SDSS-IV MaNGA: The Impact of Diffuse Ionized Gas on Emission-line Ratios, Interpretation of Diagnostic Diagrams, and Gas Metallicity Measurements
Kai Zhang, Renbin Yan, Kevin Bundy, Matthew Bershady, L. Matthew Haffner, René Walterbos, Roberto Maiolino, Christy Tremonti, Daniel Thomas, Niv Drory, Amy Jones, Francesco Belfiore, Sebastian F. Sánchez, Aleksandar M. Diamond-Stanic, Dmitry Bizyaev, Christian Nitschelm, Brett Andrews, Jon Brinkmann, Joel R. Brownstein, Edmond Cheung, Cheng Li, David R. Law, Alexandre Roman Lopes, Daniel Oravetz, Kaike Pan, Thaisa Storchi-Bergmann, Audrey Simmons
MMNRAS , 1–30 (2016) Preprint 30 January 2017 Compiled using MNRAS L A TEX style file v3.0
SDSS-IV MaNGA: The Impact of Diffuse Ionized Gas onEmission-line Ratios, Interpretation of DiagnosticDiagrams, and Gas Metallicity Measurements
Kai Zhang, (cid:63) , Renbin Yan , Kevin Bundy , Matthew Bershady , L. Matthew Haffner ,Ren´e Walterbos , Roberto Maiolino , , Christy Tremonti , Daniel Thomas , Niv Drory ,Amy Jones , Francesco Belfiore , , Sebastian F. S´anchez ,Aleksandar M. Diamond-Stanic , Dmitry Bizyaev , , Christian Nitschelm ,Brett Andrews , Jon Brinkmann , Joel R. Brownstein , Edmond Cheung ,Cheng Li , , David R. Law , Alexandre Roman Lopes , Daniel Oravetz ,Kaike Pan , Thaisa Storchi-Bergmann , , Audrey Simmons (Affiliations can be found after the references)
30 January 2017
ABSTRACT
Diffuse Ionized Gas (DIG) is prevalent in star-forming galaxies. Using a sampleof 365 nearly face-on star-forming galaxies observed by MaNGA, we demonstrate howDIG in star-forming galaxies impacts the measurements of emission line ratios, hencethe interpretation of diagnostic diagrams and gas-phase metallicity measurements.At fixed metallicity, DIG-dominated low Σ H α regions display enhanced [S ii ]/H α ,[N ii ]/H α , [O ii ]/H β , and [O i ]/H α . The gradients in these line ratios are determinedby metallicity gradients and Σ H α . In line ratio diagnostic diagrams, contamination byDIG moves H ii regions towards composite or LI(N)ER-like regions. A harder ionizingspectrum is needed to explain DIG line ratios. Leaky H ii region models can only shiftline ratios slightly relative to H ii region models, and thus fail to explain the com-posite/LI(N)ER line ratios displayed by DIG. Our result favors ionization by evolvedstars as a major ionization source for DIG with LI(N)ER-like emission.DIG can significantly bias the measurement of gas metallicity and metallicity gra-dients derived using strong-line methods. Metallicities derived using N2O2 are optimalbecause they exhibit the smallest bias and error. Using O3N2, R , N2=[N ii ]/H α ,and N2S2H α (Dopita et al. 2016) to derive metallicities introduces bias in the derivedmetallicity gradients as large as the gradient itself.The strong-line method of Blanc et al. (2015; IZI hereafter) cannot be appliedto DIG to get an accurate metallicity because it currently contains only H ii regionmodels which fail to describe the DIG. Key words: galaxies: surveys – galaxies: evolution – galaxies: fundamental parame-ters – galaxies: ISM – galaxies: abundances – galaxies: active
Diffuse ionized gas (DIG hereafter) is an important gas com-ponent in star-forming galaxies. It was first identified in ourMilky Way (MW) off the disk, and known as the Reynoldslayer (Reynolds 1984). DIG is a major part of ionized gas in (cid:63)
Contact e-mail: [email protected]
MW. In terms of mass, it is about 30% of the MW neutralhydrogen (Reynolds 1990, 1991). DIG is also found in exter-nal galaxies both in extra-plannar halos (e.g., Dettmar 1990;Rand et al. 1990; Rand 1996; Hoopes et al. 1999; Rossa &Dettmar 2003a,b) and in the disk (e.g., Monnet 1971; Zu-rita et al. 2000; Oey et al. 2007). The contribution of DIG tothe total emission line flux for face-on galaxies is substan-tial (e.g., Walterbos & Braun 1994; Ferguson et al. 1996; c (cid:13) a r X i v : . [ a s t r o - ph . GA ] J a n K. Zhang et al.
Hoopes, Walterbos, & Greenawalt 1996; Greenawalt et al.1998). For 109 star-forming galaxies in the SINGG sample,the DIG fraction in H α flux is 0.59 ± α surface brightness of the whole galaxy (Oeyet al. 2007). DIG is important in understanding the ionizedgas in star-forming galaxies.The differences between DIG and H ii regions arenot only in emission intensity, but also in emissionline ratios, indicating different physical conditions. The[S ii ] λλ α , [N ii ] λ α , ([S ii ]/H α and[N ii ]/H α hereafter, e.g. Reynolds 1985a; Hoopes & Wal-terbos 2003; Madsen et al. 2006), [O i ] λ α and[O ii ] λ β ([O i ]/H α and [O ii ]/H β hereafter, e.g. Vo-ges & Walterbos 2006; Haffner et al. 1999) are found tobe enhanced in DIG relative to H ii regions. More quanti-tively, [S ii ]/H α , [N ii ]/H α , [O ii ]/H α , [O i ]/H α ratios corre-late negatively with H α flux in the MW and other galaxies(Reynolds et al. 1998; Haffner et al. 1999; T¨ullmann et al.2000; Hausen et al. 2002; Voges & Walterbos 2006; Mad-sen et al. 2006; Blanc et al. 2009). In high spatial resolutionobservations (pc to tens of pc, such as those by The Wis-consin H-Alpha Mapper (WHAM)) the line ratio vs Σ H α relation is a gradual transition from H ii regions to DIG. Inlow spatial resolution observations (hundreds of pc or kpc,such as integral-field studies of nearby galaxies) this relationis due to the mixing of H ii regions and DIG within resolutionelements. Different fractions of DIG and H ii region contri-bution, combinned with the different line ratios for thesetwo types of regions would naturally produce the trend wesee. Observed relations between line ratios and Σ H α providean empirical method to separate DIG dominated and H ii region dominated regions (Blanc et al. 2009; Kreckel et al.2016; Kaplan et al. 2016).DIG line ratios cannot be explained by models of H ii regions. On the classical BPT diagram (Baldwin et al.1981; Veilleux & Osterbrock 1987; Kewley et al. 2001, 2006;Kauffmann et al. 2003), the location of an H ii region ismainly determined by its metallicity and ionization parame-ter (e.g., Kewley et al. 2002; Dopita et al. 2013). DIG showsa lower ionization parameter than H ii regions, which ex-plains partly if not mostly the enhancement of [N ii ]/H α ,[S ii ]/H α , and [O ii ]/H α in DIG. However, just varying thesetwo parameters cannot fully produce the line ratios seen inDIG (e.g., Galarza et al. 1999; Hoopes & Walterbos 2003;Kaplan et al. 2016). A third variable is needed. Also, theline ratios of DIG within a galaxy often vary much morethan those of H ii regions. For example, in our Milky Way,DIG [S ii ]/H α ratios display a dispersion of 0.13 (in lin-ear space), compared with a dispersion of 0.03 for H ii re-gions (Madsen et al. 2006). Temperature has been proposedas the strongest factor explaining the variations of DIGline ratios because it can explain the coherent variation of[O ii ] λ α , [N ii ] λ α , and [S ii ]/H α (Haffner etal. 1999; Mierkiewicz et al. 2006). The variation of tempera-ture is a result of the balance between heating and cooling,which also requires a physical explanation. Besides photoion-ization, an additional source of heating might be importantespecially when the density is low (e.g., Reynolds & Cox1992). We will show in this paper that a harder ionizingspectrum could easily explain the DIG line ratios we observein star-forming galaxies. The harder spectrum is capable of producing partially-ionized regions that emit strong [S ii ],[N ii ], [O ii ] and [O i ], and an increase in temperature.The DIG can impact our interpretation of the line-ratio diagnostic diagrams of galaxies for either integrated orspatially-resolved spectroscopy. The BPT diagrams, for ex-ample, are widely used to diagnose the physical propertiesof ionized gas and separate different types of galaxies (Bald-win et al. 1981; Veilleux & Osterbrock 1987; Kewley et al.2001, 2006; Kauffmann et al. 2003 ). When metallicity, ion-ization parameter, density, or ionizing spectrum change, thelocation of ionized gas on the BPT diagram changes (e.g.,Dopita et al. 2000, 2013; Kewley et al. 2002, 2013a,b). Whenthe N/O ratio is high, it is possible for a star-forming galaxyto be classified as composite galaxy (P´erez-Montero & Con-tini 2009; P´erez-Montero et al. 2013, 2016). The enhanced[N ii ]/H α and [S ii ]/H α of DIG would move the position ofa star-forming galaxy toward the composite or LI(N)ER re-gion (e.g., Sarzi et al 2006; Stasi´nska et al. 2008; Yan &Blanton 2012; Kehrig et al. 2012; Singh et al. 2013; Gomeset al. 2016; Belfiore et al. 2016a,b) on the BPT diagrams.Consequently, understanding DIG helps us understand thenature of galaxies classified as composites or LI(N)ERs bythe BPT diagrams.The study of DIG is also critical for gas-phase metallic-ity measurements in star-forming galaxies. Metallicity cal-ibrations are generally based on H ii region models whichhave certain assumptions. Some of these assumptions are notvalid for DIG, and hence lead to biased metallicity measure-ments when DIG is present. The metallicity and ionizationparameter (q= ionizing photon fluxNe = U × c ) determine theline ratios of a H ii region, and these two parameters are cor-related (Dopita et al. 2006). Metallicity also determines theshape of the ionizing spectrum. In DIG, however, the corre-lation between q and metallicity no longer holds, since DIGhas a much lower ionization parameter than H ii regions,and the ionizing spectrum shape can change. Given thevery different line ratios such as [N ii ]/H α , [O iii ] λ β ([O iii ]/H β hereafter) and [S ii ]/H α for DIG and H ii regions,biases in metallicities derived from strong line methods areinevitable. These biases potentially contribute to the largedispersion in metallicity measurements found in the litera-ture. This will influence metallicity gradient measurements(S´anchez et al. 2014; Ho et al. 2015), metallicities at the out-skirts of galaxies (Moran et al. 2012), the mass-metallicityrelation in the local universe (e.g. McClure & van den Bergh,1968; Lequeux et al. 1979; Garnett 2002; Tremonti et al.2004; Lee et al. 2006) and at high redshift (Erb et al. 2006;Maiolino et al. 2008; Mannucci et al. 2009), and the mass-metallicity-SFR fundamental plane (Mannucci et al. 2010;Yates et al. 2012; Andrews & Martini 2013). For single-fibersurveys such as SDSS, the observed emission lines come froma combination of DIG and H ii regions, and the bias intro-duced by DIG is uncertain and hard to quantify. With thehelp of integral field spectroscopy (IFS), we can study howthe presence of DIG impacts metallicity measurements indetail.In this paper, we demonstrate the prevalence of DIGin star-forming galaxies from the MaNGA survey. The largesample and full optical wavelength coverage enable us toexplore the optical line ratios for DIG for an unprece-dented number of star-forming galaxies. Section 2 describesour sample and the emission line measurements; while Sec- MNRAS000
Hoopes, Walterbos, & Greenawalt 1996; Greenawalt et al.1998). For 109 star-forming galaxies in the SINGG sample,the DIG fraction in H α flux is 0.59 ± α surface brightness of the whole galaxy (Oeyet al. 2007). DIG is important in understanding the ionizedgas in star-forming galaxies.The differences between DIG and H ii regions arenot only in emission intensity, but also in emissionline ratios, indicating different physical conditions. The[S ii ] λλ α , [N ii ] λ α , ([S ii ]/H α and[N ii ]/H α hereafter, e.g. Reynolds 1985a; Hoopes & Wal-terbos 2003; Madsen et al. 2006), [O i ] λ α and[O ii ] λ β ([O i ]/H α and [O ii ]/H β hereafter, e.g. Vo-ges & Walterbos 2006; Haffner et al. 1999) are found tobe enhanced in DIG relative to H ii regions. More quanti-tively, [S ii ]/H α , [N ii ]/H α , [O ii ]/H α , [O i ]/H α ratios corre-late negatively with H α flux in the MW and other galaxies(Reynolds et al. 1998; Haffner et al. 1999; T¨ullmann et al.2000; Hausen et al. 2002; Voges & Walterbos 2006; Mad-sen et al. 2006; Blanc et al. 2009). In high spatial resolutionobservations (pc to tens of pc, such as those by The Wis-consin H-Alpha Mapper (WHAM)) the line ratio vs Σ H α relation is a gradual transition from H ii regions to DIG. Inlow spatial resolution observations (hundreds of pc or kpc,such as integral-field studies of nearby galaxies) this relationis due to the mixing of H ii regions and DIG within resolutionelements. Different fractions of DIG and H ii region contri-bution, combinned with the different line ratios for thesetwo types of regions would naturally produce the trend wesee. Observed relations between line ratios and Σ H α providean empirical method to separate DIG dominated and H ii region dominated regions (Blanc et al. 2009; Kreckel et al.2016; Kaplan et al. 2016).DIG line ratios cannot be explained by models of H ii regions. On the classical BPT diagram (Baldwin et al.1981; Veilleux & Osterbrock 1987; Kewley et al. 2001, 2006;Kauffmann et al. 2003), the location of an H ii region ismainly determined by its metallicity and ionization parame-ter (e.g., Kewley et al. 2002; Dopita et al. 2013). DIG showsa lower ionization parameter than H ii regions, which ex-plains partly if not mostly the enhancement of [N ii ]/H α ,[S ii ]/H α , and [O ii ]/H α in DIG. However, just varying thesetwo parameters cannot fully produce the line ratios seen inDIG (e.g., Galarza et al. 1999; Hoopes & Walterbos 2003;Kaplan et al. 2016). A third variable is needed. Also, theline ratios of DIG within a galaxy often vary much morethan those of H ii regions. For example, in our Milky Way,DIG [S ii ]/H α ratios display a dispersion of 0.13 (in lin-ear space), compared with a dispersion of 0.03 for H ii re-gions (Madsen et al. 2006). Temperature has been proposedas the strongest factor explaining the variations of DIGline ratios because it can explain the coherent variation of[O ii ] λ α , [N ii ] λ α , and [S ii ]/H α (Haffner etal. 1999; Mierkiewicz et al. 2006). The variation of tempera-ture is a result of the balance between heating and cooling,which also requires a physical explanation. Besides photoion-ization, an additional source of heating might be importantespecially when the density is low (e.g., Reynolds & Cox1992). We will show in this paper that a harder ionizingspectrum could easily explain the DIG line ratios we observein star-forming galaxies. The harder spectrum is capable of producing partially-ionized regions that emit strong [S ii ],[N ii ], [O ii ] and [O i ], and an increase in temperature.The DIG can impact our interpretation of the line-ratio diagnostic diagrams of galaxies for either integrated orspatially-resolved spectroscopy. The BPT diagrams, for ex-ample, are widely used to diagnose the physical propertiesof ionized gas and separate different types of galaxies (Bald-win et al. 1981; Veilleux & Osterbrock 1987; Kewley et al.2001, 2006; Kauffmann et al. 2003 ). When metallicity, ion-ization parameter, density, or ionizing spectrum change, thelocation of ionized gas on the BPT diagram changes (e.g.,Dopita et al. 2000, 2013; Kewley et al. 2002, 2013a,b). Whenthe N/O ratio is high, it is possible for a star-forming galaxyto be classified as composite galaxy (P´erez-Montero & Con-tini 2009; P´erez-Montero et al. 2013, 2016). The enhanced[N ii ]/H α and [S ii ]/H α of DIG would move the position ofa star-forming galaxy toward the composite or LI(N)ER re-gion (e.g., Sarzi et al 2006; Stasi´nska et al. 2008; Yan &Blanton 2012; Kehrig et al. 2012; Singh et al. 2013; Gomeset al. 2016; Belfiore et al. 2016a,b) on the BPT diagrams.Consequently, understanding DIG helps us understand thenature of galaxies classified as composites or LI(N)ERs bythe BPT diagrams.The study of DIG is also critical for gas-phase metallic-ity measurements in star-forming galaxies. Metallicity cal-ibrations are generally based on H ii region models whichhave certain assumptions. Some of these assumptions are notvalid for DIG, and hence lead to biased metallicity measure-ments when DIG is present. The metallicity and ionizationparameter (q= ionizing photon fluxNe = U × c ) determine theline ratios of a H ii region, and these two parameters are cor-related (Dopita et al. 2006). Metallicity also determines theshape of the ionizing spectrum. In DIG, however, the corre-lation between q and metallicity no longer holds, since DIGhas a much lower ionization parameter than H ii regions,and the ionizing spectrum shape can change. Given thevery different line ratios such as [N ii ]/H α , [O iii ] λ β ([O iii ]/H β hereafter) and [S ii ]/H α for DIG and H ii regions,biases in metallicities derived from strong line methods areinevitable. These biases potentially contribute to the largedispersion in metallicity measurements found in the litera-ture. This will influence metallicity gradient measurements(S´anchez et al. 2014; Ho et al. 2015), metallicities at the out-skirts of galaxies (Moran et al. 2012), the mass-metallicityrelation in the local universe (e.g. McClure & van den Bergh,1968; Lequeux et al. 1979; Garnett 2002; Tremonti et al.2004; Lee et al. 2006) and at high redshift (Erb et al. 2006;Maiolino et al. 2008; Mannucci et al. 2009), and the mass-metallicity-SFR fundamental plane (Mannucci et al. 2010;Yates et al. 2012; Andrews & Martini 2013). For single-fibersurveys such as SDSS, the observed emission lines come froma combination of DIG and H ii regions, and the bias intro-duced by DIG is uncertain and hard to quantify. With thehelp of integral field spectroscopy (IFS), we can study howthe presence of DIG impacts metallicity measurements indetail.In this paper, we demonstrate the prevalence of DIGin star-forming galaxies from the MaNGA survey. The largesample and full optical wavelength coverage enable us toexplore the optical line ratios for DIG for an unprece-dented number of star-forming galaxies. Section 2 describesour sample and the emission line measurements; while Sec- MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG tion 3.1 shows the [S ii ]/H α vs Σ H α relation that illustratesthe dominance of DIG in low Σ H α regions. Section 3 demon-strates how DIG impact line ratios like [S ii ]/H α , [N ii ]/H α ,[O ii ]/H β , [O i ]/H α , [O iii ]/[O ii ], [O iii ]/H β . Section 4 teststhe leaky H ii region model as well as one in which hotevolved stars serve as the ionization source of the DIG, us-ing different line ratios and diagnostic diagrams. Section 5presents our study of how the DIG impacts metallicity de-rived using strong line methods: N2, N2O2, R , O3N2 andN2S2H α , and IZI. Section 6 contains a discussion of theseresults, summarized then in Section 7. We use a cosmol-ogy with H = 70 km s − Mpc − , Ω m = 0.3, and Ω Λ = 0.7throughout this paper. MaNGA (Mapping Nearby Galaxies at APO) (Bundy et al.2015) is one of the three core programs in the Sloan Dig-ital Sky Survey-IV (SDSS-IV). It aims at obtaining Inte-grated Field Spectroscopy (IFS) of 10,000 nearby galaxies.The survey employs the BOSS spectrographs (Smee et al.2013) on the 2.5m Sloan Foundation Telescope (Gunn et al.2006). The spectrographs provide a spectral coverage from3600˚A to 10300˚A at a resolution around R ∼ . < z < .
15, meaning the targetscover a factor of 15 in distance, and consequently the physi-cal resolution also span a range of 15. We select more lumi-nous galaxies, which are larger, at higher redshift. The typi-cal physical resolution is 1-2 kpc. The physical resolution ishighly correlated with luminosity which means that we needto be aware that any results we find which might depend onphysical resolution will be specific to a specific range of lu-minosity and vice versa. The primary sample, which coversto 1.5 effective radius ( R e ) in major axis, comprise 2/3 ofthe sample while the secondary sample, which covers to 2.5 R e comprise the remaining 1/3. Massive galaxies, which arebigger, are selected at higher redshift. All plates are exposeduntil we reach a ( S/N ) of 20 per pixel per fiber in the g-band continuum for a galactic-extinction-corrected g-bandfiber magnitude of 22, and a ( S/N ) of 36 per pixel per fiberin the i-band continuum for a galactic-extinction-corrected i-band fiber magnitude of 21 (Yan et al. 2016b). MaNGA pro-vides a benchmark of resolved ionized gas properties, stellar population, and dynamical evolution in the local universe(e.g., Belfiore et al. 2015; Li et al. 2015; Wilkinson et al.2015). In this paper, we use a sample of 81 regular survey plates ob-served before summer of 2015. There are 1391 unique galax-ies observed. This corresponds to the sample that was re-leased in DR13. The photometry data is from NASA-SloanAtlas catalog (NSA Catalog). By applying a color cut of M u − M r < M u and M r are rest-frame absolute mag-nitudes without extinction correction) we select only bluegalaxies. This resulted in a sample of 592 galaxies. To re-move edge-on galaxies we apply the cut b/a > .
5, whereb and a are the minor and major axis of the Sersic model.Edge-on galaxies suffer from strong projection effects andare also prone to severe extinction. These criteria leave uswith 365 galaxies. AGNs are not eliminated because we careabout DIG that is far away from the center. From this sam-ple we choose three galaxies that have large internal vari-ations of H α surface brightness and best spatial resolution(127 fiber IFUs) to demonstrate the impact of DIG on lineratios, interpretation of diagnostic diagrams, and metallic-ity measurements. We discuss in Section 3.4 and Section 5.7that the impact of DIG is prevalent in all star-forming galax-ies in our sample. We start with the datacubes produced by the MaNGA datareduction pipeline (Law et al. 2016). We first construrctVoronoi bins of the spectra by requiring the S/N in r bandin each bin to be greater than 30 per ˚A. The covariancebetween spaxels is not accounted for when binning. WithBC03 (Bruzual & Charlot 2003), we produce 14 continuumtemplates for SSPs with Z ∗ =0.02 and 0.008 and ages=13,7, 2, 1, 0.5, 0.25, and 0.125Gyr. For the stacked spectrumin each bin, we fit combinations of these simple stellar pop-ulations (SSPs), fitting velocity and relative amplitudes tospectra combined from spaxels in a Voronoi bin. We thenderive the stellar velocity dispersion using the vdispfit.pro inIDLSPEC2D package. 20˚A windows around [O ii ] λ β , [O iii ] λ iii ] λ i ] λ ii ] λ α ,[N ii ] λ ii ] λλ ∼ − . By doingthis, we are left with only small scale spectral variations suchas absorption lines and emission lines, which we refer to asthe ’line feature spectrum’. The same technique is appliedto the SSP templates to get the ’line feature templates’ thatonly contains small scale information. A linear regression isperformed to fit the ’line feature spectrum’ with the sum of http : .iap.fr/users/charlot/bc / http : //spectro.princeton.edu/idlspec d doc.html MNRAS , 1–30 (2016)
K. Zhang et al. ’line feature templates’ at a number of velocity offsets rel-ative to the systematic redshift of the galaxy. The velocitygrid ranges from -450 km s − to 450 km s − with an inter-val of 30 km s − . For each Voronoi bin at each velocity, theleast-square fitting yields a χ for that fit. A quadratic curveis fitted to the χ vs velocity curve to find the velocity offsetyielding the minimum χ . For each spaxel in one Voronoibin, we use the fitted stellar continuum, the combination of14 BC03 templates, as one template, and adjust the am-plitude to fit the continuum in each spaxel so that we canmeasure the emission-line ratios on a spaxel by spaxel ba-sis. The residual emission-line only spectrum is stored forrefined emission line fitting. We use single gaussians to fitthe [O i ], H β , [O iii ], [N ii ] λ α , [N ii ] λ ii ]respectively, and we use two gaussians to fit [O ii ]. The lineratio of [N ii ] λ ii ] λ α surface brightness maps forthe 3 representative galaxies are shown in Figure 1 . The er-rors of the line strength are obtained by the MPFIT packagewhich only includes the fitting errors (Markwardt 2009). AS/N cut of 5 is applied to emission lines used. H α SURFACE BRIGHTNESS3.1 Separation of regions dominated by DIG andH ii regions The first step to study DIG is separating DIG from H ii re-gions. The best way to isolate DIG is to identify individualH ii regions and subtract them out (e.g., Walterbos & Braun1994; Zurita et al. 2000, 2002; Thilker, Walterbos, & Braun2002; S´anchez et al. 2012). Due to the limited spatial reso-lution of MaNGA, we can not resolve individual H ii regionsin one galaxy. Our reconstructed PSF is about 2.5” FWHM(2” covers 1 kpc at z=0.025) while the size of a typical H ii region is a few to hundreds of pc (Kennicutt 1984; Garay& Lizano 1999; Kim & Koo 2001; Hunt & Hirashita 2009).So the light in one spaxel is always a mixture of H ii re-gion emission and the surrounding DIG. Instead we separateDIG dominated regions and H ii regions dominated regions.In Figure 2, we see that the low surface brightness regionshave [S ii ]/H α ∼ . − . ii ]/H α ∼ . − .
4, depending on the metallic-ity. Since [S ii ]/H α is sensitive to metallicity, and there aremetallicity gradients in galaxies, we separate all the spax-els in each galaxy into different radial bins: [0,0.6],[0.6,1.2],[1.2,1.8], and [1.8,2.4] R e . The [S ii ]/H α vs Σ H α relations aresimilar in all radial bins, meaning the variation of [S ii ]/H α is not caused by metallicity variation, but reflects the tran-sition from DIG dominated low Σ H α regions to H ii regiondominated high Σ H α regions. This figure illustrates that Σ H α can be used to separate the two different kinds of regions:low Σ H α DIG dominated regions and high Σ H α H ii regiondominated regions. ii ]/ H α , [N ii ]/ H α , [O ii ]/ H β , and [O i ] / H α Galaxies show metallicity variations, in particular in theform of radial gradients (e.g. S´anchez et al. 2014; Ho et al. 2015). It is crucial to control metallicity before exploring howthe line ratios vary with other parameters like H α surfacebrightness because metallicity is a major source of variationin line ratios. We plot line ratios as a function of radius, andcolor code the dots with H α surface brightness. We assumethat the metallicity is constant within annuli for each galaxybut may be changing with radius.We explore the impact of DIG on [S ii ]/H α first. Wecolor-code the dots by H α surface brightness and show how[S ii ]/H α changes with radius in Figure 3. At a fixed ra-dius, we see a rainbow pattern such that low Σ H α regionshave higher [S ii ]/H α . This is a direct demonstration that[S ii ]/H α is enhanced in DIG after controlling for metallic-ity. We then explore how DIG impacts [N ii ]/H α . [N ii ]/H α is used as a metallicity indicator since Nitrogen is a sec-ondary element and proportional to Z while H α is notsensitive to metallicity (e.g. Storchi-Bergmann et al., 1994,van Zee et al., 1998, Denicol´o et al., 2002). We color-codethe dots by H α surface brightness and show how [N ii ]/H α changes with radius in Figure 4. If we control for H α sur-face brightness by looking at the dots with the same color,[N ii ]/H α decreases towards large radius, reflecting a metal-licity gradient. Due to the presence of an H α surface bright-ness gradient, the different surface brightness bins usuallytrace different parts of the galaxy. However, at the radiuswhere they overlap, we see a rainbow pattern. At a fixedradius, DIG dominated low surface brightness region showa higher [N ii ]/H α . If we use N2 to derive the metallicity,the enhancement means that the metallicity would be over-estimated in those spaxels with a high DIG fraction andvice-versa. The impact of DIG on the metallicity derivedusing [N ii ]/H α is given in Section 5.1.We show how [O ii ]/H β changes with radius and H α sur-face brightness in Figure 5. The extinction correction is notapplied because it is not reliable when the emission line, es-pecially H β , is weak. Besides, a foreground dust screen maynot be the appropriate model for DIG. Finally, We showin Figure 6 that H β /H α does not depend on H α surfacebrightness, meaning extinction will not produce any line ra-tios change between DIG and H ii regions.[O II ]/H β generally increases with radius at fixed H α surface brightness due to a metallicity gradient. At fixedradius, the low H α surface brightness regions have higher[O ii ]/H β . It is interesting that [O ii ]/H β and [N ii ]/H α bothincrease with decreasing surface brightness at fixed radiuswhile they change reversely with radius at fixed H α surfacebrightness. The opposite variation trends of [O ii ]/H β and[N ii ]/H α with radius at fixed surface brightness are consis-tent with a metallicity variation. With [O/H] above solarvalue, [N ii ]/H α increases with metallicity due to the addi-tion of secondary Nitrogen while [O ii ]/H β drops with in-creasing metallicity because of decreasing temperature. Thepositive correlation between [O ii ]/H β and [N ii ]/H α withsurface brightness at fixed radius is seemingly consistentwith temperature variation. Mierkiewicz et al. (2006) showthat [O ii ]/H α and [N ii ]/H α correlate positively with tem-perature variation and they concluded that the variation ofline ratios is driven by temperature variations (Haffner et al.1999; Haffner et al. 2009). However, the variation of temper-ature is a result of the balance between heating and coolingwhich, itself, needs a physical explanation. Besides, some MNRAS000
4, depending on the metallic-ity. Since [S ii ]/H α is sensitive to metallicity, and there aremetallicity gradients in galaxies, we separate all the spax-els in each galaxy into different radial bins: [0,0.6],[0.6,1.2],[1.2,1.8], and [1.8,2.4] R e . The [S ii ]/H α vs Σ H α relations aresimilar in all radial bins, meaning the variation of [S ii ]/H α is not caused by metallicity variation, but reflects the tran-sition from DIG dominated low Σ H α regions to H ii regiondominated high Σ H α regions. This figure illustrates that Σ H α can be used to separate the two different kinds of regions:low Σ H α DIG dominated regions and high Σ H α H ii regiondominated regions. ii ]/ H α , [N ii ]/ H α , [O ii ]/ H β , and [O i ] / H α Galaxies show metallicity variations, in particular in theform of radial gradients (e.g. S´anchez et al. 2014; Ho et al. 2015). It is crucial to control metallicity before exploring howthe line ratios vary with other parameters like H α surfacebrightness because metallicity is a major source of variationin line ratios. We plot line ratios as a function of radius, andcolor code the dots with H α surface brightness. We assumethat the metallicity is constant within annuli for each galaxybut may be changing with radius.We explore the impact of DIG on [S ii ]/H α first. Wecolor-code the dots by H α surface brightness and show how[S ii ]/H α changes with radius in Figure 3. At a fixed ra-dius, we see a rainbow pattern such that low Σ H α regionshave higher [S ii ]/H α . This is a direct demonstration that[S ii ]/H α is enhanced in DIG after controlling for metallic-ity. We then explore how DIG impacts [N ii ]/H α . [N ii ]/H α is used as a metallicity indicator since Nitrogen is a sec-ondary element and proportional to Z while H α is notsensitive to metallicity (e.g. Storchi-Bergmann et al., 1994,van Zee et al., 1998, Denicol´o et al., 2002). We color-codethe dots by H α surface brightness and show how [N ii ]/H α changes with radius in Figure 4. If we control for H α sur-face brightness by looking at the dots with the same color,[N ii ]/H α decreases towards large radius, reflecting a metal-licity gradient. Due to the presence of an H α surface bright-ness gradient, the different surface brightness bins usuallytrace different parts of the galaxy. However, at the radiuswhere they overlap, we see a rainbow pattern. At a fixedradius, DIG dominated low surface brightness region showa higher [N ii ]/H α . If we use N2 to derive the metallicity,the enhancement means that the metallicity would be over-estimated in those spaxels with a high DIG fraction andvice-versa. The impact of DIG on the metallicity derivedusing [N ii ]/H α is given in Section 5.1.We show how [O ii ]/H β changes with radius and H α sur-face brightness in Figure 5. The extinction correction is notapplied because it is not reliable when the emission line, es-pecially H β , is weak. Besides, a foreground dust screen maynot be the appropriate model for DIG. Finally, We showin Figure 6 that H β /H α does not depend on H α surfacebrightness, meaning extinction will not produce any line ra-tios change between DIG and H ii regions.[O II ]/H β generally increases with radius at fixed H α surface brightness due to a metallicity gradient. At fixedradius, the low H α surface brightness regions have higher[O ii ]/H β . It is interesting that [O ii ]/H β and [N ii ]/H α bothincrease with decreasing surface brightness at fixed radiuswhile they change reversely with radius at fixed H α surfacebrightness. The opposite variation trends of [O ii ]/H β and[N ii ]/H α with radius at fixed surface brightness are consis-tent with a metallicity variation. With [O/H] above solarvalue, [N ii ]/H α increases with metallicity due to the addi-tion of secondary Nitrogen while [O ii ]/H β drops with in-creasing metallicity because of decreasing temperature. Thepositive correlation between [O ii ]/H β and [N ii ]/H α withsurface brightness at fixed radius is seemingly consistentwith temperature variation. Mierkiewicz et al. (2006) showthat [O ii ]/H α and [N ii ]/H α correlate positively with tem-perature variation and they concluded that the variation ofline ratios is driven by temperature variations (Haffner et al.1999; Haffner et al. 2009). However, the variation of temper-ature is a result of the balance between heating and coolingwhich, itself, needs a physical explanation. Besides, some MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 1. H α surface brightness plot and SDSS optical image for each galaxy. The cyan ellipses in the upper represent are 1.5 and 2.5times the effective radius. The hexagons in the lower row demonstrate the location of the IFU bundles on the object. Figure 2. [S ii ]/H α vs H α surface brightness plot for each galaxy. Different color represent different annuli. The solid line is the [S ii ]/H α vs H α surface brightness relation for Milky Way (Madsen et al. 2006). We see that high and low H α surface brightness regions havedifferent [S ii ]/H α ˙The red and blue bars show the typical error at log Σ H α =38.5 erg s − kpc − and log Σ H α =38 erg s − kpc − forindividual spaxels. The length of the error bars is 2 σ . signficant assumptions need to be made to use [O ii ]/H β and [N ii ]/H α as temperature tracers. The biggest assump-tion is that [O ii ] and H β are co-spatial. However, we onlyresolve galaxies to kpc scale for MaNGA survey, thus wesee the integrated emission from all layers of ISM aroundan ionizing source. The co-spatial assumption may not holdfor DIG. [N ii ]/H α line ratios are often used as a metallicityindicator for extra-galactic studies, while it may be used asa temperature indicator for galactic uses. Both metallicityand temperature are relevant to [N ii ]/H α emission. Finally,a decrease in ionization parameter and/or a harder ionizingspectrum towards lower Σ H α will also translate into enhance-ment of [O ii ]/H β and [N ii ]/H α . This is further explored inSection 4. [O I ] is an important line to diagnose the physical prop-erties of ionized gas (Dopita et al. 2000; Kewley et al. 2006).[O i ] is detected in DIG of M33 and [O i ]/H α is found to cor-relate negatively with emission measure (Voges & Walterbos2006). [O i ]/H α is a strong function of temperature and thehardness of the ionizing spectrum. [O i ] has a high criticaldensity ( n [ OI ] = 10 . cm − ), thus it is emitted in high den-sity neutral and partially ionized regions. [O i ]/H α is alsoan excellent tracer of shocks (Dopita et al. 2000; Kewley etal. 2002; Allen et al. 2008). In Figure 7, we show how the[O i ]/H α changes with H α surface brightness and radius forour three galaxies. For all galaxies, [O i ]/H α increases withradius. At a fixed radius, [O i ]/H α is enhanced for low sur-face brightness regions, indicating [O i ]/H α is higher in DIG. MNRAS , 1–30 (2016)
K. Zhang et al.
We note the contrast between DIG and H ii region is large,[O i ]/H α drops ∼ dex with an increase of 1 dex in Σ H α . iii ]/[O ii ] and [O iii ]/ H β [O III ]/[O ii ] is a good ionization parameter proxy whenmetallicity is controlled (Kewley et al. 2002, see also Fig-ure 18). We show how the [O iii ]/[O ii ] values of low surfacebrightness spaxels differ from the high surface brightnessspaxels for our sample galaxies in Figure 8. This relationhas large dispersion but there is a clear trend. At a fixed H α surface brightness, [O iii ]/[O ii ] increases as we go to largerradii due to a metallicity gradient (see also Figure 14 ). Ata fixed radius, [O iii ]/[O ii ] mostly decreases with decreasingsurface brightness. The result suggests that DIG has a lowerionization parameter than H ii regions.[O III ]/H β is another frequently used ionization pa-rameter proxy. Its dependence on metallicity and ionizingspectrum harndess is much stronger than [O iii ]/[O ii ]. How[O iii ]/H β of DIG differs from that of H ii regions dependson the specific physical properties of the ISM. DIG can showeither higher (Wang, Heckman, & Lehnert 1997; Rand 1998,2000; Collins & Rand 2001; Otte 2001, 2002; Otte, Gal-lagher, & Reynolds 2002) or lower (MW: Reynolds 1985b;M31: Greenawalt et al. 1997; Galarza, Walterbos, & Braun1999) [O iii ]/H β than H ii regions. The study of our wholesample of 395 galaxies in Section 3.4 confirms the diversebehavior of [O iii ]/H β in DIG. In the previous paragraphwe have shown that DIG has lower ionization parameterthan H ii regions. At a fixed metallicity, a decrease in ion-ization parameter will result in a decease of [O iii ]/H β forq < > iii ]/H β is roughlyconstant and independent of ionization parameter. This di-verse behavior of [O iii ]/H β means a third parameter otherthan metallicity and ionization parameter is needed, as willbe discussed in Section 4. The 3rd galaxy: 8243-12704 showsa negative [O iii ]/[O ii ] and [O iii ]/H β gradient. This is notconsistent with an inverse metallicity gradient at the centerbecause the [N ii ]/[O ii ] gradient is negative in this range.This could be due to AGN activity, but the strength of theAGN is not strong enough to produce AGN-like line ratioson the BPT diagram. These kind of sources will be exploredin future work. Σ H α relation Σ H α relation: The Whole Sample To quantify the variation of the line ratios as a functionof H α surface brightness for the whole sample, we select allspaxels that have [0.4 R e , 0.6 R e ] in each galaxy. We then nor-malize the log line ratios vs log Σ H α relation of each galaxyby subtracting from all the spaxels the median log line ratioand the median log Σ H α of the set of spaxels. All the spaxelsin a galaxy are weighted by N , where N is the number ofvalid spaxels in this galaxy. Figure 10 shows ∆ log line ratiosvs ∆ log Σ H α relation for the whole sample by combining allgalaxies. [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α are higherin DIG dominated low Σ H α regions. [O iii ]/[O ii ] is slightlylower in DIG while [O iii ]/H β is mildly higher in DIG.Furthermore, we perform linear regression to the line ratio vs H α surface brightness relation in narrow annuli forindividual galaxies. For each galaxy, we get the slope of thelinear regression at three radial bins: [0.4 R e , 0.6 R e ], [0.8 R e ,1.0 R e ], and [1.3 R e , 1.5 R e ]. The distribution of slopes for dif-ferent line ratios at three radii are shown in black, blue, andred in Figure 11. A negative slope means the line ratio isenhanced in DIG while a positive slope means the line ratiois lower in DIG. We see that the slope peaks around ∼ -0.3for log [S ii ]/H α , log [N ii ]/H α and log [O ii ]/H β . The slopesfor these three ratios are rarely positive. The slope peaks at-1.0 for log [O i ]/H α , which is the most significant amongthe line ratios explored here. The slopes for [O iii ]/[O ii ] ismostly positive, and its dispersion is larger than the dis-persion for [S ii ]/H α , [N ii ]/H α , and [O ii ]/H β . The largerdispersion tells us that the ionization parameter varies sig-nificantly. The slope distribution for ∆ log [O iii ]/H β vs ∆log Σ H α peaks at ∼
0, but the distribution is skewed to neg-ative values. There are more galaxies with higher [O iii ]/H β in DIG than galaxies with lower [O iii ]/H β in DIG. The dis-persion is even larger than that for [O iii ]/[O ii ]. This illus-trates that [O iii ]/H β is very diverse in DIG. As shown inSection 3.3, DIG shows higher [O iii ]/H β than H ii regionsin some galaxies, while it shows lower or similar [O iii ]/H β for other galaxies. Figure 11 demonstrates that DIG withhigher [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α and lower[O iii ]/[O ii ] than H ii regions is prevalent among all galax-ies in our sample. [O iii ]/H β could be significantly higher orlower than H ii regions. The contamination of DIG influencesevery star-forming galaxy. Σ H α relation: Split by Stellar Mass In Figure 12 and Figure 13, we show the ∆ log line ratiovs ∆ log Σ H α relations for galaxies with stellar mass lessthan 10 . and higher than 10 . respectively. These arethe one third least massive and the one third most mas-sive galaxies in our sample. The slopes of ∆ log [S ii ]/H α ,[N ii ]/H α , [O ii ]/H β , [O i ]/H α , and [O iii ]/[O ii ] vs ∆ log Σ H α relations do not depend much on stellar mass. However, mas-sive galaxies show a significantly negative ∆ log [O iii ]/H β vs ∆ log Σ H α relation while less massive galaxies show apositive ∆ log [O iii ]/H β vs ∆ log Σ H α relation. In otherwords, [O iii ]/H β of DIG is always enhanced relative to H ii regions in the most massive galaxies. A leaky H ii regionmodel can not produce high [O iii ]/H β relative to H ii re-gions (Section 4.2). The dependence of the ∆ log [O iii ]/H β vs ∆ log Σ H α relation on stellar mass may indicate differ-ent ionization mechanisms in galaxies of different masses.We leave the study of the physical reason behind the depen-dence of ∆ log line ratios vs ∆ log Σ H α relation on stellarmass for future studies. ii regiongrids In Figure 14, we plot the line ratios for each spaxel onthe [O iii ]/[O ii ] vs [N ii ]/[O ii ] diagram (Dopita et al. 2000).[N ii ]/[O ii ] is sensitive to metallicity because N is a sec-ondary element. At high metallicity, [O ii ] is suppressed due MNRAS000
0, but the distribution is skewed to neg-ative values. There are more galaxies with higher [O iii ]/H β in DIG than galaxies with lower [O iii ]/H β in DIG. The dis-persion is even larger than that for [O iii ]/[O ii ]. This illus-trates that [O iii ]/H β is very diverse in DIG. As shown inSection 3.3, DIG shows higher [O iii ]/H β than H ii regionsin some galaxies, while it shows lower or similar [O iii ]/H β for other galaxies. Figure 11 demonstrates that DIG withhigher [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α and lower[O iii ]/[O ii ] than H ii regions is prevalent among all galax-ies in our sample. [O iii ]/H β could be significantly higher orlower than H ii regions. The contamination of DIG influencesevery star-forming galaxy. Σ H α relation: Split by Stellar Mass In Figure 12 and Figure 13, we show the ∆ log line ratiovs ∆ log Σ H α relations for galaxies with stellar mass lessthan 10 . and higher than 10 . respectively. These arethe one third least massive and the one third most mas-sive galaxies in our sample. The slopes of ∆ log [S ii ]/H α ,[N ii ]/H α , [O ii ]/H β , [O i ]/H α , and [O iii ]/[O ii ] vs ∆ log Σ H α relations do not depend much on stellar mass. However, mas-sive galaxies show a significantly negative ∆ log [O iii ]/H β vs ∆ log Σ H α relation while less massive galaxies show apositive ∆ log [O iii ]/H β vs ∆ log Σ H α relation. In otherwords, [O iii ]/H β of DIG is always enhanced relative to H ii regions in the most massive galaxies. A leaky H ii regionmodel can not produce high [O iii ]/H β relative to H ii re-gions (Section 4.2). The dependence of the ∆ log [O iii ]/H β vs ∆ log Σ H α relation on stellar mass may indicate differ-ent ionization mechanisms in galaxies of different masses.We leave the study of the physical reason behind the depen-dence of ∆ log line ratios vs ∆ log Σ H α relation on stellarmass for future studies. ii regiongrids In Figure 14, we plot the line ratios for each spaxel onthe [O iii ]/[O ii ] vs [N ii ]/[O ii ] diagram (Dopita et al. 2000).[N ii ]/[O ii ] is sensitive to metallicity because N is a sec-ondary element. At high metallicity, [O ii ] is suppressed due MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 3. [S ii ]/H α as a function of radius. The dispersion at fixed radius is about 0.2 dex. But when the dots are color-coded byH α surface brightness as shown in the colorbar, we see a beautiful rainbow pattern. The dispersion is significantly reduced at fixed Σ H α .This is because DIG that dominates the low Σ H α region has higher [S ii ]/H α . The red and blue bars show the typical line ratio errorat logΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . We show inFigure 11 that the impact of DIG is prevalent in all star-forming galaxies in our sample, not only in the three galaxies we show. Figure 4. [N ii ]/H α as a function of radius. We see [N ii ]/H α generally drops toward large radius due to a metallicity gradient. Thedispersion at fixed radius is about 0.2 dex. But when the dots are color-coded by H α surface brightness as shown in the colorbar, wesee a beautiful rainbow pattern. This shows [N ii ]/H α is enhanced in DIG. The red and blue bars show the typical line ratio error atlogΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . Figure 5. [O ii ]/H β as a function of radius. The dots are color-coded by Σ H α . We see rainbow patterns as in Figure 4, indicating[O ii ]/H β is enhanced in DIG. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ .MNRAS , 1–30 (2016) K. Zhang et al.
Figure 6. H β /H α as a function of radius. The dots are color-coded by Σ H α . We don’t see any dependence of H β /H α on Σ H α at a givenradius, implying the extinction correction is not the reason for the different line ratios of DIG and H ii regions. The red and blue barsshow the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxel. The length of theerror bars is 2 σ . Figure 7. [O i ]/H α as a function of radius. The dots are color-coded by Σ H α . We see rainbow patterns as in Figure 4, indicating[O i ]/H α is enhanced in DIG. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . Figure 8. [O iii ]/[O ii ] as a function of radius. The dots are color-coded by Σ H α . At a given radius, DIG shows lower [O iii ]/[O ii ], meaninglower ionization parameter. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . MNRAS000
Figure 6. H β /H α as a function of radius. The dots are color-coded by Σ H α . We don’t see any dependence of H β /H α on Σ H α at a givenradius, implying the extinction correction is not the reason for the different line ratios of DIG and H ii regions. The red and blue barsshow the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxel. The length of theerror bars is 2 σ . Figure 7. [O i ]/H α as a function of radius. The dots are color-coded by Σ H α . We see rainbow patterns as in Figure 4, indicating[O i ]/H α is enhanced in DIG. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . Figure 8. [O iii ]/[O ii ] as a function of radius. The dots are color-coded by Σ H α . At a given radius, DIG shows lower [O iii ]/[O ii ], meaninglower ionization parameter. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . MNRAS000 , 1–30 (2016) DSS IV-MaNGA: Impact of DIG Figure 9. [O iii ]/H β as a function of radius. The dots are color-coded by Σ H α . At a given radius, DIG might show higher or lower[O iii ]/H β than H ii regions depending on the specific situation of a galaxy. The red and blue bars show the typical line ratio error atlogΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . Figure 10. ∆ log (line ratios) vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for all galaxies. We normalize the log line ratios vs log Σ H α relation by subtracting the median log line ratio and median log Σ H α . All the spaxels in a galaxy are weighted by N , where N is thenumber of valid spaxels in this galaxy. Contour levels are 5, 40, 70, 85, and 92 percentile. [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α arehigher in DIG dominated low Σ H α regions. [O iii ]/[O ii ] is slightly lower in DIG while [O iii ]/H β is mildly higher in DIG. to the decrease in temperature. [O iii ]/[O ii ] is a very goodproxy of the ionization parameter. We over-plot the mostup-to-date grids for H ii regions from Dopita et al. (2013)with κ = Inf , which means the electrons obey the Maxwell-Boltzmann distribution. Our data fall within the grids. How-ever, when we turn to the BPT diagrams in Figure 15 and 16, we see that while the high H α surface brightness spaxelsare still consistent with the grid prediction, DIG dominatedlow H α surface brightness regions are located outside themodel grids. Specifically, the [N ii ]/H α and [S ii ]/H α are sig-nificantly enhanced in DIG. Previous work found a similarphenomenon for DIG (e.g., Galarza et al. 1999; Hoopes & MNRAS , 1–30 (2016) K. Zhang et al.
Figure 11.
The distribution of ∆ log line ratios vs ∆ log Σ H α slope at three radii. The black line is for the radius at [0.4 R e , 0.6 R e ], theblue line is for the radius at [0.8 R e , 1.0 R e ], and the red line is for the radius at [1.3 R e , 1.5 R e ]. The green dashed line is the slope=0 linefor reference. For almost all star-forming galaxies in our sample, DIG shows higher [S ii ]/H α [N ii ]/H α [O ii ]/H β , [O i ]/H α and lower[O iii ]/[O ii ] than H ii regions. [O iii ]/H β in DIG can be higher or lower than in H ii regions. Walterbos 2003; Kaplan et al. 2016). This means that H ii region models with only a lower q can not explain the lowionization line ratios like [S ii ]/H α and [N ii ]/H α . Apart frommetallicity and ionization parameter, there must be otherparameter(s) that govern the behavior of the emission line.The location of DIG on the BPT diagrams gives us someclue to what the other parameters could be. DIG enters thecomposite or LI(N)ER regions on the [O iii ]/H β vs [N ii ]/H α and [O iii ]/H β vs [S ii ]/H α diagrams. AGN and LI(N)ERare characterized by their hard ionizing spectrum. A hardspectrum is capable of producing a large partially-ionizedregion which enhances the [S ii ]/H α and [N ii ]/H α , just likewhat we observe. Thus, a harder ionizing spectrum is onepossible answer to the line ratios enhancement. To get a quantitative idea of how hardness can change lineratios, we use CLOUDY (Ferland et al. 1998) to calculatenew grids for a series of incident spectra. We test two modelsthat are capable of producing a hard ionizing spectrum: aleaky H ii region model and a hot evolved star model. Theleaky H ii region model can have a harder ionizing spectrumbecause the lower energy part is more likely to be absorbedby the IGM (e.g. Giammanco et al. 2005). Hot evolved stars,like pAGB stars, are characterized by very high tempera- tures, yielding a hard ionizing spectrum. The input spectrato test the two models include (see Figure 17):(1) The spectrum of an O star with T eff = 42 , K ,log g=4.22, solar metallicity. (Tlusty OSTAR2002, Lanz &Hubeny 2003; magenta line)(2) Spectrum in (1) filtered through a gas cloud with n e =100 cm − and N H = 10 cm − . The resulting spectrumincludes both the transmitted spectrum and the emissionfrom the cloud itself. (orange line)(3) Spectrum in (1) filtered through a gas cloud with n e =100 cm − and N H = 10 . cm − (green line)(4) Spectrum in (1) filtered through a gas cloud with n e =100 cm − and N H = 10 . cm − (cyan line)(5) Spectrum in (1) filtered through a gas cloud with n e =100 cm − and N H = 10 . cm − (blue line)(6) The spectrum of a 13 Gyr
SSP generated with BC03.(red line)The metallicity of the ionized cloud is that of OrionNebula ( Baldwin et al. 1991; Rubin et al. 1991; Osterbrocket al. 1992; Rubin et al. 1993). The density of the cloud formodel (1) is 10 cm − , and 1 cm − for model (2), (3), (4),(5) and (6). We run the calculation for log U =-4.5 to -2.0with an interval of 0.5 dex ( log U = log q/c , c is the speedof light). MNRAS000
SSP generated with BC03.(red line)The metallicity of the ionized cloud is that of OrionNebula ( Baldwin et al. 1991; Rubin et al. 1991; Osterbrocket al. 1992; Rubin et al. 1993). The density of the cloud formodel (1) is 10 cm − , and 1 cm − for model (2), (3), (4),(5) and (6). We run the calculation for log U =-4.5 to -2.0with an interval of 0.5 dex ( log U = log q/c , c is the speedof light). MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 12.
The ∆ log (line ratios) vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for the one third most massive galaxies. We normalize thelog line ratios vs log Σ H α relation by subtracting the median log line ratio and median log Σ H α . All the spaxels in a galaxy are weightedby N , where N is the number of valid spaxels in the galaxy. [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α are higher in DIG dominated lowΣ H α regions. [O iii ]/[O ii ] is slightly lower in DIG while [O iii ]/H β is significantly higher in DIG. [O iii ]/H β in DIG of massive galaxies isdifferent from that in less massive galaxies. Model (1) represents a typical H ii region. Models (2),(3), (4), and (5) are for testing the leaky H ii region model.The hardening of the ionizing spectrum is obvious whencompared with (1). Model (6) simulates LI(N)ER-like ioniza-tion by an old stellar population which has been proposedto explain the emission seen in a large fraction of passivegalaxies (e.g., Sarzi et al 2006; Stasi´nska et al. 2008; Yan &Blanton 2012; Kehrig et al. 2012; Singh et al. 2013; Gomeset al. 2016; Belfiore et al. 2016a,b). Evolved stellar popu-lations have a very hard spectrum. In Figure 18, we showthe diagnostic diagrams for the different incident spectralisted above. The O star model lies near the Kauffmann de-marcation on the [O iii ]/H β vs [N ii ]/H α and [O iii ]/H β vs[S ii ]/H α diagrams as expected. The 13 Gyr
SSP grid showssignificant higher [S ii ]/H α , [N ii ]/H α , and [O ii ]/H α relativeto H ii regions with the same ionization parameters due to amore extended partially-ionized region. Models (2) and (3)show a harder ionizing spectrum than H ii region, but theirline ratios are similar to those for H ii regions. Model (4) ex-hibits enhancement in [N ii ]/H α , [S ii ]/H α , and [O ii ]/H α rel-ative to H ii regions. The enhancements are small, typicallyless than 0.2 dex. However, model (5), which has the hard-est spectrum blueward of 912˚A shows the lowest [N ii ]/H α ,[S ii ]/H α , and [O ii ]/H α . This is because even though thosephotons lower than 13.6 eV cannot ionize Hydrogen, they have enough energy to excite an electron in the Hydrogenatom from ground level (n=1) to n=3 or n=4 level. Whenthey jump back, they can jump to n=2 which will produceH α and H β lines. This is fluorescent production. In model(5), photons pass through a very high column density whichabsorbs more than 99.9% of all the ionizing photons. Whenthe transmitted spectrum is forced to have log U > − . α due to fluorescent production that givesrise to lower [N ii ]/H α , [S ii ]/H α , [O ii ]/H α , and [O i ]/H α .Considering the large amount of ionizing photons fromOB stars and the spatial correlation of DIG and H ii regions,leaky H ii regions are likely a major mechanism for produc-ing DIG. However, our results disfavor leaky H ii region mod-els to account for ionization of ALL DIG. DIG has a lowerionization parameter than H ii regions, a decrease in ioniza-tion parameter leads to enhancement of [N ii ]/H α , [S ii ]/H α ,[O ii ]/H α , and [O i ]/H α , but a decrease in [O iii ]/H β atthe same time. Leaky H ii region models can not produceLI(N)ER-like emission. The cyan line shows enhancementof these four line ratios, but it needs fine tuning of the filter-ing column density. DIG that shows LI(N)ER-like emissionneeds another ionization source. We favor evolved stars as amajor ionization source for DIG because only ionization by MNRAS , 1–30 (2016) K. Zhang et al.
Figure 13.
The ∆ log (line ratios) vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for the one third least massive galaxies. We normalize thelog line ratios vs log Σ H α relation by subtracting the median log line ratio and median log Σ H α . All the spaxels in a galaxy are weightedby N , where N is the number of valid spaxels in the galaxy. [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α are higher in DIG dominated lowΣ H α regions. [O iii ]/[O ii ] is slightly lower in DIG while [O iii ]/H β is slightly lower in DIG. Figure 14. [O iii ]/[O ii ]vs [N ii ]/[O ii ] diagram (Dopita et al. 2000, 2013). The grids are from Dopita et al. (2013) with κ = Inf . Thenearly-horizontal lines are for constant ionization parameters, and the nearly vertical lines are for constant metallicities. The labelsdenote the metallicities and ionization parameters (log q) for the grid. evolved stars (red line, Model 6) can produce enhancementof [S ii ]/H α , [N ii ]/H α , [O ii ]/H α , [O i ]/H α , and [O iii ]/H β even when the ionization parameter drops. Evolved stars areprevalent all over galaxies, and their contribution to ioniz-ing photons may prevail. Flores-Fajard et al. (2011) pro-posed hot low-mass evolved stars (HOLMES) as an impor- tant ionization source for the extra-planar diffuse ionizedgas in edge-on galaxies. For NGC 891, HOLMES begin tocontribute more than 50% of the ionizing photons at scaleheight | z | > MNRAS000
The ∆ log (line ratios) vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for the one third least massive galaxies. We normalize thelog line ratios vs log Σ H α relation by subtracting the median log line ratio and median log Σ H α . All the spaxels in a galaxy are weightedby N , where N is the number of valid spaxels in the galaxy. [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α are higher in DIG dominated lowΣ H α regions. [O iii ]/[O ii ] is slightly lower in DIG while [O iii ]/H β is slightly lower in DIG. Figure 14. [O iii ]/[O ii ]vs [N ii ]/[O ii ] diagram (Dopita et al. 2000, 2013). The grids are from Dopita et al. (2013) with κ = Inf . Thenearly-horizontal lines are for constant ionization parameters, and the nearly vertical lines are for constant metallicities. The labelsdenote the metallicities and ionization parameters (log q) for the grid. evolved stars (red line, Model 6) can produce enhancementof [S ii ]/H α , [N ii ]/H α , [O ii ]/H α , [O i ]/H α , and [O iii ]/H β even when the ionization parameter drops. Evolved stars areprevalent all over galaxies, and their contribution to ioniz-ing photons may prevail. Flores-Fajard et al. (2011) pro-posed hot low-mass evolved stars (HOLMES) as an impor- tant ionization source for the extra-planar diffuse ionizedgas in edge-on galaxies. For NGC 891, HOLMES begin tocontribute more than 50% of the ionizing photons at scaleheight | z | > MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 15. [O iii ]/H β vs [N ii ]/H α diagram for each galaxy. Different colors represent different Σ H α . The solid and dashed lines are thedemarcation line from Kauffmann et al. (2003) and Kewley et al. (2001). The grids are from Dopita et al. (2013) with κ = Inf . DIGcan not be covered by the grids, suggesting HII region models with variations of metallicity and ionization parameter can not producethe DIG line ratios.
Figure 16. [O iii ]/H β vs [S ii ]/H α diagram for each galaxy. Different colors represent different Σ H α . The solid lines are the demarcationlines to separate Seyfert galaxies from LINER in Kewley et al. (2006). The grids are from Dopita et al. (2013) with κ = Inf . The gridsdo not cover DIG, suggesting HII region models with variations of metallicity and ionization parameter can not produce the DIG lineratios.
We have shown in last subsection that a 13
Gyr
SSP canproduce the LI(N)ER-like emission we see for DIG. One in-teresting question is: How would the line ratios change asa SSP ages? To answer this question, we generate SSPs at1
Myr , 3
Myr , 9
Myr , 0.125
Gyr , 7
Gyr , and 13
Gyr usingBC03, and use these spectra as the incident ionizing spec-tra for CLOUDY. The spectra from 0.125
Gyr to 13
Gyr change only slightly. The metallicity is that of the OrionNebula, and the density is 1 cm − . We run the calculationfor log U =-4.5 to -2.0 with an interval of 0.5 dex. The inci-dent spectra are shown in Figure 19. The output line ratiosare plotted in the same color as their incident spectra inFigure 20. At 1 Myr , the line ratios are located at the H ii region part on the diagnostic diagrams. At 3 Myr , the lineratios are similar to those of 1
Myr . At 9
Myr , however, wesee a significant decrease in [N ii ]/H α , [S ii ]/H α , [O ii ]/H α ,[O iii ]/H β , [O iii ]/[O ii ], and an increase in [N ii ]/[O ii ]. Thisis due to the significant decrease in 200-500˚A spectrum hard- ness as OB stars age (Levesque et al. 2010). At 125 Myr ,the line ratios are already in the LI(N)ER/AGN region. Thismeans once OB stars die, the line ratios of the ionized gaswould turn into “LI(N)ER/AGN like” very fast. The line ra-tios do not change very much afterwards. This sheds light onthe interpretation of LI(N)ER/AGN like emission line ratiosin galaxies. When OB stars are alive, their light dominatesthe ionizing spectrum and the ionized gas located in thestar-forming regions dominates on the diagnostic diagrams.When OB stars cease to form, their contribution disappearsvery fast, and the light of hot evolved stars begins to dom-inate after tens of
Myr , and the ionized gas is located inthe LI(N)ER/AGN regions on the diagnostic diagrams. Wehave shown in Section 4.2 that a filtered spectrum can notreproduce the DIG line ratios with an increase of [O iii ]/H β .A higher [O iii ]/H β in DIG than in H ii regions means theionization of DIG may not be linked to OB stars directlybut linked to the hot evolved stars.We are not claiming that evolved stellar populationsare the major ionizing source of DIG. Overall, the radiative MNRAS , 1–30 (2016) K. Zhang et al.
Figure 17.
The incident spectra for our models described in Section 4.2. All spectra are normalized to the flux at 1000˚A. The magentaline is a 42,300K, log g=4.22, solar metallicity O star generated by Tlusty (Lanz & Hubeny 2003). The orange, green, cyan, and bluelines are the spectra by filtering the O star spectrum through column density of 10 cm − , 10 . cm − , 10 . cm − , 10 . cm − ,simulating the leaky H ii regions. The red line is the spectrum of a 13Gyr Simple Stellar Population generated by BC03. Figure 18.
The BPT diagrams, log [O iii ]/H β vs log [O i ]/H α diagram, log [O iii ]/[O ii ] vs log [N ii ]/[O ii ] diagram, and log [O iii ]/[O ii ]vs log [O ii ]/H α diagrams for different incident spectra shown in Figure 17. The color scheme is the same as in Figure 17. The dots withthe same color have different ionization parameters from log U=-4.5 to log U=-2 with an interval of 0.5 dex. The leaky H ii region models(orange, green, cyan lines) show very similar line ratios to the O star model (magenta line) at the same ionization parameter. We notethat DIG has lower ionization parameters than HII regions. Leaky H ii region models can not produce enhancement of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β with a decrease in ionization parameter in most cases. The cyan line shows enhancement of these three lineratios, but it needs fine tuning of the filtering column density. The blue line even shows decreases in [S ii ]/H α [N ii ]/H α [O ii ]/H α and[O iii ]/H β , because of large H α florescent production, which is unphysical. Only ionization by evolved stellar populations (red line) canproduce enhancement of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β even when the ionization parameter drops.MNRAS000
The BPT diagrams, log [O iii ]/H β vs log [O i ]/H α diagram, log [O iii ]/[O ii ] vs log [N ii ]/[O ii ] diagram, and log [O iii ]/[O ii ]vs log [O ii ]/H α diagrams for different incident spectra shown in Figure 17. The color scheme is the same as in Figure 17. The dots withthe same color have different ionization parameters from log U=-4.5 to log U=-2 with an interval of 0.5 dex. The leaky H ii region models(orange, green, cyan lines) show very similar line ratios to the O star model (magenta line) at the same ionization parameter. We notethat DIG has lower ionization parameters than HII regions. Leaky H ii region models can not produce enhancement of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β with a decrease in ionization parameter in most cases. The cyan line shows enhancement of these three lineratios, but it needs fine tuning of the filtering column density. The blue line even shows decreases in [S ii ]/H α [N ii ]/H α [O ii ]/H α and[O iii ]/H β , because of large H α florescent production, which is unphysical. Only ionization by evolved stellar populations (red line) canproduce enhancement of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β even when the ionization parameter drops.MNRAS000 , 1–30 (2016) DSS IV-MaNGA: Impact of DIG and mechanical energy from hot evolved stars falls shortof the energy budget of diffuse ionized gas in star-forminggalaxies (Reynolds 1984; Ferguson et al. 1996; Binette etal. 1994). Even if the energy emitted by hot evolved starsmeets the energy requirement of DIG, O stars can provideat least one order of magnitude more ionizing photons thanhot evolved stars in star-forming galaxies (Reynolds 1984;Ferguson et al. 1996). Indeed, DIG is found to be linkedto H ii regions both along the disk and in vertical direction(e.g. Ferguson et al. 1996; Zurita et al. 2000; 2002; Rossa& Dettmar 2003a,b). In star-forming galaxies, hot evolvedstars cannot compete with O stars in total ionizing photonsproduction. However, we see in two cases that hot evolvedstars might be a major contributor of ionizing photons. 1) Inlow surface brightness regions that are located far away fromH ii regions. The density of ionizing photons from the H ii regions has dropped significantly so the hot evolved starsbegin to dominate. One example is in regions at large verti-cal height (extra-planar gas at high | z | ). Other low surfacebrightness regions that show higher [O iii ]/H β than H ii re-gions probably belong to this category as well. 2) In galax-ies where OB stars have died: post-starburst galaxies andquiescent/passive galaxies. Quiescent galaxies are known toshow LI(N)ER emissions (Yan et al. 2006; Sarzi et al. 2006;Stasi´nska et al. 2008; Kehrig et al. 2012; Singh et al. 2013;Gomes et al. 2016; Belfiore et al. 2015, 2016a,b). We predictDIG to be prevalent in post-starburst galaxies, and they willshow LI(N)ER/AGN like emission. There are several methods to measure gas-phase metallicity in star-forming galaxies usingstrong lines in the optical, such as N2=[N ii ]/H α , R =([O ii ] λ iii ] λ iii ] λ β ,O3N2=(([O iii ]/H β )/([N ii ]/H α )) (e.g., Alloin et al.1979; Zaritsky et al. 1994; Pilyugin, 2001; Denicol´o et al.2002; Pettini & Pagel, 2004; Pilyugin & Thuan, 2005)and N2O2=([N ii ]/[O ii ]) (Dopita et al. 2000; Kewley et al.2001), N2S2H α =8.77+log [N ii ]/[S ii ] + 0.264 × log[N ii ]/H α (Dopita et al. 2016). If the line ratios of DIG are differentfrom these in H ii regions, the metallicity measurement willinevitably be biased because these metallicity indicatorsare a combination of several line ratios. The bias dependson the fraction of emission contributed by DIG, whichincreases towards low surface brightness regions. ii ]/ H α If we use N2 to derive the metallicity, the enhancement seenin DIG means that the metallicity will be overestimated inthose spaxels with a high DIG fraction. This is especiallyimportant for observation of galaxies at high redshift whenonly the emission lines near H α are available. We see a typ-ical [N ii ]/H α enhancement of 0.2 dex. We use the equationin Pettini & Pagel (2004) to convert [N ii ]/H α to metallicity:12+ log ( O/H ) = 9 . . × N . × N +0 . × N ,N2=log([N ii ]/H α ). Assuming a H ii region has N2=-0.5,Z(N2)=8.63 in a local galaxy, we obtain Z(N2)=8.865 forDIG with the same metallicity because N2=-0.3. The typical metallicity gradient for a local star-forming galaxy is about-0.1 dex R e − (S´anchez et al. 2014; Ho et al. 2015), mean-ing metallicity drops by ∼ R e . Thebias due to DIG would flatten the metallicity gradient from-0.1 dex R e − to 0 if DIG totally dominates at 2 R e . If weassume a metal poor H ii region has N2=-1.5, Z(N2)=8.08in a high redshift galaxy, DIG with the same metallicity hasN2=-1.3 and we would obtain Z(N2)=8.157. The metallic-ity gradient will be flattened as well. N2 is usually used formeasuring metallicity and metallicity gradients at high red-shift (e.g., Wuyts et al. 2016). Our result indicates that thepresence of DIG potentially flattens N2 derived metallicitygradients at high redshift. ii ]/[O ii ] N2O2 is a good metallicity indicator because it is not sensi-tive to ionization parameter or ionizing spectrum hardnessbut to metallicity (Dopita et al. 2000, 2013; Kewley et al.2002). The caveat of N2O2 is that it relies on the Nitrogen-to-Oxygen abundance ratio (N/O). Nitrogen and α elementshave different enrichment timescales, hence their ratio de-pends on the star formation history and several other pa-rameters (e.g. Vincenzo et al. 2016) as well as mixing issues(e.g. Belfiore et al. 2015), especially for low mass galaxiesand galaxy outskirts, where there are prominent variationsof N/O vs O/H relative to metal rich systems and centralregions. Moreover, these diagnostics are sensitive to metal-licity only at 12+log(O/H) > < ii ]/[O ii ] as a function of effective radius andcolor-code the dots with H α surface brightness in Figure 21.N2O2 is converted to metallicity using log ( O/H ) + 12 =log[1 . . × R + 0 . × R ] + 8 .
93, R=log[N ii ]/[O ii ] (Dopita et al. 2013). At fixed radius, Z(N2O2)in the low surface brightness bin is similar to or a little bithigher than that in the high surface brightness bin. As statedin Section 3.2, [N ii ]/[O ii ] only depends on the N/O abun-dance ratio and temperature (Dopita et al. 2000, 2013). Itis not subject to the ionization parameter and ionizing spec-trum shape variation. This makes it an excellent metallicityindicator. The weak dependence of [N ii ]/[O ii ] on H α surfacebrightness at fixed radius (Figure 21) demonstrates it is agood metallicity indicator even in the presence of DIG. How-ever, we note that N2O2 is subject to variation of N/O ratio(P´erez-Montero & Contini 2009; P´erez-Montero et al. 2013,2016), temperature variation, and uncertainty of extinctioncorrection. Z(N2O2) could be enhanced or suppressed forDIG in some regions, maybe due to N/O variation and tem-perature variation. The measurement of extinction involvesan accurate measurement of H β which is not an easy tasksince the stellar continuum shows deep H β absorption. Thisis even harder when the emission line is weak and S/N islow, such as in the outskirts of galaxies. Besides, H α andH β arise from the fully ionized region, while [N ii ] and [O ii ]are from the partially-ionized region. The extinction derivedusing H α and H β may not necessarily be the same as theextinction experienced by the partially-ionized region. FromFigure 6, we don’t see any signs of deviation of H β /H α fordifferent H α surface brightness spaxels, so the result that MNRAS , 1–30 (2016) K. Zhang et al.
Figure 19.
The incident spectra for SSP at 0.001 Gyr, 0.003 Gyr, 0.009 Gyr, 0.125 Gyr, 7 Gyr, and 13 Gyr with solar metallicity,normalized to flux at 6000˚A. We can see that OB stars dominate the ionizing spectrum in the beginning, and wane after tens of Myr.After that the ionizing spectrum is dominated by emission from evolved stellar populations and it hardens.
Figure 20.
BPT diagrams, log [O iii ]/H β vs log [O i ]/H α , log [O iii ]/[O ii ] vs log [N ii ]/[O ii ], and log [O iii ]/[O ii ] vs log [O ii ]/H α , fordifferent incident spectra of SSPs at different ages. The color scheme is the same as Figure 19. The dots with the same color have differentionization parameters from log U=-4.5 to log U=-2 with an interval of 0.5 dex. At Age < tens of Myr, the ionizing spectrum softens as OBstars evolve and die. [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β all decrease. After the OB stars die, the ionizing spectrum is dominatedby hot evolved stars, and the spectrum hardens. We see increases of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O i ]/H α . This figure shows howthe ionized gas moves on the diagnostic diagrams as the stellar population ages. MNRAS000
BPT diagrams, log [O iii ]/H β vs log [O i ]/H α , log [O iii ]/[O ii ] vs log [N ii ]/[O ii ], and log [O iii ]/[O ii ] vs log [O ii ]/H α , fordifferent incident spectra of SSPs at different ages. The color scheme is the same as Figure 19. The dots with the same color have differentionization parameters from log U=-4.5 to log U=-2 with an interval of 0.5 dex. At Age < tens of Myr, the ionizing spectrum softens as OBstars evolve and die. [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O iii ]/H β all decrease. After the OB stars die, the ionizing spectrum is dominatedby hot evolved stars, and the spectrum hardens. We see increases of [S ii ]/H α [N ii ]/H α [O ii ]/H α and [O i ]/H α . This figure shows howthe ionized gas moves on the diagnostic diagrams as the stellar population ages. MNRAS000 , 1–30 (2016) DSS IV-MaNGA: Impact of DIG DIG have similar N2O2 as H ii region is robust. We see inFigure 6 that H β /H α shows a gradient with positive slopeeven though it does not depend on Σ H α . The gradient inH β /H α will change the Z(N2O2) gradient slope. To correctthis, we suggest getting an H β /H α gradient by fitting theH β /H α of individual spaxels and apply the overall extinc-tion gradient to [N ii ]/[O ii ] to get the correct metallicitygradient. In this way, we avoid the very noisy H β /H α whenΣ H α is low when correcting extinction.A key for probing abundance is to use lines from dif-ferent elements with the same ionization potential. Theionization potential for H , O , O + , O ++ , S + and N + are 13.598eV, 13.618eV, 35.121eV, 54.936eV, 23.337eV, and29.601eV respectively. Ionization potential of [O ii ] and [N ii ]are similar: 35.121eV and 29.601eV. This makes [N ii ]/[O ii ]insensitive to ionization parameter, thus a good metallicityindicator. The complexity in using [N ii ]/[O ii ] is the extinc-tion correction and the N/O ratio variation at fixed O/H.To quantify the change of the metallicity measurementsas a function of H α surface brightness, we do a linear regres-sion fit to the derived metallicities vs H α surface brightnessrelation in a narrow annulus. The distribution of slopes fordifferent metallicity indicators at three radii are shown inblack, blue, and red in Figure 29. The metallicity is inde-pendent of surface brightness. The reason we are seeing adependence of metallicity on H α surface brightness is due tocontamination by DIG. For Z(N2O2), the slope distributionis narrow, and peaks at ∼ R R =([O ii ] λ iii ] λ iii ] λ β is acommonly used metallicity indicator (McGaugh 1991;Zaritsky et al. 1994; Pilyugin 2001; Kewley & Dopita 2002;Kobulnicky & Kewley 2004). The advantage is it does notdepend on the N/O ratio. The disadvantage is that themetallicity is double − valued at a given R . In order to use R to measure Z, one has to first determine the ionizationparameter and which branch it is on. Therefore, it alwaysneeds to be used in conjunction with [O iii ]/[O ii ] and N2O2(or similar). We use the method of Kewley & Ellison (2008)to determine which branch a spaxel is on and derive themetallicity. Figure 22 show how the Z derived from R changes with radius and H α surface brightness. One can seethat low SB regions have lower measured Z( R ) values. Wehave shown in Section 3.2 that DIG has higher [O ii ]/H β ,lower [O iii ]/[O ii ], and on average similar [O iii ]/H β as H ii regions. The combination of the three line ratios leads tothe result that Z( R ) is biased to be lower in DIG.For the whole sample, the Z( R ) vs Σ H α slope distri-bution peaks at ∼ R e − . Meanwhile, the slope of Z( R ) vs Σ H α is ∼ H α of DIG at 2 R e is about38 erg s − kpc − . Compared to a Σ H α =39 erg s − kpc − H ii region, the Z( R ) bias for DIG would be 0.2. Note the dropof metallicity from the center to 2 R e is only 0.2 dex. Themetallicity gradient would be -0.2 dex R e − if the outskirtof the galaxy is pure DIG. In other words, the metallicitymeasured using Z( R ) is biased systematically by ∼ -0.1 dex R e − . O3N2 is sensitive to oxygen abundance, and it is not im-pacted by extinction. The disadvantage of O3N2 is it de-pends on N/O (P´erez-Montero & Contini 2009; P´erez-Montero et al. 2013, 2016) and the ionization parameter. InFigure 23, we show how DIG would impact metallicity mea-surements made from O3N2=([O iii ]/H β )/([N ii ]/H α ). Wesee in Figure 23 that O3N2 is higher in DIG for some galax-ies and lower or similar in other galaxies. The contaminationby DIG could be responsible for a substantial portion of thescatter in metallicity measurements. When confined only tothe high surface brightness regions, the metallicity gradi-ent derived using O3N2 is similar to the ones using R orN2O2. To make robust metallicity gradient measurements,one has to properly isolate H ii regions and correct for DIGcontamination.Because the surface brightness generally decreases to-wards large radii, the metallicity gradient derived usingO3N2 at large radii might be flatter or steeper than thatderived using N2O2 if we include all the spaxels. For ex-ample, in MaNGA galaxy 8313-12702, we get a metallicityof -0.15 dex R e − using Z(N2O2). Using only the high sur-face brightness regions (red dots) and Z(O3N2) gives thesame result. However, if we include everything, the metal-licity gradient derived using O3N2 will be -0.1 dex R e − .The bias is significant at least for this galaxy. Z(O3N2) isbiased in the other direction for MaNGA galaxy 8603-12704.Similarly, Mast et al. (2014) found that the presence of DIGwould bias the abundance gradient when using O3N2 to de-rive metallicity. The strength of the bias depends a lot onthe assumed calibrator. We note the degree of bias is notthe same for all galaxies, and we definitely can give a robustmetallicity gradient measurement using high surface bright-ness regions. The σ of the distribution of Z(O3N2) vs Σ H α slope at fixed radius is ∼ ± R e − through a similaranalysis as in Section 5.3, consistent with the different biaseswe see in different galaxies. For a large sample of galaxies(S´anchez et al. 2014, Ho et al. 2015), the bias for differentgalaxies may cancel out when deriving average metallicitygradients. H α Dopita et al. (2016) proposed a new metallicity proxy us-ing only [N ii ] λ ii ] λλ α : 12 +log (O/H)=Z(N2S2H α )=8.77+log [N ii ]/[S ii ] + 0.264 × log[N ii ]/H α . It is especially suitable for metallicity measure-ments of high redshift galaxies whose spectral wavelengthcoverage is limited because the 4 lines used are close to eachother. This estimator is almost linear up to an abundance of12 + log (O/H) = 9.05. The caveat is the calibration of thismetallicity proxy is based on a well defined N/O and O/H re-lation. This calibration fails for any systems deviating from MNRAS , 1–30 (2016) K. Zhang et al.
Figure 21.
Metallicity derived using N2O2 as a function of radius. Z(N2O2) drops with radius due to a metallicity gradient. We see thatthe derived metallicity does not depend on the Σ H α at a fixed radius, and the dispersion at a fixed radius is similar to the measuringerror. Red and blue bars are errors at log Σ H α =39 erg s − kpc − and log Σ H α =38.5 erg s − kpc − respectively. The length of the errorbars is 2 σ . We show in Section 5.7 that the impact of DIG is prevalent in all star-forming galaxies in our sample, not confined to thethree galaxies we show. Figure 22.
Metallicity derived using R as a function of radius. Z( R ) drops with radius due to metallicity gradient. At fixed radius,Z( R ) decreases with Σ H α , indicating Z( R ) is systematically biased low for DIG. The measuring errors at log Σ H α =39 erg s − kpc − and log Σ H α =38.5 erg s − kpc − are indicated using red and blue bars. The dispersion of Z( R ) is about twice that of Z(N2O2). Themetallicity gradient would be systematically biased by -0.1 dex R e − if the bias in DIG is not accounted for. Figure 23.
Metallicity derived using O3N2 as a function of radius. Z(O3N2) drops with radius due to a metallicity gradient. At fixedradius, Z(O3N2) could either decrease or increase with Σ H α . The measuring errors at log Σ H α =39 erg s − kpc − and log Σ H α =38.5erg s − kpc − are indicated using red and blue bars. The dispersion of Z(O3N2) is similar to that of Z(N2O2). The metallicity gradientwould be biased by ± R e − if the bias in DIG is not considered. MNRAS000
Metallicity derived using O3N2 as a function of radius. Z(O3N2) drops with radius due to a metallicity gradient. At fixedradius, Z(O3N2) could either decrease or increase with Σ H α . The measuring errors at log Σ H α =39 erg s − kpc − and log Σ H α =38.5erg s − kpc − are indicated using red and blue bars. The dispersion of Z(O3N2) is similar to that of Z(N2O2). The metallicity gradientwould be biased by ± R e − if the bias in DIG is not considered. MNRAS000 , 1–30 (2016) DSS IV-MaNGA: Impact of DIG the assumed N/O − O/H relation. We explore how DIG bi-ases the metallicities derived using N2S2H α in Figure 24. Atfixed radius, we don’t see a significant offset in Z(N2S2H α )between DIG and H ii regions. This is because Z(N2S2H α ) isinsensitive to variation of ionization parameter and spectralhardness (Dopita et al. 2016), very similar to N2O2. Com-pared with Z(N2O2), the dispersion of Z(N2S2H α ) is larger.This is understandable because:(a) Four lines are used in Z(N2S2H α ) while only 2 lines areused to derive Z(N2O2). The measurement errors enter intothe derivation of metallicities.(b) The relative abundances of N, S, and O are involved inZ(N2S2H α ) while Z(N2O2) only include N/O.According to the upper right panel of Figure 29, The distri-bution of Z(N2S2H α ) vs Σ H α slope is very similar to thatof Z(O3N2) in its centroid and dispersion. The slope distri-bution for Z(N2S2H α ) has a wider wing, meaning there aresome sources that show significant biases in Z(N2S2H α ) forDIG and H ii regions. This is similar to the scatter intro-duced by DIG on O3N2. IZI (Blanc et al. 2015) uses strong nebular emission lines toderive the Bayesian posterior probability density functionfor metallicity and ionization parameter based on a seriesof H ii region models. In Figure 25, we plot the metallic-ities derived using IZI as a function of radius. We input[O ii ] λ iii ] λ iii ] λ α , [N ii ] λ ii ] λλ α /H β assuming a Milky Way extinction curve(Fitzpatrick 1999). We use the values derived with the ”out-put joint mode” in IZI. The dots are color-coded by Σ H α .For high Σ H α regions, IZI gives similar metallicity gradi-ents as the other metallicity indicators. However, at a fixedradius, different surface brightness regions (different colorsin the plots) have large metallicity discrepancies. For lowΣ H α regions, the metallicities derived from IZI can vary by0.2 dex among themselves. This is at least partly because IZIcurrently only adopts H ii region grids. These models makean inherent assumption that the ionizing spectrum shape isfixed by the temperature of the OB stars, which is deter-mined by the metallicity. However, for DIG this assumptiondoes not hold. We have shown that metallicity+low q for anH ii region model can not produce the line ratios we see inDIG. Thus, IZI is very vulnerable to contamination by DIGbecause currently it only contains H ii region models. How-ever, if one were to include a DIG model in IZI and otherBayesian codes to derive metallicities (e.g.,HII-CHI-mistry:P´erez-Montero 2014; BOND: Vale Asari et al 2016) , thenthey may be able to give accurate estimates of metallicityand ionization parameter even with contamination of DIG.Furthermore, we input only extinction − corrected [N ii ]and [O ii ] into IZI to see if IZI could give a less biased metal-licity in DIG. [N ii ] and [O ii ] can only be combined to N2O2to estimate metallicity. Figure 26 shows that the metallic-ity derived through IZI using [N ii ] and [O ii ] only is bet-ter than metallicity derived using all strong emission linesavailable. The dispersion at a fixed radius is smaller and thedependence on Σ H α at a fixed radius is weaker. However,Z(N2O2) in Figure 21 still shows a tighter metallicity gradi- ent than Z(IZI) [N ii ]+[O ii ]. This is because we do not applyextinction corrections to individual spaxels when calculatingZ(N2O2). IZI metallicity with extinction − uncorrected [N ii ]and [O ii ] input alone gives very similar result as Z(N2O2). Σ H α relation: The Whole Sample To quantify the bias introduced by DIG on strong − linemetallicity measurements, we examine the derived metallic-ities as a function of Σ H α for the whole sample. We select allspaxels in [0.4 R e , 0.6 R e ], subtract from all metallicity andlog Σ H α measurements their respective medians, then plotall galaxies together in the same plot. The spaxels in eachgalaxy are weighted by N , where N is the number of validspaxels in this galaxy. Figure 28 shows the weighted ∆ Z vs∆ log Σ H α relation for the whole sample. A non − zero slopeof the relation means metallicities are biased by DIG con-tribution, and the dispersion of this relation in y directionreflects how reliable a metallicity proxy is for an individ-ual galaxy. The ∆Z(N2O2) vs ∆Σ H α relation has a slope of ∼ R ) vs ∆ log Σ H α , ∆Z(O3N2) vs∆ log Σ H α , ∆Z(N2S2H α ) vs ∆log Σ H α ” ∆Z(N2) vs ∆ logΣ H α , relations are ∼ ∼
0, and -0.2. We define thedispersion as the interval between 90% percentiles contoursat ∆ log Σ H α =0. For Z(N2O2), the dispersion is 0.2, and forZ( R ), Z(O3N2), Z(N2S2H α ) and Z(N2), the dispersionsare 0.35, 0.23, 0.37 and 0.25. Based on slopes and disper-sions, Z(N2O2) is optimal because the slope is mild and thedispersion is smallest. Z( R ) has the most significant slopeand large dispersion. Z(O3N2) has smallest slope and a dis-persion only slightly larger than Z(N2O2). Z(N2S2H α ) hasa nearly 0 slope but the dispersion is the largest. Z(N2) isderived using [N ii ]/H α , thus very vulnerable to DIG con-tamination.We also do linear regression to the derived metallicitiesvs log Σ H α relation in narrow radial annuli so the metallic-ity is essentially fixed. The distribution of slopes for differ-ent metallicity indicators at three radii are shown in black,blue, and red in Figure 29. The intrinsic metallicity is in-dependent of surface brightness. The reason we are seeing adependence of derived metallicities on H α surface brightnessis due to contamination by DIG. A slope of 0 means the DIGdoes not impact the metallicity measurements, and a pos-itive slope means DIG biases the metallicity measurementlow. The centroid of the distribution reflects the system-atic offset of that metallicity indicator for the whole sample,while the dispersion of the distribution reflects the metallic-ity error it brings for individual galaxies. For Z(N2O2), theslope distribution is narrow, and peaks at ∼ dex − . Thebias in metallicity using N2O2 is small and the error is smallbecause N2O2 does not vary much among DIG. DIG regionshave marginally lower Z(N2O2), suggesting slightly highertemperature. The slope distribution of Z(N2S2H α ) is verysimilar to that of Z(N2O2), suggesting Z(N2S2H α ) to be arobust metallicity estimator even in the presence of DIG.For Z( R ), the slope distribution peaks at 0.2 dex − anddispersion is large. The contribution of DIG would not onlybias the metallicity measurements systematically, but alsointroduce a large metallicity measurement error. This canbe seen in the tightness of metallicity vs radius relation. For MNRAS , 1–30 (2016) K. Zhang et al.
Figure 24.
Metallicity derived using [N ii ]/[S ii ] and [O ii ]/H α proposed by Dopita et al. (2016) as a function of radius. We see that thederived metallicity does not depend on the Σ H α at a fixed radius, and the dispersion at a fixed radius is similar to the measuring error(red and blue bars are errors at log Σ H α =39 erg s − kpc − and log Σ H α =38.5 erg s − kpc − respectively). Z(N2S2H α ) performs verysimilar to Z(O3N2). Figure 25.
Metallicity derived using IZI by Blanc et al. (2015) as a function of radius. The dots are color-coded by Σ H α . For highΣ H α regions, the metallicity gradients are similar to those derived using other metallicity indicators. However, at fixed radius, differentsurface brightness regions have large metallicity discrepancies. For low Σ H α regions, the metallicities derived from IZI can vary by 0.2 dexthemselves. This demonstrates that IZI is not robust to derive the metallicity of DIG. Note the metallicity lower limit is set to 8, so thecut-off at Z ∼ Figure 26.
Metallicity derived using IZI by Blanc et al. (2015) as a function of radius using only extinction corrected [N ii ] and [O ii ].The impact of DIG is much less than when using all strong emission lines available. The red and blue bars show the typical line ratioerrors at logΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxels. The length of the error bars is 2 σ .MNRAS000
Metallicity derived using IZI by Blanc et al. (2015) as a function of radius using only extinction corrected [N ii ] and [O ii ].The impact of DIG is much less than when using all strong emission lines available. The red and blue bars show the typical line ratioerrors at logΣ H α =39 erg s − kpc − and logΣ H α =38.5 erg s − kpc − for individual spaxels. The length of the error bars is 2 σ .MNRAS000 , 1–30 (2016) DSS IV-MaNGA: Impact of DIG Figure 27.
Metallicity derived using IZI by Blanc et al. (2015) as a function of radius using only extinction corrected [N ii ], [O ii ], and[O iii ]. The result is indistinguishable from the result using only [N ii ] and [O ii ]. The impact of DIG is much less than when using allstrong emission lines available. The red and blue bars show the typical line ratio error at logΣ H α =39 erg s − kpc − and logΣ H α =38.5erg s − kpc − for individual spaxels. The length of the error bars is 2 σ . Z(O3N2), the distribution of slope is broad, and peaks at ∼ - 0.05 dex − . The magnitude of the systematic metallic-ity measurement bias is similar for Z(O3N2) and Z(N2O2),but Z(O3N2) in DIG varies much more significantly. UsingZ(O3N2) will not bias the measurement if we average a largesample of galaxies, but it will inevitably bias the metallic-ity in individual galaxies as shown in Section 5.4. The largevariation of Z(O3N2) and Z( R ) in DIG is expected becausethese two indicators involve ionization parameters. For anH ii region, this does not matter because the metallicity andionization parameter are linked to each other. This assump-tion does not hold for DIG. The variation in the ionizationparameter in DIG enters the metallicity measurement. Theimpact of DIG on using Z(N2), Z( R ) and Z(O3N2) metal-licity gradients is discussed in Section 5.1, Section 5.3 andSection 5.4 respectively. For Z(IZI), the metallicities derivedfor DIG are systematically lower. Besides, the dispersion ofthe slope distribution is broadest, with σ ∼ Σ H α relation: Split by Stellar Mass In Figure 30 and Figure 31, we show the normalized Zvs Σ H α relations for galaxies with stellar mass less than10 . and higher than 10 . respectively. They are theone third least massive and one third most massive galaxiesin our sample. The ∆Z(N2O2), ∆Z( R ), ∆Z(N2S2H α ) and∆Z(N2) vs ∆ log Σ H α relationa are similar in the most mas-sive and least massive galaxies. However, the ∆Z(O3N2) vs∆ log Σ H α relations in the most massive galaxies have pos-itive slopes while those in the least massive galaxies havenegative slopes. This is mostly due to the dependence of the∆ log [O iii ]/H β vs ∆ log Σ H α relation on stellar mass. Thedependence of DIG’s impact on stellar mass means it is cru-cial to take care of DIG contamination when comparing themetallicity and metallicity gradient of galaxies of differentmasses using Z(O3N2).In summary, (i) Metallicities derived using N2O2 are optimal becausethey exhibit the smallest bias and error.(ii) Metallicities derived using the O3N2 or N2S2H α (Do-pita et al. 2016) for DIG can be significantly higher or lowerthan those for H ii regions. Using O3N2 or N2S2H α to de-rive metallicity can bias the metallicity gradient by ± R e − for an individual galaxy if the contamination by DIGis not accounted for. For a large sample of galaxies (S´anchezet al. 2014, Ho et al. 2015), the bias for different galaxiesmay cancel out when deriving average metallicity gradients.(iii) R derived metallicities for DIG are lower than thosefor H ii regions due to a lower ionization parameter. Using R to derive metallicity will systematically bias the metal-licity gradient by ∼ -0.1 R e − because of DIG.(iv) Using N2=[N ii ]/H α to derive metallicities will sys-tematically bias the metallicity gradient by ∼ R e − , considering that DIG typically shows 0.2 dex higher[N ii ]/H α .(v) IZI works well for the H ii region dominated regions,but fails for deriving metallicities of DIG, probably becauseIZI currently only contains H ii region models. For the local universe, we usually have [O ii ], [O iii ], H β ,[N ii ], H α , and [S ii ] to derive metallicities. Recent worksfind that the local star-forming galaxies exhibit a univer-sal metallicity gradient of -0.1 dex R e − within 2 R e (e.g.S´anchez et al. 2014; Ho et al. 2015). These works employO3N2 (S´anchez et al. 2014, Ho et al. 2015) and N2O2 (Hoet al. 2015) to derive the metallicity. S´anchez et al. (2014)isolated H ii regions from DIG and used O3N2 to derive themetallicities using H ii regions. The contamination of DIG issmall. Ho et al. (2015) rejected the DIG-dominated spaxelsby imposing S/N >
3, and Z(O3N2) and Z(N2O2) give consis-tent metallicities and metallicity gradients. The metallicitygradient at R < R e is robust against DIG contamination.Using MaNGA data and the robust metallicity estimatorN2O2 to remeasure metallicities would be worthwhile to re-examine the metallicity gradient at large radius. MNRAS , 1–30 (2016) K. Zhang et al.
Figure 28. ∆ Z vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for all galaxies. We normalize the ∆ Z vs ∆ log Σ H α relation by subtractingthe median Z and median log Σ H α . All the spaxels in a galaxy are weighted by N , where N is the number of valid spaxels in the galaxy.A non-zero slope of the relation means metallicities are biased by DIG contribution, and the dispersion of this relation in y directionreflect how reliable a metallicity proxy is for an individual galaxy. At high-redshift, we need a new calibration of strongline metallicity indicators because the physical properties aredifferent in high redshift galaxies compared to local galax-ies. High redshift star-forming galaxies are systematicallyoffset towards higher [O iii ]/H β on the BPT diagram rela-tive to the local star-forming galaxy locus (Shapley et al.2005; Erb et al. 2006; Liu et al. 2008; Brinchmann et al.2008; Hainline et al. 2009; Wright et al. 2010; Trump et al.2011 Kewley et al. 2013a,b; Steidel et al. 2014). The lineratio shifts of high-redshift star-forming galaxies are simi-lar to DIG. Based on photoionization models, Steidel et al.(2014) concluded that the offset on the BPT diagram of thez ∼ iii ], H β ,[N ii ], H α for a galaxy at high redshift. In most cases, either[O iii ]+H β or [N ii ]+H α is observed. The few lines availablelimit the metallicity estimator to either R or N2, two indi-cators that are impacted most by the presence of DIG. Thismakes the metallicity measurements at high-z vulnerable tocontamination of DIG. However, we note current studies areunlikely to be seriously affected by DIG because in general MNRAS000
Figure 28. ∆ Z vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for all galaxies. We normalize the ∆ Z vs ∆ log Σ H α relation by subtractingthe median Z and median log Σ H α . All the spaxels in a galaxy are weighted by N , where N is the number of valid spaxels in the galaxy.A non-zero slope of the relation means metallicities are biased by DIG contribution, and the dispersion of this relation in y directionreflect how reliable a metallicity proxy is for an individual galaxy. At high-redshift, we need a new calibration of strongline metallicity indicators because the physical properties aredifferent in high redshift galaxies compared to local galax-ies. High redshift star-forming galaxies are systematicallyoffset towards higher [O iii ]/H β on the BPT diagram rela-tive to the local star-forming galaxy locus (Shapley et al.2005; Erb et al. 2006; Liu et al. 2008; Brinchmann et al.2008; Hainline et al. 2009; Wright et al. 2010; Trump et al.2011 Kewley et al. 2013a,b; Steidel et al. 2014). The lineratio shifts of high-redshift star-forming galaxies are simi-lar to DIG. Based on photoionization models, Steidel et al.(2014) concluded that the offset on the BPT diagram of thez ∼ iii ], H β ,[N ii ], H α for a galaxy at high redshift. In most cases, either[O iii ]+H β or [N ii ]+H α is observed. The few lines availablelimit the metallicity estimator to either R or N2, two indi-cators that are impacted most by the presence of DIG. Thismakes the metallicity measurements at high-z vulnerable tocontamination of DIG. However, we note current studies areunlikely to be seriously affected by DIG because in general MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 29.
The distribution of metallicities vs Σ( Hα ) slope at three radii. The black line is for radius within [0.4 R e , 0.6 R e ], the blue lineis for radius in [0.8 R e , 1.0 R e ], and the red line is for the radius in [1.3 R e , 1.5 R e ]. The green dashed line has slope=0 line for reference.Themetallicity is independent of surface brightness; the reason we are seeing a dependence of metallicity on H α surface brightness is due tocontamination by DIG. they are forming stars at an incredibly high rate and theyare very well represented by pure H ii region spectra. Fu-ture studies at high redshift probing low surface brightnesslimits (e.g., with JWST) may have to take into account thecontribution of DIG. One needs to be cautious when mak-ing comparisons of high-z and low-z line ratios because DIGmatters more at low z. H α surface brightness relationfor an individual H ii region Rela˜no et al. (2010) studied the line ratios and H α surfacebrightness of NGC 595, one of the most luminous H ii re- gions in M33. The scale of NGC 595 is about 300pc. Bycomparing the H α surface brightness distribution shown intheir Figure 3 and the [N ii ]/H α , [S ii ]/H α , and [O iii ]/H β maps in Figure 6, we can see that the low surface bright-ness regions, which are located far away from the ionizingstars, have high [N ii ]/H α , [S ii ]/H α , and lower [O iii ]/H β .The line ratios vs. H α surface brightness relations are de-rived on a scale of hundreds of pc around an individual H ii region. Madsen et al. (2006) saw the same trend in theirobservation of Milky Way. The resolution of the WHAMsurvey is 1 degree, which corresponds to 90 pc at 5 kpc. Sothe physical scale of WHAM is tens to hundred parsecs. Dothese local relations on scales of hundreds of parsecs extendto kpc scale probed by MaNGA? MaNGA’s spatial resolu-tion of 1 kpc means each PSF may include tens or even MNRAS , 1–30 (2016) K. Zhang et al.
Figure 30. ∆ Z vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for the one third most massive galaxies in our sample. We normalize the Zvs log Σ H α relation by subtracting the median Z and median log Σ H α . All the spaxels in a galaxy are weighted by N , where N is thenumber of valid spaxels in the galaxy. hundreds of H ii regions. If the H α surface brightness vs lineratios relation is confined to 300pc scales, when smeared bya kpc-scale PSF, the line ratio variation will be smoothedout and disappear. In other words, if DIG is dominated byregions within 10pc of each H ii region, then we would notsee the line ratios vs Σ H α relation after smoothing it witha kpc-sized PSF. The smearing effect can produce a surfacebrightness gradient but can not produce an anti-correlationbetween surface brightness and line ratios.
The fact that wesee a surface brightness vs line ratio relation on kpc-scalesmeans that these relations for individual H ii region can notbe local, but extend to kpc scales. The DIG we see is domi- nated by regions far away (kpc) from individual H ii regions.In edge-on galaxies, we do see that DIG can extend to a fewkpc above the galaxy plane (e.g, Rand 1996, 1997; Rossa &Dettmar 2000; Kehrig et al. 2012; Jones et al. 2016).Then what does the line ratio vs H α surface brightnessrelation tell us? What we need to keep in mind is that we arelooking at the average properties of the galaxy at our res-olution. As discussed in detail in Appendix A, this relationis preserved for individual H ii regions no matter what res-olution we use to observe the galaxy. However, under poorresolution, many H ii regions are mixed, and their relationsare also mixed. We don’t know how many H ii regions there MNRAS000
The fact that wesee a surface brightness vs line ratio relation on kpc-scalesmeans that these relations for individual H ii region can notbe local, but extend to kpc scales. The DIG we see is domi- nated by regions far away (kpc) from individual H ii regions.In edge-on galaxies, we do see that DIG can extend to a fewkpc above the galaxy plane (e.g, Rand 1996, 1997; Rossa &Dettmar 2000; Kehrig et al. 2012; Jones et al. 2016).Then what does the line ratio vs H α surface brightnessrelation tell us? What we need to keep in mind is that we arelooking at the average properties of the galaxy at our res-olution. As discussed in detail in Appendix A, this relationis preserved for individual H ii regions no matter what res-olution we use to observe the galaxy. However, under poorresolution, many H ii regions are mixed, and their relationsare also mixed. We don’t know how many H ii regions there MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG Figure 31. ∆ Z vs ∆ log Σ H α relation at [0.4 R e , 0.6 R e ] for the one third least massive galaxies in our sample. We normalize the Zvs log Σ H α relation by subtracting the median Z and median log Σ H α . All the spaxels in a galaxy are weighted by N , where N is thenumber of valid spaxels in the galaxy. are in one spaxel and what their line ratio vs H α surfacebrightness relations looks like. What we see is the averageproperties within 1 kpc, and how the line ratios and surfacebrightness change coherently on this scale. The curvature ofthe relation does tell us that an H ii region is not a simpleStr¨omgren sphere but includes an extra component of DIGwhich has different line ratios. At 1 kpc scale, we can modelthe ISM as a combination of DIG and H ii regions with theirrespective line ratios. This relation also illustrates the fea-sibility of using H α surface brightness to roughly separateDIG and H ii regions. Any factors that change the line ratios can influence the po-sitions on the diagnostic diagrams (e.g., Zhang et al. 2008;P`erez-Montero & Contini 2009). In Figure 15 and 16, weexplored the distribution of each spaxel on the BPT di-agram: [O iii ]/H β vs [N ii ]/H α and [O iii ]/H β vs [S ii ]/H α (Baldwin et al. 1981; Veilleux & Osterbrock 1987). In the[O iii ]/H β vs [N ii ]/H α diagram, we plot the demarcationsfrom Kewley et al. (2001) and Kauffmann et al. (2003) toidentify the ionizing source properties. We color-code the MNRAS , 1–30 (2016) K. Zhang et al. dots by H α surface brightness, and the dots in composite& LI(N)ER region have low H α surface brightness. As dis-cussed in Section 4, a hardened O star spectrum filteredby ISM can not produce LI(N)ER-like emissions, while hotevolved stars like pAGB stars can produce the LI(N)ER-likeline ratios. In Belfiore et al. (2016a), they classified galax-ies that show LI(N)ER-like emission into two classes: cLIERand eLIER. cLIER shows LI(N)ER-like emission in the cen-ter, while eLIER shows extended LI(N)ER-like emission allover the galaxy. In some galaxies in our sample, the lineratios could extend to LI(N)ER regions on the BPT dia-grams (See also Galarza et al. 1999; Kaplan et al. 2016). Wesuspect these two phenomenon may have a common origin.However, the galaxies that show eLIER emission and cLIERcould have different stellar population and gas metallicityfrom the sample we study in this paper. Therefore a lotmore work is needed to prove if they are really the same.We defer this to a future paper. The Shocked POststarburstGalaxy Survey (SPOGS, Alatalo et al. 2016a,b) explores asample of galaxies selected from SDSS Data Release 7 thatshow LI(N)ER like emission line ratios and post-starburstspectral features. A similar phenomenon is found by Yan etal. (2006). After OB stars die, the spectrum of the galaxyis dominated by spectral features of A stars, characterizingthe post-starburst signature. According to our grids, the de-crease in OB stars and increase LI(N)ER-like emission lineratios are naturally linked through increasing photoioniza-tion by evolved stars. DIG, cLIER, and SPOGs are verysimilar to each other, and the study of their similarity anddistinctions can greatly help us understand the ionized gas indifferent types of galaxies. From our figures, DIG dominatedregions can mimic composite or even AGN spectra. If theselection of AGN is purely based on the BPT diagram, star-forming galaxies with significant DIG could be incorrectlyclassified as AGN. DIG and AGNs might be distinguishedusing the WHAN diagram (Cid Fernandes et al 2010, 2011)due to their different EW(H α ) and [N ii ]/H α distributions.This will be explored further in the future. Our analysisdemonstrates the importance of considering the presence ofDIG while making optical line ratio classifications. Addi-tionally, the analysis illustrates the complexity caused byDIG in interpreting the line ratio diagnostic diagrams. Theso-called composite galaxies or LI(N)ERs could be comingfrom a variety of objects.Shocks could also produce LI(N)ER-like line ratios.Shocked gas with velocities less than 500 km s − occupy theLI(N)ER part of the diagnostic diagrams (Farage et al. 2010;Rich et al. 2011), while shocked gas with velocities greaterthan 500 km s − fall in the Seyfert part of the diagram (Allenet al. 2008; Kewley et al. 2013). Shock model can producethe enhanced [S ii ]/H α , [N ii ]/H α , [O ii ]/H β , [O i ]/H α and[O iii ]/H β we see. Additionally, the temperature of shockedgas is elevated (Allen et al. 2008). To test if shocks are indeeda major ionization source for DIG, high spectral resolutionspectra ( ∼ ∼ The best way to minimize the impact of DIG is to separateDIG and H ii regions spatially. With high spatial resolutionIFS data like CALIFA, it is possible to resolve individualH ii regions to lower the contamination of DIG. For low spa-tial resolution data like MaNGA and SAMI, the individualspaxels are covering kpc scale regions, thus the emission isa mix of H ii regions and DIG. The mixing also happens forMUSE data of high redshift galaxies. For these data, select-ing high Σ H α spaxels would mitigate the impact of DIG.As we have shown in this paper, the high Σ H α regions haveH ii region line ratios. Besides, the metallicity gradient de-rived using only the high Σ H α regions for different strong linemethod: Z(N2O2), Z( R ), Z(O3N2), Z(N2S2H α ) are con-sistent with each other. This means the high Σ H α regions,even though contaminated by DIG due to beam smearing,are H ii region dominated. A Σ H α cut is a robust and easyway to reduce the impact of DIG. The exact value of Σ H α cutdepends on the spatial resolution. For our MaNGA survey,Σ H α >
39 erg s − kpc − select reliable H ii region dominatedspaxels. An equivalent width (EW) cut is not recommendedsince EW depends on metallicity (e.g., Tresse et al. 1999).For low metallicity regions, for example in low metallicitygalaxies or the outskirt of a galaxy, EW will be high due tothe low metallicity. A EW cut suitable for the center of agalaxy will select DIG contaminated spaxels at the outskirtof this galaxy. So selecting high Σ H α regions is a reliable andconvenient way to minimize DIG contamination. We selected a sample of 365 blue face-on galaxies from 1391galaxies observed by MaNGA, and illustrated the impact ofDIG on line ratios, interpretation of diagnostic diagrams,and metallicity measurements. We find that H α surfacebrightness is a good indicator to separate H ii regions fromDIG. DIG shows distinct properties as listed below:(i) [S II ]/H α , [N ii ]/H α , [O ii ]/H β , and [O i ]/H α are en-hanced in DIG relative to H ii regions.(ii) DIG has lower [O iii ]/[O ii ], indicating lower ioniza-tion parameter. [O iii ]/H β of DIG can be higher or lowerthan H ii regions.(iii) On BPT diagrams, contamination by DIG moves H ii regions towards composite or LI(N)ER-like regions. A harderionizing spectrum is needed to explain DIG line ratios.(iv) Leaky H ii region models only shift the line ratiosslightly relative to H ii region models, thus fail to explaincomposite/LI(N)ER line ratios displayed by DIG. Leaky H ii region models cannot explain the [O iii ]/H β but do prettywell for the other line ratios.(v) Our result favors ionization by evolved stars as a ma-jor ionization source for DIG with LI(N)ER-like emission.(vi) Metallicities derived using N2O2=[N ii ]/[O ii ] are op-timal because they exhibit the smallest bias and scatter.(vii) Metallicities derived using theO3N2=([O iii ]/H β )/([N ii ]/H α ) or N2S2H α =8.77+log[N ii ]/[S ii ] + 0.264 × log[N ii ]/H α (Dopita et al. 2016) forDIG can be significantly higher or lower than those forH ii regions. Using O3N2 or N2S2H α to derive metallicitiescan bias the metallicity gradient by ± R e − for MNRAS000
39 erg s − kpc − select reliable H ii region dominatedspaxels. An equivalent width (EW) cut is not recommendedsince EW depends on metallicity (e.g., Tresse et al. 1999).For low metallicity regions, for example in low metallicitygalaxies or the outskirt of a galaxy, EW will be high due tothe low metallicity. A EW cut suitable for the center of agalaxy will select DIG contaminated spaxels at the outskirtof this galaxy. So selecting high Σ H α regions is a reliable andconvenient way to minimize DIG contamination. We selected a sample of 365 blue face-on galaxies from 1391galaxies observed by MaNGA, and illustrated the impact ofDIG on line ratios, interpretation of diagnostic diagrams,and metallicity measurements. We find that H α surfacebrightness is a good indicator to separate H ii regions fromDIG. DIG shows distinct properties as listed below:(i) [S II ]/H α , [N ii ]/H α , [O ii ]/H β , and [O i ]/H α are en-hanced in DIG relative to H ii regions.(ii) DIG has lower [O iii ]/[O ii ], indicating lower ioniza-tion parameter. [O iii ]/H β of DIG can be higher or lowerthan H ii regions.(iii) On BPT diagrams, contamination by DIG moves H ii regions towards composite or LI(N)ER-like regions. A harderionizing spectrum is needed to explain DIG line ratios.(iv) Leaky H ii region models only shift the line ratiosslightly relative to H ii region models, thus fail to explaincomposite/LI(N)ER line ratios displayed by DIG. Leaky H ii region models cannot explain the [O iii ]/H β but do prettywell for the other line ratios.(v) Our result favors ionization by evolved stars as a ma-jor ionization source for DIG with LI(N)ER-like emission.(vi) Metallicities derived using N2O2=[N ii ]/[O ii ] are op-timal because they exhibit the smallest bias and scatter.(vii) Metallicities derived using theO3N2=([O iii ]/H β )/([N ii ]/H α ) or N2S2H α =8.77+log[N ii ]/[S ii ] + 0.264 × log[N ii ]/H α (Dopita et al. 2016) forDIG can be significantly higher or lower than those forH ii regions. Using O3N2 or N2S2H α to derive metallicitiescan bias the metallicity gradient by ± R e − for MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG an individual galaxy if the contamination by DIG is notaccounted for. R derived metallicities for DIG are lowerthan those for H ii regions due to a lower ionization param-eter. Using R to derive metallicities will systematicallybias the metallicity gradient by ∼ -0.1 dex R e − because ofDIG. Using N2=[N ii ]/H α to derive metallicities will sys-tematically bias the metallicity gradient by ∼ R e − , considering that DIG typically shows 0.2 dex higher[N ii ]/H α .(viii) The metallicities in high redshift galaxies are mostlyderived using R or N2, rendering their metallicity andmetallicity gradient measurements most vulnerable to theimpact of DIG. Using Z(N2S2H α ) for high redshift galaxiesis more robust to prevent the contamination by DIG. Formost of the recent high − z observations, the contaminationby DIG is probably not severe because we only see high Σ H α regions. When comparing the metallicities of high-z galax-ies with local galaxies, one needs to use caution since DIGmight impact metallicity measurements of local galaxies. ACKNOWLEDGEMENTS
REFERENCES
Alatalo K., et al., 2016, arXiv, arXiv:1604.01122Alatalo K., et al., 2016, arXiv, arXiv:1601.05085Allen M. G., Groves B. A., Dopita M. A., Sutherland R. S., Kew-ley L. J., 2008, ApJS, 178, 20-55Alloin D., Collin-Souffrin S., Joly M., Vigroux L., 1979, A&A, 78,200Andrews B. H., Martini P., 2013, ApJ, 765, 140Baldwin, J. A., Phillips, M. M., & Terlevich, R. 1981, PASP, 93,5Baldwin, J. A., Ferland, G. J., Martin, P. G., et al. 1991, ApJ,374, 580Belfiore, F., Maiolino, R., Bundy, K., et al. 2015, MNRAS, 449,867Belfiore F., et al., 2016a, MNRAS, 461, 3111Belfiore F., et al., 2016b, arXiv, arXiv:1609.01737Binette L., Magris C. G., Stasi´nska G., Bruzual A. G., 1994, A&A,292, 13Blanc, G. A., Heiderman, A., Gebhardt, K., Evans, N. J., II, &Adams, J. 2009, ApJ, 704, 842-862Blanc, G. A., Kewley, L., Vogt, F. P. A., & Dopita, M. A. 2015,ApJ, 798, 99Brinchmann, J., Pettini, M., & Charlot, S. 2008, MNRAS, 385,769Bruzual, G., & Charlot, S. 2003, MNRAS, 344, 1000Bundy, K., Bershady, M. A., Law, D. R., et al. 2015, ApJ, 798, 7Collins, J. A., & Rand, R. J. 2001, ApJ, 551, 57Cid Fernandes R., Stasi´nska G., Schlickmann M. S., Mateus A.,Vale Asari N., Schoenell W., Sodr´e L., 2010, MNRAS, 403,1036Cid Fernandes R., Stasi´nska G., Mateus A., Vale Asari N., 2011,MNRAS, 413, 1687Denicol´o, G., Terlevich, R., & Terlevich, E. 2002, MNRAS, 330,69Dettmar R.-J., 1990, A&A, 232, L15Dopita, M. A., Kewley, L. J., Heisler, C. A., & Sutherland, R. S.2000, ApJ, 542, 224Dopita, M. A., Fischera, J., Sutherland, R. S., et al. 2006, ApJs,167, 177Dopita, M. A., Sutherland, R. S., Nicholls, D. C., Kewley, L. J.,& Vogt, F. P. A. 2013, ApJs, 208, 10Dopita, M. A., Kewley, L. J., Sutherland, R. S., & Nicholls, D. C.2016, ApJS, 361, 61Drory, N., MacDonald, N., Bershady, M. A., et al. 2015, AJ, 149,77Erb, D. K., Shapley, A. E., Pettini, M., et al. 2006, ApJ, 644, 813Farage C. L., McGregor P. J., Dopita M. A., Bicknell G. V., 2010,ApJ, 724, 267Ferguson, A. M. N., Wyse, R. F. G., Gallagher, J. S., III, &Hunter, D. A. 1996, AJ, 111, 2265Ferland G. J., Korista K. T., Verner D. A., Ferguson J. W., King-don J. B., Verner E. M., 1998, PASP, 110, 761Fitzpatrick E. L., 1999, PASP, 111, 63Flores-Fajardo N., Morisset C., Stasi´nska G., Binette L., 2011,MNRAS, 415, 2182Galarza, V. C., Walterbos, R. A. M., & Braun, R. 1999, AJ, 118,2775Garay, G., & Lizano, S. 1999, PASP, 111, 1049Garnett, D. R. 2002, ApJ, 581, 1019Giammanco C., Beckman J. E., Cedr´es B., 2005, A&A, 438, 599Gomes J. M., et al., 2016, A&A, 588, A68Greenawalt, B., Walterbos, R. A. M., & Braun, R. 1997, ApJ,483, 666MNRAS , 1–30 (2016) K. Zhang et al.
Greenawalt, B., Walterbos, R. A. M., Thilker, D., & Hoopes, C. G.1998, ApJ, 506, 135Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ,131, 2332Haffner, L. M., Reynolds, R. J., & Tufte, S. L. 1999, ApJ, 523,223Haffner, L. M., Dettmar, R.-J., Beckman, J. E., et al. 2009, Re-views of Modern Physics, 81, 969Hainline, K. N., Shapley, A. E., Kornei, K. A., et al. 2009, ApJ,701, 52Hausen, N. R., Reynolds, R. J., & Haffner, L. M. 2002, AJ, 124,3336Ho, I.-T., Kudritzki, R.-P., Kewley, L. J., et al. 2015, MNRAS,448, 2030Hoopes, C. G., Walterbos, R. A. M., & Greenwalt, B. E. 1996,AJ, 112, 1429Hoopes C. G., Walterbos R. A. M., Rand R. J., 1999, ApJ, 522,669Hoopes, C. G., & Walterbos, R. A. M. 2003, ApJ, 586, 902Hunt, L. K., & Hirashita, H. 2009, A&A, 507, 1327Jones, T., Ellis, R. S., Richard, J., & Jullo, E. 2013, ApJ, 765, 48Jones, T., Wang, X., Schmidt, K. B., et al. 2015, AJ, 149, 107Jones A., et al., 2016, arXiv, arXiv:1612.03920Kaplan K. F., et al., 2016, arXiv, arXiv:1606.05713Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MN-RAS, 346, 1055Kennicutt, R. C., Jr. 1984, ApJ, 287, 116Kewley, L. J., Heisler, C. A., Dopita, M. A., & Lumsden, S. 2001,ApJs, 132, 37Kewley, L. J., & Dopita, M. A. 2002, ApJs, 142, 35Kewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006,MNRAS, 372, 961Kewley, L. J., & Ellison, S. L. 2008, ApJ, 681, 1183Kewley L. J., Dopita M. A., Leitherer C., Dav´e R., Yuan T., AllenM., Groves B., Sutherland R., 2013, ApJ, 774, 100Kim, K.-T., & Koo, B.-C. 2001, ApJ, 549, 979Kobulnicky, H. A., & Kewley, L. J. 2004, ApJ, 617, 240Kreckel K., Blanc G. A., Schinnerer E., Groves B., Adamo A.,Hughes A., Meidt S., 2016, arXiv, arXiv:1603.08009Kehrig C., et al., 2012, A&A, 540, A11Lanz T., Hubeny I., 2003, ApJS, 146, 417Law, D. R., Yan, R., Bershady, M. A., et al. 2015, AJ, 150, 19Law D. R., et al., 2016, AJ, 152, 83Lee, H., Skillman, E. D., Cannon, J. M., et al. 2006, ApJ, 647,970Leethochawalit N., Jones T. A., Ellis R. S., Stark D. P., RichardJ., Zitrin A., Auger M., 2016, ApJ, 820, 84Lequeux, J., Peimbert, M., Rayo, J. F., Serrano, A., & Torres-Peimbert, S. 1979, A&A, 80, 155Levesque, E. M., Kewley, L. J., & Larson, K. L. 2010, AJ, 139,712Li, C., Wang, E., Lin, L., et al. 2015, ApJ, 804, 125Liu, X., Shapley, A. E., Coil, A. L., Brinchmann, J., & Ma, C.-P.2008, ApJ, 678, 758-779Madsen, G. J., Reynolds, R. J., & Haffner, L. M. 2006, ApJ, 652,401Maiolino, R., Nagao, T., Grazian, A., et al. 2008, A&A, 488, 463Mannucci, F., Cresci, G., Maiolino, R., et al. 2009, MNRAS, 398,1915Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci,A. 2010, MNRAS, 408, 2115Markwardt, C. B. 2009, Astronomical Data Analysis Software andSystems XVIII, 411, 251Mast D., et al., 2014, A&A, 561, A129Masters D., et al., 2014, ApJ, 785, 153McClure, R. D., & van den Bergh, S. 1968, AJ, 73, 313McGaugh, S. S. 1991, ApJ, 380, 140 Mierkiewicz, E. J., Reynolds, R. J., Roesler, F. L., Harlander,J. M., & Jaehnig, K. P. 2006, ApJl, 650, L63Monnet, G. 1971, A&A, 12, 379Moran S. M., et al., 2012, ApJ, 745, 66Oey, M. S., Meurer, G. R., Yelda, S., et al. 2007, ApJ, 661, 801Osterbrock, D. E., Tran, H. D., & Veilleux, S. 1992, ApJ, 389,305Otte, B. 2001, Ph.D. Thesis, 3233Otte, B., Reynolds, R. J., Gallagher, J. S., III, & Ferguson,A. M. N. 2001, ApJ, 560, 207Otte, B., Gallagher, J. S., III, & Reynolds, R. J. 2002, ApJ, 572,823Pettini, M., & Pagel, B. E. J. 2004, MNRAS, 348, L59P´erez-Montero, E., & Contini, T. 2009, MNRAS, 398, 949P´erez-Montero E., et al., 2013, A&A, 549, A25P´erez-Montero E., 2014, MNRAS, 441, 2663P´erez-Montero E., et al., 2016, arXiv, arXiv:1608.04677Pilyugin, L. S. 2001, A&A, 369, 594Pilyugin, L. S., & Thuan, T. X. 2005, ApJ, 631, 231Rand R. J., Kulkarni S. R., Hester J. J., 1990, ApJ, 352, L1Rand, R. J. 1996, ApJ, 462, 712Rand R. J., 1997, ApJ, 474, 129Rand, R. J. 1998, ApJ, 501, 137Rand, R. J. 2000, ApJl, 537, L13Rela˜no, M., Monreal-Ibero, A., V´ılchez, J. M., & Kennicutt, R. C.2010, MNRAS, 402, 1635Reynolds R. J., 1984, ApJ, 282, 191Reynolds, R. J. 1985, ApJ, 294, 256Reynolds, R. J. 1985, ApJl, 298, L27Reynolds R. J., 1990, ApJ, 349, L17Reynolds R. J., 1991, ApJ, 372, L17Reynolds R. J., Cox D. P., 1992, ApJ, 400, L33Reynolds, R. J., Hausen, N. R., Tufte, S. L., & Haffner, L. M.1998, ApJl, 494, L99Rich J. A., Kewley L. J., Dopita M. A., 2011, ApJ, 734, 87Rossa, J., & Dettmar, R.-J. 2003, A&A, 406, 493Rossa, J., & Dettmar, R.-J. 2003, A&A, 406, 505Rubin, R. H., Simpson, J. P., Haas, M. R., & Erickson, E. F.1991, ApJ, 374, 564Rubin, R. H., Dufour, R. J., & Walter, D. K. 1993, ApJ, 413, 242S´anchez, S. F., Rosales-Ortega, F. F., Marino, R. A., et al. 2012,A&A, 546, A2S´anchez, S. F., Rosales-Ortega, F. F., Iglesias-P´aramo, J., et al.2014, A&A, 563, A49Sanders R. L., et al., 2016, ApJ, 816, 23Sarzi, M., Falc´on-Barroso, J., Davies, R. L., et al. 2006, MNRAS,366, 1151Shapley, A. E., Coil, A. L., Ma, C.-P., & Bundy, K. 2005, ApJ,635, 1006Shapley A. E., et al., 2015, ApJ, 801, 88Singh, R., van de Ven, G., Jahnke, K., et al. 2013, A&A, 558, A43Smee, S. A., Gunn, J. E., Uomoto, A., et al. 2013, AJ, 146, 32Stasi´nska, G., Vale Asari, N., Cid Fernandes, R., et al. 2008, MN-RAS, 391, L29Steidel, C. C., Rudie, G. C., Strom, A. L., et al. 2014, ApJ, 795,165Storchi-Bergmann, T., Calzetti, D., & Kinney, A. L. 1994, ApJ,429, 572Swinbank, A. M., Sobral, D., Smail, I., et al. 2012, MNRAS, 426,935Thilker, D. A., Walterbos, R. A. M., Braun, R., & Hoopes, C. G.2002, AJ, 124, 3118Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004,ApJ, 613, 898Tresse L., Maddox S., Loveday J., Singleton C., 1999, MNRAS,310, 262Trump, J. R., Weiner, B. J., Scarlata, C., et al. 2011, ApJ, 743,144 MNRAS000
Greenawalt, B., Walterbos, R. A. M., Thilker, D., & Hoopes, C. G.1998, ApJ, 506, 135Gunn, J. E., Siegmund, W. A., Mannery, E. J., et al. 2006, AJ,131, 2332Haffner, L. M., Reynolds, R. J., & Tufte, S. L. 1999, ApJ, 523,223Haffner, L. M., Dettmar, R.-J., Beckman, J. E., et al. 2009, Re-views of Modern Physics, 81, 969Hainline, K. N., Shapley, A. E., Kornei, K. A., et al. 2009, ApJ,701, 52Hausen, N. R., Reynolds, R. J., & Haffner, L. M. 2002, AJ, 124,3336Ho, I.-T., Kudritzki, R.-P., Kewley, L. J., et al. 2015, MNRAS,448, 2030Hoopes, C. G., Walterbos, R. A. M., & Greenwalt, B. E. 1996,AJ, 112, 1429Hoopes C. G., Walterbos R. A. M., Rand R. J., 1999, ApJ, 522,669Hoopes, C. G., & Walterbos, R. A. M. 2003, ApJ, 586, 902Hunt, L. K., & Hirashita, H. 2009, A&A, 507, 1327Jones, T., Ellis, R. S., Richard, J., & Jullo, E. 2013, ApJ, 765, 48Jones, T., Wang, X., Schmidt, K. B., et al. 2015, AJ, 149, 107Jones A., et al., 2016, arXiv, arXiv:1612.03920Kaplan K. F., et al., 2016, arXiv, arXiv:1606.05713Kauffmann, G., Heckman, T. M., Tremonti, C., et al. 2003, MN-RAS, 346, 1055Kennicutt, R. C., Jr. 1984, ApJ, 287, 116Kewley, L. J., Heisler, C. A., Dopita, M. A., & Lumsden, S. 2001,ApJs, 132, 37Kewley, L. J., & Dopita, M. A. 2002, ApJs, 142, 35Kewley, L. J., Groves, B., Kauffmann, G., & Heckman, T. 2006,MNRAS, 372, 961Kewley, L. J., & Ellison, S. L. 2008, ApJ, 681, 1183Kewley L. J., Dopita M. A., Leitherer C., Dav´e R., Yuan T., AllenM., Groves B., Sutherland R., 2013, ApJ, 774, 100Kim, K.-T., & Koo, B.-C. 2001, ApJ, 549, 979Kobulnicky, H. A., & Kewley, L. J. 2004, ApJ, 617, 240Kreckel K., Blanc G. A., Schinnerer E., Groves B., Adamo A.,Hughes A., Meidt S., 2016, arXiv, arXiv:1603.08009Kehrig C., et al., 2012, A&A, 540, A11Lanz T., Hubeny I., 2003, ApJS, 146, 417Law, D. R., Yan, R., Bershady, M. A., et al. 2015, AJ, 150, 19Law D. R., et al., 2016, AJ, 152, 83Lee, H., Skillman, E. D., Cannon, J. M., et al. 2006, ApJ, 647,970Leethochawalit N., Jones T. A., Ellis R. S., Stark D. P., RichardJ., Zitrin A., Auger M., 2016, ApJ, 820, 84Lequeux, J., Peimbert, M., Rayo, J. F., Serrano, A., & Torres-Peimbert, S. 1979, A&A, 80, 155Levesque, E. M., Kewley, L. J., & Larson, K. L. 2010, AJ, 139,712Li, C., Wang, E., Lin, L., et al. 2015, ApJ, 804, 125Liu, X., Shapley, A. E., Coil, A. L., Brinchmann, J., & Ma, C.-P.2008, ApJ, 678, 758-779Madsen, G. J., Reynolds, R. J., & Haffner, L. M. 2006, ApJ, 652,401Maiolino, R., Nagao, T., Grazian, A., et al. 2008, A&A, 488, 463Mannucci, F., Cresci, G., Maiolino, R., et al. 2009, MNRAS, 398,1915Mannucci, F., Cresci, G., Maiolino, R., Marconi, A., & Gnerucci,A. 2010, MNRAS, 408, 2115Markwardt, C. B. 2009, Astronomical Data Analysis Software andSystems XVIII, 411, 251Mast D., et al., 2014, A&A, 561, A129Masters D., et al., 2014, ApJ, 785, 153McClure, R. D., & van den Bergh, S. 1968, AJ, 73, 313McGaugh, S. S. 1991, ApJ, 380, 140 Mierkiewicz, E. J., Reynolds, R. J., Roesler, F. L., Harlander,J. M., & Jaehnig, K. P. 2006, ApJl, 650, L63Monnet, G. 1971, A&A, 12, 379Moran S. M., et al., 2012, ApJ, 745, 66Oey, M. S., Meurer, G. R., Yelda, S., et al. 2007, ApJ, 661, 801Osterbrock, D. E., Tran, H. D., & Veilleux, S. 1992, ApJ, 389,305Otte, B. 2001, Ph.D. Thesis, 3233Otte, B., Reynolds, R. J., Gallagher, J. S., III, & Ferguson,A. M. N. 2001, ApJ, 560, 207Otte, B., Gallagher, J. S., III, & Reynolds, R. J. 2002, ApJ, 572,823Pettini, M., & Pagel, B. E. J. 2004, MNRAS, 348, L59P´erez-Montero, E., & Contini, T. 2009, MNRAS, 398, 949P´erez-Montero E., et al., 2013, A&A, 549, A25P´erez-Montero E., 2014, MNRAS, 441, 2663P´erez-Montero E., et al., 2016, arXiv, arXiv:1608.04677Pilyugin, L. S. 2001, A&A, 369, 594Pilyugin, L. S., & Thuan, T. X. 2005, ApJ, 631, 231Rand R. J., Kulkarni S. R., Hester J. J., 1990, ApJ, 352, L1Rand, R. J. 1996, ApJ, 462, 712Rand R. J., 1997, ApJ, 474, 129Rand, R. J. 1998, ApJ, 501, 137Rand, R. J. 2000, ApJl, 537, L13Rela˜no, M., Monreal-Ibero, A., V´ılchez, J. M., & Kennicutt, R. C.2010, MNRAS, 402, 1635Reynolds R. J., 1984, ApJ, 282, 191Reynolds, R. J. 1985, ApJ, 294, 256Reynolds, R. J. 1985, ApJl, 298, L27Reynolds R. J., 1990, ApJ, 349, L17Reynolds R. J., 1991, ApJ, 372, L17Reynolds R. J., Cox D. P., 1992, ApJ, 400, L33Reynolds, R. J., Hausen, N. R., Tufte, S. L., & Haffner, L. M.1998, ApJl, 494, L99Rich J. A., Kewley L. J., Dopita M. A., 2011, ApJ, 734, 87Rossa, J., & Dettmar, R.-J. 2003, A&A, 406, 493Rossa, J., & Dettmar, R.-J. 2003, A&A, 406, 505Rubin, R. H., Simpson, J. P., Haas, M. R., & Erickson, E. F.1991, ApJ, 374, 564Rubin, R. H., Dufour, R. J., & Walter, D. K. 1993, ApJ, 413, 242S´anchez, S. F., Rosales-Ortega, F. F., Marino, R. A., et al. 2012,A&A, 546, A2S´anchez, S. F., Rosales-Ortega, F. F., Iglesias-P´aramo, J., et al.2014, A&A, 563, A49Sanders R. L., et al., 2016, ApJ, 816, 23Sarzi, M., Falc´on-Barroso, J., Davies, R. L., et al. 2006, MNRAS,366, 1151Shapley, A. E., Coil, A. L., Ma, C.-P., & Bundy, K. 2005, ApJ,635, 1006Shapley A. E., et al., 2015, ApJ, 801, 88Singh, R., van de Ven, G., Jahnke, K., et al. 2013, A&A, 558, A43Smee, S. A., Gunn, J. E., Uomoto, A., et al. 2013, AJ, 146, 32Stasi´nska, G., Vale Asari, N., Cid Fernandes, R., et al. 2008, MN-RAS, 391, L29Steidel, C. C., Rudie, G. C., Strom, A. L., et al. 2014, ApJ, 795,165Storchi-Bergmann, T., Calzetti, D., & Kinney, A. L. 1994, ApJ,429, 572Swinbank, A. M., Sobral, D., Smail, I., et al. 2012, MNRAS, 426,935Thilker, D. A., Walterbos, R. A. M., Braun, R., & Hoopes, C. G.2002, AJ, 124, 3118Tremonti, C. A., Heckman, T. M., Kauffmann, G., et al. 2004,ApJ, 613, 898Tresse L., Maddox S., Loveday J., Singleton C., 1999, MNRAS,310, 262Trump, J. R., Weiner, B. J., Scarlata, C., et al. 2011, ApJ, 743,144 MNRAS000 , 1–30 (2016)
DSS IV-MaNGA: Impact of DIG T¨ullmann R., Dettmar R.-J., 2000, A&A, 362, 119Vale Asari N., Stasi´nska G., Morisset C., Cid Fernandes R., 2016,MNRAS, 460, 1739van Zee, L., Salzer, J. J., & Haynes, M. P. 1998, ApJl, 497, L1Veilleux, S., & Osterbrock, D. E. 1987, ApJs, 63, 295Vincenzo F., Belfiore F., Maiolino R., Matteucci F., Ventura P.,2016, MNRAS, 458, 3466Voges, E. S., & Walterbos, R. A. M. 2006, ApJl, 644, L29Walterbos, R. A. M., & Braun, R. 1994, ApJ, 431, 156Wang J., Heckman T. M., Lehnert M. D., 1997, ApJ, 491, 114Wilkinson, D. M., Maraston, C., Thomas, D., et al. 2015, MN-RAS, 449, 328Wright, S. A., Larkin, J. E., Graham, J. R., & Ma, C.-P. 2010,ApJ, 711, 1291Wuyts, E., Wisnioski, E., Fossati, M., et al. 2016,arXiv:1603.01139Yan, R., Newman, J. A., Faber, S. M., et al. 2006, ApJ, 648, 281Yan, R., & Blanton, M. R. 2012, ApJ, 747, 61Yan, R., Tremonti, C., Bershady, M. A., et al. 2016, AJ, 151, 8Yan R., et al., 2016, arXiv, arXiv:1607.08613Yates R. M., Kauffmann G., Guo Q., 2012, MNRAS, 422, 215Yuan, T.-T., Kewley, L. J., Swinbank, A. M., Richard, J., & Liv-ermore, R. C. 2011, ApJl, 732, L14Zaritsky, D., Kennicutt, R. C., Jr., & Huchra, J. P. 1994, ApJ,420, 87Zhang K., Wang T., Dong X., Lu H., 2008, ApJ, 685, L109Zurita, A., Beckman, J. E., Rozas, M., & Ryder, S. 2002, A&A,386, 801Zurita, A., Rozas, M., & Beckman, J. E. 2000, A&A, 363, 9 Department of Physics and Astronomy, University ofKentucky, 505 Rose Street, Lexington, KY 40506, USA Kavli Institute for the Physics and Mathematics of theUniverse (Kavli IPMU, WPI), Todai Institutes for Ad-vanced Study, the University of Tokyo, Kashiwa 277-8583,Japan Department of Astronomy, University of Wisconsin-Madison, 475 N. Charter Street, Madison, WI, 53706, USA Department of Astronomy, New Mexico State University,Las Cruces, NM 88003, USA Cavendish Laboratory, University of Cambridge, 19 J. J.Thomson Avenue, Cambridge CB3 0HE, United Kingdom Kavli Institute for Cosmology, University of Cambridge,Madingley Road, Cambridge, United Kingdom Institute of Cosmology & Gravitation, University ofPortsmouth, Dennis Sciama Building, Portsmouth, PO13FX, UK Department of Astronomy, University of Texas at Austin,Austin, TX 78712, USA Max-Planck-Institut f¨ur Astrophysik, Karl-Schwarzschild-Str. 1, D-85748 Garching, Germany Instituto de Astronom´ıa, Universidad Nacional Autonomade Mexico, A.P. 70-264, 04510, Mexico, D.F., Mexico Apache Point Observatory and New Mexico State Uni-versity, P.O. Box 59, Sunspot, NM, 88349-0059, USA Sternberg Astronomical Institute, Moscow State Univer-sity, Moscow, Russia Unidad de Astronom´ıa, Universidad de Antofagasta,Avenida Angamos 601, Antofagasta 1270300, Chile PITT PACC, Department of Physics and Astronomy,University of Pittsburgh, Pittsburgh, PA 15260, USA Department of Physics and Astronomy, University ofUtah, 115 S. 1400 E., Salt Lake City, UT 84112, USA Tsinghua Center of Astrophysics & Department of Physics, Tsinghua University, Beijing 100084, China Shanghai Astronomical Observatory, Nandan Road 80,Shanghai 200030, China Space Telescope Science Institute, 3700 San Martin Drive,Baltimore, MD 21218, USA Departamento de F´ısica, Facultad de Ciencias, Universi-dad de La Serena, Cisternas 1200, La Serena, Chile Universidade Federal do Rio Grande do Sul, IF, CP15051, Porto Alegre 91501-970, RS, Brazil Laborat´orio Interinstitucional de e-Astronomia - LIneA,Rua Gal. Jos´e Cristino 77, Rio de Janeiro, RJ - 20921-400,Brazil
APPENDIX A: BEAM SMEARING EFFECT
Most of the H ii reigon is smaller than 1 kpc, and the sizedepends on the density of the gas (e.g. Hunt & Hirashita2009). Under MaNGA resolution, the mixing of H ii regionwith DIG and other H ii region is unavoidable. What’s theeffect of this mixing? We have seen in Section 3.1 that in oneannulus, the line ratio vs surface brightness relation is tight.What would happen if we mix 2 regions on the line ratio vssurface brightness relation? From an analytical calculation,the mixed point is still on the same relation. However, ifwe mix two or more H ii regions together, what would theline ratio vs H α surface brightness relation like? We showin Figure A1 for a simulation of observing our galaxies withpoorer resolution. The left panel is the original Σ(H α ) map,and the middle panel is the smeared Σ(H α ) map. In theright panel, we show the comparison of original (black dots)and smeared (red dots) line ratio vs H α surface brightnessrelation. The H α flux map and [S ii ] flux map is convolvedwith a 2D gaussian with σ pixels listed in the middle panelsof each row. We see several fact from this exercise:(i) The beam smearing effect does not change the lineratio vs SB relation for one H ii region, so the curve wederive using MaNGA data is likely to preserve when higherresolution data is obtained. In other word, this relation isintrinsic, reflecting the relationship between line ratio andsurface brightness. Our resolution of ∼ α flux peak in our map is probably a mixture of several H ii regions.(ii) It is impossible to disentangle DIG from H ii regionfrom the curve alone.(iii) The smearing makes the relation tighter and shorter.(iv) When mixing two or more H ii regions together, theirline ratio vs SB relations also merge to an intermediate one.The smearing process is an averaging process. We are wit-nessing more and more of the average properties of the H ii region and DIG of the galaxy as we go to poorer resolution.The overall smeared line ratio vs H α surface brightness re-lation is tighter than before. So the tightness of the relationwe see should be partly due to the beam smearing effect. MNRAS , 1–30 (2016) K. Zhang et al.
Figure A1.
Left panel: the original Σ(H α ) map, middle panel: the smeared Σ(H α ) map. Right panel: the comparison of original (blackdots) and smeared (red dots) line ratio vs H α surface brightness relation. The H α flux map and [N ii ] flux map is convolved with a 2Dgaussian with σ pixels listed in the middle panels of each row.This paper has been typeset from a TEX/L A TEX file prepared bythe author. MNRAS000