THE PHYSICAL PROPERTIES OF THE RED SUPERGIANT WOH G64: THE LARGEST STAR KNOWN?^*

Red supergiants (RSGs) are an evolved He-burning phase in the life of a moderately massive star (10–25 M_☉). Their location in the Hertzsprung–Russell (H–R) diagram has, in the past, been at odds with the predictions of stellar evolutionary theory, occupying a position that was too cold and too luminous to agree with the position of the evolutionary tracks. Levesque et al. (2005, 2006; hereafter Papers I and II) and Massey et al. (2009) have used the newest generation of the MARCS stellar atmosphere models (Gustafsson et al. 1975, 2003, 2008; Plez et al. 1992; Plez 2003) to fit model spectral energy distributions (SEDs) to moderate-resolution spectrophotometry of RSGs in the Milky Way, Magellanic Clouds, and M31. The new physical parameters derived from this fitting shifted the properties of RSGs to higher effective temperatures and lower luminosities, bringing them into generally excellent agreement with the evolutionary tracks.

However, the RSGs include several oddball members that require a more detailed analysis. One classic example is the case of VY CMa, a well-studied dust-enshrouded RSG in the Milky Way. Massey et al. (2006) fit a spectrum of VY CMa with the MARCS models and derived an effective temperature (T_eff) that was much warmer than previous estimates (Le Sidaner & Le Bertre 1996) and brought this previously troublesome star into much better agreement with the evolutionary tracks; however, determination of VY CMa's stellar luminosity remains difficult due to the challenges posed by the large structured dust-reflection nebula that surrounds the star (Monnier et al. 1999; Smith et al. 2001; Smith 2004; Humphreys et al. 2005; Massey et al. 2008) and its uncertain distance (Choi et al. 2008). Another example is the unusual Magellanic Cloud RSGs with variable effective temperatures and dust production as described in Levesque et al. (2007) and Massey et al. (2007), and henceforth referred to as Levesque–Massey variables. These stars show considerable changes in their effective temperatures and V magnitudes and a common set of variations—when they are at their warmest they are also brighter, dustier, and more luminous. They are recognizable in their cooler states by occupying the "forbidden" region on the H–R diagram to the right of the Hayashi track, where stars are no longer expected to be in hydrostatic equilibrium (Hayashi & Hoshi 1961). They are currently believed to be in an unstable (and short-lived) evolutionary phase not previously observed in RSGs.

WOH G64 (α₂₀₀₀ = 04:55:10.49, δ₂₀₀₀ = −68:20:29.8, from Buchanan et al. 2006) is another unusual RSG in the Large Magellanic Cloud (LMC), originally discovered by Westerlund et al. (1981), with a number of properties that set it apart from the rest of the LMC RSG population. It is an IRAS source (IRAS 04553–6825), and is surrounded by an optically thick dust torus (Elias et al. 1986; Roche et al. 1993; Ohnaka et al. 2008). Studies of the optical spectra by Elias et al. (1986) and van Loon et al. (2005) show a spectrum dominated by very strong TiO bands, which has led to assignments of spectral types as late as M7–8, by far the latest spectral type assigned to an LMC RSG (e.g., Paper II), with corresponding extreme physical properties, which would make WOH G64 the star with the largest known radius, and the RSG with the largest known luminosity and mass (van Loon et al. 2005). It is also a known source of OH, SiO, and H₂O masers (Wood et al. 1986; van Loon et al. 1996, 1998, 2001), with two different maser components, suggesting a substantial mass outflow and two different expanding dust shells (Marshall et al. 2004). Finally, Elias et al. (1986) report detections of Hα, [O i], and [N ii] emission in the optical spectrum. WOH G64 is neither isolated nor in a well-populated region. The star is not in any of the Lucke–Hodge OB associations (Lucke & Hodge 1970). Inspection of the UBVR CCD images of the LMC described by Massey (2002) reveals that although there are bright neighboring blue stars in the vicinity (such as [M2002] 22861, located 70'' away, or the star BI 23, classified as B0 II by Conti et al. 1986, located 110'' away), the region is not rich in bright, blue stars.

We have investigated this unique RSG in more detail, applying a similar analysis used in Papers I and II to examine the physical properties that can be derived from the optical spectrum. We obtain moderate-resolution spectrophotometry of WOH G64 using IMACS on the Baade 6.5 m Magellan telescope at Las Campanas Observatory (Section 2), confirm this star's membership in the LMC (Section 3), and compare the parameters that we determine from our model fitting to those presented in past work and predictions of evolutionary theory (Section 4). Finally, we discuss the intriguing issue of emission line features in WOH G64 (Section 5) and summarize our findings on this unusual star (Section 6).

A flux-calibrated spectrum of WOH G64 was obtained on UT 2008 December 10 using the IMACS spectrometer on the Baade 6.5 m Magellan telescope during an observing run primarily devoted to an unrelated project. A 300 line mm⁻¹ grating was used in first order for spectral coverage from 3500 Å to 9500 Å. A 09-wide long slit was used, resulting in a spectral resolution of 4.3 Å. The observation consisted of 3 × 600 s exposures, and were obtained under photometric conditions. The seeing, measured in the visible off the guide camera (which is off-axis) registered 05–06 during the exposure. We measure 07 full width at half-maximum intensity on our spectra. The airmass was approximately 1.5 at the time. The slit was oriented to the parallactic angle; we note that IMACS also includes an atmospheric dispersion compensator, so our relative fluxes from the red to the blue should be good despite the modestly large airmass.

Flat fielding was performed by means of a series of flats obtained on a projector screen; wavelength calibration (He–Ne–Ar lamps) was obtained immediately after the three exposures. A large number of bias frames (60) were averaged and used to remove residual two-dimensional structure after overscan correction and trimming.

In addition to the observation of WOH G64, observations of three LMC RSGs with known radial velocities from Massey & Olsen (2003) were obtained immediately prior to the WOH G64 observation in order to check on LMC membership for WOH G64 (see Section 3). Those exposures were 60 s long, and obtained at the parallactic angle. Each observation was also followed by an He–Ne–Ar comparison arc.

Throughout the course of this and several other nights, we also observed spectrophotometric standard stars from the list of Hamuy et al. (1994). The spectrograph slit was oriented to the parallactic angle for these observations as well. Some of those data were obtained under nonphotometric conditions, but all of the standards data were combined, with a gray shift calculated such that the average throughput agreed with the observation with the highest throughput. The resulting residuals from a high-order fit were a few hundredths of a magnitude, as expected. The sensitivity curve we derived was applied to the WOH G64 observation.

All data were processed using IRAF.⁶ The spectra were extracted using an optimal extraction algorithm, with deviant pixels identified and rejected based upon the assumption of a smoothly varying spatial profile.

Before investigating the physical properties of WOH G64, we first wished to confirm that it was indeed an RSG member of the Magellanic Clouds, rather than an unusual foreground giant. Elias et al. (1986) measure a radial velocity of 315 km s⁻¹ from several nebular emission lines, but state that the velocity zero point is uncertain by 50 km s⁻¹, and cite this as evidence of membership in the LMC. However, this really only demonstrates that the star is not a foreground dwarf; a halo giant will have a similar velocity to the LMC's systematic velocity, as it is due mainly to the reflex motion of the Sun. This value can also be compared to the ∼270–278 km s⁻¹ velocity of the maser emission reported by Marshall et al. (2004), which is more in line with our expectations of the LMC's rotation (Olsen & Massey 2007). Finally, van Loon et al. (1998) use the Ca ii triplet to measure a heliocentric velocity of ∼300 km s⁻¹.

We can use our spectra to determine if the star's radial velocity is consistent with the kinematics of other RSGs in the LMC. We measured the radial velocity of WOH G64 by cross correlating our spectra against those of three LMC RSGs with accurate velocities measured by Massey & Olsen (2003), [M2002] LMC 022204, 024014, and 024987. For this, we used the fxcor task in the IRAF rv package, restricting the cross-correlation region to 8450–8700 Å, which contains the very strong Ca ii triplet lines at λλ8498, 8542, 8662 (Figure 1). The spectra were first normalized to unity, and then a value of 1 subtracted so that only spectral features add signal to the cross correlation. We used the three individual WOH G64 spectra in order to determine the internal consistency of our measurements; in addition, we cross correlated each of the radial velocity "standards" against each other. These tests revealed that the maximum difference seen in cross correlating any two calibrating stars against themselves was 10 km s⁻¹, giving us assurance that the velocity we determine with the combination of the three would be a few km s⁻¹. We expect to be able to centroid each line to about 10% of the spectral resolution element or better, and so this is consistent since the 4.3 Å resolution translates to about 150 km s⁻¹. We obtained a radial velocity of 294 ± 2 km s⁻¹, where the error refers to the formal standard deviation of the mean, and reflects the good agreement in cross correlating the stars used for the velocity determinations against themselves. Given this star's location in the LMC, we expect a heliocentric velocity of 263 km s⁻¹ due to contributions from the LMC's space motion relative to the Sun and its internal rotation, using a kinematic model fit to LMC carbon stars (Olsen & Massey 2007). Although our measured velocity for WOH G64 is ∼30 km s⁻¹ higher than that of the carbon stars, it is still entirely consistent with LMC membership, as seen in Figure 1. The difference in velocity can be attributed to the fact that the LMC's RSGs appear to exhibit ∼45 km s⁻¹ faster rotation than the carbon stars (whether this is because either the LMC's RSGs or its carbon stars have been kinematically disturbed by external forces is not yet clear; see Olsen & Massey 2007). In any case, WOH G64's velocity is indeed typical for an LMC RSG at its position.

Zoom In Zoom Out Reset image size

4.1. Previous Work

4.2. Properties from Spectral Fitting

Based on the strengths of the TiO bands in our WOH G64 optical spectrum, we assign a spectral type of M5 I by comparison with other RSGs we have classified (Papers I and II; Massey et al. 2009). This type confirms WOH G64 as the latest-type RSG in the LMC, although it is considerably earlier than the M7.5 types discussed in Elias et al. (1986) and van Loon et al. (2005).

The observed SED for WOH G64 was compared to MARCS stellar atmosphere models of metallicity Z/Z_☉ = 0.5, corresponding to the metallicity of the LMC (Westerlund 1997; Massey et al. 2004). The MARCS model SEDs available ranged from 3000 to 4500 K in 100 K increments, and were interpolated for intermediate temperatures at 25 K increments. The log g values ranged from −1 to +1 in increments of 0.5 dex. When fitting the models to the data, they were reddened adopting the Cardelli et al. (1989) R_V = 3.1 reddening law.

The reddening and T_eff were determined by finding the best by-eye fit between the MARCS models and the WOH G64 SED. The reddening was based on the agreement between the model and observed continuum, while T_eff was based on the strengths of the TiO bands at λλ6158,6658,7054 (Jaschek & Jaschek 1990), with the bluer TiO bands at λλ5167,5448,5847 used as secondary confirmations of the fit quality; this ensured that there was minimal degeneracy in determining the best model fit. The λ8433 and λ8859 TiO bands are present in the spectrum as well; however, we have found in past work that these features are not generally well matched by the MARCS models (see Papers I and II). Considering WOH G64's strong TiO band strengths and corresponding late spectral type, our precision for this fit was ±25 K. The extinction value A_V is determined to approximately 0.15 mag precision.

From fitting the 5000–9000 Å optical spectrum, we find a T_eff of 3400 ± 25 K and A_V = 6.82 ± 0.15 mag, adopting a log g value of −0.5 dex. Our best model fit is shown in Figure 2, and alternative model fits demonstrating the precision of these derived properties are shown in Figure 3. In order to confirm that this is the appropriate log g value for WOH G64, we must first determine the star's radius, which in turn requires calculation of the bolometric luminosity. While RSGs in the LMC are known to be variable by up to a magnitude or more in V (Levesque et al. 2007), their variability at K is much smaller (∼0.2 mag; Josselin et al. 2000). For this reason, we determine the bolometric luminosity based on WOH G64's K magnitude. There are a number of K-band observations for this star, including K_s = 6.85 from Two Micron All Sky Survey (2MASS), K = 6.85 from Buchanan et al. (2006), K = 6.88 from Elias et al. (1986), and K = 6.91 from DENIS (Epchtein et al. 1997). The consistency of these magnitudes confirms that the K magnitude of WOH G64 has stayed quite constant with time. We adopt the 2MASS K_s magnitude for this work, correcting K_s to K by the relation K = K_s + 0.04 (Carpenter 2001) and adopting an error of ±0.2 mag (Josselin et al. 2000). We adopt the T_eff-dependent bolometric correction at K for the LMC (Paper II):

$$\begin{equation} {\rm BC}_K = 5.502-0.7392\left(\frac{T_{\rm eff}}{1000\;\rm {K}}\right). \end{equation} \tag{ 1 }$$

By taking these parameters, calculating A_K = 0.12 × A_V (Schlegel et al. 1998), and adopting a distance modulus of 18.5 (50 kpc) for the LMC (van den Bergh 2000) with an error of ±0.1 mag, we determine a physical log g of −0.7 ± 0.1 dex, which agrees with our model surface gravity. With these values, and taking M_bol,☉ = 4.74 (Bessell et al. 1998) we also determine an M_bol = −9.4 ± 0.3 (log(L/L_☉) = 5.65 ± 0.14), or ∼0.6 mag more luminous than the value obtained by Ohnaka et al. (2008). We also measure a radius of R/R_☉ = 1970, with an uncertainty of ∼5%. While van Loon et al. (2005) mention some disparity in spectral-type determinations between different TiO bands, we find no such disagreement when assigning our spectral type and T_eff.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

4.3. Properties from (V − K)₀

As a self-consistency check, we also determine these physical properties based on the (V − K)₀ color of WOH G64, as we have done in past papers (Papers I and II; Levesque et al. 2007; Massey et al. 2009). For a V magnitude, we adopt V = 18.63 from the MACHO photometric data (Allsman & Axelrod 2001). We deredden the photometry using (V − K)₀ = V − K − (0.88 × A_V) (Schlegel et al. 1998). From this, we find that (V − K)₀ for WOH G64 is 5.74. In Paper II, we determined the theoretical relation between T_eff and (V − K)₀ based on the MARCS models and the assumptions of Bessell et al. (1998) concerning the effective broadband bandpasses, and found

$$\begin{eqnarray} &&\hspace{-24pt} T_{\rm eff} = 7621.1-1737.74(V-K)_0 + 241.762(V-K)_0^2\nonumber\\ && - 11.8433(V-K)_0^3. \end{eqnarray} \tag{ 2 }$$

With these we determined a T_eff of 3372 K. Once again using the bolometric correction at K and A_K and propagating our errors for these values, these numbers in turn produce a log g = −0.7 ± 0.1 dex and M_bol = −9.4 ± 0.3 (log(L/L_☉) = 5.65 ± 0.13). Based on these parameters, we also calculate a radius of R/R_☉ = 1990 with an uncertainly of ∼5%. This shows excellent agreement between the parameters derived from spectral fitting as compared to the (V − K)₀ colors.

The effective temperatures from both spectral fitting and (V − K)₀ colors agree with the predictions of the T_eff scale for the LMC (Paper II); WOH G64 is the only M5 I RSG known in the LMC, and this additional data point agrees nicely with the general trend that the LMC T_eff scale is ∼50 K cooler than the Galactic scale (the single M5 I RSG in the Milky Way, α Her, has a T_eff of 3450 K; Paper I).

The parameters determined from past work, our spectral fitting, and our (V − K)₀ colors are summarized in Table 1, where we also compare our values to those derived by others.

4.4. The Physical Parameters of WOH G64 and the H–R Diagram

In our analysis of WOH G64, we find an unusually high bolometric luminosity, M_bol = −9.4 (log L/L_☉ = 5.65) compared to the other most luminous RSGs in the Magellanic Clouds (M_bol = −8.8). As shown in Table 1, our value is consistent with that of Elias et al. (1986) and van Loon et al. (2005), but about 0.5 mag more luminous than that found by Ohnaka et al. (2008).

As is obvious from the large extinction, WOH G64 is surrounded by a dense circumstellar envelope. If this envelope were spherical, then the flux absorbed along the line of sight would be re-radiated toward us in the NIR, but with an asymmetrical circumstellar environment (either a torus or a disk), we may receive more (or less) than what is absorbed. If we adopt the geometrical model proposed by Ohnaka et al. (2008), that we view a torus almost head on, then we are receiving more flux form the combined system (star plus torus) than is being absorbed. Based on Figure 5 in Ohnaka et al. (2008), we estimate the torus contributes approximately 0.5 mag at K. In order to derive the stellar flux, we then need to correct our M_bol derived from K (−9.4) by this amount, and arrive at an M_bol = −8.9. We list our final adopted parameters in Table 1.⁷

We believe our value for the extinction is to be preferred over that of the Ohnaka et al. (2008) analysis, as they adopt a blackbody flux distribution for the star. The use of a blackbody overestimates the stellar flux in the optical, which leads to an overestimate of the amount of attenuation required to reproduce the observed data. In the end, though, our estimates for the bolometric luminosities are in excellent agreement, as shown in Table 1. Ohnaka et al. (2008) did adopt a somewhat lower T_eff than what we found (3200 K versus 3400 K), but fortuitously they also applied their analysis with a 3400 K T_eff and derive essentially the same bolometric luminosity (K. Ohnaka 2009, private communication).

With our determination of T_eff = 3400 K and M_bol = −8.9, we derive a stellar radius of 1540R_☉. Given the minimum formal errors (ΔT_eff = 25 K and ΔM_bol = 0.1 mag), the uncertainty of the radius is 5%. This value of 1540R_☉ is significantly larger than the largest Magellanic Cloud stars (1240–1310R_☉), as expected from its cooler temperature and higher bolometric luminosity. But, while the star may be a behemoth by Magellanic Cloud standards, it is about the same size as the largest RSGs we found in Milky Way (Paper I), where KW Sgr, Case 75, KY Cyg, and μ Cep all have essentially the same large radii as WOH G64.

In Figure 4, we show the position of the star in the H–R diagram, compared to the location of the Geneva evolutionary tracks. We have included too the locations of the star from van Loon et al. (2005) and Ohnaka et al. (2008). We see that the star is cooler than the evolutionary tracks allow, but that other LMC stars also suffer from this problem; we agree with the Ohnaka et al. (2008) conclusion that this star is on the edge (or slightly to the right) of the Hayashi limit. The luminosity is in good agreement with the prediction of stellar evolutionary theory, and we see that the mass corresponding to this luminosity is about 25 M_☉ (note that although the masses labeled are the stars' initial masses, mass loss is expected to remove only a few percent of a star's mass during its earlier evolution for stars in this mass range). Combined with our radius estimate, this leads to a log g value of −0.5, consistent with the surface gravity adopted in our fit (Section 4.2). Given the 5% uncertainty in the radius, and adopting a 20% uncertainty in the mass, leads to an uncertainty of 0.1 dex in log g.

Zoom In Zoom Out Reset image size

Elias et al. (1986) report detection of [O i] λ6300, Hα, [N ii] λ6548, and [N ii] λ6584 emission lines in their ∼6 Å resolution spectrum of WOH G64, along with a tentative detection of [S ii] λλ6717,6731 emission. Using the [O i], Hα, and [N ii] lines they measure a radial velocity of 315 km s⁻¹, which implies a redshift compared to the 270–278 km s⁻¹ value found in the maser analysis by Marshall et al. (2004) and our 294 km s⁻¹ measurement from the Ca ii lines (Section 3), given that the Elias et al. (1986) value may have a zero-point error as large as 50 km s⁻¹. Given that the Elias et al. (1986) values have as much as 50 km s⁻¹ zero-point error, the significance of this redshift was not large, but we confirm the redshift below, using more accurate measurements.

In the red part of our spectrum, we detect [O i] λ6300, Hα, [N ii] λ6548, [N ii] λ6584, and [S ii] λ6731 emission (Figure 5(a)). The [S ii] λ6717 doublet line is not measurable, as its wavelength is coincident with one of the numerous absorption features in our spectrum. In addition, we detect Hβ, [N i] λλ5198,5200, and [O iii] λ5007 emission in the bluer region of the spectrum (Figure 5(b)). We also note that some of the TiO band heads are in weak emission. This is also seen in at least one other RSG with an extreme circumstellar environment, e.g., VY CMa (Hyland et al. 1969; Wallerstein & Gonzalez 2001). The lines are clearly spatially coincident with the stellar spectrum, and there is no evidence of extended (nebular) emission in our good seeing spectra. We would readily detect an extension of 0.3 pixels, corresponding to 033, or 0.08 pc at the distance of the LMC. We have measured the radial velocities of the emission lines by fitting Gaussians to them, and find an average (heliocentric) value of 344 ± 9 km s⁻¹, where the error refers to the standard deviation of the mean; radial velocities for each of these features are given in Table 2. This value is significantly larger than 294 ± 2 km s⁻¹ we measure for the star itself (Section 3). It is possible that the redshift of these emission lines can be explained in part by scattering of the spectrum by an extended, expanding dust shell, a phenomenon that has previously been observed and described in RSGs (Romanik & Leung 1981, and references therein).

Zoom In Zoom Out Reset image size

Table 1. Physical Parameters of WOH G64

Reference	Spectral Type	Teff (K)	$M_{\rm bol} \left({\rm log} \frac{L}{L_{\odot }}\right)$	$\frac{R}{R_{\odot }}$	AV	log g
Elias et al. (1986)	M7.5	⋅⋅⋅	−9.6 (5.75)	⋅⋅⋅	∼3	⋅⋅⋅
van Loon et al. (2005)	M5, M7–8a	3008	−9.5 (5.70)	2575b	⋅⋅⋅	⋅⋅⋅
Ohnaka et al. (2008)	⋅⋅⋅	3200c	−8.9 (5.45)	1730	∼10d	⋅⋅⋅
Ohnaka et al. (2008)	⋅⋅⋅	3400c	−8.9 (5.45)e	1730e	?e	⋅⋅⋅
This work, (V − K)0 colors	M5	3372	−9.4 (5.65)	1990	6.82	−0.7
This work, spectral fitting	M5	3400	−9.4 (5.65)	1970	6.82	−0.7
This work, final preferred parameters	M5	3400	−8.9 (5.45)	1540	6.82	−0.5

Notes. ^aDifferent spectral types determined based on different measured line depths. ^bCalculated based on listed M_bol and T_eff. ^cAdopted based on previously assigned spectral types. ^dFormally 9.8 ± 2.2, but the high value comes from assuming a blackbody distribution for the star with a low T_eff; see the text. ^eAdopted based on reportedly good agreement with parameters derived from T_eff = 3200 K; Ohnaka et al. (2008).

Download table as:  ASCIITypeset image

We measure the integrated fluxes of these emission lines, and correct these values for extinction based on the Hα/Hβ emission line ratio, assuming the Balmer decrement for case B recombination (Hα/Hβ = 2.87, following Osterbrock 1989) and the Cardelli et al. (1989) reddening law. We calculate an A_V = 1.90, very different than the value determined for the star from our spectral fitting; however, it is not surprising that the gas we observe has a net extinction that is lower than the star, and this is in fact consistent with it forming in the outer part of the dust shell. The observed and dereddened integrated fluxes are included in Table 2.

With these dereddened fluxes, we are able to calculate several common emission line diagnostic ratios, notable log([N ii] λ6584/Hα) = 0.05, log([O i] λ6300/Hα) = −0.42, and log([O iii] λ5007/Hβ) = −0.58), and a lower limit of log([S ii]/Hα) = −0.45 (see Baldwin et al. 1981 for a detailed discussion of these diagnostics). The value of the [O i]/Hα ratio is much higher than the minimum value at which [O i] is considered "present," defined in Baldwin et al. (1981) as log([O i] λ6300/Hα) = −1.3. This high relative strength of the [O i] line implies that the dominant source of ionization is likely shock heating. This is also implied by the relatively high lower limit of the log([S ii]/Hα) = −0.45 ratio in our spectrum (see, for example, Allen et al. 1999, 2008), and we thank the anonymous referee for pointing this out. However, this is not the only explanation, as called to our attention in thoughtful comments by N. Smith (2009, private communication) and P. Bennett (2009, private communication). Bennett argues in particular that such nebular emission is typical of VV Cep-like systems, i.e., formed by ionization by a hot companion, but see discussion below.

We also find that nitrogen is highly enhanced in the nebula; if we use the standard galaxy diagnostics of Pettini & Pagel (2004), which rely upon a constant N/O ratio, we would derive a 12 + log(O/H) = 8.9, considerably in excess of the 12 + log(O/ H) = 8.4 that is typical of H ii regions in the LMC (Russell & Dopita 1990). But, far more likely is the possibility that N is simply enhanced in this gas. Without other line diagnostics (such as [O ii] λ3727, [O iii] λ4363, or [S ii] λ6717), we lack a quantitative knowledge of the oxygen abundance, electron density, and electron temperature. However, such N enrichment is a signature of many shells and rings around massive stars, such as Sher 25 (Brandner et al. 1997; Hendry et al. 2008). Sher 25 is a blue supergiant, but is thought to be in a blue loop phase, with the high N-enriched gas ejected during a prior evolutionary phase (Brandner et al. 1997; see also discussion in Smartt et al. 2002; Hendry et al. 2008). Other such examples are also known (Hunter et al. 2008). We note that [N i] λλ5198, 5200 is strongly present. Like the [O i] λ6300 line, this doublet is also collisionally excited (Osterbrock 1989), and we take its strength as further evidence of shocks and nitrogen enrichment, consistent with the above.

We have used the MARCS stellar atmospheres and moderate-resolution spectrophotometry and broadband photometry to investigate the physical properties of the enigmatic supergiant WOH G64. In addition, our study conclusively shows that the star is a supergiant member of the LMC, rather than (say) an unusual giant star in our own galaxy's halo.

Our analysis finds an effective temperature of 3400 ± 25 K, considerably warmer than previous studies have adopted. With our parameters from spectral fitting, and taking into account the additional flux contribution at K from the asymmetric circumstellar dust envelope as seen in Ohnaka et al. (2008), we determine M_bol = −8.9 for WOH G64. While WOH G64 has previously been cited as having a T_eff and M_bol that would make it the largest RSG known, those parameters have also placed it at odds with the position of the evolutionary tracks on the H–R diagram. By contrast, our parameters place WOH G64 closer to the predictions of stellar evolutionary theory and in good agreement with other RSGs in the LMC. We determine a stellar radius of R/R_☉ = 1540 which, while still the largest in the LMC, is now in good agreement with the size of other large RSGs (∼1420–1520 R_☉; Paper I), given that such radii have 5% errors or larger.

The question still remains as to whether or not WOH G64 is a Levesque–Massey variable. Its extreme physical properties and position on the H–R diagram are similar to those observed in Levesque–Massey variables. However, there is currently no evidence for spectral variability based on past observations; spectra obtained by Elias et al. (1986), van Loon et al. (2005), and this work all show WOH G64 remaining in a consistent, cool, late-M-type state. The disagreement in type between M5 and M7–8 is almost certainly a result of differing methods rather than changes in the spectrum. However, more frequent observations of WOH G64 could well uncover variations in its V magnitude and T_eff; data from the All Sky Automated Survey (ASAS) project (Pojmanski 2002) demonstrate that the star's I magnitude shows considerable variability, but there is no V- or K-band data available. On the other hand, the star's very high reddening, well-defined dusty torus, and considerable maser activity would be unique among the Levesque–Massey variables; these properties suggest WOH G64's similarity to the Milky Way's VY CMa, as well as other dust-enshrouded RSGs with maser activity such as NML Cyg, VX Sgr, and S Per (Schuster et al. 2006).

Finally, we detect a number of emission lines in the WOH G64 optical spectrum, including Hα and Hβ, [O i], [N i], [S ii], [N ii], and [O iii]. The radial velocity of these features, 344 ± 9 km s⁻¹, is considerably greater than the star's radial velocity, 294 ± 2 km s⁻¹, which may be due to scattering by the expanding dust region. We also note the presence of some TiO band heads in emission. This has been previously seen in the spectrum of VY CMa, which is due to the extreme circumstellar environment. Inspection of our data shows that the emission is not spatially extended, and the strength of the [O i] feature suggests that the dominant source of ionization for these features is shock heating.

We cannot reject the possibility that WOH G64 is a binary system, with the nebular component contributed by a hot companion, but we do not favor this interpretation. First is the issue of the nebular radial velocities. We find a significant (50 km s⁻¹) velocity difference between the nebular emission and the stellar component, and, at first blush, this might be taken as an indication of binarity; i.e., with the nebula emission (somehow) tracking the motion of the hot component. However, a similar velocity was also noted by Elias et al. (1986). While this could be a coincidence, it certainly does not argue in favor of a binary hypothesis. Besides, in a short-period system we would not expect the ionized gas to be "attached" to the hot star. Rather, we prefer the explanation that the nebular redshift is the result of scattering within the expanding dust shell. We do note that if the putative hot companion shared the extinction measured from the nebular lines, rather than that of the RSG, its flux would dominate even in the red part of the spectrum, which is clearly not the case. If instead the companion shared the same high extinction as the RSG, but its H ii region did not, we would still expect that the companion would dominate the flux in the blue part of the spectrum. A better spectrum going further into the blue should be able to substantiate or refute the possibility that a hot companion is present with the same extinction, as a late O-type dwarf would have a bolometric luminosity of log L/L_☉ ∼ 5 (M_bol ∼ −7.5). Accounting for the bolometric corrections involved would lead to M_V = −4 for the companion, and −6 for the RSG. Thus, at B, the two stars would be equally bright, as (B − V)₀ ∼ −0.3 for a late-type O star, while B − V ∼ 2 for an RSG (see Massey 1998a, 1998b; Paper I). However, further investigation into the source of these emission lines will likely require further spectral analyses, as well as observations of WOH G64 in the bluer regions of the spectrum.

We gratefully acknowledge LCO's hospitality and assistance provided during our observations for this paper. This paper made use of data from the 2MASS, which is a joint project of the University of Massachusetts and the Infrared Processes and Analysis Center, California Institute of Technology, funded by the National Aeronautics and Space Administration and the National Science Foundation. We thank Phil Bennett, George Herbig, Deirdre Hunter, Eric Josselin, Scott Kenyon John Monnier, Keiichi Ohnaka, Nathan Smith, Jacco van Loon, George Wallerstein, and the anonymous referee for helpful comments and discussion regarding this manuscript. E.L.'s participation was made possible in part by a Ford Foundation Predoctoral Fellowship. This work was supported by the National Science Foundation through AST-0604569 to P.M.