DISCOVERY OF NEW, DUST-POOR B[e] SUPERGIANTS IN THE SMALL MAGELLANIC CLOUD

Optical spectra of our three newly identified sgB[e] stars, along with that of the previously known object, are shown in Figure 1, and their properties are given in Table 1. Table 2 gives their photometry. Our optical spectra clearly show strong emission features, particularly in the Balmer series. R38 exhibits strong P Cygni profiles in its Balmer lines as does the previously known sgB[e] star, AzV 154, indicating intense mass loss. In accordance with the standard definition of the B[e] phenomenon presented in, e.g., Zickgraf (1998) and Lamers et al. (1998), our stars also show numerous other emission lines. These lines are primarily permitted and forbidden low-ionization lines from metals, such as Fe ii, [Fe ii], Ti ii, Cr ii, and V ii. We identified emission lines in the spectral range 4000–4750 Å, based on the work of Zickgraf et al. (1989), Jaschek & Andrillat (1998), Andrillat et al. (1997), and Jaschek et al. (1996a, 1996b). Four to twelve forbidden emission lines are confirmed in each of the three new stars at these wavelengths, including [Fe ii] λλ4288, 4360, 4414, and [Ni ii] λ4462, which are detected in all objects. We list the heliocentric radial velocities (RVs) in Table 1, together with the root-mean-square velocity residual 〈v_rms〉 measured for the fitted RV from both permitted and forbidden lines, as well as that measured from only the forbidden lines 〈v_{fb, rms}〉. Given that there are fewer forbidden lines, the values of 〈v_{fb, rms}〉 are fully consistent with those of 〈v_rms〉.

Zoom In Zoom Out Reset image size

The absorption lines of B-star spectra are also present, for example, He i, Si ii, and Mg ii. The classification criteria of Fitzpatrick (1991) and Walborn & Fitzpatrick (1990) were used for spectral typing of our stars (Table 1). The presence in absorption of Si ii λλ4128-30, and the relative strength of Mg ii λ4481 compared to He i λ4471 indicate the spectral types of all three of our newly discovered sgB[e] stars to be in the mid- to late-B range. These spectral types are unexpectedly cooler than those derived for normal OB stars in the RIOTS4 survey. While it is possible that a cooler binary companion dominates the optical spectra, it turns out that higher R-band fluxes caused by the strong Hα emission in the decretion disks actually lead to lower Q_UBR values, mimicking a bluer continuum. This reveals a selection bias for the RIOTS4 survey which thus includes other R-excess objects, in addition to the target blue stars, thereby also selecting a large number of classical Be stars.

We can use the photospheric absorption lines to estimate the projected rotational velocity of our stars (Table 1) by Gaussian fitting of the Si ii, Mg ii, and He i lines. The values given in the table are rough estimates that correspond directly to the measured FWHM, since the accurate measurement of vsin i is difficult; effects such as macroturbulence can cause significant, additional line broadening for supergiants (Simón-Díaz & Herrero 2007). Of course, another important factor is the inclination angle with which we are viewing our stars. Two of our stars (R15 and R38) show single-peaked Balmer emission lines. If the Balmer emission comes from a disk, single-peaked lines imply that we are viewing these two objects close to pole-on (Stee & de Araujo 1994), so their rotational velocities could be significantly higher. This therefore suggests that macroturbulence significantly affects our measurements of the projected rotational velocity in Table 1.

We calculated the stellar luminosities from their V-band fluxes, using photometry from Massey (2002; Table 2), along with bolometric corrections for B supergiants from Drilling & Landolt (2000). These luminosities may be upper limits since there may be some contribution from the circumstellar disk. Photometry from the All Sky Automated Survey (Pojmanski 1997) hints that our three new sgB[e] stars may be variable at a level of ∼0.1 mag in V, and the known sgB[e] star AzV 154 shows variability in V at a level of 0.25 mag. Lamers et al. (1998) note that sgB[e] generally have low photometric variability. Extinction values were estimated using the extinction maps of Zaritsky et al. (2002), and a distance modulus of 18.9 was used (Harries et al. 2003). These quantities are also given in Table 2.

We now use the data we have compiled to constrain the evolutionary state of our three newly discovered B[e] stars. The selection criteria of the RIOTS4 survey only identify luminous, early-type stars. This means that we can immediately rule out the B[e] phenomenon due to planetary nebula sources or symbiotic stars. Thus, the stars must either be pre-main-sequence objects or evolved supergiants. At most, about 25% of massive stars in the SMC form outside clusters (Oey et al. 2004), and therefore the RIOTS4 field star criteria select against the pre-main-sequence Herbig B[e] stars. To further verify that our stars are not present in star-forming regions, we searched for dense nebulae by checking images of our stars in Hα, [S ii], and [O iii] using data from the Magellanic Clouds Emission-Line Survey (MCELS; Smith & MCELS Team 1998; Pellegrini et al. 2012). Figure 2 shows that although there is nebular emission present near R15, none of the stars are inside regions clearly suggestive of current, active star formation.

Zoom In Zoom Out Reset image size

We then compare our stars to the criteria of sgB[e] stars laid out by Lamers et al. (1998). Our stars are all supergiants, having luminosities above the limit suggested by Lamers of log(L/L_☉) ⩾ 4 (Table 1). AzV 154 and R38 both show P Cygni profiles in the Balmer recombination lines, which is a secondary criterion of sgB[e] stars.

We directly compare the optical spectra of our three newly identified sgB[e] stars to that of the previously discovered sgB[e] star, AzV 154. While the emission lines of AzV 154 have a much higher intensity than for our stars, they show many of the same lines, particularly the Fe ii and [Fe ii] lines typical of sgB[e] stars (Figure 1). Figure 3 shows a color–magnitude diagram (CMD) for our three stars (plus signs), along with AzV 154 (asterisk). The crosses show the other known SMC sgB[e] stars from the literature. Figure 3 indicates that our stars are clearly off the main sequence, as shown by the evolutionary tracks of Charbonnel et al. (1993). The proximity of these other SMC sgB[e] stars to our stars suggests that they are near the same evolutionary phase. Thus, our stars are clearly evolved supergiants.

Zoom In Zoom Out Reset image size

An important property of B[e] stars is a large infrared excess due to circumstellar dust. SgB[e] stars are noted for their combination of high luminosity and strong infrared excess. An example of this is Bonanos et al. (2010) who used data from the Spitzer SAGE SMC survey to study the properties of massive stars in the SMC. In all infrared CMDs and color–color diagrams, the sgB[e] stars form a clearly distinct group of massive stars. We obtained JHK photometry of all the B-type stars, including the Be and sgB[e] stars, in the RIOTS4 survey from the Two Micron All Sky Survey (2MASS; Skrutskie et al. 2006), and 3.6–24 μm photometry from the SAGE SMC survey (Gordon et al. 2011). The IR photometry for our objects is given in Table 2 and plotted in Figures 4 and 5 relative to these other groups of SMC B stars. The figures also show the previously known sgB[e] stars in the SMC and LMC (Zickgraf 2006; Wisniewski et al. 2007), which show a large IR excess, whereas our stars are on average 0.6 mag bluer than AzV 154 in H − K, 0.3 mag bluer in J − K, and 2.4 mag bluer in J −[3.6]. The modest IR excess for our stars, R15, R38, and R48, is similar to that for classical Be stars. In contrast, B stars that do not show evidence of a circumstellar disk have IR colors around zero, with the exception of some outliers which are just evolving off the main sequence.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Figure 6 shows the infrared spectral energy distributions (SEDs) for the sgB[e] stars, using the same data sets. For comparison, we also show the SEDs for three other stars from the RIOTS4 survey including a representative classical Be star, smc 37502 (Massey 2002); and OB supergiant, smc 19728, along with a late-type B supergiant (AzV 65) taken from Bonanos et al. (2010). Again, our three stars are very clearly different from the previously known sgB[e] star, AzV 154. However, they are also different from the B supergiants, whose SEDs are steeper. Interestingly, the infrared properties of our three stars closely match the infrared properties of classical Be stars. This is apparent in all of Figures 4–6.

Zoom In Zoom Out Reset image size

Through the infrared data obtained from the 2MASS and SAGE SMC surveys, we see that our stars show an infrared excess that is distinctly lower than that of previously discovered sgB[e] stars. This suggests a lack of warm dust in the circumstellar material around our three stars. An excess in K −[24] ⩾ 5 indicates the presence of an optically thick, dusty disk (Hernández et al. 2006). Our stars fall well below that limit, around K −[24] = 2.8, indicating that a dusty circumstellar disk is lacking, but indicating the presence of an optically thin circumstellar disk, which is characterized by K −[24] ⩾ 0.3.

The modest infrared excess of our stars is similar to that of classical Be stars (Figures 4 and 5), and their IR SEDs are also similar (Figure 6). Figure 6 shows that R15, R38, and R48 lack the IR bump from hot dust emission typical of sgB[e] stars like AzV 154. While our stars are much brighter, being supergiants, the general shape of the SEDs of our stars more closely follows that of classical Be stars. Thus, the IR emission in our objects may not be due to dust emission as is typical of sgB[e] stars, but may instead result from free–free emission as in classical Be stars (Gehrz et al. 1974). In the IR, free–free opacity increases at longer wavelengths, leading to an inverse power-law SED (Lamers & Cassinelli 1999). For example, Kastner et al. (2010) obtained IR spectra of several sgB[e] stars and found evidence of free–free emission in the near-IR. Thus our stars appear to have optically thin circumstellar disks, since they show weak IR excesses.

The lack of circumstellar dust could be due to a couple of different factors. First, the circumstellar disks around our stars could be ultra metal-poor. The SMC is already known to be a metal-poor environment, and the normal sgB[e] stars in the SMC show a lower infrared excess on average than those in either the Large Magellanic Cloud (LMC) or the Galaxy. Alternatively, our sgB[e] stars simply could have much smaller, or less massive, circumstellar disks, possibly because they are just transitioning into or out of the sgB[e] phase. If there is a lack of material in the disks around our stars, then the disks would be optically thin, meaning that dust would not be able to form, and the only infrared excess would be due to free–free emission. If our stars have such low-mass circumstellar disks, this would manifest in the optical spectra through the presence of weak emission lines, as opposed to the very strong emission lines seen in other sgB[e] stars; this is exactly what is seen in our spectra.

We have referred to R15, R38, and R48 as sgB[e] stars in this work. However, as described earlier, the current definition of B[e] stars includes a "strong" infrared excess (e.g., Conti 1997; Zickgraf 1998). This has been straightforward, since the optical spectral properties and infrared emission have always appeared to occur together. Now, however, our discovery of several objects having optical B[e] spectra, while lacking the strong IR excess, presents a complication for classification. While Lamers et al. (1998) specify the criteria for B[e] stars to include the strong IR excess, he also states that the spectroscopic criteria were defined for the optical spectrum. On the advice of Peter Conti and Henny Lamers, we classify these IR-weak objects as B[e] stars, based on the original, optically based definition of the stars (P. Conti 2012, private communication; H. J. G. L. M. Lamers 2012, private communication). Our classification therefore reverts to a definition of B[e] stars based exclusively on the existence of B-star properties together with optical permitted and forbidden emission lines (Conti 1976). This is reinforced by the fact that the B[e] nomenclature references only the optical spectral features.

Our three objects have optical and infrared properties that are quantitatively more similar to each other than to other sgB[e] stars. This is not the first detection of objects showing all of the properties of sgB[e] stars, but lacking the characteristic dusty infrared excess. Recently, Dunstall et al. (2012) reported the detection of a similar object in 30 Doradus in the LMC. This object, VFTS 698, shows optical [Fe ii] emission and IR properties similar to the three stars that we have found, and to classical Be stars. As also found in our stars, the photospheric emission is visible in VFTS 698, as seen in the absorption-line spectrum of He i and Si ii, whereas the photosphere is only occasionally seen in sgB[e] stars. We include the data for this star in Table 2 and Figures 4–6. There are likely to be a number of other such known stars, for example, MWC 314 in the Milky Way (Miroshnichenko 1996; Miroshnichenko et al. 1998). We caution that both VFTS 698 and MWC 314 are known binaries, which may complicate the comparison with our newly found SMC objects. However, one of our objects, R48, shows an anomalously high RV (Table 1) which may be evidence of binarity. Thus, dust-poor sgB[e] stars are found in widely different environments: VFTS698 in 30 Doradus, our three stars in the field of the SMC, and MWC 314 in the Milky Way. We therefore propose that such stars may be a distinct subset of sgB[e] stars, which may be transitioning to or from the full-fledged B[e] phenomenon.