— Comparison of derived McGaugh and T e abundances. Black points within circles are H II regions from galaxies in the SHIELD sample with T e derived abundances. Red diamonds representgalaxies from Berg et al. (2012); filled diamonds are those used to calibrate the empirical offsetbetween the two abundance methods, open diamonds are either obvious outliers or outside of theabundance range that we considered for the offset calculation. 21 –Table 4:: “Corrected” SHIELD Galaxy AbundancesGalaxy log(O/H)+12 NoteAGC 112521 7.33 ± XMD galaxy
AGC 110482 7.79 ± a (corrected) and b (T e )AGC 111946 7.86 ± ± ± XMD galaxy
AGC 182595 7.75 ± e abundanceAGC 731457 8.00 ± ± a, b, and c (all corrected)Corrected abundances for SHIELD galaxies (see text for correction). For AGC 110482 and AGC749237 a weighted mean of the abundances for each HII regions was used. Both AGC 112521 andAGC 111164 qualify as XMD galaxies. 22 –dwarf galaxies.Low luminosity galaxies (e.g., M B & − M B on the luminosity axis. On Figure 6 b and c we have also included comparisongalaxies from the literature, described in more detail in the paragraph below. The SHIELD galaxies(black points) within an outer circle represent galaxies where a T e abundance was measured. TheT e abundance in Table 3 has been adopted for AGC 110482 b and the weighted mean of the 3 H II regions in AGC 749237 has been adopted for that system.Our comparison sample is described in Table 5. It has been broken into two pieces: the “primary”sample (plotted as filled symbols) is composed of only those galaxies with velocity-independentdistances and T e abundances from Lee et al. (2006); van Zee & Haynes (2006), and Berg et al.(2012). The “expanded” sample (plotted as open symbols) includes galaxies from Berg et al. (2012)and van Zee & Haynes (2006) with velocity-based distances or strong-line abundances as well asthe ADBS sample from Haurberg et al. (2013). The ADBS sample is of particular interest as it waschosen in a similar manner to the SHIELD sample (e.g., ADBS is a blind H I survey) and thus hassimilar selection biases. The ADBS galaxies have velocity-derived distances and the abundanceswere derived using the same techniques presented in this work with a mix of both the T e andMcGaugh methods used to derive the abundances (see Haurberg et al. 2013 for details).Figure 6 a shows just the SHIELD galaxies, while Figure 6 b and c include comparison samples fromthe literature. Figure 6 b includes the primary and expanded comparison samples while Figure 6 c shows only the primary comparison sample and the SHIELD galaxies. As indicated, the abundancesfrom the SHIELD galaxies in Figure 6 c have been corrected, meaning the McGaugh abundanceshave been adjusted by the offset calculated in Section 5 in order to be consistent with the T e scale.The data shown in Figure 6 were fit using a bivariate linear regression method weighting the errorsin both x and y. The method we used is based on that outlined in Akritas & Bershady (1996) andreferences therein. The fit to the primary comparison sample is shown on all three panels in Figure6 as a solid line. The dash-dot line shown in Figure 6 b is the fit to all of the comparison sampleplus the SHIELD galaxies and the dashed line on Figure 6 c is the fit to the primary comparisonsample plus the SHIELD galaxies (using the corrected abundances). The XMD outlier, UGC 5340( M B = − .
83, log(O/H)+12 = 7 . b was obtained by fitting only the H I - selected galaxies (SHIELD and 23 –Table 5:: Comparison SampleSample Reference Symbol Primary
Stellar based distances and T e abundances Berg et al. (2012) Filled Red DiamondsLee et al. (2006) Filled Blue Pointsvan Zee & Haynes (2006) Filled Orange Squares
Expanded
Velocity determined dist. and/or strong-line abundance
Haurberg et al. (2013) Open Green TrianglesBerg et al. (2012) Open Red Diamondsvan Zee & Haynes (2006) Open Orange Squares 24 –ADBS samples) and is interesting since it is notably different from the other fits. The fit parametersand samples which were used to obtain them are described in Table 6. The reduced χ values listedin that table are calculated for each fit and compared to the sample plotted in the indicated figure; χ was calculated assuming abundance to be the dependent variable. We chose not to includea fit to the SHIELD sample alone because it is a statistically small sample that covers a limitedluminosity range, thus it produced significant uncertainties in the quality of the fit.The various fits are in general agreement within the errors (with the exception of the “H I - sample”fit which is discussed in the following paragraph). In all cases, the inclusion of the SHIELD sampleresults in a more shallow slope, as could easily be predicted from inspection of Figure 6. TheSHIELD galaxies clearly lie above literature galaxies of similar luminosity even when the correctedabundances are used. This is certainly an intriguing trend, but since we have a relatively smallsample it is possible that this is due to intrinsic scatter in the relationship owing to differing mass-to-light ratios in the B -band or uncertainty in measurement of both luminosity and abundance.However, it may represent something more significant such as a fundamental difference in thesamples or a flattening of the L-Z relationship at low-luminosity.The fit obtained from the H I - selected samples (dotted line) is interesting as it is significantlydifferent from the other fits. It has a slope that is substantially shallower than the average slopeof the other fits, but provides the best fit to the SHIELD sample according the reduced χ test.This indicates that the sample selection process may be a contributing factor and may suggestunderlying physical differences between samples selected from H I surveys and those selected fromoptical catalogs or emission line-surveys. For consistency, we used the uncorrected abundancesfor Figure 6 b and when fitting this H I selected sample since the ADBS galaxies have a mix ofstrong-line and T e derived abundances. When the corrected abundances are used (Fig. 6 c ) theobvious discrepancy between the SHIELD sample and the fit from the comparison sample is reduced.However, even when the corrected abundances are used, all but one of the SHIELD galaxies lieabove the fit line indicating some level of systematic difference that is unlikely to be due solely tointrinsic scatter in the relationship. These issues are further discussed in Section 7.Figure 7 a shows another version of the L-Z diagram. This version uses the corrected abundancesfor the SHIELD galaxies, like Figure 6 c , but uses the NIR luminosity (4.5 µ m band; M . ) instead ofthe B -band luminosity as the luminosity metric. Since we do not have reliable 4.5 µ m luminositiesfor one of the SHIELD galaxies for which we measured an abundance (AGC 731457), this version ofthe L-Z relationship only features 7 of the galaxies in the SHIELD sample. The SHIELD galaxiesare more coincident with the comparison data in the NIR version of the L-Z diagram than they arein Figure 6. The fits derived from fitting the comparison sample (solid line) are nearly identical tothose calculated when the SHIELD data are added to the set (dashed line). Despite the similarityof the fits, the majority of the SHIELD galaxies still appear above the fit line on this diagram,again suggesting the possible presence of a systematic difference from the comparison sample. Thefit parameters for Figure 7 are given in Table 6.In Figure 7 b the mass metallicity (M-Z) relationship is shown. For the comparison sample, stellar 25 –Fig. 6.— Luminosity-metallicity diagrams. The SHIELD sample is displayed with filled blackpoints; dwarf irregular galaxies from the literature are shown with various symbols, filled pointsrepresent the “primary” comparison sample and open points the “expanded” comparison sam-ple (see Table 5): Lee et al. (2006) = blue points, Haurberg et al. (2013) = green triangles,van Zee & Haynes (2006) = orange squares, and Berg et al. (2012) = red diamonds. Figure 6 b shows the entire comparison sample and the uncorrected SHIELD abundances (Table 3); c includesgalaxies from only the primary comparison sample and the “corrected” SHIELD abundances (Table4). The lines represent fits described in the text and Table 6. 26 – Table 6:: L-Z and M-Z Relationship Fit ParametersSample Slope Intercept Symbol Goodness of Fit ( χ ) log(O/H)+12 vs. M B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6a 6b 6c Primary Comparison -0.125 ± ± ± ± · · · · · · HI-Selected (SHIELD + ADBS) -0.061 ± ± · · · · · · SHIELD † + Primary -0.113 ± ± · · · · · · log(O/H)+12 vs. M . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7a Primary Comparison -0.112 ± ± † + Primary -0.110 ± ± log(O/H)+12 vs. log( M ⋆ ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Primary Comparison 0.29 ± ± † + Primary 0.30 ± ± † indicates that corrected McGaugh abundances wereused for the SHIELD galaxies. The “goodness of fit” parameter is the reduced χ value for the given fit whencompared with the data plotted on the indicated panel or figure. See text for details of the fitting process anddifferent samples.
27 –mass was derived using the 4.5 µ m flux and the m . − K ( K -band magnitude) color followingLee et al. (2006). The SHIELD IR observations do not include the K-band, so stellar masses for theSHIELD sample had to be derived in a different manner. We used the 3.6 and 4.5 µ m magnitudes andfollowed the method of Eskew et al. (2012). Since there can be significant uncertainties associatedwith stellar mass calculations, we performed an independent derivation to check against the massobtained with the method of Eskew et al. (2012). As a check we used a straight-forward conversionof 3.6 µ m luminosity into mass assuming a mass-to-light ratio of 0.5 (McGaugh & Schombert 2013)and M ⊙ , . µ m = 3 .
24 (Oh et al. 2008). The difference between the mass derived using Eskew et al.(2012) and that derived from our “straight-forward” conversion of 3.6 µ m luminosity was very small;the average difference was only 0.008 dex. While we can not confirm that our mass estimates arefully consistent with the Lee et al. (2006) method used by Berg et al. (2012), we do feel confidentthat the stellar-mass estimates from Eskew et al. (2012) for the SHIELD galaxies given in thispaper are consistent with other IR-based mass derivation methods. Due to the use of differentstellar-mass calculation methods and the inherently difficult nature of determining stellar masses,we proceed with some caution concerning the results derived using these mass estimates.Similar to the NIR-LZ relationship, the SHIELD galaxies appear consistent with the literaturesamples in the M-Z relationship shown in Figure 7 b and the inclusion of the SHIELD data hasvery little effect on the results of the fit. The fit parameters and χ for the M-Z relationship areincluded in Table 6.
7. DISCUSSION
Various physical mechanisms have been suggested to explain the origin of the M-Z relationship fordIrrs, but the ejection of enriched gas from low-mass galaxies via supernova and stellar winds isone of the most commonly invoked (e.g., Tremonti et al. 2004). In this scenario, the relationshiparises as an effect of an enriched-gas-retention sequence where more massive galaxies are betterable to retain enriched ejecta thus producing higher gas phase metallicities. While the retentionof metal-enriched gas is likely a contributing factor, most observational studies that indicate thepresence of significant gas outflows from isolated dwarf galaxies are based primarily on BCDs (e.g.,Papaderos et al. 1994; Marlowe et al. 1995; Martin 2005; Oey et al. 2007) and may not apply todIrrs with much less active modes of star-formation. Stinson et al. (2007), among others, haveproposed models for the evolution of dwarf galaxies where star-formation occurs in patterns ofrelatively strong starbursts that are quenched by supernova feedback and then followed by periodsof more moderate star formation (the so called “breathing” model). This “bursty” star forma-tion history may supply a mechanism to remove enriched gas while allowing for periods of morequiescent evolution. However, it is unclear if the majority of dIrrs ever undergo the massive star-bursts that are necessary to drive such a model as photometric studies of BCDs suggest that thestructural properties of these systems are fundamentally different than more “normal” low-surfacebrightness dIrrs (Papaderos et al. 1996; Doublier et al. 1999; Marlowe et al. 1999; Salzer & Norton 28 –Fig. 7.— NIR-L-Z relationship and M-Z relationship. Symbols and colors are the same as Figure6. The two fits shown are the fit to the primary comparison sample only (solid line) and fits to thesample including the SHIELD galaxies (dashed line). 29 –1999; Janowiecki & Salzer 2014). Thus other evolutionary scenarios may need to also be consid-ered to explain the observed trends at the low-mass (and low-luminosity) end of the L-Z and M-Zrelationships.As an alternative to bursty star-formation histories for isolated dIrrs, Gavil´an et al. (2013) suggestthat many observational features of isolated dIrrs can be explained without invoking winds or largescale outflows, but instead rely on continuous infall or cooling of primordial gas and continuousstar formation throughout the lifetime of the galaxy. This model provides interesting implicationsin the context of the results presented in this paper.In the scenario proposed by Gavil´an et al. (2013) the galaxies are modeled as a single region and thestar formation rate is controlled only by the amount of gas available and an assumed star formationefficiency. This model does not allow for “bursts” of rapid star formation but instead produces asmooth, continuous (though not constant) star-formation history for the galaxy. Their resultssuggest that with moderate to low star-formation efficiencies this continuous type of evolutioncan plausibly reproduce the observed dIrr trends and the luminosity-metallicity relationship. TheGavil´an et al. (2013) models imply that the observed dispersion in abundance seen at a given B -band luminosity is a result of differing star-formation histories, owing to different star-formationefficiencies and/or different gas-collapse timescales. This is intriguing in the context of the workwe have presented here, because it suggests the offset of the SHIELD sample seen in the B -bandL-Z relationship may be indicative of different global star-formation history compared to opticallyselected samples. Since the SHIELD sample has specifically been chosen to represent some of thelowest-luminosity and lowest-surface brightness galaxies, it is not unreasonable to assume that ourselection may be biased to select galaxies with different star-formation histories than those selectedfrom optical catalogs.Tassis et al. (2008) produced similar results to Gavil´an et al. (2013) suggesting that galaxian out-flow winds are not necessary to reproduce the observed L-Z trends and attribute the inefficientstar-formation in low-mass systems to inefficient gas cooling (i.e. longer collapse timescales) be-cause of the more extended neutral gas distributions often observed in lower mass systems. Sincethere are only a limited number of low-mass dIrr samples where high quality optical and H I datacurrently exist, it is very difficult to analyze whether this scenario is observationally consistent.However, we suggest this is an idea worthy of further investigation. Further studies with theSHIELD project, and similar projects focusing on low-mass ALFALFA sources, should lead to abetter understanding of the nature of the gas distributions in low-mass systems and the role thatgas dynamics play in their chemical evolution.The SHIELD galaxies all seem to be “offset” above the fit line from the comparison sample inthe B -band L-Z relationship. When the “corrected” abundances are used, the offset, with respectto the fit from the comparison sample, is reduced, yet all but one of the SHIELD galaxies stillappears above the fit line. This offset becomes even less pronounced when M B is replaced with M . and disappears when stellar mass ( M ⋆ ) is used instead. This indicates that, on average, themass-to-( B -band) light ratio for the SHIELD galaxies differs from the comparison sample. This is 30 –plausibly consistent with our previous suggestion, that our selection criteria may have biased oursample toward galaxies with a certain star-formation history. However, the general trend simplysuggests that the SHIELD galaxies are too massive for their B -band luminosity. This may arisebecause at least some portion of the SHIELD galaxies do not approach the quasi-continuous star-formation history that has been suggested for more massive dIrrs by van Zee (2001). A stochasticstar-formation history that is similar to that suggested in van Zee (2001) but not as continuous(owing to the lower mass of the system) could cause the B -band luminosity to be less reflective ofthe current stellar mass in these galaxies.The results presented in this paper do not necessarily clarify any of the issues associated withthe chemical evolution of dIrrs, but they do suggest that varied star-formation histories need tobe considered along with the effects of enriched gas outflows in order to adequately explain theobserved trends. Additionally, our data suggest that while the M-Z relationship may be morefundamental, understanding the relationship of the mass-to-light ratio could reveal details of thestar-formation and chemical evolution in these systems. This sample is just a first step in beginningto fill in the very low-luminosity range of dwarf irregulars.
8. CONCLUSIONS
We have used long-slit spectra from the Mayall 4m telescope at KPNO to determine the oxygenabundances for 8 of the 12 SHIELD galaxies. For two galaxies we were able to measure thetemperature sensitive [O
III ] λ II region. We calculated abundances for the other 6 galaxies using the strong-linemethod of McGaugh (1991). We calculated an empirical correction to the McGaugh abundancescale using the dIrr sample from Berg et al. (2012), along with the two H II regions for whichwe detected [O III ] λ e . The galaxies in our sample have well-constraineddistances (HST; McQuinn et al. 2014), optical luminosities (WIYN 3.5m BVR; N. Haurberg et al. inpreparation), and NIR luminosities (Spitzer 3.6 and 4.5 µ m; Cannon, Marshall et al. in preparation).We compared the L-Z and M-Z relationships derived from our results with a substantial comparisonset from the literature that also have well-determined abundances, distances, and optical and NIRluminosities (Lee et al. 2006; van Zee & Haynes 2006; Berg et al. 2012, and references therein). Weadditionally included dIrr galaxies with well-determined abundances that were selected from theADBS blind H I - survey (Haurberg et al. 2013) in our comparison sample. While more luminousthan our data set, the latter sample is particularly interesting because the method of selection wasvery similar to that used for the SHIELD data set.When the L-Z relationship was examined using the B -band luminosity the SHIELD galaxies appearoffset from the comparison sample, but that offset disappears when derived stellar mass is usedinstead. We suggest that this may indicate that a range of star-formation histories and chemicalevolution scenarios for dwarf irregulars should be considered. The possible effects that such scenar- 31 –ios may have on the the luminosity-metallicity and mass-metallicity relationships warrant furtherexamination and exploration. While the M-Z relationship may be more well-constrained, the L-Zrelationship may prove observationally important for disentangling the origins of that relationship.A larger and more comprehensive sample is needed and a more exhaustive comparison betweenan array of models and multiple observational parameters should be done before any model is tooheavily favored. Such a comparison is beyond the scope of this work.In this work, we identified two new XMD galaxies in the local universe, AGC 112521 and AGC111164, and showed that blind H I - surveys like ALFALFA are effective in providing candidateXMD dIrrs (e.g., Skillman et al. 2013). As follow-up studies of the low-H I mass sources from theALFALFA survey continue, a larger sample of low-luminosity gas-rich dIrrs will become availableand allow us to start to unravel some of the puzzles concerning the chemical evolution of dwarfgalaxies. Additionally, current work on resolved stellar populations in the SHIELD galaxies andthe Leo P dwarf promise to further illuminate the star-formation histories in these very low-massgalaxies and lead to a deeper understanding of galaxian chemical evolution.NCH and JJS acknowledge financial support for this project from Indiana University, including aDissertation Year Fellowship to NCH from the College of Arts and Sciences. NCH also acknowledgesthe financial support from Knox College and, in addition, NCH received support from the IndianaSpace Grant Consortium in the form of a graduate fellowship. JMC is supported by NSF grantAST-1211683. We are grateful for the professional support provided by the staff of Kitt PeakNational Observatory during our two observing runs. 32 – REFERENCES
Adams, E. A. K., Giovanelli, R., & Haynes, M. P. 2013, ApJ, 768, 77Akritas, M. G., & Bershady, M. A. 1996, ApJ, 470, 706Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93, 5Berg, D. A., Skillman, E. D., Marble, A. R., et al. 2012, ApJ, 754, 98Brown, W. R., Kewley, L. J., & Geller, M. J. 2008, AJ, 135, 92Cannon, J. M., Giovanelli, R., Haynes, M. P., et al. 2011, ApJ, 739, L22Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245Dopita, M. A., & Evans, I. N. 1986, ApJ, 307, 431Doublier, V., Caulet, A., & Comte, G. 1999, A&AS, 138, 213Eskew, M., Zaritsky, D., & Meidt, S. 2012, AJ, 143, 139Gavil´an, M., Ascas´ıbar, Y., Moll´a, M., & D´ıaz, ´A. I. 2013, arXiv:1306.6565Giovanelli, R., Haynes, M. P., Kent, B. R., et al. 2005, AJ, 130, 2598Giovanelli, R., Haynes, M. P., Adams, E. A. K., et al. 2013, AJ, 146, 15Guseva, N. G., Izotov, Y. I., Stasi´nska, G., et al. 2011, A&A, 529, A149Haurberg, N. C., Rosenberg, J., & Salzer, J. J. 2013, ApJ, 765, 66Haynes, M. P., Giovanelli, R., Martin, A. M., et al. 2011, AJ, 142, 170Henry, R. B. C., Liebert, J., & Boroson, T. A. 1989, ApJ, 339, 872Hu, E. M., Cowie, L. L., Kakazu, Y., & Barger, A. J. 2009, ApJ, 698, 2014Izotov, Y. I., Stasi´nska, G., Meynet, G., Guseva, N. G., & Thuan, T. X. 2006, A&A, 448, 955Izotov, Y. I., Thuan, T. X., & Guseva, N. G. 2012, A&A, 546, A122Janowiecki, S., & Salzer, J. J. 2014, submittedJohnson, M. D., Levitt, J. S., Henry, R. B. C., & Kwitter, K. B. 2006, in IAU Symp. 234, Plan-etary Nebulae in our Galaxy and Beyond, ed. M. J. Barlow & R. H. Me’ndez (Cambridge:Cambridge Univ. Press), 439Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MNRAS, 346, 1055Kakazu, Y., Cowie, L. L., & Hu, E. M. 2007, ApJ, 668, 853Kennicutt, R. C., Jr. 1998, ApJ, 498, 541 33 –Kewley, L. J., & Ellison, S. L. 2008, ApJ, 681, 1183Kniazev, A. Y., Grebel, E. K., Hao, L., et al. 2003, ApJ, 593, L73Lee, H., Skillman, E. D., Cannon, J. M., et al. 2006, ApJ, 647, 970Lee, J. C., Salzer, J. J., & Melbourne, J. 2004, ApJ, 616, 752Lequeux, J., Peimbert, M., Rayo, J. F., Serrano, A., & Torres-Peimbert, S. 1979, A&A, 80, 155Marlowe, A. T., Heckman, T. M., Wyse, R. F. G., & Schommer, R. 1995, ApJ, 438, 563Marlowe, A. T., Meurer, G. R., & Heckman, T. M. 1999, ApJ, 522, 183Martin, A. M., Papastergis, E., Giovanelli, R., et al. 2010, ApJ, 723, 1359Martin, C. L. 2005, ApJ, 621, 227Massey, P., Strobel, K., Barnes, J. V., & Anderson, E. 1988, ApJ, 328, 315McGaugh, S. S. 1991, ApJ, 380, 140McGaugh, S., & Schombert, J. 2013, arXiv:1303.0320McQuinn, K. B. W., Cannon, J. C., Dolphin, A. E., et al. 2014, ApJ, 785, 3Melbourne, J., Phillips, A., Salzer, J. J., Gronwall, C., & Sarajedini, V. L. 2004, AJ, 127, 686Melbourne, J., & Salzer, J. J. 2002, AJ, 123, 2302Oey, M. S., Meurer, G. R., Yelda, S., et al. 2007, ApJ, 661, 801Oh, S.-H., de Blok, W. J. G., Walter, F., Brinks, E., & Kennicutt, R. C., Jr. 2008, AJ, 136, 2761Oke, J. B., & Gunn, J. E. 1983, ApJ, 266, 713Pagel, B. E. J., Simonson, E. A., Terlevich, R. J., & Edmunds, M. G. 1992, MNRAS, 255, 325Papaderos, P., Fricke, K. J., Thuan, T. X., & Loose, H.-H. 1994, A&A, 291, L13Papaderos, P., Loose, H.-H., Fricke, K. J., & Thuan, T. X. 1996, A&A, 314, 59Papaderos, P., Guseva, N. G., Izotov, Y. I., & Fricke, K. J. 2008, A&A, 491, 113Peimbert, M. 1967, ApJ, 150, 825Pilyugin, L. S. 2001, A&A, 374, 412Pustilnik, S. A., Kniazev, A. Y., & Pramskij, A. G. 2005, A&A, 443, 91Pustilnik, S. A., Martin, J.-M., Tepliakova, A. L., & Kniazev, A. Y. 2011, MNRAS, 417, 1335Rhode, K. L., Salzer, J. J., Haurberg, N. C., et al. 2013, AJ, 145, 149Rosenberg, J. L., & Schneider, S. E. 2000, ApJS, 130, 177 34 –Salzer, J. J., & Norton, S. A. 1999,
The Low Surface Brightness Universe, ASP Conference Series170 , (Astronomical Society of the Pacific p. 253)Salzer, J. J., Gronwall, C., Lipovetsky, V. A., et al. 2000, AJ, 120, 80Salzer, J. J., Jangren, A., Gronwall, C., et al. 2005, AJ, 130, 2584Salzer, J. J., Lee, J. C., Melbourne, J., et al. 2005, ApJ, 624, 661Skillman, E. D., Kennicutt, R. C., & Hodge, P. W. 1989, ApJ, 347, 875Skillman, E. D., Terlevich, R., & Melnick, J. 1989, MNRAS, 240, 563Skillman, E. D., Salzer, J. J., Berg, D. A., et al. 2013, AJ, 146, 3Stinson, G. S., Dalcanton, J. J., Quinn, T., Kaufmann, T., & Wadsley, J. 2007, ApJ, 667, 170Tassis, K., Kravtsov, A. V., & Gnedin, N. Y. 2008, ApJ, 672, 888Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004, ApJ, 613, 898Ugryumov, A. V., Engels, D., Pustilnik, S. A., et al. 2003, A&A, 397, 463van Dokkum, P. G. 2001, PASP, 113, 1420van Zee, L. 2000, ApJ, 543, L31van Zee, L. 2001, AJ, 121, 2003van Zee, L., & Haynes, M. P. 2006, ApJ, 636, 214Veilleux, S., & Osterbrock, D. E. 1987, ApJS, 63, 295Xia, L., Malhotra, S., Rhoads, J., et al. 2012, AJ, 144, 28Zahid, H. J., Bresolin, F., Kewley, L. J., Coil, A. L., & Dav´e, R. 2012, ApJ, 750, 120