Two Circumstellar Nebulae Discovered with the Wide-field Infrared Survey Explore and Their Massive Central Stars

Spectra of BD+60° 2668A&B were obtained with the TWIN spectrograph attached to the Cassegrain focus of the 3.5 m telescope in the Observatory of Calar Alto (Spain) on 2012 July 12. Three exposures of 60 s were taken. The setup used for TWIN consisted of the gratings T08 in the first order for the blue arm (spectral range of 3500–5600 Å) and T04 in the first order for the red arm (spectral range of 5300–7600 Å), providing a reciprocal dispersion of 72 Å mm⁻¹ for both arms. The resulting FWHM spectral resolution measured on strong lines of the night sky and reference spectra was 3.1–3.7 Å. The slit of 240 × 2.1 arcsec² was oriented at a PA selected in a way to observe both stars simultaneously. The seeing during the observations was not stable, ≃1.0–1.5 arcsec. Spectra of the He–Ar comparison arcs were obtained to calibrate the wavelength scale, and spectrophotometric standard star BD+33° 2642 (Bohlin 1996) was observed at the beginning of the night for the flux calibration.

The primary data reduction was done using the iraf package. The data for each CCD detector were trimmed, bias-subtracted, and flat-corrected. The subsequent long-slit data reduction was carried out in the way described in Kniazev et al. (2008). The two-dimensional (2D) spectra were averaged, and one-dimensional (1D) spectra for both components were then extracted using the iraf apall task. The resulting normalized 1D spectra are shown in Figure 4.

Zoom In Zoom Out Reset image size

Unfortunately, the optical counterpart to WS3 was discovered after the spectra of the central stars were obtained. The short total exposure used for the spectroscopy of these bright stars did not allow us to detect signatures of the nebular emission in the 2D spectrum.

Spectra of ALS 19653 and the optical shells around it were taken with the Robert Stobie Spectrograph (RSS; Burgh et al. 2003; Kobulnicky et al. 2003) mounted on the Southern African Large Telescope (SALT; Buckley et al. 2006; O'Donoghue et al. 2006). Observations were carried out in the long-slit mode on three occasions during 2016 (logged in Table 2). The PG900 grating was used for the first observation (PA = 0°) to cover the spectral range of 4200–7300 Å with a final reciprocal dispersion of 0.97 Å pixel⁻¹. The spectral resolution FWHM was 4.43 ± 0.18 Å. We call this spectrum low resolution hereafter. A Xe lamp arc spectrum was taken immediately after the science frame. Spectrophotometric standard stars were observed during twilight time for the relative flux calibration. To study the radial velocity and Hα intensity distributions in WS29 and its surroundings, we obtained two additional spectra of higher resolution with the PG2300 grating and two different orientations of the slit. The slit was placed on ALS 19653 and oriented in such a way to cross the brightest sides of the shell (PA = 33°) and the protrusions in the shell (PA = 120°). The obtained spectra cover the spectral range of 6035–6870 Å with a final reciprocal dispersion of 0.26 Å pixel⁻¹ and the spectral resolution FWHM of 1.85 ± 0.11 Å. We call these spectra high resolution hereafter. A Ne lamp arc spectrum was taken immediately after the science frames. The spatial scale of all three RSS spectra was 0.51 arcsec pixel⁻¹.

Primary reduction of the RSS data was done in the standard way with the SALT science pipeline (Crawford et al. 2010). The subsequent long-slit data reduction was carried out in the way described in Kniazev et al. (2008). The resulting reduced RSS spectrum of ALS 19653 is shown in Figure 5, while that of WS29 is presented and discussed in Section 5.

Zoom In Zoom Out Reset image size

To study ALS 19653 in more detail and to search for possible radial velocity variations, we observed this star with the SALT High Resolution Spectrograph (HRS; Barnes et al. 2008; Bramall et al. 2010, 2012; Crause et al. 2014) on two occasions, 2016 May 27 and 2017 October 5, with single exposures of 1500 s and a seeing of about 3 arcsec (see Table 2). The HRS is a dual-beam, fiber-fed échelle spectrograph. It was used in the medium-resolution mode (R = 40,000–43,000 and 2.23 arcsec diameter for both the object and sky fibers) to obtain a spectrum in the blue and red arms over the total spectral range of ≈3700–8900 Å. Both the blue and red arm CCDs were read out by a single amplifier with a 1 × 1 binning. Three arc spectra of the ThAr lamp and three spectral flats were obtained for each observation in this mode during a weekly set of HRS calibrations.

Primary reduction of the HRS spectra was performed with the SALT science pipeline (Crawford et al. 2010). The subsequent reduction steps, including background subtraction, order extraction, removal of the blaze function, identification of the arc lines, and merging of the orders for the object spectra, were carried out using the midas HRS pipeline described in detail in Kniazev et al. (2016). Parts of the 2016s HRS spectrum are shown in Figure 7.

The spectra of BD+60° 2668A&B and ALS 19653 are dominated by H i and He i absorption lines (see Figures 4 and 5). Two of these stars (BD+60° 2668A and ALS 19653) show the He ii λ4686 line, implying that they are B stars of spectral types earlier than B0.7 (Walborn & Fitzpatrick 1990). The HRS spectrum of ALS 19653 also shows the presence of the weak He ii λ4541 line. The only emission line detected is that of Hα in the spectrum of ALS 19653. We attribute this emission to the circumstellar material around the star (see the next subsection). Note the presence of a prominent diffuse interstellar band (DIB) at 6379 Å in the spectra of BD+60° 2668A&B, creating the illusion of asymmetry in the Hβ line profile (see Figures 6 and 7).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Using the classification criteria from Evans et al. (2004), we classify BD+60° 2668A and ALS 19653 as B0.5,⁹ and BD+60° 2668B as B1.5. The measured equivalent widths (EWs) of the Hγ line in the spectra of BD+60° 2668A, BD+60° 2668B, and ALS 19653 of 2.06 ± 0.09, 2.90 ± 0.11, and 1.42 ± 0.04 Å, respectively, and the EW(Hγ)–absolute magnitude calibration by Balona & Crampton (1974) suggest for these stars the luminosity classes of Ib, III, and Ia. The luminosity classes derived for BD+60° 2668A and ALS 19653, however, are inconsistent with the absolute visual magnitudes of these stars of M_V ≈ −5.4 mag and −5.5 mag (see Section 6), which instead imply the luminosity classes of II and Ib, respectively (e.g., Humphreys & McElroy 1984). We therefore regard BD+60° 2668A and ALS 19653 as B0.5 II and B0.5 Ib stars.

The stellar atmosphere analysis of the three stars is rooted on a large grid of synthetic stellar atmosphere models built using the atmosphere/line formation code fastwind (Santolaya-Rey et al. 1997; Puls et al. 2005; Rivero González et al. 2012). The fastwind stellar code takes into account non-local thermodynamic equilibrium effects in spherical symmetry with an explicit treatment of the stellar wind. The stellar grid was designed to accomplish the analysis of late O- and B-type stars with an effective temperature, T_eff, between 34,000 and 12,000 K in 1000 K steps, and a surface gravity, $\mathrm{log}\,g$, between 4.4 and 2.0 dex in steps of 0.1 dex. The helium abundance was fixed at the solar value. Explicit atomic models for H i; He i,ii; N ii,iii; O ii,iii; C ii,iii; Si ii,iii,iv; and Mg ii were included in the determination of the fundamental stellar parameters and chemical abundances. The rest of the chemical species were treated in an implicit way to account for blanketing/blocking effects. For further details, see Puls et al. (2005).

First, several key lines in the spectra¹⁰ were simultaneously compared with the grid looking for the set of stellar parameters that best reproduces the spectrum, following the routines and the line list described in Castro et al. (2012; see also Lefever et al. 2010). Subsequently, and based on the best predicted effective temperatures and gravities, new synthetic subgrids were built for each star with chemical abundances varying around the cosmic abundance standard (CAS) in the solar neighborhood (Nieva & Przybilla 2012) in steps of 0.2 dex and microturbulence velocities, ξ, spanning from 1 to 15 $\mathrm{km}\,{{\rm{s}}}^{-1}$ in steps of 1 $\mathrm{km}\,{{\rm{s}}}^{-1}$. The best combination of abundances and microturbulence that reproduces the observations were found through an optimized genetic algorithm. The main chemical transitions modeled in this study are labeled in Figures 6 and 7 (see also Table 4 in Castro et al. 2012). The derived stellar parameters and chemical abundances are listed in Tables 3 and 4.

Table 3.  Stellar Parameters for BD+60° 2668A&B and ALS 19653

	BD+60° 2668A	BD+60° 2668B	ALS 19653
${T}_{\mathrm{eff}}$ (kK)	${26.0}_{-1.8}^{+1.7}$	${23.0}_{-1.2}^{+1.4}$	${26.0}_{-1.7}^{+1.3}$
$\mathrm{log}g$	${3.30}_{-0.15}^{+0.14}$	${3.60}_{-0.12}^{+0.17}$	${3.10}_{-0.28}^{+0.26}$
$\xi (\mathrm{km}\,{{\rm{s}}}^{-1})$	9 ± 2	8 ± 2	7 ± 2
$v\sin i$ ($\mathrm{km}\,{{\rm{s}}}^{-1})$	130	150	28
${v}_{{\rm{r}},\mathrm{hel}}$ ($\mathrm{km}\,{{\rm{s}}}^{-1}$)	−20.9 ± 2.1	−20.9 ± 3.6	see Table 5

Download table as:  ASCIITypeset image

Table 4.  Elemental Abundances (by Number) in BD+60° 2668A&B and ALS 19653

	BD+60° 2668A	BD+60° 2668B	ALS 19653	CAS	Brott et al. (2011)
log(C/H)+12	${7.7}_{-0.2}^{+0.2}$	${7.5}_{-0.2}^{+0.2}$	${7.9}_{-0.2}^{+0.5}$	8.33 ± 0.04	8.13
log(N/H)+12	${8.3}_{-0.2}^{+0.2}$	${7.9}_{-0.2}^{+0.1}$	${7.9}_{-0.2}^{+0.2}$	7.79 ± 0.04	7.64
log(O/H)+12	${8.7}_{-0.1}^{+0.1}$	${8.8}_{-0.1}^{+0.1}$	${8.5}_{-0.2}^{+0.2}$	8.76 ± 0.05	8.55
log(Mg/H)+12	${7.6}_{-0.5}^{+0.4}$	${7.7}_{-0.2}^{+0.4}$	${7.5}_{-0.1}^{+0.3}$	7.50 ± 0.05	7.32
log(Si/H)+12	${7.6}_{-0.3}^{+0.2}$	${7.5}_{-0.1}^{+0.2}$	${7.5}_{-0.1}^{+0.2}$	7.56 ± 0.05	7.41

Note. The cosmic abundance standard (CAS; Nieva & Przybilla 2012) in the solar neighborhood and initial abundances adopted in the evolutionary models by Brott et al. (2011) are given for reference. Note the enhanced nitrogen abundance in BD+60° 2668A.

Download table as:  ASCIITypeset image

The high resolution of the HRS spectrum of ALS 19653 allowed us to determine the projected rotational velocity, $v\sin i$, through the Fourier transform of the Si iii λ4552 line profile using the iacob-broad code (Simón-Diaz & Herrero 2007, 2014). The much lower spectral resolution of the TWIN spectra prevents us from applying the same technique for BD+60° 2668A&B. For these stars, after a first guess of the best stellar parameters, we convolved the synthetic spectra until we reproduced the FWHM of the metallic lines. This procedure was repeated, updating $v\sin i$ until it converged and we reached the best combination of stellar parameters and $v\sin i$ that fits the observations. The obtained values of $v\sin i$ are given in Table 3. We note that the actual values of $v\sin i$ derived for BD+60° 2668A&B could be smaller because the measured velocities are around the limit imposed by the low spectral resolution.

We also estimated the heliocentric radial velocities, ${v}_{{\rm{r}},\mathrm{hel}}$, of all three stars using the ulyss (University of Lyon Spectroscopic analysis Software) program (Koleva et al. 2009) with a medium spectral resolution library (Prugniel et al. 2011). For BD+60° 2668A&B, we derived almost equal velocities (see Table 3). For ALS 19653, we used all five RSS and HRS spectra and found significant radial velocity variations (${\rm{\Delta }}{v}_{{\rm{r}},\mathrm{hel}}\approx 50$ $\mathrm{km}\,{{\rm{s}}}^{-1}$; see Table 5). These changes suggest that ALS 19653 is a close binary system, which was in the periastron passage between 2016 April 25 and May 27.

Table 5.  Heliocentric Radial Velocity Changes with Time in the Spectrum of ALS 19653

Date	${v}_{{\rm{r}},\mathrm{hel}}\,(\mathrm{km}\,{{\rm{s}}}^{-1})$	Spectrograph
2016 Apr 5	−29 ± 3	RSS GR900
2016 May 27	+9.25 ± 0.33	HRS
2016 May 28	+10.2 ± 0.5	RSS GR2300
2016 Jun 17	+23.7 ± 1.3	RSS GR2300
2017 Oct 5	+12.18 ± 1.33	HRS

Download table as:  ASCIITypeset image

To get an idea of the possible parameters of the binary system, we assume that its orbit is circular and fit the data points to a sine curve using the χ² algorithm. The best-fitting result (χ² = 2.33 × 10⁻³) is shown in Figure 8 and implies an orbital period of 164.20 ± 0.02 days, an amplitude of 33.33 ±0.04 $\mathrm{km}\,{{\rm{s}}}^{-1}$, and a systemic velocity of −9.02 ± 0.05 $\mathrm{km}\,{{\rm{s}}}^{-1}$. We realize that this solution is not unique because of the limited number of data points. Time-series spectroscopy and photometry of ALS 19653 are required to determine the actual parameters of the system.

Zoom In Zoom Out Reset image size

Still, using the above parameters of the binary and the mass of ALS 19653 of $\approx 20\,{\text{}}M\odot$ (see Section 6.2), one can estimate the mass of the companion star to be $\approx 8\,{\text{}}M\odot$. Assuming that this star is on the main sequence, one finds that its V-band brightness should be a factor of ∼20–30 lower than that of ALS 19653. This is consistent with the non-detection of double-lined structures in the HRS spectra of ALS 19653, although both spectra were obtained at the time favorable for their detection, i.e., when the difference in the radial velocities of the binary components was close to maximum (see Figure 8 and Table 5).

The presence of the Hα emission line in the spectrum of ALS 19653 (see Figure 5) could indicate that this star possesses a strong wind. In Figure 9, we compare the observed Hα line profile (from the HRS spectrum obtained on 2016 May 27) with synthetic ones obtained in fastwind models with four different mass-loss rates. In these models, we adopted a wind velocity law exponent β = 2 and a terminal wind velocity of 1500 $\mathrm{km}\,{{\rm{s}}}^{-1}$, based on the effective temperature of ALS 19653 and the empirical calibration in Kudritzki & Puls (2000). The intensity of the observed line matches a mass-loss rate of ≈4 × 10⁻⁷ ${\text{}}M\odot \,{\mathrm{yr}}^{-1}$. However, the synthetic lines have a P-Cygni-type profile (typical of lines formed in powerful winds or expanding gaseous shells), while in the observed line, the intensity peak is shifted to the violet side with respect to the peaks of the model lines, and the line itself has an asymmetric shape with an absorption dip on the red side (see also Figure 10). This suggests that the Hα emission is instead formed in the circumstellar material around the star.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The circumstellar origin of the Hα line is also suggested by changes in the profile of this line. Figure 10 shows that in 2016, the Hα line was dominated by a sharp asymmetric emission and also shows the presence of a second, much weaker emission component at ≈90 $\mathrm{km}\,{{\rm{s}}}^{-1}$, separated from the first one by a shallow absorption dip at ≈80 $\mathrm{km}\,{{\rm{s}}}^{-1}$. In 2017, the two-peak shape of the Hα line became more obvious after the absorption dip had extended below the continuum and shifted bluewards to ≈40 $\mathrm{km}\,{{\rm{s}}}^{-1}$. This change in the absorption was accompanied by a ≈10% decrease of the intensity of the main emission component and a blueward shift of the second emission peak. A similar behavior of the Hα emission line was detected in late-B and early-A supergiants (e.g., Kaufer et al. 1996; Markova & Valchev 2000; Verdugo et al. 2000), where the shape of this line changes from blue- or redshifted asymmetric emission to a double-peaked or inverse P Cygni profile (see, e.g., Figure 1 in Markova & Valchev 2000 for a good example of these changes). This behavior of the line profile was interpreted in terms of the deviation of the stellar wind from spherical symmetry or the presence of a flattened (disk-like) circumstellar structure (e.g., Kaufer et al. 1996; Petrenz & Puls 1996; Fullerton et al. 1997). The presence of an elongated (IR) nebula around ALS 19653 and the detection of intrinsic polarization of this star (with the polarization angle aligned with the elongated nebula) make it natural to assume that the Hα line originates in a flattened circumstellar structure (viewed nearly edge-on) and that its variability is caused by a large-scale density asymmetry in this structure (see Section 6.2).

The similar distances to and radial velocities of BD+60° 2668A&B, and the small angular (or projected linear) separation between these stars of ≈3 arcsec (or ≈0.05 pc) strongly suggest that they are members of the same stellar system. We note that both stars are located within the boundaries of the Cas OB5 association (Humphreys 1978) and that the Gaia DR2 parallaxes of a large part of stars listed in Humphreys (1978; see there Table 8) as the association members are similar to those of BD+60° 2668A&B. This implies that BD+60° 2668A&B reside in Cas OB5 and justifies why we adopt the same distance to both stars (see Section 2).

Using the adopted distance of 3.6 kpc, the B and V photometry from Table 1, T_eff from Table 3, ${(B-V)}_{0}$ colors and bolometric corrections from Fitzgerald (1970), and assuming that the total-to-selective absorption ratio equals 3.1, we calculate the absolute visual and bolometric magnitudes: M_V ≈ −5.4 mag and M_bol ≈ −7.9 mag for BD+60° 2668A and M_V ≈ −4.6 mag and M_bol ≈ −6.8 mag for BD+60° 2668B. The derived values of M_bol translate into bolometric luminosities of BD+60° 2668A&B of $\mathrm{log}(L/{\text{}}\,L\odot )\approx 5.0$ and 4.6, which, along with the effective temperatures of these stars, imply that both of them are still on the main sequence or only recently left it, and that their initial (zero-age main-sequence) masses were, respectively, M_ZAMS ≈20 and 15 M_⊙ (e.g., Brott et al. 2011).

The inference that both BD+60° 2668A&B are relatively unevolved is supported by the surface elemental abundances measured for these stars. The comparison of the CNO abundances of BD+60° 2668A given in Table 4 with those predicted for 20 M_⊙ rotating main-sequence stars by the stellar evolutionary models of Brott et al. (2011) shows that they agree with each other well enough if the initial rotational velocity of BD+60° 2668A was ∼400 $\mathrm{km}\,{{\rm{s}}}^{-1}$, meaning that the enhanced nitrogen abundance of this star could be caused by rotational mixing. The only difference is that Brott et al.'s models predict a 0.2 dex lower O abundance, which could be due to the lower initial abundance for this element (as compared to the CAS) adopted in these models (see Table 4). Brott et al.'s models also suggest that the age of BD+60° 2668A is ≈7–8 Myr. Similarly, we found that the CNO abundances of BD+60° 2668B would reasonably agree with the model predictions if the initial rotational velocity of this star was ∼200 $\mathrm{km}\,{{\rm{s}}}^{-1}$. The models also suggest that the age of BD+60° 2668B is ≈11 Myr, i.e., a factor of ≈1.5 higher than that of BD+60° 2668A.

The age estimates for BD+60° 2668A&B are in seeming contradiction with the possibility that the two stars are members of the same stellar system. The age discrepancy, however, could be understood if BD+60° 2668A is a blue straggler, i.e., a rejuvenated product of binary mass transfer or merger, implying that this star is or was a binary system. This possibility appears reasonable because the majority of massive stars are formed in binary and multiple systems and because their evolution is dominated by various binary-interaction processes (Sana et al. 2012). Moreover, the binary population synthesis modeling shows (de Mink et al. 2014) that about 30% of massive stars undergo binary interaction (i.e., mass transfer, common-envelope evolution, and/or merger) during the main-sequence stage. Enhancement of the nitrogen abundance by a factor of several on the surface of BD+60° 2668A could be the result of binary interaction as well (see Langer 2012).

Figure 1 shows that the circumstellar nebula around BD+60° 2668A&B is composed of two distinct components: a diffuse halo and an inner shell. Similar two-component nebulae were also detected around two candidate LBVs, GAL 079.29 + 00.46 (Gvaramadze et al. 2010b; see their Figure 2(j)) and [GKF2010] MN112 (Gvaramadze et al. 2010a; see their Figure 1). The morphology of these nebulae might be interpreted as indicating that their central stars experienced a brief episode of enhanced mass loss, leading to the formation of a compact region of dense material (visible in Hα as a density-bounded H ii region—the halo), and that afterwards the stellar wind increases its speed and sweeps up the material of the preceding slower wind, and thereby creates a shell within the halo.

In the life of a single massive star, such changes in the mass-loss rate and wind velocity can occur if after the red supergiant stage the star becomes a Wolf-Rayet star or undergoes a blue loop evolution. The CNO abundances derived for BD+60° 2668A&B, however, suggest that these stars have not yet gone through the red supergiant stage. Besides this, the episode of enhanced mass loss and subsequent increase of the wind velocity could be caused by the bi-stability jump (Pauldrach & Puls 1990; Lamers & Pauldrach 1991) if in the course of stellar evolution the effective temperature falls below some critical value (≈21–22 kK) and then again increases above it. According to the stellar evolutionary models by Brott et al. (2011), for 15–20 M_⊙ stars, this situation takes place during the transition from core-H to core-He burning. The duration of this transition of ∼10⁵ years, however, is too long, and the wind velocity is too high (∼1000 $\mathrm{km}\,{{\rm{s}}}^{-1}$) to result in a compact (parsec-size) circumstellar nebula.

On the other hand, as discussed above, BD+60° 2668A might be the result of a merger of two stars, which causes a temporary inflation of the resulting single star and a decrease of its T_eff below the temperature of the bi-stability jump. As a result of this, the stellar wind velocity drops by a factor of 2, while the mass-loss rate increases by a factor of 10 (Vink 2018), leading to a factor of 20 increase of the stellar wind density and formation of a halo around the star. Later on, after the thermal adjustment of the merger product on its Kelvin–Helmholtz timescale (∼10⁴ years), the effective temperature of the star increases above the temperature of the bi-stability jump, and the stellar wind velocity increases to its initial value, which in turn leads to the formation of a shell within the halo. Since the (relaxed) merger product is expected to be a fast-rotating star, the temperature distribution across its surface could be inhomogeneous due to gravity-darkening. Correspondingly, the wind velocity along the stellar rotational axis might be higher than in the equatorial region, which could be responsible for the observed elongated shape of the shell (see Figure 1). We speculate also that the postmerger relaxation of the newly formed single star might be accompanied by photometric and spectroscopic variability typical of LBVs and that in the recent past BD+60° 2668A probably looked like an LBV.

The possibility that LBV activity (including giant, η Car-like, eruptions) is triggered by the interaction between companion stars in binary or triple systems (e.g., through the merger of two stars, tidal interaction, or interchange of components) has been widely discussed in the last two decades (e.g., Justham et al. 2014; Portegies Zwart & van den Heuvel 2016 and references therein). Since this interaction can take place at both early and late stages of evolution of binary/triple systems, it is natural to expect that the LBV phenomenon (accompanied by the formation of circumstellar nebulae) can be found among massive stars at both early and advanced evolutionary stages. It should also be noted that the detection of possible companion stars in wide orbits around several (candidate) LBVs (e.g., Martayan et al. 2016) does not contradict the possibility that the LBV activity of these stars is caused by binary-interaction processes. In such cases, the (candidate) LBV stars might be unresolved binaries or the merger products of two stars, while the nearby stars might represent the tertiary stars dynamically scattered in wide orbits (see Reipurth & Mikkola 2012) or former (unbound) members of dissolved triple systems (see Gvaramadze & Menten 2012). BD+60° 2668B might be such a tertiary star.

In summary, the spectral analysis of BD+60° 2668A&B indicates that both stars are either still on the main sequence or only recently left it, and that the surface nitrogen abundance of BD+60° 2668A is enhanced by a factor of several. Also, it appears that BD+60° 2668A is a factor of two younger than BD+60° 2668B despite strong indications that both stars are members of the same stellar system and, therefore, most likely were formed simultaneously. The age discrepancy could be naturally understood if BD+60° 2668A is a rejuvenated product of binary interaction, in which case the enhanced nitrogen abundance on the surface of the star is a direct consequence of this interaction. The binary-interaction scenario also provides a framework for understanding the origin of circumstellar nebula, which can hardly be produced by a single unevolved star.

For ALS 19653, we derive M_V ≈ −5.5 mag and $\mathrm{log}(L/{\text{}}{L}_{\odot })\approx 5.1$, which, along with the effective temperature, imply an initial mass and age of ALS 19653 of, respectively, ≈20 M_⊙ and ≈7–8 Myr, meaning that this star is either still on the main sequence or has just left it (e.g., Brott et al. 2011).

The chemical abundances derived for ALS 19653 (see Table 4) support the inference that this star is relatively unevolved and agree with Brott et al.'s models if the initial rotational velocity of this star was ∼200 $\mathrm{km}\,{{\rm{s}}}^{-1}$. Note that the uncertainties in the measurements allow the possibility that the surface nitrogen abundance is enhanced by factors of 2 or 3 in comparison with the CAS or the initial nitrogen abundance adopted in Brott et al. (2011), respectively. This enhancement, if real, could be caused not only by rotationally induced mixing, but also by binary interaction, e.g., because of accretion of CNO-processed material from the companion star or mixing caused by the merger of the binary components (e.g., Langer 2012).

The detection of two nested shells around ALS 19653 indicates that this star has experienced at least two episodes of enhanced mass loss.¹¹ At the adopted distance to ALS 19653, the characteristic radius of the second (outer) shell is ≈1.2 pc, which along with the expansion velocity of this shell of ∼100 $\mathrm{km}\,{{\rm{s}}}^{-1}$ gives its kinematic age of ∼10⁴ years. Similarly, assuming that the optical shell of WS29 is expanding with a velocity of 20–30 $\mathrm{km}\,{{\rm{s}}}^{-1}$ (see Figure 12) and given its characteristic radius of ≈0.2 pc, one finds that this shell was formed shortly after the outer shell. WS29 is especially curious because it consists of an optical circular shell with protrusions in the northwest and southeast directions, and an inner elongated IR nebula, stretched in the direction perpendicular to the axis defined by the protrusions. Below we discuss the possible origin of WS29.

The changes in the Hα line profile suggest that ALS 19653 is surrounded by a disk-like structure viewed nearly edge-on (see Section 4.2), while the polarization angle of the stellar light of 121° (see Section 2) implies that this structure is oriented in the same direction as the IR nebula. We therefore interpret this nebula as a flattened outflow or ejecta from ALS 19653 viewed nearly edge-on. Since it is unlikely that a single unevolved star can produce a circumstellar nebula, it is reasonable to assume that its formation is due to a binary-interaction process, e.g., because of the merger of two stars or common-envelope evolution. Correspondingly, the spin axis of the merger product or the angular momentum of the post-common-envelope binary orbit should coincide with the line connecting the protrusions in the optical shell of WS29 (PA ≈ 120°). If ALS 19653 is indeed the result of merger of two stars, then the inclination angle of its rotational axis to our line of sight should be close to 90°, while its equatorial rotational velocity should be almost equal to the projected rotational velocity of 28 $\mathrm{km}\,{{\rm{s}}}^{-1}$. The Hα line variability in ALS 19653 could be understood if the density structure of the (inner part of the) flattened circumstellar nebula is non-axisymmetric, e.g., because of the tidal force from the companion star (see Okazaki et al. 2002; Oktariani & Okazaki 2009). The possible binary nature of ALS 19653 (see Section 4.2) makes this explanation plausible. It should be noted, however, that the binarity of ALS 19653 does not exclude the possibility that this star is a merger product because originally it might have been a triple system (see Pasquali et al. 2000; Schneider et al. 2016).

The strong differential rotation originating in the course of the merger of a binary system or common-envelope evolution might be responsible for the generation of a strong large-scale magnetic field in the newly formed single star (Langer 2012; Wickramasinghe et al. 2014; Schneider et al. 2016) or in the ejected envelope (Regös & Tout 1995; Tout & Regós 2003), respectively. This magnetic field can in turn effectively spin down the merger product through dynamical mass loss (e.g., Langer 2012) and may affect (or even determine) the geometry of the ejected material (Chevalier & Luo 1994; Różyczka & Franco 1996; Garcia-Segura et al. 1999; Nordhaus & Blackman 2006). Correspondingly, the low rotational velocity derived for ALS 19653 might be the direct result of magnetic braking of this star, while the origin of protrusions in the northwest and southeast directions might be caused by the deflection of ejected material toward the polar directions by the tension of the toroidal magnetic field. The gravity-darkening of the initially fast-rotating merger remnant may also contribute to the origin of the polar protrusions.

Finally, we note that the distribution of the Hα radial velocity along the slit with PA = 33° (i.e., along the slit aligned with the IR nebula) suggests that the outer parts of the equatorial ejecta might be rotating around the axis defined by the orientation of the polar protrusions. If real, this rotation would imply a very efficient transfer of angular momentum from the merger remnant (possibly rotating at break-up velocity at the equator) or from the shrinking binary system to a tiny fraction of the equatorial ejecta by means of magnetic torque. A follow-up study of the Hα velocity field in WS29 with the Multi-Unit Spectroscopic Explorer (MUSE; Bacon et al. 2010) could potentially clarify this issue.