Carbon- and Oxygen-rich Asymptotic Giant Branch (AGB) Stars in the Bulge Asymmetries and Dynamical Evolution (BAaDE) Survey

The sequence of regions in the IRAS color–color diagram presented by van der Veen and Habing (1988, their Figure 2) is commonly thought of as a sequence of increasing CSE optical-depth, and the regions laid out in that work are often used as the basis of IR color source selection (e.g., see te Lintel Hekkert et al. 1991). The sources in the BAaDE survey were selected to give a high chance of SiO maser activity, corresponding to section IIIa of the van der Veen and Habing IRAS color–color diagram, which is the region inhabited by variable AGB stars with dusty circumstellar shells and the 9.7 μm silicate feature in emission (van der Veen & Habing 1988). Unfortunately, the IRAS survey is highly confused in the crowded area of the Galactic plane, so the actual selection criteria used for BAaDE were determined by mapping sections of the IRAS color–color diagram to MSX colors (SCC09). In the mapping, each IRAS two-color region (designated by uppercase roman numerals) is assigned an MSX counterpart (designated by lowercase roman numerals). MSX has a higher angular resolution (18'') than IRAS (∼1'), allowing for a color-based target selection in the Galactic plane. BAaDE sources were chosen from the MSX Point Source Catalog version 2.3 (Egan et al. 2003) based on their MSX [A] − [D], [A] − [E], or [C] − [E] color⁷ to belong to SCC09 region iiia using the following criteria for uncorrected MSX magnitudes: −0.6 < [A] – [D] < 0.4, −0.7 < [A] – [E] < 0.4, or −0.75 < [C] – [E] < 0.1. A vast majority of sources (96.7%) were selected by their [A] − [D]value. However, as shown by SCC09, the relatively narrow range of wavelengths covered by MSX (8.28−21.34 μm) prevented a direct mapping from the IRAS color–color diagram. C-AGBs are preferentially identified using the longer 60 μm IRAS wavelengths and, thus, cannot easily be separated from O-rich AGB stars in the MSX data. The BAaDE sample has, therefore, not actively deselected C-AGBs, even though we do not expect SiO detections in these sources (SCC09).

All BAaDE sources have been cross matched with the Two Micron All Sky Survey (2MASS), AKARI, Wide-field Infrared Survey Explorer (WISE), and MIPSGAL IR surveys, with the most matches present in 2MASS (27,050 2MASS matches, 24,821 AKARI matches, 8925 WISE matches, and 12,875 MIPSGAL matches out of the 28,062 total BAaDE sources). The cross matching is discussed in Hilburn (2018) and follows the general scheme of searching within a specific radius, which is dependent on the positional accuracies of the surveys, and assigning the reddest source as a match in the case of multiple source matches. The cross matching for the full sample of 28,026 stars is not completely finalized, because of, for example, insufficient reddening corrections, but at this time, we estimate that less than 3% of the cross-matches between MSX and 2MASS may be incorrect counterparts. For most sources in the BAaDE sample, there, thus, exists a set of cataloged IR magnitudes in the wavelength range 1.2–24 μm, as well as known maser line fluxes and ratios. The work presented here relies on MSX data as well as 2MASS [K_s] magnitudes (centered at 2.2 μm) and AKARI [18] magnitudes (centered at 18 μm) that were obtained through this cross-matching system. In order to compare these magnitudes across studies, zero-magnitude flux densities of 55.810, 26.423, 18.267, 8.666, 665, and 12.001 Jy are applied to MSX [A], [C], [D], [E], 2MASS [K_s], and AKARI [18] magnitudes, respectively (SCC09; Cohen et al. 2003; Tanabé et al. 2008). No other corrections (i.e., reddening corrections) were applied to any of the IR magnitudes or colors used in this work.

The initial IR-selection for the BAaDE survey contained color ranges for luminous sources embedded in dusty material where AGB stars, but also YSOs, can be found. Here, we show that the YSO population in our sample is easily identified by its MSX [D] − [E] color, with "DE-red" sources ([D] − [E] > 1.35) comprising the YSOs. This population differentiates itself from the "DE-blue" sample in several ways: (1) YSOs occupy a distinct space in various IR color–color diagrams (Section 3.1.1), (2) they are spatially clumped in and around molecular clouds in the plane (Section 3.1.2), (3) their spectra show a lack of SiO detections and a high rate of CS detections (Section 3.1.3), and (4) when imaged, the CS emission is often extended rather than point-like (Section 3.1.4; Stroh et al. 2019). It should be noted that although the positions of the DE-red sources in color–color diagrams and in the Galactic plane lead us to identify the population as dominated by YSOs, some sources may be post-AGB sources, planetary nebulae, red super-giants,⁸ or H ii regions, as there is overlap between these types of sources in IR color–color diagrams and not all of our DE-red sources are associated with clouds. The term "YSO population" as used in the rest of this paper refers to the DE-red population even though we acknowledge that there may be other types of objects among the DE-red sources.

3.1.1. IR Color–Color Diagrams

Lumsden et al. (2002) use MSX [A]−[D] versus [D]−[E] diagrams to separate a sequence of AGB stars from YSOs. In our sample, 6656 sources (45%) have reliable [E]-magnitudes and BAaDE spectra and so can be plotted in a diagram (Figure 1) corresponding to that of Lumsden et al. (2002, their Figure 10). Although the Lumsden et al. (2002) sample and our BAaDE sample were selected very differently, there are some common aspects to their distributions in this color–color diagram. Two populations can be seen in our data, separated at [D]–[E] = 1.35. This value was determined by fitting a pair of Gaussians to the [D]−[E] values of all of the BAaDE sources with [E]-magnitudes, and determining where the Gaussian curves intersect.⁹ Comparison of our sample with that of Lumsden et al. (2002; first panel of their Figures 3–5) indicates that "DE-blue" sources with [D]–[E] < 1.35 are primarily O-rich AGB stars, while DE-red sources are primarily YSOs.

Zoom In Zoom Out Reset image size

3.1.2. Spatial Distribution

Compared to the AGB sample, the DE-red sources are distributed in a much narrower latitude range close to the Galactic plane (Figure 2). DE-red sources are also clumped in and around molecular clouds. An illustrative example of this is shown in Figure 3, where, by placing DE-red and DE-blue sources on their corresponding Digital Sky Survey backdrop, it can be seen that the DE-red sources show a clear tendency to clump toward the molecular clouds while the DE-blue are more uniformly distributed. Both of these patterns in the distribution of DE-red objects suggest that these objects were likely formed in molecular clouds in the plane and have not had time to disperse. Therefore, both the narrow latitude range and the spatial clustering coinciding with molecular clouds are in agreement with the classification of DE-red sources as primarily YSOs.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.1.3. Detection Rate

3.1.4. CS Emission Distribution

With DE-red sources (mostly YSOs) removed from the sample, we can discuss the AGB population in the survey with very little contamination from non-AGB objects. SiO detections in conjunction with IR colors can also be used to separate C-rich from O-rich AGB stars. SCC09 found that one cannot divide O- and C-rich stars based solely on MSX colors because of the narrow total wavelength range covered by MSX. Our sample shows similar results with no clear division on MSX color–color diagrams (see Section 4.2). Therefore, following Lumsden et al. (2002), we utilize both 2MASS and MSX colors to separate O- and C-rich AGB stars. We find, DE-red sources excluded, that there is a clear division between a population with a high SiO detection rate (∼73%) and one with a low rate (∼12%) in the $[{K}_{s}]-[A]$ versus [A] − [E] color–color diagram (Figure 4). These populations can be interpreted to represent O-rich and C-rich AGB stars, respectively. This division scheme closely matches the populations found in the same color–color space as in Lumsden et al. (2002; first panel of their Figures 9–11), with candidate C-rich AGB stars occupying a space that is redder in $[{K}_{s}]-[A]$ color and bluer in [A] − [E] color than the O-rich AGB. The YSO population discussed in Section 3.1, has not been removed from Figure 4 to show that the YSOs occupy the reddest (in both colors) section of the diagram and are easy to visually separate from both C and O-AGBs, which also agrees with Lumsden et al. (2002). Empirically, we define

$$\begin{eqnarray}&&[{K}_{s}]-[A]=4.5([A]-[E])-1.2\end{eqnarray} \tag{ 1 }$$

as the division between the populations. Based on this division, the C-rich region is characterized more by a low detection rate of SiO than a high rate of any C-bearing molecule (see Figure 5). This is due mostly to the fact that VLA sources make up 90% of the sources examined in this work, and that there are no certain detections of any molecule other than SiO in any VLA spectrum. We also lack IR spectroscopy, including the Si- or C-containing dust bands often used to identify C- versus O- rich stars. We do not assume that an SiO non-detection in the survey is solely indicative of a non-O-AGB source, as stellar variability, the sensitivity of our observations, and the fact that some O-AGB stars may not have CSE conditions conducive to maser activity all play a role.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

3.2.1. Candidate C-rich AGB Stars

Most CS-detected objects lie either with the C-AGB or YSO populations, which is in keeping with the scheme that O-rich objects do not have many C atoms outside of CO molecules with which to form C-rich molecules like CS. Similarly SiO-detected sources lie almost exclusively in the O-rich AGB section. The exceptions (CS detections in the O-AGB region and SiO detections outside of that region) are listed in Tables 1–3 with brief suggested explanations for each. It is not unexpected that our IR cuts sometimes show sources bleeding over into the wrong color region because (a) our IR colors have not been reddening corrected, and (b) the 2MASS and MSX data for these variable objects were not taken simultaneously leading to small uncertainties in any calculated color value. Any photometric errors in the 2MASS or MSX catalogs would also add to magnitude and color uncertainties. More detailed explanations follow.

3.3.1. CS Detections in the O-AGB Color Region

3.3.2. SiO Maser Detections in the C-AGB Color Region

3.3.3. SiO Maser Detections in the YSO Region

[E]-magnitude information is critical in identifying the YSO population. The AGB and YSO populations are not separable in MSX colors that do not utilize [E]-magnitude, including [A] − [D] (the primary color used for BAaDE selection), which is why these sources were not deselected initially. Relying more heavily on selection criteria from SCC09 involving the [E]-magnitude (picking a greater percentage of sources based on [A]−[E] or [C]−[E] as opposed to [A]−[D] colors) could have excluded most of these sources.

Unfortunately, the [E]-band sensitivity is lower than that of the other MSX bands ([A]-band is 10 times more sensitive), and only ∼45% (∼6000) of our sample have reliable (quality 3 and higher) [E]-magnitudes. This means that even though the distinction between AGB stars and other sources is clear, we can only definitively separate these two groups for about half of our sample.

4.1.1. Alternatives to [E] and the 18 micron Feature

Because many BAaDE sources lack reliable [E]-magnitudes (centered at 21 μm), we here explore IR-band options from other surveys that may separate YSO and AGB sources. The cross matching between these surveys is discussed in Section 2.1. Both [WISE4] and MIPSGAL [24] (centered at 22 and 24 microns, respectively) are similar in center wavelength and bandwidth to MSX [E]; however, when applied to our [D]−[E] selection, neither [D]−[WISE4] nor [D]−[24] show significant separation of a red-colored low-detection-rate population from the bulk of the sources. In contrast, although AKARI [18] (centered at 18 μm) is much wider than [E] and is centered at a shorter wavelength, [D]−[18] does show a significant red-colored low-detection-rate population that closely matches the MSX-defined DE-red population (see top panel of Figure 6). Although [WISE4], MIPSGAL [24], and [E] are all similar, [E] and AKARI [18] are the only two of the four bands that cover the 18 μm silicate feature; the other two filters fall off shortward of around 20 μm. This indicates that the 18μm silicate feature likely determines the DE-red/DE-blue and YSO/AGB division.¹⁰

Zoom In Zoom Out Reset image size

As only 47% of the complete BAaDE sample has reliable [E]-magnitudes, using AKARI [18] as well nearly doubles the number of sources in our sample that can be classified. However, although a significant difference between red and blue populations can be seen, the actual division at a specific color value is less evident in [D]−[18] than in the original DE separation, and therefore, drawing a line to separate the two is more straightforward in [D]−[E]. Additionally, sources that have [18]-magnitudes but no [E]-magnitudes are strongly preferentially [D]−[18] blue and so appear to be AGB stars based on [D]−[18] color (see Figure 6). Separating our sample by sources with or without reliable [E] data, we find [18]-magnitudes are typically clearly 1–2 magnitudes fainter for the sources without [E] (in the range 1 < [18] < 4.5 versus −1 < [18] < 3.5 mag for sources with [E]). It seems that most sources that are not detected in [E] are not detected because their [E]-magnitude emission is weak compared to their [D]-magnitude emission, and that these sources are, therefore, not likely to be YSOs as defined by our color-cut. We, therefore, define all DE-blue and all sources without [E]-magnitudes as likely AGB stars and only exclude DE-red sources from the AGB population.

4.1.2. Extension to the Entire MSX PSC

The DE-red population is easily identifiable in the entire MSX sample (Figure 7), not only in the BAaDE subset, and was seen as an offset population in an [A] − [D] versus [A] − [E] diagram by SCC09. SCC09 suggested that the offset might have been caused by an error in the [E]-band photometry, but we find that, because the SiO detection rate is strongly linked to whether a source is DE-red or not, the offset likely has a physical basis. As we do not have SiO/CS emission information of MSX sources outside of 0.66 < [A] – [D] < 1.66, we cannot confirm that all MSX sources with [D] – [E] > 1.35 are related to a YSO population. A study with a less limited [A] − [D] color range is needed to determine how the BAaDE DE-red population relates to the entire MSX sample.

Zoom In Zoom Out Reset image size

The O- and C-rich populations are not clearly separable in MSX colors alone. This was a major conclusion of SCC09 and can now be seen through the consistency of the SiO detection rate throughout the AGB sample on MSX color–color diagrams. For example, there is a lack of a clear division even in the [A] − [D] versus [D] − [E] diagram in Figure 1, despite the fact that this color–color diagram has shown a clear distinction between O- and C-stars in smaller samples in the literature (Lumsden et al. 2002; Ortiz et al. 2005; Suh & Kwon 2011). SCC09 define an "OC-index" in their Figure 5 as the distance from the OC-dividing line (shown as the diagonal line in Figure 1), which is a predictor of whether a source from the Ortiz et al. (2005) sample will be an O- or C-rich AGB. Negative indices indicate a more likely C-star, and positive indices indicate more likely O-rich AGB stars. SCC09 found that the OC-index is not a secure predictor of whether cross-matched MSX and IRAS sources appear in a C- or O-dominated region of the IRAS diagram, but that the index is more suited to divide stars based on their temperature. Indeed, in the SCC09 study, many IRAS-defined O-AGBs are found on the C-side of the OC-dividing line. Conversely, OC-index defined C-stars include most sources with [12] – [25] ≤ −1.5, not only those in regions VIb and VII (designated for C-stars), showing that the OC-index is more closely related to [12] – [25] color than to source chemistry (see SCC09 Section 3.2 for details).

In Figure 8, the SiO detection rate is relatively consistent across OC-index with only a small decrease toward the more likely C-rich stars, especially as compared to Figure 5. Therefore, in our sample, the OC-dividing line is ineffective; we find that the OC-index is not a reliable predictor of AGB chemistry within the BAaDE survey, which is in agreement with SCC09.

Zoom In Zoom Out Reset image size

Our primary explanation as to why this separation is so distinct in other studies but lacking in our sample lies in the way the samples were collected. The BAaDE sample is pseudo-blind when it comes to O/C-AGBs (although we may cut out proportionately more C-AGBs than O-AGBs with our [A]−[D] > 0.66 cut, see Section 4.4). The Lumsden et al. (2002), Ortiz et al. (2005), and Suh & Kwon (2011) samples were collected from the literature. Lumsden et al. (2002) uses a compilation from Alksnis et al. (2001) as their identified C-stars. These sources were mostly optically selected and are, therefore, likely biased toward thin-shell, bluer C-stars. For identified O-rich AGB stars, Lumsden et al. (2002) use the OH/IR stars from Chengalur et al. (1993), which biases their O-stars toward the thicker-shell redder OH/IR population. Ortiz et al. (2005) suffer similar biases; their carbon stars come from the Loup et al. (1993) sample that is biased toward strong CO emitters and includes mainly sources for which [12] – [25] < −0.3 (see Loup et al. Figure 3), and their O-stars are again primarily OH/IR sources. In short, these samples have selected thin-shell, blue-biased carbon sources, and thick-shell (OH/IR) red-biased oxygen sources, which has exaggerated the distinction between the two groups in MSX color–color space. Suh & Kwon (2011) include some SiO-detected O-AGBs in their sample, which lessens this bias by including more thin-shell O-AGBs (as the thinner shells are more conducive to SiO masers), and indeed, they do see cross-over of O-rich objects into the C regime, although their sample still separates reasonably well along the OC-dividing line. BAaDE focuses strongly on thin-shell O-AGBs, as this is where SiO emission is most common, and the sample shows almost no separation by the OC-index. As such, we conclude that the OC-index is not suitable for separating thin-shell O-AGBs from C-AGBs.

Within the BAaDE sample, a better division between O- and C-AGBs is found when including [K_s]-band measurements. Comparing Figures 4 and 1 shows the strong improvement that including [K_s]-band measurements makes on the dependability of color-based predictors of SiO maser activity and, therefore, AGB chemistry.

With a tightly defined region of our color–color diagram designated for O-AGBs from Figure 4 the effects of (a) sensitivity and variability, (b) CSE conditions, and (c) non-O-AGB objects in the O-AGB section on our detection rate can be better constrained.

Sources that host consistently weak masers throughout the stellar cycle, or masers that temporarily were very weak at the time of the BAaDE observations may be non-detections in the BAaDE survey. It is likely that such a population exists, which would be part of the O-AGB region non-detections. In the O-AGB region of Figure 4, the SiO detection rate is 73%. This rate depends on the MSX [A]-magnitude (Figure 9), with a peak rate around [A] ≈ 2.5. The sources with magnitudes of 4.6 and fainter should be treated separately; these weaker sources are found in specific, deeper raster scans performed by MSX in select regions of the plane (Egan et al. 2003). Thus, in some regions, the BAaDE sample contains a set of fainter sources in addition to the typical ones. A lower SiO detection rate in those regions may be expected due to sensitivity if most of these sources are fainter because they are farther from the Sun. We conclude that sensitivity negatively affects the detection rate of these fainter sources. Removing the fainter sources provides a uniformly estimated SiO detection rate of 75% for the O-rich AGB stars with [A] magnitudes brighter than 4.6 (the main part of our sample).

Zoom In Zoom Out Reset image size

To address O-rich sources that do not host SiO masers at any time, we rely on IR data again as IR data reflects the conditions in the CSE. If we exclude sources with [K_s] – [A] < 1.5 (which show a lower detection rate, see Figure 4) our detection rate jumps to 80%. This could mean that O-AGB region sources with [K_s] – [A] < 1.5 are O-AGBs with CSE conditions less favorable to maser conditions (perhaps too thin-shelled), while their [K_s] – [A] ≥ 1.5 counterparts are almost exclusively SiO maser sources that turn on and off as a function of stellar cycle.

For comparison, Stroh et al. (2018) report a 78% detection rate (66/86) when re-observing a sample of known, bright SiO maser sources from the BAaDE survey. This indicates that even in a sample of previously detected sources, a ∼78% detection rate is expected due to variability effects. Similarly, Pihlström et al. (2018) redetected 26 SiO masers out of a sample of 33 BAaDE sources that had previously shown emission, again showing that a detection rate slightly below 80% is to be expected. As such, we conclude that variability and weak masers below our sensitivity may be responsible for most of the non-detections reported in our O-AGB sample and that, at most, only a small percentage of non-detections can be attributed to sources that are contaminants in our color regions. In short, we postulate that nearly all BAaDE-selected O-AGB color-region sources are SiO emitters but that some are just in a part of their stellar phase where their masers are off or below our moderate sensitivity.

O-AGBs were the intended targets of the BAaDE survey, as their maser emission provides line-of-sight velocities with which to study Galactic dynamics. Of the 28,062 sources covered by the entire BAaDE survey, only 2335 (8%) do not fall into the O-AGB region based on their [D] − [E], [A] − [E], and $[{K}_{s}]-[A]$ colors. All other BAaDE sources either fall into this region (10,084 sources), or do not have reliable data for one or more of these colors (15,643 sources). Most of the sources that cannot be categorized based on these colors are missing [E]-magnitude information but can be categorized as likely AGB stars based on [D] – [18]. The original selection of sources for the survey was, therefore, very effective and the dynamics goal of BAaDE is not compromised by the small number of candidate C-AGBs and non-AGBs in the sample. At our current detection rate, we expect ∼14,000–18,000 SiO maser detections from the survey.

Our C/O division was constructed by examining BAaDE detections and non-detections but ultimately only relies on IR colors. We can, therefore, apply the C/O division to all of the sources in the BAaDE sample regardless of whether they have been observed and/or analyzed yet. Within the entire BAaDE sample, of the 13,120 sources that have both reliable [E] and [K_s]-magnitudes, we find that our color criteria would yield 697 C-AGBs, 10,084 O-AGBs, and 1638 YSO/non-AGB sources. This gives a fraction of C-AGBs of 6%, but this fraction is both approximate and only applicable to the color and latitude ranges of the BAaDE survey. First, our division between O- and C-AGBs is empirical and slight shifts of the division can alter this fraction by several percentage points. Second, a small percentage of non-detections in the O-AGB or YSO regions may belong to C-AGBs whose derived colors place them in the wrong region of our diagrams. Additionally, although we find that the OC-index is not a useful predictor of chemistry within our color selection, it is possible that outside of our color selection, specifically toward lower (bluer) [A] − [D] values, the fraction of C-AGBs could increase, as would be consistent with Ortiz et al. (2005). Therefore, our C/O fraction is not only approximate but also only representative of AGB stars satisfying 0.66 < [A] – [D] < 1.66, and other surveys with wider color-coverage are required to fully understand the effect introduced by the BAaDE color-selection. Using BAaDE sources to determine this ratio also introduces spatial effects; the BAaDE sample is representative of sources that lie close to the Plane, most of which are likely to lie in the Galactic bulge.

Regardless uncertainties and the limits of our sample, we can compare our fractional distributions among source types to several other surveys in the Galaxy in order to start to quantify the ratio of C-AGBs. Here, we highlight studies that use differing methods for chemical identification. Kwok et al. (1997) categorize 11,224 IRAS Low-Resolution Spectra (LRS) and compare many of their classifications to those in the literature. Based on the literature identifications collected in their work, they find that 13% (or 571 out of 4277) of AGB stars are C-rich. The identifications by Kwok et al. (1997) based on IRAS LRS show 715 C-type and 3931 combined A- and E-type spectra (where A- and E-type sources show the 9.7 μm silicate feature in absorption and emission, respectively—both indicative of O-rich AGB stars). Excluding F-type, featureless, spectra that have more ambiguous chemical types, this puts the total number of identified AGB stars at 4646 and the fraction of C-AGBs at 15% (715/4646). Le Bertre et al. (2003) find that using O- and C- identifications from the InfraRed Telescope in Space (IRTS) LRS yields 18% (126/689) C-AGBs and find that their biases should equally effect C- and O-AGBs. A study by Ishihara et al. (2011) that uses IR color cuts to define the chemical types of AGB stars (similar to this study) finds 33% (5537/16,953) of their sample to be C-rich but puts their completeness for C-AGBs at more than double their completeness for O-AGBs (64% and 29%, respectively) suggesting that the percentage of C-AGBs could be about two times too high. By these comparisons, our estimated percentage of C-AGBs is relativity low and may imply that the number of carbon stars increases at lower values of [A] − [D]. Physically, we may also expect our sample to yield a low number of C-AGBs because we select mainly for thin-shelled sources. This means that many of our sources may be too early in their evolution or too low-mass to be C-AGBs. Furthermore, the BAaDE ratio is dominated by sources in the plane (Figure 2) while some of the ratios listed are based on all-sky surveys or surveys that significantly cover areas away from the plane (IRAS and IRTS) and the C/O-AGB ratio may be lower in the metal-rich plane than at higher latitudes. Finally, BAaDE likely includes many Bulge sources that may also contribute to the low ratio of C-AGBs, as the Bulge is known to include mostly O-rich AGB stars. We find a 6% fraction of C-AGBs reasonable for our Bulge-dominated population.

Comparing the distributions in the Galactic plane of each of the populations as found in the entire BAaDE sample generally supports the identification of each population (Figure 10). As seen in Section 3.1, the BAaDE identified YSOs are found primarily close to the plane (Figure 2); their distribution is more clumped than the AGB stars and shows a tendency toward molecular clouds (Figure 3), all of which are consistent with YSOs.

Zoom In Zoom Out Reset image size

Multiple studies have shown that C-stars are more uniformly distributed across the plane than their O-rich counterparts, which are preferentially found near central longitudes (e.g., Noguchi et al. 2004; Ishihara et al. 2011 and references therein). Our sample also shows this trend. Figure 10 shows that BAaDE color-selected O-AGB sources show a strong tendency toward the center of the Galaxy while their C-counterparts span a large range of longitudes at a fairly constant number density. The dependence of the C/O ratio on longitude is likely a metallicity effect (Brewer et al. 1995).

4.5.1. C-AGBs in the Bulge