METAL TRANSPORT TO THE GASEOUS OUTSKIRTS OF GALAXIES

In 2005 April, over the course of four clear nights, we imaged 27 galaxies from the H i Rogues catalog (Hibbard et al. 2001) in Hα and R-band continuum with the MDM 2.4 m Hiltner telescope. The MDM 2.4 m telescope direct images, with the Echelle CCD, have a field of view of 95 × 95 and a plate scale of 0275 pixel⁻¹. The average seeing over the course of these four nights was 12. These observations were made in a similar fashion to the Hα imaging of the Survey of Ionization in Neutral Gas Galaxy (SINGG) survey (Meurer et al. 2006). We used four of the SINGG narrowband filters with bandpasses that corresponded to the Hα emission line at the target galaxy's velocity. The central wavelengths, in angstroms, of the filters used are 6568, 6596, 6605, and 6628, each with an FWHM of ∼30 Å (Meurer et al. 2006) contains a full description of the SINGG filters). Exposure times were 3 × 600 s in Hα and 3 × 120 s in R. Targets were selected from the full H i Rogues catalog on the basis of their visibility from MDM observatory and their recessional velocities, such that the redshifted Hα emission line is encompassed by the SINGG Hα filters' narrow bandpasses. Galaxies were also selected to have angular optical and H i extents such that they were contained in the 95 field of view. Observations of several different spectrophotometric standards were used to flux-calibrate the data.

We performed a basic reduction of the data using IRAF CCD processing routines for bias-subtraction and flat-field division. Images were then combined and astrometry was performed using the SINGG IDL pipeline routines (see Meurer et al. 2006) modified specifically for the MDM 2.4 m telescope. Photometry was performed on the images based on observations of spectrophotometric standards calibrated to the AB magnitude system. We followed the basic SINGG procedure detailed in Appendix A of Meurer et al. (2006). The average 5σ limiting Hα flux for a point source is ∼1 × 10⁻¹⁶ erg cm⁻² s⁻¹ and the average 5σ limiting R-band magnitude is ∼24.

We searched the reduced, calibrated MDM Hα and R-band images for ELdots (emission-line dots; see Werk et al. 2010b) using the basic selection criteria outlined in Werk et al. (2010b), with the exception of the requirement that the ELdot lie beyond 2 × R₂₅. For reference, ELdots are defined to be strong emission-line point sources in narrowband images that lie well outside the broadband optical emission of nearby galaxies (Werk et al. 2010b). Generally, ELdots can be outlying Hα emitting H ii regions at a similar velocity to the target galaxy or they can be background galaxies emitting a different line ([O iii] λ5007, [O ii] λλ3727, or Hβ) that is redshifted into the narrow filter passband used for the observations. For this study, we selected by-eye sources that appeared to lie beyond the main R-band optical extents of the potential host galaxies. The reason for relaxing our projected distance requirement is that R₂₅ is not well defined for merging, interacting, and/or irregular, asymmetric galaxies. For comparison to the sample in Werk et al. (2010b), and an estimate of the ELdots' projected distances from the main optical extents of their host galaxies, we did calculate an approximate R₂₅ at a later time. In order to obtain the approximate location of R₂₅ in these messy systems, we used the IRAF task ELLIPSE found in the STSDAS package, fitting elliptical isophotes to the target Rogue galaxies in the 2.4 m MDM R-band images. In one case (NGC 3227), two merging galaxies had to be fit with a single ellipse. In other cases, incidences of tails, plumes, and warps caused the fixed-center elliptical isophotes to be significantly larger than they otherwise would have been. The laxity of the R₂₅ fitting should be noted, and the quantity itself should be treated as an optical radial reference point when discussing oxygen abundance gradients. All of the Rogue ELdots, originally identified by eye, lie beyond this optical radius.

Thirteen of the 27 H i Rogues imaged with the MDM 2.4 m were found to have over 40 ELdots in total. We began the process of spectroscopically confirming their association with the target galaxy, with longslit spectroscopy at the MDM 2.4 m telescope (Section 3.2). Various difficulties with blind offsets and large observational overhead forced us to discontinue the follow-up longslit spectroscopy before it was complete. Yet, of the dozen ELdot spectra that were obtained, all were confirmed as outlying H ii regions (the Hα emission line, instead of any of the redshifted strong blue emission lines, falls in the SINGG narrow bandpass). This high confirmation rate led us to pursue deeper multi-slit spectroscopy with the Gemini 8 m telescope, as described in Section 3.1.

Figure 1 shows the Hα continuum-subtracted images for the 13 H i Rogue fields in which we found ELdots. For nine of these galaxies, we also show H i column density contours generated from reduced H i data cubes kindly provided by several project PIs (see Table 1). These images highlight the inner- and outer-galaxy SF and often do not show the full MDM field of view or the full extent of the H i distribution. We show R₂₅ on the images, in addition to all of the H ii regions (outer + inner) for which we were able to obtain GMOS multi-slit spectra (see the next section). Regarding the few galaxies for which H i data cubes were not available to us, we refer the reader to the online H i Rogues catalog and references therein (Hibbard et al. 2001). Table 1 lists the optical, H i, and derived properties of this sample. Column 7 provides an estimate of the total oxygen abundance for each galaxy from the literature. In most cases, the oxygen abundance measurement was made with various strong emission-line ratios for H ii regions in the central parts (within R₂₅ in all cases) of each galaxy. Column 8 lists the H i morphology code from the H i Rogues catalog for each target. The codes are defined in the caption of Table 1, and the H i morphology of each target is described in greater detail in Section 5.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 1. Properties of the Galaxies in the H i Rogues Sample

Name	R.A.	Decl.	d	R25	E(B − V)	Abun	[N ii]/Hα	H i Morph	Beam Size	H i Reference	M*	$M_{\rm H\,\mathsc{i}}$	SFRtot
(1)	(2)	(3)	(4)	(5)	(6)	(7)	(8)	(9)	(10)	(11)	(12)	(13)	(14)
NGC 2146	06 18 38.1	+78 21 26.3	22.45	22.15	0.096	8.681	0.471	mMR	77 × 49	Taramopoulos et al. (2001)	6.02	2.10	2.2
NGC 2782	09 14 05.1	+40 06 48.9	37.30	12.93	0.016	8.801	0.471	mMR	9.4 × 8.7	Smith (1994)	1.55	0.42	1.2
NGC 3227	10 23 29.0	+19 52 42.3	20.85	20.57	0.023	8.602	0.302	E−Sp	80 × 47	Mundell (2001)	5.17	0.20	0.47
NGC 3239	10 25 04.3	+17 09 11.5	8.10	6.26	0.032	8.191	0.071	Sp+ −Sp+	64.3 × 54.6	Iyer et al. (2001)	0.122	0.12	0.29
NGC 3310	10 38 44.3	+53 30 07.9	18.10	9.17	0.022	8.401	0.231	mMR	20 × 20	Iyer et al. (2001)	0.798	0.52	3.0
NGC 3359	10 46 37.7	+63 13 25.5	17.80	13.05	0.008	8.603	0.153	H ic	30 × 30	Boonyasait et al. (2001)	1.31	0.45	0.89
NGC 3432	10 52 31.2	+36 37 06.9	12.98	12.46	0.013	8.301	0.121	mM	30 × 30	Swaters et al. (2002)	0.313	0.60	0.50
NGC 3718	11 32 35.0	+53 04 05.2	17.00	12.24	0.014	8.481	0.221	WARP	30 × 30	Verheijen & Sancisi (2001)	2.94	1.00	0.25
NGC 3893	11 48 38.6	+48 42 40.8	18.13	12.81	0.021	8.601	0.331	mM	30 × 30	Verheijen & Sancisi (2001)	2.45	0.60	1.60
NGC 5774	14 53 42.5	+03 34 56.8	26.80	12.18	0.042	8.554	0.264	Sp0 −Sp0	29 × 23	Irwin (1994)	1.10	0.54	0.47
NGC 5775	14 53 57.2	+03 32 48.5	26.80	20.04	0.042	8.704	0.364	Sp0 −Sp0	29 × 23	Irwin (1994)	1.36	0.91	0.70
NGC 6239	16 50 05.9	+42 44 18.9	22.30	8.92	0.018	8.404	0.154	mM	18.2 × 16.7	Hogg & Roberts (2001)	0.403	0.70	0.94
UGC 5288	09 51 17.2	+07 49 37.1	6.03	1.37	0.034	8.085	0.075	ENVL	20.2 × 16.9	van Zee & Haynes (2001)	0.00852	0.023	0.0063
UGC 9562	14 51 14.4	+35 32 32.0	23.80	3.81	0.013	8.646	0.226	Sp0 −Sp0	22.5 × 20.0	Cox et al. (2001)	0.095	0.082	0.15

Notes. (1) NGC name of the galaxies. (2) R.A. (J2000) in hms. (3) Decl. (J2000) in dms. (4) Redshift-independent distances from NED, in Mpc. (5) Length of the 25th magnitude elliptical isophote (R₂₅) semimajor axis, in kpc. (6) Color excess E(B − V) from Schlegel et al. (1998). (7) and (8) Oxygen abundances (average or integrated) as 12 + log(O/H) and [N ii]/Hα ratios from various sources: 1. Moustakas & Kennicutt (2006): strong-line R23 abundances and [N ii]/Hα ratios from integrated spectroscopy. 2. Lisenfeld et al. (2008): strong-line R23 abundances and [N ii]/Hα ratios from one H ii region at 0.8 × R₂₅. 3. Martin & Roy (1995): average of R23 oxygen abundances and [N ii]/Hα ratios for disk H ii regions located between 4 and 9.5 kpc in projected distance from the optical center. 4. Márquez et al. (2002): average of strong-line [N ii]/Hα abundances and ratios (Pettini & Pagel 2004) for a number of disk H ii regions. 5. van Zee & Haynes (2006): average strong-line R23 oxygen abundances and [N ii]/Hα ratios for six H ii regions within R₂₅. 6. Shi et al. (2005): average strong-line R23 oxygen abundances and [N ii]/Hα ratios for H ii regions within R₂₅. (9) H i morphology code from the online H i Rogues catalog—mMR: minor merger remnants; E−Sp: interacting doubles, only one with H i; Sp⁺ −Sp⁺: interacting doubles, both with H i, both with tails; H ic: detached H i clouds; mM: minor mergers; WARP: galaxies with two-sided warps; Sp⁰ −Sp⁰: interacting doubles, both with H i, no tails; ENVL: galaxies with extended H i envelopes. (10) Beam size of H i observations, in arcseconds. (11) Reference for the H i data. (12) Stellar masses derived from B-, V-, and sometimes R-band photometry and the luminosity/color and stellar mass functions of Bell et al. (2003) in units of 10¹⁰ ${\cal M}_\odot$. (13) H i masses for the galaxy in units of 10¹⁰ ${\cal M}_\odot$, obtained from the literature (see the text for the original sources) and corrected for the galaxy distances we use in Column 5 of this table. (14) Derived SFRs in ${\cal M}_\odot$ yr⁻¹ from the total Hα luminosity obtained from the MDM 2.4 m images.

Download table as:  ASCIITypeset image

For the calculations we perform in Section 6, we derive the total baryonic (stellar + gas) and global star formation rate (SFR) for each galaxy in our sample. To obtain a value of the stellar mass, we use the published total integrated broadband magnitudes in B- and V-bands available on NED in conjunction with the stellar mass–luminosity/color relation from Bell et al. (2003). Most of the total B-band and V-band apparent magnitude estimates come from the 1991 De Vaucoleurs Catalog, using a combination of new and previously published data, and are corrected for internal and Milky Way extinction. UGC 5288 is an exception in this regard (not included in the De Vaucoleurs Catalog), and therefore we used the broadband optical B- and R-band photometry from van Zee (2000). To obtain the total stellar mass from the B-band luminosity and the B−V color (sometimes B − R), we use Table 7 of Bell et al. (2003) which gives coefficients for their derived stellar mass–luminosity/color relation for Sloan Digital Sky Survey (SDSS) galaxies. These stellar masses are accurate to within 50% and are given in Column 11 of Table 1. In some cases, we are able to compare our B − V results with other calibrations of the Bell et al. (2003) mass–luminosity/color relations using different broadband filters, which give values that are consistent within the errors. The H i masses in Column 12 of Table 1 are obtained from the literature, and references are given in Column 10.

We calculate SFRs using the relation SFR (${\cal M}_\odot$ yr⁻¹) = 5.3 × 10⁻⁴² L(Hα) (erg s⁻¹) (Calzetti 2008) and obtain total Hα luminosities from the MDM 2.4 m images within an elliptical aperture at 2 × R₂₅. These total Hα luminosities are corrected for Galactic extinction (Schlegel et al. 1998) and for the contribution of [N ii]λ6583 using the [N ii]/Hα emission-line ratios from the literature (listed in Column 9 of Table 1) and convolving this contamination with the SINGG narrowband filter transmission curve (see Appendix A of Meurer et al. 2006 for details). This [N ii]/Hα emission-line ratio for each galaxy represents a global average (from integrated spectra or averages of individual H ii regions in the galaxy) over its optical extent. The only exception is NGC 3227, in which only one measurement of an H ii region at ∼0.8 × R₂₅ was available. We did not attempt to add in the potential effects from metallicity gradients across the galaxy since such effects would be negligible in light of our rather high photometric errors. Considering errors in the continuum subtraction, occasional Hα filter reflections/diffuse patterns, and uncertainties in the extinction correction, we estimate that these total Hα luminosities, and thus the total SFRs, are accurate to within ∼20%.

Table 2 lists the remaining 14 targets in which we did not find any ELdots, along with their H i morphologies, distance, and R₂₅. Although their properties are not fundamentally different from those of the H i Rogues with ELdots, we do note that the only three optical early-type elliptical galaxies with H i that we imaged with the MDM 2.4 m appear in this sample.

The H i Rogues sample is neither complete nor representative. It was compiled as a sample largely dominated by interacting or otherwise "weird-featured" gas-rich galaxies. Of the 13 galaxies (or galaxy pairs) studied here, 10 are mergers or remnants of mergers with extended gaseous features. Both NGC 3359 and UGC 5288 are galaxies with visible SF contained in a much larger, more extended concentration of H i gas which may or may not be the result of some past interaction despite very regular kinematic structures. NGC 3718 exhibits a large warp in its H i distribution of unknown origin. As we discuss in Section 5, each individual system is at a different stage in the interaction process and the nature of the interaction ranges from minor to major.

To compare our Galaxies with a sample of "H i-normal" galaxies, we use the H i-selected sample of the SINGG (Meurer et al. 2006). The SINGG sample was selected from the H i Parkes All-Sky Survey (HIPASS; Barnes et al. 2001) purely on the basis of H i mass and recessional velocity. The primary goal of SINGG is to uniformly survey the SF properties of H i-selected galaxies across the entire H i mass function sampled by HIPASS (log($M_{\rm H\,\mathsc{i}}$/M_☉) ∼ 8.0–10.6) in a way that is blind to previously known optical properties of the sources.

Hanish et al. (2006) used the SINGG Release 1 data to calculate the luminosity densities in Hα and the R band of the local universe for H i-selected galaxies in the SINGG sample. They then derive the fractional contribution to these luminosity densities as a function of various additional parameters (see their Figure 5 and Table 5). Compared to our minimum ${\cal M}_{\rm H\,\mathsc{i}}$, about 12% and 22% of the R and Hα luminosity density, respectively, comes from galaxies with lower ${\cal M}_{\rm H\,\mathsc{i}}$ values, while about 6% of both luminosity densities comes from galaxies having ${\cal M}_{\rm H\,\mathsc{i}}$ larger than our maximum value. The median ${\cal M}_{\rm H\,\mathsc{i}} = 5.3\times 10^9\,{\cal M}_\odot$ is slightly higher than the value ${\cal M}_{\rm H\,\mathsc{i}}\approx 3.6\times 10^9\,{\cal M}_\odot$ corresponding to the midpoint contributing to both these luminosity densities. Hence, the ${\cal M}_{\rm H\,\mathsc{i}}$ values for our sample correspond well to that of SINGG, but are skewed perhaps to slightly larger values. We can also compare ${\cal M}_\star$ values, if we convert the SINGG R-band luminosities to stellar mass. For this rough calculation, we assume ${\cal M}_\star /L_R \approx 1$ and adopt M_R = 4.61 ABmag for the Sun's absolute magnitude. Then we find that at most 3% of the local R and Hα luminosity densities have ${\cal M}_\star$ lower than our minimum value, while 30% of the R luminosity density and 13% of the Hα luminosity density come from galaxies more massive than our largest ${\cal M}_\star$. The median ${\cal M}_\star = 1.31\times 10^{10}\, {\cal M}_\odot$ for our sample, while the midpoint for the R and Hα luminosity densities is ${\cal M}_\star = 3.3\times 10^{10}\,{\cal M}_\odot$ and $1.4 \times 10^{10}\,{\cal M}_\odot$, respectively. Hence our sample covers the ${\cal M}_\star$ range of the SINGG sample fairly well, but may be deficient in the highest ${\cal M}_\star$ galaxies.

In the spring of 2008, we obtained multi-slit spectra of numerous H ii regions associated with the 13 H i Rogues using the Gemini-North Multi-Object Spectrograph (GMOS-N) in order to make metallicity measurements of outer-disk and/or stripped gas. We were somewhat limited by the 5' × 5' field of view of GMOS, which restricted us to partial coverage of the Rogue galaxies and their outskirts. The generation of the GMOS slit mask required that we obtain pre-imaging with GMOS-N in queue mode prior to our spectroscopic observations. We used a 90 Å wide Hα filter for GMOS to image the 13 fields in our sample, and in concert with the R-band continuum-subtracted MDM images were then able to pick out and obtain precise instrumental positions for outlying and galactic H ii regions. In only a few cases, we selected potential outlying H ii regions from the Gemini pre-imaging that were not detected in the MDM images. Four to five alignment star holes per-field and slitlet lengths of at least 10'' allowed us to fit ∼25 science slitlets per mask. In practice, however, we had an average of eight usable science slitlets per mask (minimum = 4; maximum = 17) due to H ii region position/alignment conflicts, very faint H ii regions without multiple emission-line detections (discovered post-reduction), several blank "sky slitlets" (see below), constraints on the instrument rotator angle due to available guide stars, and an increased slitlet length for the brightest central H ii regions (for proper night-sky emission-line subtraction).

Our spectral observations were taken in queue mode under the following conditions: a gray lunar phase, an image quality of 15 (occurring 85% of the time), and some cloud cover (better than that occurring 50% of the time). We used the B600-G5303 600 l mm⁻¹ grating with various grating angles optimized for each individual slit mask so that key emission lines would not fall in either of the two CCD chip gaps. Central wavelengths range between 4100 and 4800 Å. The above grating configuration results in a spectral coverage of 2760 Å and was chosen so that we could detect emission lines between ∼3700 and 5007 Å for the vast majority of the H ii regions on our masks. Several H ii regions falling close to the image edges did not achieve this range, and, for a few of those H ii regions, we were able to detect the Hα and [N ii] λ6583 emission lines. However, we generally did not obtain spectral data above 5500 Å for two reasons: (1) the low-resolution grating (R150, which would have given a larger spectral coverage) is not sensitive enough in the blue to obtain a detection of [O ii]λλ3727 for our faint objects and (2) an additional grating configuration for red spectral coverage would have doubled our observing time and therefore would have reduced the number of targets we could observe with Gemini. Since we opted for only blue spectral coverage, we used the R23 method for obtaining abundances. We use additional longslit data with red spectral coverage to break the degeneracy of the R23 relation (see Section 4.2 for details). 10 wide slitlets and 4 (spatial) × 2 (spectral) binning yield a spectral resolution of 1.8 Å pixel⁻¹ and a spatial resolution of 0.28 arcsec pixel⁻¹. Our science exposure times were 3 × 20 minutes for each field.

Arc-lamp (CuAr) and spectral flat-field calibrations were done between and after each of the science observations. Baseline Gemini calibrations additionally include the observation of one standard star per grating configuration, HZ-44 in our case, over the execution of the entire science program. The main dangers of using a single standard not taken on the same night as the targets are that seeing variability combined with differential refraction will alter the effective sensitivity function, and that the atmospheric extinction curve will change. Fortunately, in every case, our targets (and the standard star) were observed close to zenith, minimizing the effect of seeing variability with differential refraction. Additionally, the variation in atmospheric extinction (or overall throughput) is minimal on Mauna Kea at optical wavelengths (Krisciunas et al. 1987). Nonetheless, our single relative sensitivity function probably subjects our relative line fluxes to higher errors than if we had used at least one standard star on every night data were taken.

For the reduction of the GMOS multi-slit data, we used the IRAF Gemini/GMOS software package, specifically designed for GMOS multi-slit data reduction. Once the basic reduction is performed (GPREPARE, GSFLAT, GSREDUCE), we apply the wavelength solution from GSWAVELENGTH, using GSTRANSFORM, the fits to which had a typical rms of 0.8–1.0 Å. The short slit lengths complicated night-sky line subtraction for several of the bright or more extended H ii regions in our sample. Several blank sky slitlets on every mask allowed us to properly subtract the sky for these bright/extended objects whose emission lines occupied the entire length of the slitlet. We therefore did two passes of sky subtraction for each mask data set, one using GSSKYSUBTRACT, and the other using the manual sky subtraction method. Using GEMCOMBINE and GSEXTRACT, we obtained one-dimensional reduced spectra, which we flux calibrated using GSCALIBRATE based on our sensitivity function derived from HZ-44.

As a test of our manual sky subtraction, we compared the two methods using calibrated one-dimensional spectra of the compact, fainter H ii regions and found that line fluxes were identical. Typical rms noise in the final, reduced spectra ranges from ∼9 × 10⁻¹⁸ erg s⁻¹ cm⁻² at 3730 Å, to ∼2 × 10⁻¹⁸ erg s⁻¹ cm⁻² at 5000 Å, to ∼7 × 10⁻¹⁹ erg s⁻¹ cm⁻² at 6500 Å (when data were redder due to objects being on image edges). These values do vary somewhat for different slit masks/fields, depend on the aperture size, and are noted only to illustrate that the emission-line Poisson noise is not a major contributor to the error. The rms of the fitted sensitivity function was 6.5%, and we estimate that the addition of flux calibration uncertainty, read noise, sky noise, and flat-fielding errors brings our total errors in line flux measurements up to ∼10%.

Over the course of three observing runs, one at the MDM 2.4 m in the winter of 2007, one at the Keck 10 m in the spring of 2010, and one at the Shane 3 m telescope at Lick Observatory in the spring of 2011, we obtained longslit, low-resolution spectra for outlying H ii regions in our sample of H i Rogues with extended SF. We use these spectra in the subsequent analysis to obtain [N ii] λ6583 and Hα emission lines, since these lines were generally absent in the GMOS multi-slit data. At MDM, we used the low-resolution Mark III spectrograph with the Echelle CCD, employing the 300/5400 grating and a 0.824 arcsec slit width. The 300/5400 grating centered at 6500 Å gives a spectral coverage of 3120–9879 Å, a range containing all of the optical emission lines we could hope to detect, and a resulting dispersion of 3.3 Å pixel⁻¹. Blind offsets from nearby stars were required for positioning the often very faint ELdots in the slit. As mentioned in Section 2.2, the run on the MDM 2.4 m telescope was intended as a spectroscopic confirmation run, and we obtained spectra for only a dozen ELdots. Exposure times ranged from 300 s to 2 × 600 s, depending on the narrowband flux of the ELdot.

At Keck, we were able to use the low-resolution imaging spectrometer (LRIS) for longslit spectroscopy of three H i Rogues with ELdots, during an observing run which had several target list "holes" at the very start and end of the nights. The data were taken with the 600 l mm⁻¹ grism (blue side) and 400 l mm⁻¹ grating (red side) which gives a spectral coverage between 3000 and 9300 Å. Binning the data 2 × 2 resulted in a dispersion of 1.2 and 2.3 Å pixel⁻¹ for the blue and red cameras, respectively. ELdots were positioned in the 1'' slit precisely by aligning the offset position angle between it and a bright star with the slit angle. In one case, NGC 6239, we were able to align several other galactic H ii regions with the slit position angle as well. A problem with an amplifier on the red camera results in an unusable lower CCD chip on the red side (each side has two chips). Exposure times were 2 × 600 s in the blue and 3 × 400 s in the red.

At Lick, we used the Kast spectrometer on the Shane 3 m telescope to observe nine remaining outlying H ii regions without red spectral coverage. We used Grism 1(600/4310) on the blue side, the D55 dichroic, and the 600/7500 grating on the red side, which gave full spectral coverage between 3300 and 8800 Å. Exposures were taken through light clouds in most cases, and exposure times ranged from 30 minutes to 1.5 hr. The same observing strategy of bright-star position angle alignment used at the Keck telescope was employed here, and our overall success rate of acquiring the faint outlying H ii regions was 100%.

Basic reduction steps, which included bias-subtraction, flat-fielding, and flux calibration with a single spectrophotometric standard star (one for each telescope), were performed with the IRAF longslit package for MDM data and with the LowRedux⁶ IDL software package for the Keck and Lick data. We use these longslit data only to obtain [N ii]/Hα ratios in order to break the R23 degeneracy (as described in Section 4.2) for our blue GMOS multi-slit data. Because these two emission lines ([N ii] λ6583 and Hα at λ = 6563 Å) are so close in wavelength, we did not require an accurate flux (or wavelength) calibration or reddening correction. Table 3 lists the outlying H ii regions for which we were able to observe the [N ii]/Hα ratio using the longslit data. Table 3 also lists the outlying H ii regions so far to the edge of the CCD in the GMOS data for NGC 2146 and NGC 2782 that we were able to obtain [N ii]/Hα ratios.

The H ii regions in our final sample and marked in Figure 1 were included based on the presence of the relevant emission lines for R23 strong-line abundances. A direct measurement of the H ii region's oxygen abundance from the oxygen emission lines present in its spectrum requires a measurement of the temperature-sensitive line ratio [O iii] λλ4959, 5007/λ4363 (Osterbrock 1989). In these spectra, the critical emission line [O iii] λ4363 is not detected. Therefore, we use the R₂₃ = ([O ii] λ3727 + [O iii] λλ4959, 5007)/Hβ (Pagel et al. 1979) calibration of McGaugh (1991) to determine the oxygen abundance for numerous H ii regions in 13 H i Rogue galaxies. The majority of the multi-slit spectra contained Hγ as well, which we used to correct for interstellar reddening (see below). The highest signal-to-noise spectra on every mask often contained additional emission lines, including Hδ, He i λ4026, and He ii λ4687, occasionally [Ne iii] λ3869 and λ3970, and rarely the very faint, temperature-sensitive [O iii] λ4363 auroral emission line which provides a direct oxygen abundance measure. Using the IRAF task splot, and fitting Gaussian profiles we obtain integrated line fluxes. For the inner H ii regions with substantial underlying continuum, we corrected for absorption from underlying stellar populations by subtracting Gaussian fits to the Balmer line absorption.

We apply a correction for interstellar reddening to all line measurements from the observed Hγ to Hβ ratio for case B recombination where Hγ/Hβ = 0.459 at an effective temperature of 10,000 K and electron density of 100 cm⁻³ (Hummer & Storey 1987). These lines have been corrected for underlying Balmer absorption. We use a reddening function normalized at Hβ from the Galactic reddening law of Cardelli et al. (1989) assuming R_v = A_v/E(B − V) = 3.1. The mean E(B − V) from the Balmer decrement for these H ii regions is 0.252. We do note additionally that the reddening has very little impact on the calculated strong-line abundances, which have very large errors. Tables 4–16 present the reddening-corrected strong-line measurements for the H ii regions in all 13 Rogue galaxies. All listed line fluxes are relative to Hβ = 100.

We use the McGaugh (1991, henceforth M91) calibration of the R23 relation to obtain the oxygen abundances of all the H ii regions in our sample of 13 Rogue systems. The R23 relation exhibits a well-known degeneracy, with a turnover in the relation at Z ∼ 0.3 Z_☉ (12+log(O/H) ∼ 8.35). The most robust way to place an H ii region on the upper or lower branch of the R23 relation is to use the [N ii] λ6583 to [O ii] λλ3727 ratio. When log [N ii]/[O ii] < −1.0, H ii regions lie on the lower metallicity branch of the R23 relation (Kewley & Ellison 2008). Since our multi-slit data are bluer than 5500 Å in most cases, we cannot use this method. Instead, we use the [N ii]/Hα ratio as described in Kewley & Ellison (2008) from our longslit spectra and from the literature. The primary benefit of this line ratio is that it does not require an accurate (or absolute) flux calibration or reddening correction due to the proximity in wavelength of [N ii] λ6583 and Hα. This benefit is important for our data redward of 5600 Å, since we are using spectra from four different telescopes taken under varying conditions.

Kewley & Ellison (2008) discuss how to use the [N ii]λ6583/Hα ratio (N2 index) to break the degeneracy in the R23 relation. If log N2 < −1.3 ([N ii]/Hα < 0.05), there is a high degree of certainty (∼86%) that the oxygen abundance is on the lower branch of the R23 relation. Whereas if log N2 > −1.1 ([N ii]/Hα > 0.08), the oxygen abundance is on the upper branch. Between −1.1 and −1.3 (0.05 and 0.08), log (N2) (([N ii]/Hα) does not accurately discriminate between the upper and lower branches of R23 because the oxygen abundance is very close to the turnover at 12 + log (O/H) = 8.3. This method of breaking the R23 degeneracy works because the N2 index is sensitive to and monotonic with the oxygen abundance to within 0.3 dex accuracy at a 95% confidence level up to 12 + log(O/H) = 8.8 (Pettini & Pagel 2004, henceforth PP04). The primary drawback of using this method, then, is the large uncertainty in the calibration (∼0.4 dex), and the limited range over which it is useful. It is apparent from the [N ii]λ6583/Hα ratios given in Tables 1 and 3 that most of the outlying and central H ii regions lie on the upper branch of the R23 relation, except for a few cases which we discuss below.

We do not have any measurements of the N2 index for the outlying H ii regions in NGC 2146 and therefore cannot definitively place them on either branch. Furthermore, the integrated [N ii]λ6583/Hα ratios of NGC 3239 and the average [N ii]λ6583/Hα ratio of the H ii regions in UGC 5288 put both galaxies in the ambiguous region of the N2 index. Fortunately, for UGC 5288 and NGC 3239, there are published values of the log [N ii]/[O ii] that allow us to clear the N2 index ambiguity in the central regions. The integrated spectroscopy of the inner R = 3 kpc (∼0.5 × R₂₅) of NGC 3239 indicates a log [N ii]/[O ii] of −1.05 (Moustakas & Kennicutt 2006). UGC 5288 has an average log [N ii]/[O ii] of −1.16 within R₂₅ (van Zee & Haynes 2006). Therefore, the inner H ii regions of both galaxies lie on the lower branch of the R23 relation.

The outlying H ii regions of NGC 3239 and UGC 5288 have N2 indices in the ambiguous region as well. In the case of UGC 5288, the upper branch M91 R23 values for its outer H ii regions would put them at super-solar oxygen abundance, which is highly unlikely given that it is a dwarf galaxy and its central regions have an average oxygen abundance of 0.1 solar. We therefore assume that these outlying H ii regions are on the lower branch of the R23 relation along with their more central counterparts. In the case of NGC 3239, the situation is less clear, since both upper- and lower-branch values of the oxygen abundance are near the R23 turnover region. We cannot make any assumptions for the outlying H ii regions in this galaxy and therefore report both upper- and lower-branch values.

The H i gas densities at the locations of the outlying H ii regions range from (3–500) × 10¹⁹ H i atoms cm⁻², corresponding to a range in gas surface density of (0.25–80) M_☉ pc⁻². The median value, which is typical of most of the gas column densities at the outlying H ii region locations, is 4 × 10²⁰ cm⁻² or ∼3 M_☉ pc⁻². This gas surface density is at the traditional "threshold" level for SF, usually cited between 2 and 8 M_☉ pc⁻² (Schaye 2004). Thus, we are truly examining SF at low average gas surface densities, in the same outer-disk regime explored by Fumagalli & Gavazzi (2008) and Bigiel et al. (2010). The outlying H ii regions presented here cover a range of Hα luminosities, from 10^35.5 to 10^37.5 erg s⁻¹, corresponding to SFRs that range from 2 × 10⁻⁴ to 2× 10⁻⁶ M_☉ yr⁻¹. Most of the outlying H ii regions are isolated within surface areas of roughly a kpc², so these SFRs can be viewed roughly as SFR surface densities in units of M_☉ yr⁻¹ kpc⁻². These values of SFRs and H i gas surface densities place the outlying H ii regions in the same, or lower, "downturn" portion of the Kennicutt–Schmidt law (Wyder et al. 2009), as expected. One caveat to keep in mind is that these stars may be forming in places where the local gas density is higher than the average surrounding gas densities, on a size scale which would be unresolved in the H i synthesis maps. We will present a full analysis of the relationship between the atomic hydrogen content and SF activity in the gaseous outskirts of H i Rogues in a future paper.

First, we discuss previous work on the subject of radial abundance distributions and attempt to reconcile it with the flat gradients observed here. Declining radial abundance gradients, often out to, and/or slightly beyond R₂₅, are commonly observed in spiral galaxies (Oey & Kennicutt 1993; Zaritsky et al. 1994; Kennicutt et al. 2003). The broadest study to date is that of Zaritsky et al. (1994). They present metallicity gradients between 0.1 and 1.0 R/R₂₅ using at least five H ii regions in each of 39 local spiral galaxies with a wide range of morphologies (Sab–Sm; seven galaxies have bars) and luminosities (−21 < M_B < −17). Because the scatter in oxygen abundance at a given galactic radius (0.4 dex at most) for each galaxy is much smaller than the mean abundance range for their entire sample (8.34–9.31), they conclude that these gradients are determined globally, rather than set by local conditions (i.e., the local SFR and gas fraction). Zaritsky et al. (1994) find a wide range of slopes in the 39 radial oxygen distributions, from flat to very steeply declining, at −1.45 dex/R₂₅ corresponding to −0.231 dex kpc⁻¹. Their average slope is −0.59 dex/R₂₅ or −0.07 dex kpc⁻¹. Therefore, we do not see declining abundance gradients similar to those of Zaritsky et al. (1994) where our average gradient is −0.02 ± 0.18 dex/R₂₅ or −0.001 dex kpc⁻¹. Furthermore, our data rule out decreasing radial oxygen abundance gradients steeper than −0.30 dex over 15 kpc (−0.02 dex kpc⁻¹) in these outer regions.

Several studies of metallicity gradients in the outer parts of relatively quiescent spiral galaxies have found a flattening beyond 0.5 R/R₂₅ (Martin & Roy 1995; Bresolin et al. 2009). Similarly, studies of low surface brightness disk galaxies (de Blok & van der Hulst 1998) and dwarf galaxies (Croxall et al. 2009) are suggestive of flat oxygen abundance gradients out to R₂₅. And, the metallicity gradients of interacting pairs appear to be consistently flatter than those of spiral galaxies (Kewley et al. 2010). Yet, this study is the first to measure radial oxygen abundance gradients in the outer regions of a modest sample of galaxies beyond R₂₅. The majority of our Galaxies are interacting, which may partially explain a flattening in the abundance gradients (Rupke et al. 2010; Kewley et al. 2010). Given the wide range of systems (including minor mergers and warps) over which we measure flat gradients, our results suggest that galaxy interactions are not the sole explanation.

There is a fundamental, global relation between galaxy mass and metallicity in the sense that lower mass galaxies have lower overall heavy-element content. Larson (1974) first predicted that the average stellar metal abundance in a galaxy would depend on its mass owing to more significant gas loss from the energy supplied by supernovae in lower mass galaxies. His notion turned out to fit well with observations of nebular oxygen abundances of irregular galaxies several years later by Lequeux et al. (1979) and has become known as the mass–metallicity relation for galaxies in the 36 years since (Skillman et al. 1989; Tremonti et al. 2004). This relation is often parameterized in terms of effective yields and total (gas + stellar) mass to elucidate its origin, where effective yield measures how much a galaxy's metallicity deviates from its "closed-box" value, given a certain gas mass fraction (Dalcanton 2007 provides a more detailed, clear explanation). Systematically, low effective yields of dwarf galaxies have been taken as evidence that low-mass galaxies have evolved less as "closed boxes" than their more massive counterparts (Tremonti et al. 2004).

Although supernova-driven metal expulsion is the most popular and classic explanation for the low effective yields of dwarf galaxies (Larson 1974; Tremonti et al. 2004), there are several issues that complicate the picture. Dalcanton (2007) shows, for instance, that metal loss itself does not have a strong dependence on galaxy mass and that any subsequent SF following an episode of mass loss quickly pushes the effective yield up to its closed-box value. In the calculations presented by Dalcanton (2007), low SF efficiencies in gas-rich dwarfs between episodes of mass loss are what keep their effective yields low. Dilution by infalling low-metallicity gas may also play some role, but alone cannot explain the mass–metallicity relation (Dalcanton 2007). An explanation offered by Tassis et al. (2008) is that the metals are not expelled via supernova-driven winds, but rather efficiently mixed via gravitational processes into the hot gas, where they are unobservable by standard ground-based methods. Coupled with low SF efficiencies of low-mass, gas-rich dwarfs, Tassis et al. (2008) can reproduce the observed mass–metallicity relation without including winds from supernova in their model. Finally, if low-mass galaxies preferentially do not make as many O-type stars as more massive galaxies (Meurer et al. 2009; Lee et al. 2009), as in the IGIMF theory, then their effective yields will appear to be low (Köppen et al. 2007).

We calculate the effective oxygen yield for each galaxy, Z/ln(μ⁻¹), where μ is the gas fraction and Z is the metallicity by mass, using the mean H ii region oxygen abundance as a proxy for total metal content, the stellar masses listed in Table 1, and total gas masses computed from the H i masses given in Table 1 plus a 36% contribution from helium. We do not include molecular gas in these calculations, although it most certainly is present. Although the molecular-to-neutral gas content varies widely within and among galaxies (Bigiel et al. 2008; Leroy et al. 2008), the total galaxy molecular gas mass is rarely, if ever, larger than its neutral gas mass. If we assume that, on galaxy-wide scales, the molecular gas content scales to first order with the neutral gas content, then the effect on the overall mass–metallicity relation would be negligible.

Figure 3 (upper panel) plots the effective oxygen yield versus total baryonic mass for our Galaxies (filled circles; NGC 5774 and NGC 5775 we consider separately, though they are technically part of the same H i envelope) along with the empirical relation of Tremonti et al. (2004). To include both NGC 2146 and NGC 3239 in this figure, we have assumed upper-branch metallicities for the H ii regions. While this assumption may not be valid, the points in the upper panel of Figure 3 do not change appreciably with branch assumption. With the exception of NGC 3227, every galaxy in our sample appears to lie along the empirical Tremonti et al. (2004) relation, within the 95% contours (dash-dotted lines). Tremonti et al. (2004) explain the decrease in metal yields with decreasing galaxy baryonic mass by invoking metal-rich supernova-driven winds that preferentially escape from smaller potential wells of less-massive galaxies.

Zoom In Zoom Out Reset image size

The red asterisks in Figure 3 (upper panel) directly above or below the filled-circle galaxy points mark an "expected" oxygen yield determined by using the galaxy's total stellar mass and assuming a net oxygen yield of 0.01 (average value from Maeder 1992; the mass of oxygen ejected by all stars per unit mass of matter locked up in stars). All galaxies, with the exception of UGC 9562, have yields consistent with those predicted from a rough estimate that simply accounts for all metals using the existing stellar population.

As was the case with NGC 2915, considered in Werk et al. (2010a), there is only a low level of ongoing and previous SF at the locations of the outermost H ii regions. In this sense, the lack of radial oxygen abundance gradients beyond the optical "edges" of galaxy disks seems to necessarily imply some type of efficient metal transport. In the context of the mass–metallicity relationship, we can consider the yields for the outer and inner gaseous regions separately. We show this separation in Figure 3 (lower panel) with the assumption that 95% of the stellar mass is contained within R₂₅, and calculating the H i mass within R₂₅ from the H i 21 cm synthesis data, when it was available to us. In the four cases that it was not, we estimated by eye the total H i mass within and outside of R25 from the H i column density maps available from the online the H i Rogues Gallery. The percentage of H i mass contained in R₂₅ varies from 10% to 60% for the galaxies.

First, keeping metallicity constant while lowering the stellar mass contribution to the effective yields relative to the gas mass contribution (outer H ii regions) shifts the black points in Figure 3 (upper panel) to the left and up, shown as downward-facing green triangles in Figure 3 (lower panel). Note that the green triangles for NGC 2146 and NGC 3239 move down by 0.7 dex and 0.4 dex, respectively, if we instead assume lower-branch metallicities for their outer H ii regions. This figure shows that the effective yields for the outer disk lie above their "closed-box" values. Conversely, the general trend of lowering the gas mass contribution relative to the stellar mass contribution while keeping metallicity constant (inner H ii regions) has the effect of moving the points to the left and down, shown as blue triangles in Figure 3 (lower panel). For the most part, the effective yields for the inner disk fall below their "closed-box" values. Including molecular gas in this analysis would only make the effect more striking, since, in general, the fraction of molecular gas decreases with galactocentric radius (Leroy et al. 2008). Outer-galaxy oxygen abundances are generally more metal-enriched than their gas fractions otherwise suggest, while inner-galaxy oxygen abundances lie below what is expected based on these simple stellar-population-based assumptions. This duality appears to underlie the flat radial abundance gradients of gas-rich, interacting galaxies and apparently isolated extended gaseous disk dwarf galaxies (UGC 5288 and NGC 2915; Werk et al. 2010b).

As with the flat radial oxygen abundance gradient of NGC 2915, a number of physical mechanisms could be at work in various combinations to distribute metals in these H i Rogue galaxies. Very generally, our results indicate that there is efficient metal mixing out to large galactocentric radii, facilitated by galaxy interactions. Nonetheless, there are a few galaxies in our sample that show no evidence of a recent merger or interaction, yet are still well mixed out to large radii. The rest of our sample represents a myriad of interaction types. If indeed interactions are the primary driver of the metal mixing seen here, then the mechanism by which it acts is independent of the interaction details.

The accretion of gas following a merger event or tidal interaction results in tidal torques that drive the gas to the system center (Barnes & Hernquist 1996) and subsequently mix metals into the outer regions (Rupke et al. 2010). Other than tidal torques, metal mixing can be driven by a number of processes: magnetorotational instabilities (Sellwood & Balbus 1999), infall of gas clouds (Santillán et al. 2007), self-gravity in conjunction with differential rotation (Wada & Norman 1999), quasi-stationary spiral structure interacting with a central bar (Minchev & Famaey 2010), viscous flows generated by gravitational instability or cloud–cloud collisions (Ferguson & Clarke 2001), and thermal instability triggered self-gravitational angular momentum transport (McNally et al. 2009). In general, the sound speed, ∼10 km s⁻¹, is a robust upper limit to the speed at which mixing occurs in cold, neutral gas. Measurements of velocity dispersions in the outer H i disks of galaxies are roughly consistent with this number, generally no less than 8 km s⁻¹ (Tamburro et al. 2009). In some strongly interacting galaxies, the velocity dispersions may be higher due to these gravitational perturbations, up to 30 km s⁻¹ in some rare cases (e.g., NGC 1533; Ryan-Weber et al. 2003). Furthermore, in the extended H i disk of NGC 2915, Elson et al. (2010) find evidence for radial flows of 5–17 km s⁻¹ that may be driving central material outward into the extended disk. They note that these flows are consistent with the finding of higher-than-expected oxygen abundances in the outer gaseous disk, as observed by Werk et al. (2010a).

A simple calculation of mixing timescales, assuming that there is mixing at the sound speed over 15 kpc, yields 1.5 Gyr as the lower limit to the timescale over which mixing occurs in our extended H i features. Yet, the metals generated by massive stars are generally returned to the ISM on much shorter timescales, on the order of 100 Myr (Tenorio-Tagle 1996). Even if the velocity dispersion in the cold gas is three times higher (30 km s⁻¹; the most extreme end) due to interactions, the lower limit to the cold-gas mixing timescale is 500 Myr. By assuming a net oxygen yield of 0.01 (Maeder 1992; the mass of oxygen ejected by all stars per unit mass of matter locked up in stars), we can calculate this required average outward metal "momentum" that would reproduce a flat oxygen distribution. Taking an average inner-R₂₅ SFR of 0.5 M_☉ yr⁻¹, and an average separation between inner and outer SF of 15 kpc, we find that there needs to be a constant outward movement of 0.0025 M_☉ of oxygen at 150 km s⁻¹. We can repeat this calculation for the most compact galaxies in our sample (NGC 3239, UGC 5288, and UGC 9562), now taking an average inner-R₂₅ SFR of 0.15 M_☉ yr⁻¹ and an average separation between inner and outer SFs of 4 kpc. The corresponding "metal momentum" needed to wipe out any abundance gradient is 0.0007 M_☉ of oxygen at 40 km s⁻¹. While metal mixing in cold neutral gas probably happens, it most likely does not happen on fast enough timescales to wipe out a metallicity gradient that would arise in a typical star-forming spiral galaxy from a radial dependence of the SFR (i.e., SF is concentrated in the central regions, therefore metals are concentrated in central regions). With the advent of high-resolution radio telescopes such as the EVLA and ALMA, future work will be able to address the kinematics and motions of cold gas in considerably more detail.

Given the fast mixing speeds required, we consider that the metal transport may be occurring predominantly in a hot gas component as proposed by Tassis et al. (2008) and others. Although the dominant driver of the mixing of metals in the hot-phase gas is unclear, suggested sources include diffusion, gravitational instabilities, cold accretion flows, and galaxy interactions (Kereš et al. 2005; Tassis et al. 2008). The recent finding of O vi absorption over the entire Milky Way disk, irrespective of circumstellar environments and spiral arms (Bowen et al. 2008), may provide some support for the idea that there can be efficient mixing in the hot gas phase of the ISM. Furthermore, the hot X-ray gas generated by metal-enriched supernova-driven blowouts (i.e., winds) is able to propagate to the outer 100 kpc of a galaxy's halo (Strickland et al. 2004). Finally, the three-dimensional adaptive mesh refinement code of de Avillez & Mac Low (2002) finds that inhomogeneities in the ISM can be erased over a couple hundred Myr (dependent on supernova rate) when gas cycles between the disk and the hot halo in a galactic fountain process. Whether or not enough metals can be transported in the hot gas and quickly cooled at large enough radii to reproduce the flat oxygen distribution seen in these 13 galaxies is unclear. However, our data generally support a scenario in which galaxy interactions induce efficient metal mixing in hot gas over the full scale of the galaxy.

Efficient metal mixing has important implications for the origin of the mass–metallicity relationship. Large-scale and universal metal outflow may not be the dominant cause of the low effective yields in some gas-rich dwarf galaxies as advanced by Tremonti et al. (2004) and others. Instead, these low-mass gas-rich galaxies on the whole would have fewer metals than expected from the simple nucleosynthetic yield because they are stirring those centrally generated metals throughout their gaseous outskirts. Coupled with the idea that SF is increasingly inefficient in these low-mass, gas-rich galaxies (Dalcanton 2007), mixing could account for lower effective yields in lower mass galaxies with no real need for large-scale metal blow out.