ALFALFA DISCOVERY OF THE MOST METAL-POOR GAS-RICH GALAXY KNOWN: AGC 198691

Spectroscopic observations of AGC 198691 were performed using the KOSMOS spectrograph on the Mayall 4 m telescope at KPNO on 2015 April 19. Our program utilized both blue-sensitive and red-sensitive instrumental setups that provided wavelength coverages of ∼3500–6200 Å and ∼5000–9000 Å, respectively. This enabled the detection of emission lines blueward to [O ii]λλ3726, 3729 and redward to [S ii]λλ6716, 6731. Unfortunately, our red spectral coverage did not quite reach the [S iii] line at λ9069 Å; this prevented us from determining a reliable sulfur abundance for AGC 198691. The instrumental pixel scale is 0.29 arcsec pixel⁻¹ in the spatial direction. The dispersion for the blue spectral setup is 0.66 Å pixel⁻¹, while for the red it is 0.99 Å pixel⁻¹. The spectra were taken with a slit width of 12; the slit extends 10' in the spatial direction. The sky conditions were clear at the time of the observations. An image quality of ∼14 FWHM was derived from the measurement of stellar profiles in the object acquisition images.

AGC 198691 was observed using the blue setup with three exposures and using the red setup with two exposures, each of which were 900 s in length. The slit was positioned along the parallactic angle as closely as possible in an effort to avoid the effects of differential atmospheric diffraction (Filippenko 1982). The spectrophometric standard star Feige 34 (Massey et al. 1988) was observed to provide calibration data for flux scaling. Wavelength calibration was accomplished using observations of a HeNeAr comparison lamp.

Our initial spectroscopic observations with KOSMOS revealed that AGC 198691 was exceptionally metal-poor. However, the large uncertainties in the weak temperature-sensitive [O iii]λ4363 line motivated us to acquire a confirming spectrum. Additional spectroscopic observations of AGC 198691 were therefore obtained using the Blue Channel spectrograph on the MMT 6.5 m telescope on 2015 November 11 and 12. Our program utilized an overall wavelength coverage of 3690–6810 Å, again enabling detection of emission lines blueward to [O ii]λλ3726, 3729 and redward to [S ii]λλ6716, 6731. The pixel scale along the slit is 0.30 arcsec pixel⁻¹ and the 500 lines mm⁻¹ grating yielded a dispersion of 1.19 Å pixel⁻¹. Our spectra were taken with a slit width of 10 on the first night of observations and 15 on the second night. The increase in slit width on the second night allowed for the inclusion of more nebular emission. Seeing values were estimated to be 08 on both nights.

Five exposures of 1200 s were taken of AGC 198691 on night 1 of the MMT run, with six more obtained on night 2. The spectrophotometric standard stars BD+28 4211, G191-B2B, and GD 50 (Oke 1990) were observed to provide calibration data for flux scaling. Wavelength calibration was accomplished using observations of a HeNeAr comparison lamp.

The data reduction for both the KPNO 4 m and MMT 6.5 m observations were carried out with the Image Reduction and Analysis Facility (IRAF).¹⁸ For both data sets all of the reduction steps mentioned below were carried out on the individual spectral images independently. The fully reduced 1D spectra were then combined into a single high signal-to-noise ratio (S/N) spectrum for the KPNO run and for each night of the MMT run prior to the measurement stage.

Processing of the two-dimensional spectral images followed standard methods. The mean bias level was determined and subtracted from each image using the overscan regions. A mean bias image was then created by combining 10 zero-second exposures taken on the night of the observations. This image was subtracted to correct the science images for any possible two-dimensional structure in the bias. Flat-fielding was achieved using an average-combined quartz lamp image that was corrected for the wavelength-dependent response of the system. For cosmic ray rejection, we used L.A.Cosmic (van Dokkum 2001), taking special care that no emission lines were "clipped" by the software.

One-dimensional spectra were extracted using the IRAF APALL routine. The extraction width (i.e., distance along the slit) was set to 11 pixels or 3.21 arcsec for the KOSMOS spectra and 9 pixels or 2.70 arcsec for the MMT data. The extractions were conservatively wide such that modest changes in the extraction width result in negligible changes in the spectra. Narrower spectral extractions can achieve higher signal-to-noise in the faintest lines, but at a loss of spectral fidelity. Sky subtraction was also performed at this stage, with the sky spectrum being measured in regions on either side of the object extraction window. The HeNeAr lamp spectra were used to assign a wavelength scale, and the spectra of the spectrophotometric standard stars were used to establish the flux scale.

The fully reduced KOSMOS spectrum of AGC 198691 is presented in Figure 3. The spectrum represents 45 minutes of integration with the blue setting and 30 minutes with the red setting. The two independent spectra are plotted together in the figure. The break point between the red and blue portions is located at ∼6200 Å, and is clearly evident by the change in continuum noise characteristics. The extraction regions for the red and blue spectra were carefully matched to ensure that the same portion of the galaxy was being extracted. We checked to ensure that the flux scales in both spectral regions agreed by comparing the fluxes of the [O iii]λ5007 line that was located in both the red and blue spectra for our instrumental setup. The fluxes agreed to within 7% on average for several objects observed during this run. We scaled the fluxes of the red spectra to account for this 7% difference. The only impact that this flux scaling has on our results for the oxygen abundance in AGC 198691 is on the determination of the reddening parameter c_Hβ based on the Hα/Hβ line ratio.

Zoom In Zoom Out Reset image size

The composite MMT spectrum of AGC 198691 is presented in Figure 4. This final spectrum was created by averaging together the combined spectra taken on the two nights of the MMT observations. The nightly combined spectra were in turn the averages of the multiple individual spectra taken on the two nights. This final composite spectrum represents a total of 220 minutes (3.67 hr) of integration time.

Zoom In Zoom Out Reset image size

For both the KOSMOS and MMT spectra the emission lines were measured using SPLOT by directly summing the flux under each line. All spectra were measured independently by two of the authors as a consistency check. In all cases the individual line measurements agreed within the errors.

For an initial estimate of the internal reddening for AGC 198691 we calculate c_Hβ from the Hα/Hβ line ratio. The value of c_Hβ is then used to correct the measured line ratios for reddening, following the standard procedure (e.g., Osterbrock & Ferland 2006),

$$\begin{eqnarray*}&&\displaystyle \frac{I(\lambda )}{I({\rm{H}}\beta )}=\displaystyle \frac{F(\lambda )}{F({\rm{H}}\beta )}\ {10}^{-[{c}_{{\rm{H}}\beta }f(\lambda )]},\end{eqnarray*}$$

where f(λ) is derived from studies of absorption in the Milky Way (using values taken from Rayo et al. 1982). The final c_Hβ value used for the reddening correction was determined from a simultaneous fit to the reddening and underlying stellar absorption in the Balmer lines, using all available Balmer line ratios. Under the simplifying assumption that the same EW of underlying absorption applies to each Balmer line, an absorption line correction was applied to the spectrum that ranged in value from 0 to 5 Å. The value for the underlying absorption was varied in 0.5 Å increments until a self-consistent value of c_Hβ for the Balmer line ratios Hα/Hβ, Hγ/Hβ, and Hδ/Hβ was found. This process led to the determination of a characteristic value for the underlying absorption of 2.0 ± 0.5 Å.

Due to the velocity of the source (V_⊙ = 514 km s⁻¹), the Hg λ4358 night-sky line falls midway between Hγ and the crucial [O iii]λ4363 temperature-sensitive auroral line (see inset box in Figures 3 and 4). The placement of the night-sky line does not interfere with the measurement of either line, but residual flux left over from imperfect sky subtraction does increase the noise at this location.

The distance to AGC 198691 is not well constrained at this time. In this section we present all of the information currently available that can be used to pin down the distance to this galaxy. We also discuss briefly the environment within which it resides.

The distance to AGC 198691 obtained using the velocity flow model employed by the ALFALFA team (Masters 2005) is 7.7 Mpc. This galaxy is located in the direction of the "local velocity anomaly" (e.g., Faber & Burstein 1988), so its flow-model distance is likely to be highly uncertain. If AGC 198691 is in fact at this distance, it is located within the so-called Leo Spur (e.g., Karachentsev et al. 2015), a sparsely populated structure at a distance of roughly 7–10 Mpc. Since this is the only significant grouping of galaxies along the line of sight to AGC 198691 within a distance of ∼20 Mpc, it is tempting to assume that it is in fact located within this structure.

It is common practice to look for possible connections between dwarf galaxies and other nearby galaxies that lie along the same line of sight. There are a number of galaxies with non-velocity-based distances that are located close to AGC 198691 on the plane of the sky that possess comparable redshifts. In Table 3 we list nine such galaxies that have angular separations of ∼6° or less from AGC 198691. There are four galaxies that have distance estimates of between 7.1 and 8.3 Mpc, consistent with the flow-model distance to AGC 198691. Unfortunately, only one of these has a reliable tip of the red-giant branch (TRGB) distance. The others have less accurate distances based on the Tully–Fisher relation or the brightest stars methods. The galaxy with the smallest angular separation from AGC 198691 is included in this group: UGC 5186 is located just 130 away, which corresponds to a minimum separation of 31 kpc if the two galaxies are both located at the distance indicated for UGC 5186. In fact, UGC 5186 is detected in our WSRT H i map of AGC 198691. The other three galaxies in this apparent grouping have minimum separations of between 300 and 820 kpc from AGC 198691, all consistent with the ensemble representing a galaxy group comparable in size to the Local Group at an average distance of 8 ± 1 Mpc.

The second closest galaxy to AGC 198691 on the sky is UGC 5209, located 752 away at a distance of 10.42 ± 0.35 Mpc from the Milky Way (Karachentsev et al. 2015). This object appears to be located on the far side of the Leo Spur. The remaining galaxies in Table 3 have distances beyond 12 Mpc and all have minimum projected separations of 1.0 Mpc or greater from AGC 198691. One of these, UGC 5340, is also known as DDO 68, which is one of the XMD galaxies referred to in Section 1 (12+log(O/H) = 7.20 ± 0.05; Pustilnik et al. 2005; Berg et al. 2012). We note the interesting coincidence that AGC 198691 and DDO 68 are separated by only 54 on the sky.

The HST imaging data described in Section 2 were obtained as part of a larger effort to ascertain the distances and constituent stellar populations for 18 SHIELD galaxies. The color–magnitude diagram (CMD) for AGC 198691 has been analyzed following the methodology of McQuinn et al. (2014) with the goal of deriving a TRGB distance. The short exposure times obtained for this program (one orbit per object split between two filters) result in a relatively shallow CMD. Thus, our TRGB analysis is unable to yield a reliable distance estimate. Either AGC 198691 is nearby (distances less than 10–12 Mpc) and its upper RGB is too sparse to yield an accurate TRGB distance, or it is too far away for the TRGB to be detected in our shallow data (distances greater than 12 Mpc). As we show in Section 4.2, if AGC 198691 is located at a distance of ∼8 Mpc then it likely has a stellar mass comparable to or less than that of Leo P (McQuinn et al. 2013, 2015b; Rhode et al. 2013). Consideration of the deep HST CMD shown in McQuinn et al. (2015b) suggests that a galaxy with a stellar mass significantly lower than Leo P might well have a RGB that is so sparsely populated that it may not be detectable in our shallow data even at modest distances. Whether or not AGC 198691 falls into this category is currently unclear. At this time we believe that the existing HST data cannot rule out a distance to AGC 198691 as low as 8 ± 1 Mpc.

To summarize, we feel that one can make a strong case for AGC 198691 being located within the Leo Spur at a distance of 8 ± 1 Mpc, based primarily on a group association with UGC 5186, UGC 5272, UGC 5427, and AGC 205590. This distance estimate may or may not be supported by our shallow HST CMD. The latter cannot immediately rule out any distance in the range of approximately 7 Mpc, to as high as 20 Mpc. We note that there are galaxies over this full distance range listed in Table 3.

Figure 6 presents three plots showing the galaxy distribution in this region of space. The upper two plots illustrate the distribution of galaxies with reliable distances in supergalactic coordinates, while the lower plot is a simple Hubble diagram and shows how the galaxies within 30° of AGC 198691 on the sky are distributed in the velocity–distance plane. For the sake of clarity we only include galaxies in the upper two panels with SGZ less than −2 Mpc. The green line with stars at 8, 12, and 16 Mpc in all three plots shows the location of AGC 198691 for the expected range of distances discussed above. Nearly all of the galaxies plotted in Figure 6 have secure distances based on the TRGB method or other primary distance estimates. The exceptions are UGC 5186 and UGC 5272, two of the galaxies that, along with UGC 5427 and AGC 205590 (not plotted), make up the putative group at 8 ± 1 Mpc. The circled points in the upper two panels represent galaxies that have minimum projected separations from AGC 198691 of less than 1.0 Mpc.

Zoom In Zoom Out Reset image size

The uncertainty in the distance to AGC 198691 makes it difficult to place this system into the context of its local environment with any certainty. Nonetheless, it would appear that this galaxy is probably located in a fairly low-density region of the nearby universe. If AGC 198691 is in fact located within the Leo Spur it could still be fairly isolated unless it were in close proximity to one of the galaxies mentioned above. As seen in Figure 6, if AGC 198691 is located at a distance of 11 Mpc or beyond then it is extremely isolated (no galaxies within 1 Mpc). We note that there have been suggestions that low-metallicity galaxies are preferentially found in voids (e.g., Pustilnik et al. 2006, 2011). If AGC 198691 followed this trend then distances larger than 11 Mpc would be preferred. Regardless of its actual distance, it would appear that AGC 198691 (V_⊙ = 514 km s⁻¹) possesses a large negative peculiar velocity and is participating in the same velocity flow as the other galaxies in this region of the local universe (Karachentsev et al. 2015).

In this section we look at the physical characteristics of AGC 198691. Since the distance is only poorly constrained at this time, we present distance-dependent quantities derived for a range of distances that should bracket the possible values.

Table 4 presents relevant observed and derived quantities. The location of AGC 198691 on the sky places it within the boundaries of the constellation Leo Minor. For that reason, our group has been referring to it as the Leoncino Dwarf. The BVR magnitudes and colors are obtained from the SDSS photometry for this galaxy (Ahn et al. 2012), using the unpublished conversion relations of R. Lupton.²¹ Galactic absorption corrections from Schlafly & Finkbeiner (2011) have been applied for the computation of absolute magnitudes and luminosities. The uncertainties listed include the formal photometric errors from SDSS as well as the scatter in the regression relations. The Hα flux is measured from our WIYN 0.9 m narrowband images, which were reduced and calibrated following the procedures described in Van Sistine et al. (2016). The angular size of the galaxy was measured from our HST images and takes into account faint, extended low surface brightness emission that lies beyond the central core containing the bright young stars and the H ii region. The heliocentric velocity (V_⊙), H i profile velocity width at 50% of the peak (W₅₀), and H i flux density (S_{H i}) are all from the ALFALFA survey measurements of this source.

Table 4.  AGC 198691—Observed and Derived Properties

Observed Quantities		Value
R.A. (J2000)		9:43:32.4
decl. (J2000)		+33:26:58
mV		19.53 ± 0.03
B–V		0.29 ± 0.04
V–R		0.14 ± 0.04
Angular Diameter (arcsec)		8.1 ± 0.2
F(Hα) (erg s−1 cm−2)		(8.49 ± 0.38)$\;\times \;{10}^{-15}$
V⊙ (km s−1)		514 ± 2
W50 (km s−1)		33 ± 2
SH i (Jy km s−1)		0.53 ± 0.04
Derived Quantities
Distance (Assumed) (Mpc)	8.0	12.0	16.0
MV	−10.03	−10.91	−11.53
MB	−9.75	−10.63	−11.25
Diameter (pc)	320	480	640
L(Hα) (erg s−1)	$0.64\times {10}^{38}$	1.44 $\;\times \;{10}^{38}$	$2.56\times {10}^{38}$
SFR (M⊙ yr−1)	0.00051	0.00114	0.00202
H i Mass (M⊙)	$0.80\times {10}^{7}$	1.80 $\times {10}^{7}$	$3.20\times {10}^{7}$
Stellar Mass (M⊙)	1.6 $\;\times \;{10}^{5}$	3.6 $\;\times \;{10}^{5}$	6.4 $\;\times \;{10}^{5}$

Download table as:  ASCIITypeset image

We present derived quantities in the lower section of Table 4 for three distances: 8, 12, and 16 Mpc. We do not quote uncertainties in these parameters since they are dominated by the uncertainty in the distance. The star-formation rate (SFR) is computed from the Hα luminosity using the standard Kennicutt (1998) conversion factor. The stellar mass estimates are derived using SED fitting methods as described in Janowiecki (2015) and S. Janowiecki et al. (2016, in preparation). These values should be taken as only being representative. Unfortunately, there are currently no NIR flux points for this galaxy that would allow for a more robust stellar mass estimate via the SED fitting method. AGC 198691 was not detected in either the 2MASS or WISE surveys (e.g., Skrutskie et al. 2006; Wright et al. 2010). Furthermore, the uncertainties due to the presence of strong emission lines in the broadband filter bandpasses complicate the determination of the stellar mass via this method. It is worth noting that the value for M_* presented in Table 4 is consistent with estimates obtained using a stellar mass-to-light ratio method (e.g., Bell & de Jong 2001).

The color and luminosity of AGC 198691 reveal it to be an actively star-forming dwarf galaxy. Furthermore, its small size and high surface brightness appearance (Figure 2) allow us to classify it as a BCD (see Gil de Paz et al. 2003). The blue absolute magnitude range of −9.75 to −11.25 makes it one of the lowest-luminosity examples of a BCD known. Most previously identified examples of this class of object possess B-band absolute magnitudes in the range of −13 to −17 (e.g., Janowiecki & Salzer 2014). The SFR, while low in an absolute sense, is quite high for a galaxy with this luminosity (e.g., Lee et al. 2009). This is again consistent with our proposed classification as a BCD.

One of the parameters that stands out with AGC 198691 is its very large H i to stellar mass ratio. The ratio implied by the values in Table 4 is M_{H i}/M_* = 50, which makes it one of the most gas-rich dwarf galaxies known. It is worth stressing that this parameter is distance independent; the extreme value we derive will not change as the distance to this source is refined. For comparison, the M_{H i}/M_* ratio for Leo P is 2.6 (Rhode et al. 2013). Similarly, all of the SHIELD galaxies from the Cannon et al. (2011) sample have M_{H i}/M_* ratios less than two, and most are less than unity (McQuinn et al. 2015a). On the other hand, the M_{H i}/M_* value for AGC 198691 is similar to those for the extreme star-forming galaxies I Zw 18 and SBS 0335–052E (M_{H i}/M_* ≈ 75 for both; Hunt et al. 2014). While the M_{H i}/M_* ratio in AGC 198691 is lower than those of some rare massive systems revealed by ALFALFA (e.g., Janowiecki et al. 2015), it is reasonable to hypothesize that the large H i content of this system may very well play a role in its extreme metallicity.

The location of AGC 198691 in a L–Z diagram in relation to other XMDs provides a powerful tool for understanding the physics at work with this star-forming system. Figure 7 shows a set of nearby galaxies with direct abundances taken from the study of Berg et al. (2012; gray points, her "Combined Select" sample) plus the five XMD galaxies with 12+log(O/H) < 7.20. AGC 198691 is plotted assuming distances of 8, 12 and 16 Mpc. The solid line is a linear fit to the data points from the Berg et al. study (her Equation (9)).

Zoom In Zoom Out Reset image size

The exceptionally metal-poor systems SBS 0335–052W, I Zw 18, and DDO 68 all lie substantially below the fit to the rest of the data. As discussed by Ekta & Chengalur (2010) this may imply that the star-formation mechanism at work in these systems is atypical, being affected by the infall of significant amounts of pristine gas. The infall could both drive the observed star-formation event as well as dilute the gas resulting in under-enrichment relative to their luminosities. In contrast, the locations of Leo P and AGC 198691 lie far closer to the extrapolation of the fit line defined by the galaxies of the Berg et al. sample. This might be interpreted as an indication that these two systems are forming stars in a manner more similar to expectations for typical dwarf irregular galaxies. The fact that AGC 198691 does exhibit a modest offset from the extrapolated fit line as well as from the relatively quiescent Leo P could be due to its BCD-like nature (i.e., a modest enhancement in the B-band luminosity moves it to the left in the diagram). Alternatively, the gas-rich nature of this object might be an indication that it too has experienced recent gas infall that has reduced its observed metallicity. In either case, the impact of these putative evolutionary mechanisms is clearly far less dramatic than whatever has impacted the three more luminous XMD galaxies shown in Figure 7. The determination of an accurate distance to AGC 198691 would provide valuable insight into the possible mechanisms at play, since it would allow for a much better understanding of the environment within which this system resides.

As mentioned in Section 1 of this paper, there have been significant efforts made to detect XMD galaxies through the years. A large fraction of these efforts focused on discovering and observing star-forming dwarf galaxies with the strongest levels of activity (e.g., I Zw 18 and SBS 0335–052W). If the premise of Ekta & Chengalur (2010) that these XMDs are strongly affected by infall of pristine gas is correct, we can now appreciate why these objects are so rare. Existing surveys like KISS could not have failed to discover additional examples of objects like I Zw 18 and SBS 0335–052W if they were present in the nearby universe with even modest volume densities. The implication is that the physical processes that shape galaxies like I Zw 18 and SBS 0335–052W occur only rarely. It is also likely that the strong starbursts that are generated by the proposed gas infall produce sufficient amounts of new metals that the observed abundances increase back to "normal" levels fairly quickly.

It is worth noting that AGC 198691 is located in the area of the sky covered by the third installment of the Case Low-dispersion Northern Sky Survey (Pesch & Sanduleak 1986). The Case survey utilized objective-prism spectra to select galaxies that had detectable emission lines (usually [O iii]), blue continua, or both. Most of the objects cataloged in the Case survey were brighter than m_B = 18, suggesting that AGC 198691 (m_B = 19.8) was simply too faint to be detected by the survey.

A second method for searching for XMDs has been to obtain spectra of H ii regions in nearby gas-rich dwarf irregular galaxies (e.g., Skillman et al. 1988, 1989a, 1989b, 2003; van Zee & Haynes 2006; Berg et al. 2012; James et al. 2015). The existence of the L–Z relation implies that the observation of H ii regions in the lowest-luminosity galaxies should result in the discovery of numerous XMDs. However, the shortcomings of this approach are readily evident in Figure 7. The L–Z relation shown (solid line) only reaches 12+log(O/H) of less than 7.20 for luminosities fainter than M_B ∼ −9. Galaxies with this level of luminosity have only been discovered in or very near the Local Group. Furthermore, their low masses imply that even if they have retained an appreciable gas reservoir, the probability of catching them with current star formation is probably fairly low. The obvious example here is Leo P, which falls very close to the regression line in the L–Z diagram. Leo P contains a single H ii region that is ionized by a single O star (Rhode et al. 2013; McQuinn et al. 2015b). If Leo P were observed in 5–10 million years would it possess an H ii region that would allow a derivation of its metal abundance?

This second method of trying to detect XMDs can be effective but is mainly limited by our ability to catalog large numbers of low-luminosity, low-mass galaxies in the nearby universe. The work to date has been extremely effective at filling in the L–Z diagram at luminosities brighter than M_B ∼ −13, but has resulted in the measurement of abundances that are typically higher than 12+log(O/H) = 7.5.

The fact that the two most recently discovered XMDs, Leo P and AGC 198691, were both discovered in the ALFALFA blind H i survey immediately suggests a third approach for searching for low-metallicity galaxies. The extreme sensitivity of the ALFALFA survey has allowed for the detection of hundreds of low H i mass galaxies beyond the Local Group (e.g., Martin et al. 2010). Naturally, any dwarf galaxy contained within the ALFALFA catalog will have the raw materials necessary for forming stars (i.e., cold gas). The SHIELD project (Cannon et al. 2011) was initiated to study in detail ALFALFA detections with the lowest H i masses that also possess a clear stellar population. In the original SHIELD sample, 9 of the 12 galaxies are fainter than M_B = −13, which puts them into the regime of being interesting targets for possible XMDs.

A first attempt to derive metal abundances for the initial sample of SHIELD galaxies is described in Haurberg et al. (2015). Only 8 of the 12 galaxies in Cannon et al. (2011) had Hα fluxes strong enough to make them viable spectroscopic targets. Of these, two yielded direct measurement abundances, while for the remaining six strong-line methods were used to estimate O/H. One SHIELD galaxy was found to have an abundance of 12+log(O/H) = 7.41, while the remaining seven had 12+log(O/H) between 7.67 and 8.08. In other words, the abundances measured for the original SHIELD galaxies are similar to those measured for other nearby gas-rich dwarf irregulars. AGC 198691 was observed with KOSMOS on the KPNO 4 m as part of the second round of SHIELD galaxy spectroscopy. The same observing run has yielded several additional metallicity estimates, including two with 12+log(O/H) between 7.3 and 7.5 (A. Hirschauer et al. 2016, in preparation). Many additional low H i mass targets remain to be observed. The use of the ALFALFA survey to pre-select low-luminosity dwarf galaxies as targets for follow-up spectroscopy is a remarkably fruitful approach for expanding the number of known XMD galaxies.