Massive Stars in the SDSS-IV/APOGEE2 Survey. III. New OB Stars in the Direction of the Sagittarius Spiral Arm

The O, B, and supergiant stars presented in this study were selected from the 2MASS Point Source Catalog (PSC; Skrutskie et al. 2006) applying an NIR color selection criteria adapted from the work of Roman-Lopes et al. (2016), with the H-band magnitudes of the point sources satisfying a slightly different magnitude range: (a) H-band magnitudes in the range of 8.0 < H < 12.15 with associated errors < 0.1 magnitude (only sources with AAA flags) and (b) NIR colors satisfying 0.3 < (J − K_S) < 1.65 and 1.5 < [(J − H)/(H − K_S)] < 2.1. About 330 sources are selected from the mentioned criteria. However, for a given sample, the number of targets that can be observed in a plate is limited not only by the available fibers (265 fibers assigned to science targets, excluding those used to perform telluric and sky corrections on the APOGEE data), but it is also dependent on the fiber collision limit, which, in turn, is set by the physical size of the optical fiber connector. In the case of the southern spectrograph, the effective fiber collision radius is about 56″. In order to observe point sources in high-density stellar fields, a solution is to dedicate more visits to a given field, each with a distinct plate design, allowing us to observe more sources whose relative distances are smaller than the effective collision radius. On the other hand, due to vignetting at the border of the plate’s field of view (FOV, about 1° radius for APOGEE2-S), only targets positioned at less than 095 from the plate center are observed. Finally, the central part of the plate’s FOV is not covered/used by fibers due to a circular gap at the center of the plate.

Because of the mentioned physical constrains, about 30% of the initial sample was discarded from our science target list. In order to provide additional targets for the remaining science fibers, we looked for sources in the catalog of Mohr-Smith et al. (2017; MS sample), keeping those with assigned effective temperatures in the 30,000–45,000 K range. Unfortunately, not all these sources could be accommodated because of the same physical limitations, and the remaining science fibers were then used to observe stars from the main APOGEE2-S disk program.

The coordinates of the selected targets were crossmatched with those of known O- and B-type stars found in the SIMBAD⁸ and VIZIER⁹ databases. Some of them were found to be already known O and B stars, but they were also observed as part of the project. In Figure 1, we show the relative positions of the massive stars identified in this work. Most of the targets are seen spread around the NGC3603 and NGC3576 cluster centers and, as expected, distributed mainly along the Galactic plane direction.

From the plates and epochs listed in Table 1, a total of 265 individual targets were observed during the three visits of our SDSS-IV/APOGEE2-S CNTAC Program CN2017B-11. From visual examination of the features present in the associate spectra, and the careful measuring of the equivalent widths (EW) and the FWHM of the respective hydrogen spectral lines, we found that from the 265 individual sources observed in our CNTAC/APOGEE2-S program, the massive star subsample is composed by 95 O-type stars, 38 B-type stars, 16 blue supergiants (BSGs), and 16 yellow supergiants (YSGs). A small portion of the associated APOGEE2-S spectra are presented in the next section. The entire sample of APOGEE2-S O-, B-, BSG- and YSG-stars are public available and can be accessed using the SDSS DR16 Science Archive Server (SAS).¹⁰ Besides the mentioned types, another 32 stars present emission line features typical of Wolf-Rayet and Be-type stars, as well as young stellar objects (YSOs). Those sources will be the subject of an upcoming paper. Additionally, the remaining 68 sources are compound by late-type stars of the F, G, K, and M types.

Figure 1 shows the relative positions (solid colored circles) of the O, B, BSG, and YSG types, with the colors indicating the distinct flavors as follows: O stars (cyan), B stars (light blue), BSGs (light green), and YSGs (gold). Their positions in the [(J − H)/(H − K_S)] versus (J − K_S) diagram are shown in Figure 2, with the same color coding. In this diagram, we also show the position of a red supergiant star (red dot) identified by us as part of the main APOGEE2-S disk sample (not treated in this work). Finally, Table 2 presents the list of O (#1–#95), B (#96–#133), BSG (#134–#149), and YSG (#150–#165) stars. Column (1) contains the internal ID, column 2 contains the APOGEE ID, column 3 contains the source ID as available in the literature, column 4 contains the 2MASS H-band magnitudes, column 5 contains the spectral type and luminosity class (if any) taken from the literature, and column 6 contains the reference from which the spectral types and luminosity classes were taken.

Zoom In Zoom Out Reset image size

3.2.1. Spectral Line Parameter Measurements

3.2.2. Luminosity Classes and Spectral Types of the New O- and B-type Stars

H-band spectra of O stars usually show He ii (7–12 and the 7–13) absorption lines, as well as relatively weak hydrogen absorption lines of the Bracket series (Figure 3). On the other hand, in the case of H-band spectra of early- to mid-B stars (Figure 4) of classes iv–v, the 7–12 and the 7–13 He ii lines vanish, with the hydrogen lines of the Brackett series being much stronger and deeper (EW[Br11+Br13] > 3.6 Å) than those seen in the spectra of the O stars. Roman-Lopes et al. (2018) showed that the EW of the Br11 and Br13 hydrogen lines, combined (when possible) with the corresponding values for the He ii (7–12 and the 7–13) line transitions, enable reliable spectral type estimation for mid- to late-O stars, as well as for early- to mid-B stars. The appropriate linear relations can be chosen based on the observed position of the source in an EW[Br11+Br13] versus mean FWHM[Br11–Br13] diagram, as shown in Figure 5. There the O-type stars in our sample (indicated by the solid cyan circles) that occupy a well-defined region on its bottom left side, while the B-type stars (class iv–v, represented by the solid blue triangles) are found spreading from the top middle to the right part of the diagram.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (10 panels)

Tarball of figure set images

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (5 panels)

Tarball of figure set images

In order to estimate the spectral types of the O- and B-type stars in our sample, the first step was to determine their most probable luminosity classes using the line parameter values shown in Table 3, following the methodology described by Roman-Lopes et al. (2018). Then, by using the appropriate linear relations and the hydrogen and helium EW values, spectral types for each O and B star in Table 3 were computed, considering the numerical values in intervals following the sequences: O stars [(5.0–5.5), (5.5–6.0),…, (9.0–9.5)] corresponding to [(O5–O5.5), (O5.5–O6),..., (O9–O9.5)] and B stars [(10–10.5), (10.5–11),…, (14.5–15)] corresponding to [(B0–B0.5), (B0.5–B1),…, (B4.5–B5)]. In cases in which the spectra of the stars also contain the helium lines, the spectral types were defined by taking the average of the values computed from the equations associated to the two transitions. In case of sources with no He ii lines, we derived lower and upper limits to their spectral types, and in the cases in which both results agree, only the resulting spectral type (not a range) was assigned (Roman-Lopes et al. 2018). The results obtained in the spectral classification process are shown in the last columns of Table 3 with the O stars corresponding to sources #1–#95 and the B stars (classes iv–v) corresponding to sources #96–#134.

Zoom In Zoom Out Reset image size

3.2.3. BSG and YSG Stars

Besides the new O- and B-type stars presented above, in the APOGEE2-S sample, we also identified new Galactic BSG and YSG stars. Figures 6 and 7 show some APOGEE2-S spectra of the mentioned type. The remaining spectra of BSG and YSG stars can be accessed using the SDSS DR16 SAS.¹²

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

View figure set (2 panels)

Tarball of figure set images

From visual inspection of the APOGEE2-S spectra of BSGs and YSGs, it is possible to see that the hydrogen lines of the Brackett series are narrower and deeper than those found in the spectra of B-type stars of the class iv–v. Furthermore, in the case of YSG stars, their spectra also contain atomic transition lines produced by heavier elements, like C, Mg, and Fe. On the other hand, concerning the positions occupied by BSG and YSG stars in the EW[Br11+Br13] versus mean FWHM[Br11–Br13] parameter space (Figure 5), we can see that, like in the cases of O and B stars (classes iv–v), BSGs and YSGs are also seen forming separate groups in areas of the diagram, as follows: BSGs—10 Å < mean FWHM[Br11–Br13] < 23 Å and 3.8 Å <EW[Br11+Br13] < 7 Å; YSGs: 10 Å <mean FWHM[Br11–Br13] < 23 Å and 10 Å <EW[Br11+Br13] < 16 Å.

Zoom In Zoom Out Reset image size

In case of the new BSG stars, we computed their spectral types using the associated EW values (sources #134–#149 in Table 3), and the spectral type (“Sp. Type”) versus EW linear relation for early- to mid-B supergiant stars obtained by Roman-Lopes et al. (2018, their Equation (1)). The results are listed in the last columns of Table 3, and as can be seen there, they are mostly of the B1–B2 types. On the other hand, for the YSG stars, the approach we used in the spectral type estimation was different. As this type of star was not previously studied by us with the APOGEE spectrographs, we made use of the template spectra of supergiant stars of Meyer et al. (1998). In Figure 8, there are six exemplars of H-band spectra of supergiant stars from their sample: HR1017 (B8 Ia), HR3975 (A0 Ib, HR1865 (F0Ib), HR1017 (F5 I), HR7796 (F7/F8 -band spectra of c Ib), and HR7479 (G1 II). Although the spectral resolution of the Meyer’s spectra is lower than that of APOGEE2-S, it is still possible to easily recognize several spectral features common to both samples: (a) three hydrogen lines of the Brackett series; (b) two lines at 15745 and 15753 Å, possibly corresponding to Mg i; and (c) three narrow lines at 15965, 16010, and 16893 Å that are probably due to Fe i transitions. Indeed, for each of the mentioned lines, we found from inspection of the Fe i line lists of the NIST Atomic Spectra Database¹³ that several Fe i lines, taking into account the spectrograph’s resolution, should probably appear blended at the mentioned wavelengths. (d) One line at 16022 Å was possibly due to C i. Although it is desirable to have the mentioned spectral features properly identified in the template spectra, for the purpose of this work, it is useful just to use them as H-band NIR indicators for the YSG spectral types.

Zoom In Zoom Out Reset image size

In Figure 9 (in black), we present the H-band spectra of the supergiant stars HR1865 (F0Ib) and HR1017 (F5Ib), together with the APOGEE2-S spectra (in red) of the sources 2M11100987-6123014 and 2M11104420-6127504. In order to highlight the similarity between spectral types of the APOGEE2-S supergiant stars with those from the template spectra of supergiants of Meyer et al. (1998), the APOGEE2-S spectra were degraded in their resolution by using a Gaussian filter. We can see that, despite the small differences in the shape of the continuum of the two spectral samples, as well some differences on the intensity of some atomic lines, the observed similarity enables us to provide reliable spectral type estimates for the program stars.

Zoom In Zoom Out Reset image size

Based on the comparison of the YSG template spectra with those of the new YSG stars, spectral types were assigned, and the results for sources #150–#165 are shown in Table 3. There we can see that from the 16 YSG stars, most are probably of the F type (nine), with three others being of the A0 type. For the remaining four exemplars in Table 3, they are probably intermediate A supergiants, but due to the lack of A-type supergiant templates (e.g., A1–A9), we assign spectral type A0–F0 for them.

Regarding the luminosity classes for O-type stars in Table 3 (stars #1–#95), we see that the majority have hydrogen mean FWHM[Br11–Br13] values lower than 25 Å, which, according to Roman-Lopes et al. (2018), does indicate a giant or supergiant nature. This result is quite different from that obtained by Roman-Lopes et al. (2019) in their similar SDSS-IV/APOGEE2 study of O stars in the direction of the W3, W4, and W5 massive star formation regions in the Perseus arm. This is illustrated in Figure 10, where we show comparative distributions (histograms with 5 Å bins) of the fraction of sources as a function of their FWHM[Br11–Br13] values for the O stars in W3/W4/W5 (Perseus sample) and of those studied here (Sagittarius sample). Clearly, both distributions differ significantly. While in the Perseus sample, only 11% of the O stars belong to luminosity classes i–iii (i.e., having FWHM[Br11–Br13] < 25Å), in the Sagittarius sample, the fraction of giants/supergiants turns out to be about 70%. In the next subsection, we discuss the possible causes for this observed behavior.

Zoom In Zoom Out Reset image size

Although the Sagittarius arm and the Perseus arm samples (Roman-Lopes et al. 2019) are basically compound by an early-type population associated to massive star-forming regions of the Galactic disk, their Galactic longitude distribution make them qualitatively different. The line of sight toward the Perseus arm sample (Galactic longitude l = 134°) intersects the Perseus spiral arm perpendicularly, at a mean heliocentric distance of about 2.8 kpc (Georgelin & Georgelin 1976). The young stellar population in such sample would then be mostly limited spatially to the width of the spiral arm of about 500 pc (Román-Zúñiga et al. 2019). This volume includes the giant complexes, W3, W4, and W5, whose population probably dominates the Perseus sample.

The line of sight toward the Sagittarius sample (Galactic longitude l = 291°) runs along the Sagittarius arm, thus covering a long distance stretch that goes from 2 kpc < d < 10 kpc, obviously including several giant young star formation regions, notably NGC3603 and NGC3576. This assumption is confirmed by the results shown in Figure 11, where we present the histogram of the heliocentric distances to the O, B, BSG, and YSG stars in Table 2 (our Sagittarius sample). The values were taken from the catalog of Gaia Data Release 2 distances of Bailer-Jones et al. (2018). As expected, it shows that the O stars in our sample are in fact distributed along a wide range of heliocentric distances, mostly in the range of 2 kpc < d < 10 kpc, with at least three main O-star groups (highlighted by the peaks of the distribution) at mean heliocentric distances of 2.8, 5.8, and 7.8 kpc. Interestingly, the last one presents a very pronounced peak that is probably associated with the large NGC3603 starburst complex, at a distance of about 7.6 kpc (Melnick et al. 1989; Crowther et al. 2010).

Zoom In Zoom Out Reset image size

As the Sagittarius sample is H magnitude bound, (8 < H < 12.2), one would expect a certain bias that favors the detection of the most luminous stars, i.e., the supergiant stars. Accordingly, besides the O-type supergiants of the Sagittarius sample, we identified a significant number of other BSG and YSG stars (see e.g., Figure 5). In Figure 12, we show a modified version of the distribution of the mean FWHM[Br11+Br13] values, in which we separate the O stars of the Sagittarius sample into three distance groups: (i) d < 4 kpc (green), (ii) 4 < d < 7 kpc (blue), and (iii) d > 7 kpc (red). It is clear that for all distance groups, most of the O stars still belong to luminosity classes I–III. Defining S/D as the ratio of number of classes I–III to classes IV–V stars, we computed the S/D values for the three groups in Figure 12. The larger S/D ratio 4 corresponds to the O stars closest to the Sun, (those with d < 4 kpc). Also, in the case of the two other distance groups (d > 4 kpc), a high S/D value of 2 is found, which is a much larger value compared to that observed for the Perseus sample (S/D = 0.1). In conclusion, we confirm that the observed behavior is not produced by differences in the heliocentric distances of the O stars.

Zoom In Zoom Out Reset image size

On the other hand, as the O stars in the Perseus sample come from a much smaller volume that is occupied mainly by the star-forming regions W3/W4/W5 (Roman-Lopes et al. 2019), the O-star population is expected to be, on average, significantly younger than that from the Sagittarius sample, which covers a much larger volume where the average stellar age of the O-star population tends to be higher. This age effect will result in average S/D ratio values much smaller in the Perseus population than in the Sagittarius sample. Additional studies in other directions and arms of the Galaxy certainly will help us to understand if this effect is a general behavior or just a localized statistical fluctuation at the top heavy end of the Galactic IMF.

Despite that fact that we only have a few sources in Table 2 with previous spectral classification, it is still possible to critically discuss some of the results obtained in this work, in light of previous studies. From the 165 sources in Table 2, 24 have some spectral type in the literature. Among them, 10 are known O-type stars for which we were able to estimate spectral types that are listed as follows (literature versus this study): sources #6 (O3.5 versus O7.5), #7 (O2 versus O7), #19 (O8 versus O7.5), #25 (O5 versus O8.5), #30 (O5.5 versus O6.5), #31 (O7.5 versus O7.5), #34 (O6 versus O6), #37 (O6 versus O7.5), #48 (O7 versus O8), and #51 (O7 versus O8). The largest discrepancies correspond to the earliest O-type stars (e.g., spectral types earlier than ∼O5). This is not a surprise as Roman-Lopes et al. (2018, 2019) already warned that for the mentioned spectral range, the EWs of the hydrogen Brackett lines stop decreasing monotonically with the temperature increase (while the EWs of the He ii lines deviate from the observed linear relation), and their equations may not deliver good results in the mentioned temperature range. Further studies are necessary in order to properly extend the usage of APOGEE’s H-band spectral window to the study of the earliest O-type stars, which corresponds to the spectral type range not covered by the work of Roman-Lopes et al. (2018).

On the other hand, in the early-B-type regime, the results obtained in case of sources #52 (B0 versus O9.5-B0), #97 (B2 versus B2), #101 (B1.5 versus B1/2), #103 (B0 versus B1), #111 (B1.5 versus B1.5), #116 (B3 versus B3), #121 (B3 versus B2/5), and #133 (B4 versus B2.5) indicate that spectral types for early-B stars obtained with the SDSS-IV/APOGEE2 empirical classification method are in good agreement with those obtained from other spectral ranges and/or methods.