ALFALFA DISCOVERY OF THE NEARBY GAS-RICH DWARF GALAXY LEO P. II. OPTICAL IMAGING OBSERVATIONS

As mentioned in the previous section, the CMD we obtained for Leo P is rather unusual, especially when compared with the CMDs of other nearby dwarf galaxies. In particular, the upper MS is much better defined than the RGB. The latter is the dominant feature in most LG dwarf galaxy CMDs (e.g., Tolstoy et al. 1998; Bellazzini et al. 2004; Irwin et al. 2007; Sand et al. 2009). The TRGB method relies on the detection of a fully populated giant branch to allow for the accurate assessment of the brightness of the RGB tip. Our WIYN observations of Leo P appear to detect the RGB in this system, but the photometric depth is not sufficient to provide a robust measurement of the TRGB. Hence, any TRGB distance we derive can only be taken as approximate.

For metal-poor stellar populations, the TRGB is located at (B − V)_o ∼ 1.3 (Ferraro et al. 1999; Stetson et al. 2005, 2011). The location of the star in Figure 6(a) with V_o = 22.66 and (B − V)_o = 1.29 (Star 7 in Table 1) is consistent with it being at or near the TRGB. Its location in Figure 6(b) ((V − R)_o = 0.74) further supports this interpretation. If we assume that Star 7 is in fact at the TRGB for Leo P, and using M_V(TRGB) = −2.5 for the most metal-poor RGB stars (Salaris & Girardi 2005), the inferred distance would be D = 1.1 Mpc. We note, however, that associating the TRGB with this particular star then requires that the RGB in this galaxy is underpopulated in a dramatic way, since the next star along the RGB lies nearly 0.7 mag below the tip. At a distance of 1.1 Mpc, Leo P would have an absolute magnitude of −8.3 and an approximate stellar mass of 1.5 × 10⁵ M_☉. It would seem highly unlikely that a stellar population with this mass could possess such an underpopulated RGB. No stellar systems with an old population, either globular clusters or dwarf galaxies, are known that are this massive and yet do not have a fully populated RGB (e.g., Okamoto et al. 2012).

A more likely possibility is that Star 7 is a foreground star or a red supergiant rather than a red giant in Leo P. While its location in our CMDs is precisely where one would expect a metal-poor RGB star, it is located very far above the next brightest RGB star (Star 11, with V_o = 23.33). We can test whether Star 7 is likely to be a foreground contaminant by using the data present in Figure 4(b). By counting the number of objects that appear in specific magnitude and color ranges of the CMD, then normalizing that number by the ratio of the area covered by the full comparison field (with Leo P removed) to the area covered by Leo P only, we arrive at an estimate of the number of contaminating objects that might appear in specific magnitude and color regions of the Leo P CMD. Applying this analysis to the area occupied by Star 7 leads to an estimate that one would expect to find a foreground star with a color and brightness comparable to Star 7 in a random field the size of Leo P about 14% of the time. This is large enough that one would not rule out the possibility of Star 7 being a foreground star. Alternatively, and perhaps more likely, Star 7 could be located in Leo P but be an asymptotic giant branch (AGB) star or possibly a red supergiant star from an intermediate age stellar population.

We note in passing that there are two other very red stars located in the Leo P field with (B − V)_o > 2.1 that we believe are foreground stars. These two stars appear in Figure 6(b) as having (V − R)_o > 1.1 (they are off the plot to the right in Figure 6(a)). Given the presence of these two likely foreground objects in the Leo P field, it seems at least conceivable that Star 7 may also fall in this category.

If we assume that Star 7 is not located at the TRGB, we focus next on the group of stars located below V_o = 23.3 and with (V − R)_o colors between 0.5 and 0.7 in Figure 6(b). The brightest of these is Star 11 in Table 1, with V_o = 23.33, (V − R)_o = 0.54, and (B − V)_o = 0.95. Associating this star with the tip of the RGB leads to an implied distance of 1.5 Mpc. We note that the (B − V) colors of Star 11 and the several stars just below it in Figure 6(a) appear to be too blue for them to located on the RGB if the distance were closer to 1.5 Mpc. However, in Figure 6(b), where the RGB shows better definition due to the smaller photometric errors of the red stars in (V − R), these same stars appear to be consistent with an RGB with a distance anywhere between 1.5 and 2.0 Mpc.

One should ask at this point whether or not the putative upper RGB is heavily contaminated by foreground stars in our data. Following the same analysis method described above, we determined the average surface density of stars located in the color and brightness regime occupied by the top of the RGB in Leo P. For the magnitude range V_o = 23.25 to 24.0 and the (B − V)_o color range 0.75 to 1.25 (the TRGB portion of the CMD), there are 0.71 stars per Leo P field in our WIYN images. The Leo P CMD has 11 objects in this same region, so the estimated contamination rate for the upper RGB region of the Leo P CMD is approximately 6%. Hence it would appear that one can trust that the current data are not too badly contaminated by foreground objects in the TRGB region of the CMD.

In order to derive a valid TRGB distance, we would require data that are roughly 1 mag deeper than our current photometry. The definition of the RGB present in our current CMD is simply not sufficient to be able to run any sophisticated edge-finding software typically used for ascertaining an accurate TRGB distance. Our estimates above are based on single stars which themselves might be red supergiants or AGB stars. Therefore, it appears that the most robust distance estimate that we can derive for Leo P based on the RGB in our CMDs is 1.5–2.0 Mpc. The lower end of this range is established by the location of the brightest star in Figure 6 that can reasonably be associated with the RGB. The upper end is somewhat softer, but is based on the location of the isochrones in both Figures 6(a) and 6(b) in color space. The vast majority of the putative RGB stars in the two CMDs are located blueward of the 2.0 Mpc isochrone. Since the isochrones we are using in these figures are matched to the measured metal abundance of Leo P (Skillman et al. 2013), and since differential reddening in the galaxy will only act to push stars to redder colors in the CMDs, we infer that the 2.0 Mpc isochrone indicates a reasonable upper limit for the distance to Leo P.

Recent ground-based imaging observations (K. McQuinn 2013, private communication) have been obtained that reach substantially deeper than our WIYN images. The RGB is definitively detected in these images at the location indicated by our data, and a very preliminary analysis of these new data produce a distance determination that is consistent with our range quoted above. A thorough analysis and presentation of these new data will be forthcoming.

Due to the lack of a definitive distance determination from the TRGB method, we explored an alternative approach to estimating the distance. As we show below, it is possible to take advantage of the single H ii region present in Leo P to constrain the distance to the galaxy. In the case of this system, nature has provided us with a rather unusual circumstance in which (1) there is one and only one H ii region in the galaxy, and (2) the resolved stellar population of luminous blue stars is well delineated by our current observations.

Stars with effective temperatures cooler than ∼20,000 K, corresponding to a spectral type of B1 or B2, are not capable of producing a significant volume of ionized gas (Osterbrock 1989). The brightest star in Leo P (Star 1) is surrounded by a H ii region, while the second brightest (second star in Table 1, hereafter Star 2) is not. Our narrowband images are sensitive enough to rule out any significant emission around any of the other bright blue stars in Leo P. For example, if Leo P were located at a distance of 1.5 Mpc (e.g., consistent with the TRGB distance found in Section 4.1.1), we would be able to detect a H ii region with an Hα luminosity of only 1.3 × 10³⁴ erg s⁻¹, which is 0.1% the Hα luminosity of the Orion Nebula (O'Dell et al. 1999). Hence, we can rule out with a high degree of confidence that there is any significant Hα emission associated with any of the blue stars in Leo P other than Star 1.

We can use the information in the paragraph above to constrain the distance to Leo P. Since it has a H ii region around it, Star 1 (V_o = 21.00) must have a spectral type of B2 or hotter, while Star 2 (V_o = 22.05) must be B3 or cooler. The apparent magnitude difference between these two stars (ΔV_o = 1.05 mag) further constrains their possible spectral types and luminosities. For example, assuming that Star 2 is a B3V star (M_V = −2.00), the observed value of ΔV_o would imply that Star 1 is a B0V or O9.5V star (M_V = −2.90 to −3.05; all absolute magnitudes for O and B stars are taken from Wegner 2000, using the "smoothed" values). The implied distance to Leo P in this scenario is then D = 650 kpc. The coolest spectral class that Star 1 could have and still possess a H ii region is B2V (M_V = −2.30). The observed value of ΔV would then lead to Star 2 having an absolute magnitude of M_V = −1.25 and a spectral type of B5V or B6V. In this case the implied distance is D = 460 kpc).

The distances derived above, 460 to 650 kpc, are substantially lower than what we found based on the TRGB method in the previous section. Examination of Figures 6(a) and 6(b) appear to rule out distances as small as ∼500 kpc with high confidence. Why is our distance estimate based on the luminous blue stars so incongruous with the TRGB distance?

Three key assumptions were made in the preceding calculations: (1) the bright blue stars are single stars, (2) they are all on the MS, and (3) there is gas in the immediate vicinity of these stars that could be ionized if the surface temperatures of the stars were hot enough. In reality, all three of these assumptions could be incorrect. For example, the stars may not be single stars. It is not uncommon for massive stars to be formed in binary or multiple-star systems (e.g., all four of the Trapezium stars in the Orion Nebula Cluster are multiple-star systems). If we assume that Stars 1 and 2 are both in equal-mass binaries, the inferred distance range would increase to 650–910 kpc.

The second assumption—that all the bright blue stars are on the MS—might also be erroneous. Recall that while Star 1 is located close to the 10 Myr isochrones in Figure 5, the five blue stars below it (with V_o between 22.0 and 22.6), including Star 2, are all located to the right of the MS isochrones. These five stars may well be post-MS stars. If this is true, then they may be significantly older than Star 1. Under the assumption that Star 2 is a B3 giant (B3III) rather than a MS star, the inferred distance range for Leo P increases to 590–850 kpc if Stars 1 and 2 are both single, and to 830 kpc–1.20 Mpc if they are equal-mass binaries. Alternatively, if Star 2 were a B3 bright giant (B3II), the corresponding distance range is 780 kpc–1.15 Mpc for single stars, and 1.10–1.62 Mpc for equal-mass binaries. Hence, by simply invoking the assumption that the bright blue stars are evolved stars rather than MS stars, one can derive distances that are consistent with the TRGB distance. Unfortunately, the precise nature of these bright blue stars is unknown, meaning that this method is subject to unreasonably large uncertainties.

Finally, we point out that this distance estimation method relies on the assumption that there is a significant amount of gas in close proximity to the bright blue stars that is capable of being ionized. While this is clearly the case for Star 1, it may not be a valid assumption for the other blue stars, like Star 2. In particular, if the five stars located between V_o = 22.0–22.6 with blue colors are all post-MS stars, as suggested in Section 3 and inferred above, then it would seem likely that they have pushed away the gas associated with their formation episode via stellar winds. It is even possible that they are located in small associations that once included higher-mass stars which ended their lives as supernovae and therefore cleared out the interstellar medium (ISM) in the immediate vicinity of the groups.

It seems clear that we do not know enough about these individual stars to be able to generate a reliable distance from them using this method. The fact that the luminosity classes of these stars are unknown, combined with the possibility of binarity or multiplicity, yields distances that range over a factor of nearly four (460 kpc to 1.62 Mpc). It is nevertheless reassuring that the range covered by these possible distances overlaps the distance range inferred from the TRGB method.

We summarize the current status of the distance estimate for Leo P as being uncertain. We adopt a distance of 1.5 to 2.0 Mpc, based on the apparent location of the tip of the RGB in our CMDs. Our attempt to use the bright blue stars that have robust photometry to constrain the distance leads to a distance range that is unacceptably large, although it is consistent with our TRGB estimate under the assumption that at least some of the blue stars present in the galaxy are post-MS objects. We conclude that an unequivocal distance determination is not possible with the current data, and that deeper high-resolution imaging data are required to arrive at a more precise value of the distance. Such data have been obtained recently (K. McQuinn 2013, private communication), and appear to corroborate our distance estimate.

It is appropriate to note that Giovanelli et al. (2013) used a completely independent method to estimate the distance to Leo P. Using the baryonic Tully–Fisher relation (e.g., McGaugh 2012) and the observed rotation velocity of the galaxy derived from their H i data, Giovanelli et al. derive a distance to Leo P of 1.3 (+0.9, −0.5) Mpc. This value is consistent with our distance estimates derived using the TRGB method.