Mass Outflows in RY Scuti

# A Spectroscopic Study of Mass Outflows in the Interacting Binary RY Scuti

Erika D. Grundstrom, Douglas R. Gies11affiliation: Visiting Astronomer, Kitt Peak National Observatory and Cerro Tololo Interamerican Observatory, National Optical Astronomy Observatory, operated by the Association of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation. , Todd C. Hillwig11affiliation: Visiting Astronomer, Kitt Peak National Observatory and Cerro Tololo Interamerican Observatory, National Optical Astronomy Observatory, operated by the Association of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation. 22affiliation: Current address: Department of Physics and Astronomy, Valparaiso University, Valparaiso, IN. , and M. Virginia McSwain11affiliation: Visiting Astronomer, Kitt Peak National Observatory and Cerro Tololo Interamerican Observatory, National Optical Astronomy Observatory, operated by the Association of Universities for Research in Astronomy (AURA), Inc., under cooperative agreement with the National Science Foundation. 33affiliation: Current address: Department of Astronomy, Yale University, New Haven, CT. 44affiliation: NSF Astronomy and Astrophysics Postdoctoral Fellow. Center for High Angular Resolution Astronomy and
Department of Physics and Astronomy, Georgia State University, Atlanta, GA; erika@chara.gsu.edu, gies@chara.gsu.edu, todd.hillwig@valpo.edu, mcswain@astro.yale.edu
Nathan Smith Astronomy Department, University of California, Berkeley, CA; nathans@astro.berkeley.edu Robert D. Gehrz Astronomy Department, University of Minnesota, Minneapolis, MN; gehrz@astro.umn.edu Otmar Stahl Landessternwarte Heidelberg, Heidelberg, Germany; o.stahl@lsw.uni-heidelberg.de Andreas Kaufer European Southern Observatory, Santiago, Chile; akaufer@eso.org
###### Abstract

The massive interacting binary RY Scuti is an important representative of an active mass-transferring system that is changing before our eyes and which may be an example of the formation of a Wolf-Rayet star through tidal stripping. Utilizing new and previously published spectra, we present examples of how a number of illustrative absorption and emission features vary during the binary orbit. We identify spectral features associated with each component, calculate a new, double-lined spectroscopic binary orbit, and find masses of for the bright supergiant and for the hidden massive companion. Through tomographic reconstruction of the component spectra from the composite spectra, we confirm the O9.7 Ibpe spectral class of the bright supergiant and discover a B0.5 I spectrum associated with the hidden massive companion; however, we suggest that the latter is actually the spectrum of the photosphere of the accretion torus immediately surrounding the massive companion. We describe the complex nature of the mass loss flows from the system in the context of recent hydrodynamical models for Lyr, leading us to conclude RY Scuti has matter leaving the system in two ways: 1) a bipolar outflow from winds generated by the hidden massive companion, and 2) an outflow from the bright O9.7 Ibpe supergiant in the region near the L2 point to fill out a large, dense circumbinary disk. This circumbinary disk (radius  1 AU) may feed the surrounding double-toroidal nebula (radius  2000 AU).

binaries: close — binaries: eclipsing — circumstellar matter — stars: individual (RY Scuti) — stars: mass loss — stars: winds, outflows
slugcomment: Submitted to ApJ

## 1 Introduction

RY Scuti (HD 169515) is a distant (1.8 0.1 kpc; Smith et al. 2002) and massive eclipsing binary system with an orbital period currently estimated at 11.12445 days (Kreiner, 2004). Analysis of the light curve indicates that at least one of the components fills its Roche lobe. The present configuration may be semi-detached (Cowley & Hutchings, 1976), overcontact (Milano et al., 1981; Djurasĕvić, Eshankulova, & Erkapić, 2001), or one with the more massive component embedded in an opaque thick disk (King & Jameson, 1979; Antokhina & Kumsiashvili, 1999). The binary appears to be in an advanced stage of evolution, and it has ejected gas into a young (130 year old; Smith, Gehrz, & Goss 2001), double-toroidal emission nebula. The 2000 AU nebula is angularly resolved in radio images (Gehrz et al., 1995), infrared images (Gehrz et al., 2001), and in Hubble Space Telescope H images (Smith et al., 1999, 2001, 2002). Since the nebula is so young and the star system is still in an active Roche-lobe overflow phase, RY Scuti is a powerful laboratory for studying non-spherical mass and angular momentum loss in interacting binaries.

Only one of the component stars is readily visible in optical spectra and it has a spectral classification of O9.7 Ibpe var (Walborn, 1982). We will refer to this component as the “supergiant.” The other star appears to be enshrouded since it is very difficult to detect in the spectrum. Most investigators agree that this second star is the more massive of the two, and we will refer to the second component as the “massive companion.” However, the actual masses of the stars in this system are debatable. The radial velocity curve of the supergiant is reasonably well established, but the results for the massive companion depend critically on what spectral features one assumes are associated with that star. As these features are difficult to observe and interpret, the estimated mass range has been huge. For example, Popper (1943) arrived at a total system mass in excess of . Later investigators found lower values: Cowley & Hutchings (1976) estimated masses of 36 and , Skul’skii (1992) found 8 and , and Sahade, West, & Skul’sky (2002) estimated 9 and (for the supergiant and massive companion, respectively, in each case). There are a number of important photometric studies (e.g., ranging from the discovery by Gaposchkin 1937 through photoelectric investigations by Giuricin & Mardirossian 1981 and Milano et al. 1981, and up to the most recent multi-color work by Djurasĕvić et al. 2001), however the results differ with regard to the assumed binary configuration and depend sensitively on the mass ratio adopted from spectroscopy.

This unique system is representative of the short-lived, active mass transfer stage in the evolution of massive binaries. Theoretical models (Petrovic, Langer, & van der Hucht, 2005) indicate most of the mass transfer occurs during a brief ( yr), and thus rare, phase in which the mass donor transfers most of its mass to the mass gainer star. Mass transfer will shrink the orbital dimensions until the mass ratio is reversed (and the gainer becomes the more massive star), and then the system will enter a slower (and longer lived) mass transfer phase as the binary expands. The massive binaries that are just emerging from the rapid mass transfer phase probably belong to the observed class of W Serpentis binaries (Tarasov, 2000). Only the mass donor star is visible in the spectra of these binaries and the more massive gainer star is hidden in a thick accretion disk (one source of emission lines in the spectra). The mass transfer process is complex and leaky, and a significant fraction of the mass loss leaves the system completely (as described by Harmanec 2002 for the best known object of the class,  Lyr). The mass donor may eventually lose its entire hydrogen envelope and emerge as a Wolf-Rayet star. Therefore a system like RY Scuti may be the progenitor of a WR+O binary system (Giuricin & Mardirossian, 1981; Antokhina & Cherepashchuk, 1988; Smith et al., 2002).

In a prior paper (Smith et al., 2002), several of us presented a detailed study of the spectral features formed in the surrounding double-toroidal nebula through an examination of a set of high dispersion spectra obtained with the ESO FEROS spectrograph. Here we use the same set of spectra supplemented by additional optical spectra to explore the spectral features associated with the central binary and its immediate circumstellar environment. Our primary goal is to determine how the binary ejects the gas that ultimately flows into the dense outer double-toroidal nebula. We describe our observations and data reduction methods in §2 and we present in §3 examples of the orbital phase-related spectral variations we observed. The radial velocity curve and a new orbital solution for the supergiant mass donor are described in §4. Then in §5 we review the spectral clues about the nature of the enshrouded mass gainer, and we present a preliminary radial velocity curve and orbital solution for the massive companion. Both radial velocity curves are used in §6 to make a Doppler tomographic reconstruction of the optical spectra of the individual components. We describe in §7 a model for the mass outflows in RY Scuti that is based upon recent hydrodynamical simulations and that explains many of the observed spectral variations. Our results are summarized in §8.

## 2 Observations and Reductions

Our analysis is based upon spectra collected from three telescopes. The highest dispersion spectra (17 in number) were obtained in 1999 with the Fiber-fed Extended Range Optical Spectrograph (FEROS) mounted on the 1.52 m telescope of the European Southern Observatory (ESO) at La Silla, Chile (see Smith et al. 2002). Also in 1999, we obtained 40 moderate dispersion spectra (in the red region surrounding H) using the Kitt Peak National Observatory coudé feed 0.9 m telescope. Finally, in 2004 we used the CTIO 1.5 m telescope and Cassegrain spectrograph to obtain ten blue spectra of moderate resolution covering one orbital period. Table 1 contains run number, dates, spectral coverage, spectral resolving power, number of spectra, telescope, spectrograph grating, and CCD detector used in each case. Exposure times were generally limited to 30 minutes or less. Each set of observations was accompanied by numerous bias, flat-field, and ThAr comparison lamp calibration frames. Furthermore, we obtained multiple spectra each night of the rapidly rotating A-type star Aql for removal of atmospheric water vapor and O bands in the red spectra made with the KPNO coudé feed.

The spectra from all the telescopes were extracted and calibrated using standard routines in IRAF555IRAF is distributed by the National Optical Astronomical Observatory, which is operated by the Association of Universities for Research in Astronomy, Inc. (AURA), under cooperative agreement with the National Science Foundation.. All the spectra were rectified to a unit continuum by fitting line-free regions. We removed the atmospheric lines from the red coudé feed spectra by creating a library of Aql spectra from each run, removing the broad stellar features from these, and then dividing each target spectrum by the modified atmospheric spectrum that most closely matched the target spectrum in a selected region dominated by atmospheric absorptions. In a few cases this resulted in the introduction of spectral discontinuities near the atmospheric telluric lines, and these were excised by linear interpolation. We did not attempt any removal of atmospheric lines for the FEROS spectra, but some problem sections in the echelle-overlap regions were excised via linear interpolation. The spectra were then transformed to a common heliocentric wavelength grid for each of the FEROS, KPNO coudé feed, and CTIO 1.5 m runs. We show several examples of the final spectra in the next section.

## 3 Spectral Variations with Orbital Phase

The optical spectrum of RY Scuti is very complex and time variable, and in this section we describe the appearance and orbital phase-related variations of several representative line features. A depiction of the time-average of all the FEROS spectra in the range from 3600 to 9200 Å appears in Smith et al. (2002) (their Fig. 16). This illustration is ideal for identifying the many sharp, quadruple-peaked, emission lines formed in the surrounding double-toroidal nebula, but photospheric lines appear broad and shallow as the average was made over the full range of orbital Doppler shifts. A large fraction of the circumstellar emission lines are species that do not have a stellar absorption counterpart (such as [Fe iii] and many other forbidden lines), but there are several emission lines that are superimposed upon important stellar spectral features (for example, in the H Balmer and He i lines). Since the nebular features are analyzed in great detail elsewhere (Golovatyi & Skul’skii, 1992; Skul’skii & West, 1993; Smith et al., 2002), we will not discuss them here. We will focus on those spectral features formed in the photospheres of the stellar components and in the rapidly moving gas immediately surrounding them. Here we introduce the different kinds of patterns of variability observed, and in the following sections we analyze in detail the Doppler velocity shifts related to features associated with the supergiant (§4) and its massive companion (§5).

We begin with some examples from the high resolution FEROS spectra. The Si iv absorption feature is one of only a small number of spectral lines formed in the photosphere of the supergiant that is not filled in or affected by nebular emission. We show the orbital phase variations of this line (and the nearby N iii and H lines) in Figure 1 as a function of heliocentric radial velocity for Si iv . In this and the next figures we adopt the orbital period from Kreiner (2004) of  d and the epoch of phase zero as the supergiant superior conjunction at (derived in §4). The upper portion of Figure 1 shows the Si iv profiles with their continua aligned with the phase of observation (increasing downwards) while the lower portion shows the spectra as a gray-scale image interpolated in phase and velocity. Features moving with the radial velocity curve of the supergiant will have a characteristic “S” shape in this image. The break in the continuity of the “S” curve near phase is due to the unfortunate gap in our phase coverage near there and to the simplicity of the interpolation scheme. The Si iv feature appears to be useful for the radial velocity measurement of the supergiant, although we find that the depth of the line varies with phase, weakening at and strengthening at . There is no obvious evidence of a reverse “S” feature that would correspond to the motion of the massive companion (although a weak feature is present; §6).

The orbital variations in two helium lines, the He i singlet and the He i triplet, are illustrated in Figures 2 and 3, respectively. Both features show sharp nebular emission peaks from the surrounding double-torus (Smith et al., 2002) superimposed on the stellar absorption feature. Both lines show an absorption component that follows the radial velocity curve of the supergiant. However, the He i feature strengthens at both conjunctions and almost disappears at phase . It also shows a blueshifted absorption feature that appears at and lasts for nearly half the orbit. Other He i singlets () exhibit the same behavior. On the other hand, the He i triplet displays a much stronger blueshifted component between that develops into a very sharp and blueshifted ( km s) absorption line that lasts for greater than half the orbit. The other He i triplets () also show these same features.

Figure 4 shows the orbital variations in the triplet Si iii in the velocity frame of Si iii . The supergiant component appears to be present and undergoes the same kind of strengthening at conjunctions seen in the He i lines. However, each of the triplet members also show a broad shallow feature that moves in the manner expected for the massive companion. Thus, our observations confirm the detection of this second component that was discovered by Skul’skii (1992).

There are two very broad emission lines present that are important to the study of this binary system, H and He ii (Sahade et al., 2002). The H feature consists of a broad emission feature that is spatially coincident with the central binary plus a strong but narrow component formed in the double-toroidal nebula (Smith et al., 2002). Therefore, in order to isolate the emission component near the binary, we had to remove the nebular components of H and the nearby [N ii] lines. This removal process was done by scaling the [N ii] line in each spectrum to the appropriate size (of either H or [N ii] ) using the equivalent widths of these lines from Smith et al. (2002), shifting the rescaled line to the location of the line to be removed, and then subtracting it from the spectrum. This process was done interactively and included small adjustments in the scaling and shifting parameters to optimize the subtraction. The resulting subtracted H profiles based upon the large set of KPNO coudé feed spectra are illustrated as a function of orbital phase in the left panel of Figure 5. We also show the one spatially resolved HST spectrum of the central binary from the work of Smith et al. (2002) (which has no nebular emission present) that verifies that our line subtraction technique creates difference profiles with the appropriate shape. The emission strength appears to be much stronger at the conjunction phases, but this is due mainly to the drop in the continuum flux at those eclipse phases and our normalization of the emission strength to this varying continuum level. The right panel in Figure 5 shows a representation of the H profiles relative to a constant flux continuum we made by rescaling the emission flux by a factor

where is the -band magnitude at the orbital phase of observation (found by interpolation in the light curve data from Djurasĕvić et al. 2001) and corresponds to the maximum brightness of the system at quadrature phases. The H emission feature appears broad (spanning over 1000 km s) and approximately constant in strength and position in this version. There is often a weak, blueshifted absorption component present (with a radial velocity of km s) that gives the profile a P Cygni appearance. Similar results were seen in the H difference profiles formed from the smaller set of FEROS spectra.

Finally we show in Figure 6 the orbital phase variations in the weak He ii emission line observed in the FEROS spectra that we smoothed with a Gaussian of FWHM = 45 km s to improve the otherwise noisy appearance. This feature is found only in very hot plasmas, and there is no corresponding nebular feature in this case. The He ii emission is almost as broad as the net H emission, but is significantly weaker. The wings of the line appear to exhibit a slight anti-phase motion that was documented by Sahade et al. (2002).

## 4 Radial Velocity Curve of the Supergiant

The radial velocity shifts of the supergiant are readily apparent in many lines, and given the quality and number of the new spectra we decided to measure the radial velocities and reassess the orbital elements. We measured relative radial velocities by cross-correlating each spectrum with a template spectrum. For the FEROS and CTIO spectra, this template spectrum was generated from the non-LTE, line-blanketed model atmosphere and synthetic spectra grid from Lanz & Hubeny (2003) using kK and , which are appropriate values for an O9.7 Ib star (Herrero et al., 1995). This template was rotationally broadened (§5) and also smoothed to the instrumental resolution of the FEROS or CTIO spectra. First, we removed certain interstellar features by forming an average interstellar spectrum from the mean of the entire set and then dividing each spectrum by the average interstellar spectrum. Because the spectrum of RY Scuti contains so many stationary nebular lines and possible features from the massive companion, we restricted the wavelength range for the cross-correlation to regions surrounding a set of absorption lines that appeared to be free of line blending and that were clearly visible throughout the orbit. The main features in the selected regions are summarized in Table 2.

For the KPNO coudé feed spectra, we selected one spectrum from the set of 40 to serve as the template since the Lanz & Hubeny models did not match the emission lines we measured (Table 2). This spectrum (made on HJD 2451425.784) has relatively high signal-to-noise, exhibits well-defined spectral features, and was made near quadrature when the supergiant Doppler shifts are large and the features reasonably well separated from any component from the massive companion. Once again, we first removed interstellar features then selected regions free from nebular lines for the cross-correlation. Once we performed the cross-correlation, we determined the absolute radial velocity of the template spectrum by fitting the core of the emission lines (N ii and Si iv ) with a parabola and determined the shift from the rest wavelengths. We added the radial velocity of the template to all the relative velocities to place them on an absolute scale.

Since the velocities from the different runs are based upon measurements of different lines in the template and individual spectra, we anticipated that there might be systematic differences in the zero-points of each set. We began by making independent circular orbital fits for each set, and indeed we found the systemic velocity for the FEROS set was offset by km s from the resulting systemic velocities for the KPNO and CTIO sets. Since the latter were closer to the nebular systemic velocity ( km s; Smith et al. 2002) and based upon more observations, we arbitrarily adjusted the FEROS measurements by adding km s to bring them into consistency with the other measurements. Note that this decision adds an additional uncertainty to the real error in the final determination of the systemic velocity but otherwise has no effect on the other orbital elements. Our final radial velocities are collected in Table 3 that lists the date of observation, orbital phase (see below), radial velocity, calculated error in radial velocity, observed minus calculated residual, and the telescope where the spectrum was made (Table 1). The quoted radial velocity error for the blue spectra is the standard deviation of the cross-correlation measurements from the three line regions. We measured only two line regions in the red spectra, so the velocity error is estimated as the larger of the difference between the two measurements or the mean value of from closely spaced pairs of observations.

We determined revised orbital elements using the the nonlinear, least-squares, fitting code of Morbey & Brosterhus (1974). Since the errors associated with each run are different, we assigned each measurement a weight proportional to the inverse square of the measurement error. The orbital period was fixed at days as found by Kreiner (2004) from contemporary eclipse timing observations over a long time base. We made both eccentric and circular orbital solutions, but we think the eccentric solution, while formally statistically significant according to the test of Lucy & Sweeney (1971), is probably spurious. Harmanec (1987) found that gas streams and circumstellar matter (both present in RY Scuti) can distort spectroscopic features in mass-transferring binaries. Such distortions can lead to skewed radial velocity results and thus artificial eccentricities (Lucy, 2005). A zero eccentricity is consistent with predictions for Roche lobe overflow systems where the tidal effects are expected to circularize the orbit and synchronize the rotational and orbital periods. The final orbital elements from both the circular and eccentric solutions are presented in Table 4 together with solutions from Sahade et al. (2002) and Skul’skii (1992). The epoch refers to the time of supergiant superior conjunction (close to the time of photometric minimum light) while is the epoch of periastron in the eccentric solution. The radial velocities and orbital velocity curves are plotted in Figure 7. Our results are in reasonable agreement with earlier determinations of the elements, with the possible exception of the larger non-zero eccentricity found by Sahade et al. (2002). We suspect the difference is due to the lack of orbital phase coverage in the observations of Sahade et al. (2002) in the ascending portion of the velocity curve.

## 5 Nature of the Massive Companion

Very little is known about the massive companion because it is difficult to find its associated features in the spectrum (probably because the massive companion is enshrouded in a thick accretion disk; §7). We can make an approximate estimate for the expected semiamplitude of motion for the massive companion using a geometrical argument to find the mass ratio, . Let us assume the supergiant is filling its Roche lobe and is synchronously rotating (reasonable in the case of RY Scuti where we see so much spectroscopic evidence of active mass transfer). Gies & Bolton (1986) showed that for such Roche-filling stars, the projected rotational velocity provides a measure of the stellar radius while the semiamplitude is related to the mass ratio and semimajor axis , and the ratio of these quantities is a function that solely depends on the mass ratio,

 Vsini/K = ΩRSGsiniΩaSGsini = (1+MSGMMC)RSGa=(1+Q)f(Q)

where is the orbital angular rotation speed, , and is the fractional Roche filling radius (Eggleton, 1983). Thus, a measurement of the supergiant’s projected rotational velocity leads directly to an estimate of the mass ratio. We show below that this geometrical mass ratio estimate is consistent with the mass ratio derived from orbital velocity measurements for both components.

We determined for the supergiant through a comparison of the photospheric line width with models for a grid of projected rotational velocity. For this purpose, we selected the Si iv line since it is free of nebular emission and line blending with other features and its shape is dominated by rotational broadening. We measured the FWHM of Si iv in three of the FEROS spectra obtained near supergiant maximum velocity (to avoid the unusual strengthening seen at conjunctions and any line blending problems with absorption from the massive companion). We then used the model profiles from the non-LTE, line blanketed models of Lanz & Hubeny (2003) that were rotationally broadened by a simple convolution of the zero-rotation model profiles with a rotational broadening function (Gray, 1992) using a linear limb-darkening coefficient from the tables of Wade & Rucinski (1985). From the Lanz & Hubeny grid, we selected a model profile for  kK and and adopted a linear limb darkening coefficient at this wavelength of . Finally, the resulting models were convolved with an instrumental broadening function to match the resolution of the FEROS spectra. The projected rotational velocity for the supergiant derived from the resulting relation is km s. The error represents the standard deviation of as derived from the three measurements of line width and does not account for any systematic errors associated with the choice of model atmosphere parameters.

We caution that this is actually an upper limit for since the line may also be broadened by macroturbulence in the stellar atmosphere (Ryans et al., 2002). If we include an estimate of km s (a mid-range value for B-supergiants; Ryans et al. 2002), then we find km s (where the error range accounts for the derived error in and a range in macroturbulent velocity from 10 to 50 km s). Thus, using the method described above, we find an estimate of the mass ratio, . We can use this mass ratio estimate to offer some guidance about the expected orbital semiamplitude of the massive companion, km s.

Previous researchers (Popper, 1943; Cowley & Hutchings, 1976; Skul’skii, 1992; Sahade et al., 2002) have associated a variety of spectral features with the massive companion, and we inspected the orbital variations of all these proposed lines in the FEROS spectra (examples shown in Fig. 14). Many of these lines have complicating factors related to nebular emission and blending with other features. For example, a number of the He i lines appear to show a blueshifted absorption feature in the phase range where we would expect to find the absorption component from the massive companion, but these blueshifted features first appear at when any gas leaving the supergiant in the L2 region would also appear blueshifted (§7). Consequently, an identification of these features with the massive companion is ambiguous.

Only three of the proposed spectral features for the massive companion clearly fit the mass ratio argument described above and display the expected anti-phase velocity shifts: H, He ii and Si iii . Here we present radial velocity measurements for each of these. The H emission profile is not Gaussian in shape (see Fig. 5) and therefore could not be measured as such. Instead we determined the radial velocity of the line wings (which represent the fastest-moving gas and are unaffected by nebular emission) based upon a bisector position found using the method of Shafter, Szkody, & Thorstensen (1986). We sampled the line wings using oppositely signed Gaussian functions and determined the mid-point position between the wings by cross-correlating these Gaussians with the profile. We used Gaussian functions with FWHM = 137 km s at sample positions in the wings of km s for spectra from both the coudé feed and the FEROS instrument. An inspection of radial velocity measurements closely spaced in time indicates an average error of 1.7 km s. Our results are given in column 3 of Table 5, and the values are plotted as a function of orbital phase in Figure 7.

The He ii emission profile is also broad (FWHM km s) and not Gaussian-shaped. Therefore, we used spectral templates and cross-correlation functions to measure radial velocities. We formed templates from the averages of the entire run for both the FEROS and CTIO data sets since the He ii emission strength may have changed between 1999 (FEROS) and 2004 (CTIO). We then transformed the relative velocities to an absolute scale by adding the radial velocity derived by fitting each template with a broad Gaussian. Our results are given in column 4 of Table 6 and are plotted in Figure 7. Based on visual estimates of the goodness of the fit of this very broad line, we estimate errors in the velocities are 50 km s.

We also measured equivalent widths for both these emission lines by a direct numerical integration (over the range  Å for H and  Å for He ii ). Our results for H are given in column 4 of Table 5 and for He ii are given in column 5 of Table 6. Errors in H measurements are 0.5 Å based upon examining results closely spaced in time. The He ii equivalent width errors are 0.02 Å for the FEROS data but are about three times worse for the CTIO data (due to the lower resolution of the CTIO spectra). The orbital variations of these equivalent widths are discussed below (§7).

The absorption line Si iii displays narrow and broad components that move with the expected orbital Doppler shifts of the supergiant and massive component, respectively (Fig. 4). In order to avoid any line blending problems between these components, we selected a subset of FEROS and CTIO spectra observed near the quadrature phases where the the broad component could be measured unambiguously, and we made Gaussian fits to determine the radial velocities. Using a visual estimate of goodness of fit to this broad, shallow line, we estimate errors in these velocities are 30 km s. Our results are given in column 3 of Table 6 and are plotted in Figure 7.

The plots of the radial velocities in Figure 7 show that all three features exhibit the anti-phase motion expected for the massive companion. We made circular fits of each set by fixing the orbital period and epoch from the solution for the supergiant (Table 4) and then solving for the semiamplitude and systemic velocity . The results are listed in Table 7. The semiamplitudes for each feature are within the expected range, however, the values for the two emission lines are smaller than that for the Si iii absorption line. Furthermore, the systemic velocities for the two emission lines are larger than that found for the supergiant and the nebular emission. We suspect this is due to the P Cygni shape of both features that will bias an emission line velocity measurement to a larger value. We argue below (§7) that these emission features are probably formed in gas flows and are thus probably less representative of the orbital motion of the massive companion.

Thus, if we adopt the Si iii velocity fits as the most representative of the massive companion, we can make a double-lined solution of the spectroscopic masses, , using eq. 2.52 in Hilditch (2001),

 M1,2sin3i=(1.0361×10−7)(1−e2)3/2(K1+K2)2K2,1P M⊙,

where is the semiamplitude in km s and is the period in days. Furthermore, if we set the orbital inclination equal to that for the double-toroidal nebula, then (Smith et al., 1999) (in substantial agreement with the light curve solutions; Milano et al. 1981; Djurasĕvić et al. 2001), and we can estimate the component masses directly. We find the masses of the components are and . The former is very low for a normal O-supergiant (Martins, Schaerer, & Hillier, 2005), but we should bear in mind this star is in the process of losing its outer envelope to the massive companion and therefore its properties will be very different from those of a single star of comparable temperature and gravity. The mass of the massive companion is like that of a hot O-dwarf or an evolved B-supergiant and we will explore these and other possibilities in the next sections. Our mass results are quite similar to those first obtained by Skul’skii (1992) but are lower than those determined by Sahade et al. (2002) (given in Table 4).

## 6 Tomographic Reconstruction of Spectra

Once the orbital solution for RY Scuti was found, we used a tomographic reconstruction technique (Bagnuolo et al., 1994) with the FEROS data to separate the individual spectra of the components. Tomographic reconstruction is an iterative scheme that uses the combined spectra and their associated radial velocities to determine the appearance of each star’s spectrum. RY Scuti presents a special difficulty due to the stationary sharp nebular features. Therefore, before reconstruction, all nebular features listed by Smith et al. (2002) were excised via linear interpolation. We also removed the interstellar lines from each spectrum prior to reconstruction to avoid creating spurious reconstructed features in their vicinity. The reconstruction was based upon a subset of seven FEROS spectra that were obtained near the velocity extrema at the quadrature phases so that we might avoid introducing artifacts due to the line strengthening at the conjunctions and any eclipse effects. Figures 8 and 9 show plots of the reconstructed spectra in two different regions along with identifications of the principal lines. Also plotted are two comparison spectra from the Valdes et al. (2004) Indo-U.S. Library of Coudé Feed Stellar Spectra that have a lower resolving power () than that of the FEROS spectra ().

The supergiant has a classification of O9.7 Ibpe var (Walborn, 1982) and we see the spectrum of the normal single star, HD 188209 (O9.5 Iab; Walborn 1976), provides a reasonably good match. The He ii features appear to be slightly weaker in the RY Scuti supergiant, which is consistent with its subtype difference in spectral type. We used the ratios of the equivalent widths of several weaker He i lines for the RY Scuti supergiant and HD 188209 spectra to set the monochromatic flux ratio for the RY Scuti binary components, and we find the supergiant contributes of the total flux. The error is based on scatter between the results from different spectral features. The reconstructed spectra are plotted in Figures 8 and 9 normalized to their respective continuum fluxes, so individual lines appear deeper than in the composite spectra where each is diluted by the flux of the other star. There are several obvious differences between the supergiant component and single star spectrum in some of the stronger lines such as H and H but we do not ascribe any particular significance to these as such features are clearly distorted by emission from circumstellar gas. On the other hand, there does appear to be a significant weakness in the C lines and an enhancement in the N lines in the spectrum of the RY Scuti supergiant compared to that of HD 188209. For example, the C iii lines that are present in the spectrum of HD 188209 are conspicuously absent in the RY Scuti supergiant’s spectrum. This suggests the atmosphere of the supergiant is enhanced with CNO-processed gas (as is the surrounding nebula; Smith et al. 2002). Perhaps this is not surprising given that a large portion of the supergiant’s mass must have already been lost to reveal gas from deeper layers closer to the source of core H-fusion. This result suggests that the bright O9.7 supergiant is the ultimate source of the processed gas that now forms the outer double-toroidal nebula.

Skul’skii (1992) suggested the spectral features moving with Doppler shifts of the massive companion have the appearance of a B2 star. We find the lines in the reconstruction for the orbital shifts of the massive companion more closely resemble that of the B0.5 Ia star, HD 185859 (Morgan, Code, & Whitford, 1955). However, there are several features in the reconstruction that make us caution against assuming this spectrum forms in the photosphere of a supergiant star. First, the luminosity sensitive lines of Si iv and Si iii appear very strong in the reconstructed spectrum relative to those in HD 185859 which would imply the lines form in a very low density plasma (like those in the photospheres of the most luminous stars; Walborn & Fitzpatrick 1990). Secondly, if we suppose that the gravity of a B0.5 Ia star is (see the case of the like star  Ori; McErlean, Lennon, & Dufton 1999), then using our derived mass, the massive companion star would have a radius of , which is 83% of its Roche radius. A radius this big would have two observational consequences: first, the flux from this component would be twice the flux from the bright 09.7 supergiant, and second, a star this large would cause the spectral lines of the bright 09.7 supergiant to weaken or disappear during primary eclipse (). Neither of these two predictions are seen: the massive companion flux is less than that of the supergiant and the supergiant’s absorption lines do not weaken at primary eclipse (Fig. 14). Also, we estimate the projected rotational velocity of the massive companion spectrum by measuring the widths of the Si iii feature in five FEROS spectra where the feature was well separated from any component of the bright supergiant. We used these widths to obtain km s in the same way as described above for the primary (§5) by comparing them with rotationally broadened model spectra for  K, , and a limb darkening coefficient . This result is unreasonably large since Ryans et al. (2002) find that no early B-type supergiant has a greater than 60 km s. Therefore, we suggest these characteristics are best explained if we place the origin of the spectrum in the photosphere of a thick accretion disk surrounding the massive companion rather than in the photosphere of the star itself and we explore this idea further in the next section. Note that the very broad He ii emission is assigned to the massive companion in the reconstruction, but we argue below that its site of origin may not be exactly co-spatial with that of the absorption line spectrum.

## 7 Mass Outflows in RY Scuti

Many of the spectral features discussed above originate in gas outflows from the binary, and in this section we explore their sites of origin with reference to recent hydrodynamical simulations of gas flows in Roche lobe overflow binaries. The luminous supergiant in RY Scuti can potentially be losing mass by a radiatively driven wind and/or tidal streams along the axis joining the stars. Friend & Castor (1982) show that a continuum of states can exist between these idealized cases. When a luminous star has a radius that is significantly smaller than the critical Roche lobe, the wind will be fast and spherically symmetric (as for single stars). As the ratio of stellar to Roche radius increases, the star becomes tidally distorted and its wind becomes asymmetric with a higher mass loss rate and slower outflow along the axis joining the stars (a “focused wind”). Finally, once the star essentially attains a size filling the Roche surface, most of the mass loss will occur in slow outflows directed along the axis joining the stars. All of the photometric light curve studies confirm the supergiant is very tidally distorted and must therefore be close to Roche filling. Thus, we will explore the consequences of tidally formed outflows in reviewing the observational clues about the mass loss.

The discussion begins with the dimensions of the system, and in Figure 10 we present a cartoon illustration of the cross section of the system as viewed from above the orbital plane. The direction of the observer at different orbital phases is indicated by numbers in the periphery of the diagram. As before, we assume the supergiant fills its critical Roche surface and thus has an equivalent volume radius of approximately (Eggleton, 1983). The supergiant is depicted on the left side of Figure 10. If we assume the massive companion is a main sequence star, then for its derived mass we expect it is an O6.5 V star with a polar radius of approximately (see Table 4 in Martins et al. 2005). The equatorial radius could be significantly larger if the star is rapidly rotating, but we assume it is a spherical star in its depiction on the right hand side of Figure 10. Either way, the radius of the massive companion is probably significantly less than its critical Roche radius, and therefore we begin our analysis assuming that the system is semi-detached.

The kinds of gas outflows expected in systems like RY Scuti are explored in a series of recent papers by Nazarenko & Glazunova (2003, 2006a, 2006b). They present hydrodynamical simulations in two and three-dimensions to model the case of the interacting binary  Lyr (Harmanec et al., 1996; Harmanec, 2002). RY Scuti and  Lyr share many of the properties common to the W Serpentis class, and we can obtain considerable insight about the gas flows in RY Scuti from these papers (although we caution that the stars in RY Scuti are both hotter and more massive than those in  Lyr). These numerical models assume the mass donor is a low gravity object that fills its Roche surface, and the authors use the models to follow the time evolution of the gas flows until they reach a semi-static configuration. There are several features of these models that are relevant to the case at hand. First, Nazarenko & Glazunova find that the donor loses mass through both the inner L1 region and the outer L2 region (facing away from the companion). Because the gas is so loosely bound to the donor, these gas streams are not compressed into a classical stream but instead fan out across a significant portion of the stellar surface facing towards and away from the companion. The approximate ranges of these outflows are indicated by the dashed lines to the left and right of the supergiant in Figure 10 and are based upon simple ballistic trajectories. The thicker line to the right of the supergiant indicates the classical L1 gas trajectory (Lubow & Shu, 1975). The models also show an accretion disk does form but that gas may be lost from the binary near the outer L3 point (to the right of the massive companion in Fig. 10). We have simply illustrated the disk in Figure 10 assuming it extends to of the Roche radius of the massive companion (). The three-dimensional models (Nazarenko & Glazunova, 2006a, b) show the inner portions of the disk attain a significant height above and below the orbital plane, and therefore the gas structure is referred to as a torus rather than a disk. If the massive companion is a main sequence star, then the height of the torus may well exceed the stellar radius and block it from view for all but extremely low orbital inclinations. Finally, Nazarenko & Glazunova (2006b) consider the effect that a stellar wind from the mass gainer will have on the surrounding torus. They find the wind breaks out in bipolar outflows and that the hottest gas temperatures occur where the wind strikes the inner torus. Such a bipolar outflow may explain the elongated radio emission surrounding  Lyr (Umana et al., 2000). If the massive companion in RY Scuti is an O6.5 V star, then we expect it will be hot and luminous (perhaps exceeding the luminosity of normal dwarfs due to the addition of accretion luminosity) and thus it will be a source of a significant radiatively-driven wind.

We will use the framework of gas features illustrated in Figure 10 to offer a plausible origin for some of the spectral features discussed in earlier sections. We begin with the expected outflows from the supergiant towards L1 and L2. Most of the absorption lines associated with the supergiant appear to strengthen considerably in depth at the conjunction phases and this change is particularly striking in the case of the He i lines illustrated in Figures 2 and 3. The absorption preferentially strengthens on the blue side of the profile, and in the case of supergiant inferior conjunction (), remains blueshifted for a while longer. For triplet transitions like He i (Fig. 3), the blueshifted component transforms from a broad feature into a narrow, almost nebular absorption line that lasts for more than half the orbit. We suggest these changes are best explained by absorption from the outflows from the supergiant that emerge in the L1 and L2 directions. The blueshifts we observe with projected speeds of 200 km s are consistent with an outflow towards L1 (at ) and towards L2 (at ). The disappearance of the broad absorption shortly after is expected since this flow terminates only a short distance from the supergiant where the gas strikes the accretion disk. On the other hand, the gas leaving the L2 region forms a long trailing plume that may occult the supergiant for a considerable time following and thus may account for the presence of blueshifted absorption until . The appearance of the narrow blueshifted absorption in the He i triplets may be due to low optical depth gas associated with the L2 outflow that extends well above and below the orbital plane.

The gas that enters the accretion disk from the L1 outflow is probably very dense and extended from the orbital plane. In such a situation, the massive companion may be wholly obscured from our point of view and its photospheric spectrum will be invisible. However, we noted in §5 and §6 that there are a number of spectral features that do display radial velocity patterns opposite to those of the supergiant and may plausibly be associated with the immediate environment of the massive companion. In particular, we suggest the Si iii and other lines found in the spectral reconstruction for orbital motion of the massive companion (§6) are probably formed in the extended accretion disk “photosphere” surrounding the massive companion. These features are very broad (indicative of the large Keplerian motions in the disk) and appear similar to those in early B-supergiants (suggesting a gas density that is relatively low compared to most stellar photospheres). Furthermore, the fact that this spectral component appears slightly cooler (like a B0.5 I star) than the supergiant (O9.7 Ibpe) is consistent with idea that the gas has cooled slightly since leaving the supergiant and is shielded from the flux from the presumably hot massive companion by the denser regions of the accretion disk. Similar kinds of disk photospheric features have recently been detected in the spectrum of  Lyr (Ak et al., 2007).

Minimum light according to the recent ephemeris from Kreiner (2004) occurs at spectroscopic phase (Table 4). The fact that the deeper eclipse occurs when the supergiant is behind the massive component is consistent with the somewhat cooler disk photosphere we find from the line patterns in the reconstructed spectrum.

The large mass of the massive companion suggests the star probably has a vigorous radiatively-driven stellar wind. Hydrodynamical models for  Lyr indicate the radiatively-driven wind of the mass gainer will break out of the accretion torus in bipolar flows that are the suspected source of the H emission in  Lyr (Harmanec et al., 1996; Ak et al., 2007). If bipolar wind outflows are also the source of some of the H emission in RY Scuti, then they should extend far away from the orbital plane and be minimally eclipsed at conjunction phases. We may test this idea by comparing the orbital phase variations in the H equivalent width () with the -band light curve. We measure the total emission flux relative to a time variable continuum flux and if the emission flux is constant, then the product of the emission equivalent width and continuum flux will also be constant (i.e., the measured equivalent width will vary inversely as the continuum flux). In Figure 11, we show the orbital phase variations in (H) normalized to their mean at the quadrature phases. We see the measured increases during the decreases in flux that occur surrounding the times of eclipse at and 0.5. We also show in Figure 11 the inverse continuum flux variation derived from the -band observations of Djurasĕvić et al. (2001) and transformed by the equation described in §3. The fact the broad H equivalent width variations follow the inverse light curve so closely indicates the H emission source is not substantially eclipsed by either the supergiant or the accretion disk, and therefore probably forms over a volume that is either very large and/or extended far from the orbital plane (we discuss this further below). We showed in §5 that the bisector velocities of the H emission wings (corresponding to the fastest moving gas) show approximately the same velocity pattern as expected for the massive companion and this anti-phase velocity pattern is seen in Figure 5. The combined evidence of orbital motion and the lack of eclipses suggests the emission line wings of the H feature (and probably part of the main core) form in the bipolar outflows from the disk-constrained wind of the massive companion. The highest projected speeds observed ( km s) are consistent with outflow velocities normal to the orbital plane of km s, a typical value among O-dwarf winds.

The central part of the H profile appears as a quasi-P Cygni profile (with an emission peak at redshifted velocities and a weak absorption at blueshifted velocities) that usually implies a gas outflow. This part of the H profile is generally stationary, showing neither the motion associated with the supergiant nor that of the massive companion. This is very different from the case of the wind emission of the comparable supergiant in the massive X-ray binary system Cygnus X-1, where the H emission is much weaker and follows the Doppler motions of the supergiant (Gies et al., 2003). Furthermore, the wind emission in single OB-supergiants is generally an order of magnitude weaker than observed in RY Scuti (Morel et al., 2004), which indicates that the dense outflow in RY Scuti occurs over a volume much larger than that in the region immediately surrounding the supergiant. The lack of any orbital velocity variation suggests either the source is close to the center of mass or the flux originates far from the center of mass where any orbital motions disappear due to conservation of angular momentum. The first possibility appears to be ruled out by the fact that the main core shows no evidence of eclipses (Fig. 11). Taken together, these properties suggest that a large fraction of the emission flux originates in outflowing gas that surrounds the binary as a whole. This circumbinary disk is probably fed by the gas flows leaving the binary in the vicinity of the L2 and L3 regions. These broad flows may have sufficient vertical extent to account for the blueshifted absorption observed at the quadrature phases. The absence of any substantial decrease in emission flux around (Fig. 11) suggests that the inner radius of the circumbinary disk is larger than , i.e., approximately twice the binary separation. On the other hand, since the gas density will decline with increasing radius and since the H emission depends on gas density squared, we expect that most of the emission comes from the inner part of the circumbinary disk. Thus, the H emitting region of the circumbinary disk probably has a characteristic radius of approximately 1 AU. The circumbinary disk flux contribution appears to decline as expected among the profiles of the higher members of the Balmer sequence (see Fig. 16 in Smith et al. 2002). We note for completeness that the broad asymmetrical shape of the H emission from the inner circumbinary disk appears somewhat like a velocity-expanded version of the profile from the outer double-toroidal nebula (Smith et al., 2002). Since the latter is due to a spatial gas asymmetry in the outer expanding ring, it is possible that there may also be asymmetries in the mass loss into the inner circumbinary disk.

The He ii emission found in the spectrum of RY Scuti has a radial velocity curve that is similar to that of the massive companion (Fig. 7) and this suggests it is formed in the same vicinity. Furthermore, the equivalent width variations of the He ii emission (Fig. 11) indicate a drop in relative emission strength at when the supergiant is in the foreground, consistent with an origin near the massive companion. The hydrodynamical models (Nazarenko & Glazunova, 2006b) suggest that the hottest gas is located at the place where some of the incoming gas stream through L1 climbs over the accretion torus vertical extensions and collides with the high speed wind from the massive companion. Thus, the hot gas is predicted to have a velocity curve for a position slightly offset from the centroid of the massive companion towards the binary center of mass. This appears to be consistent with the He ii velocity curve (Fig. 7) that has a slightly smaller semiamplitude than that for the Si iii lines that probably form in the disk photosphere (which is approximately centered on the massive companion).

Our results indicate the mass outflow from RY Scuti has at least two components, a bipolar outflow of hot stellar wind gas from the massive companion and an equatorial outflow of cooler gas fed by the plume from the L2 region and disk leakage from the L3 region. There is no direct evidence of the high-speed, bipolar outflow in H images made with HST (shown in Smith et al. 1999), but future optical and radio emission maps made with much higher angular resolution may reveal these jet-like features. We suspect most of the cool gas outflow in the orbital plane ends up in a circumbinary disk that slowly transports mass and angular momentum away from the binary and may eventually be a source of gas for the outer double-toriodal nebula. This draining of angular momentum from the binary may explain the observed shrinking of the orbital period and semimajor axis (Šimon, 1999). It is possible the density and spatial extent of the circumbinary disk is sufficient to become a significant opacity source blocking the ionizing flux from the central binary. We speculate that this may be one of the reasons why the surrounding H II double-toroidal nebula seen in HST images has a darker equatorial zone (shaded from ionizing radiation by the intervening circumbinary disk).

## 8 Conclusions

Our spectroscopic investigation of RY Scuti has led to a reassessment of the component masses, a new understanding of the properties of the spectral features related to each star, and a sketch of how the binary is ejecting gas from the system. Our new radial velocities and orbital elements for the supergiant improve upon the past work and we argue that the orbit is probably circular as expected for Roche-filling systems. We determined radial velocities for the Si iii absorption line, a feature that is probably formed in the vicinity of massive companion, and by combining the semiamplitudes of the supergiant and this feature, we estimate the mass ratio is . If we adopt an orbital inclination from that for the surrounding double-toroidal nebula, then the stellar masses are and . These are consistent with the idea that the massive companion is an O-star hidden within a dense accretion disk and that the mass-losing supergiant is destined to become a Wolf-Rayet star. In the distant future, this system may resemble the well known O + WR binary  Vel that has component masses of and (North et al., 2007).

We used a Doppler tomographic reconstruction scheme to separate the component optical spectra of the stars. The line patterns of the the supergiant dominate the combined spectrum and a comparison with a similarly classified star indicates that the supergiant contributes about of the visible flux. The optical spectrum of the supergiant displays unusually weak carbon lines, suggesting the photosphere (like the surrounding nebula) contains CNO-processed gas. The reconstructed spectrum for the massive companion has a superficial resemblance to that of a rapidly rotating B-supergiant, but we argue this spectrum probably forms in the accretion disk photosphere surrounding the massive companion.

We found several lines of evidence indicating mass loss from the binary. The apparent strengthening of blueshifted parts of many absorption lines at both conjunctions and especially following supergiant inferior conjunction suggest that the supergiant is losing mass in broad streams directed both towards and away from the massive companion. The strong H feature appears as broad P Cygni line indicating outflow. The radial velocity curve of the H wings indicates the highest velocity emission originates near the massive companion, probably in a bipolar outflow where the stellar wind of the massive companion breaks out of the thick surrounding accretion disk. The core H emission is stationary and uneclipsed, and we suggest it forms mainly in an outflowing equatorial region (a circumbinary disk). The other important emission line, He ii , also has a radial velocity curve consistent with an origin near the massive companion, and we suggest it forms in hot regions where gas from the L1 stream flows over the thicker portions of the disk and strikes the fast wind from the massive companion.

We argue that the gas lost from the supergiant through the outer L2 region and also lost from the outer parts of the accretion disk near L3 ends up in a relatively cool and dense circumbinary disk with a characteristic H emitting radius of 1 AU. We suspect this disk gas will continue to flow outwards and may help channel gas into the surrounding double-toroidal nebula. Smith et al. (2001) studied the proper motions of the outer double-torus nebula and they conclude the gas in this H II region was probably ejected from the central binary around the year . It is possible this time corresponded to an epoch with an especially high binary mass transfer rate and associated outflow into the circumbinary disk. We note, for example, that the first light curve for the binary from Gaposchkin (1937) (covering the years 1924 to 1934) shows a secondary minimum that is only 0.3 mag deep compared to the current day observed 0.5 mag drop (Djurasĕvić et al., 2001) while the primary eclipse depth is the same. This suggests the disk surrounding the massive companion was cooler in the early part of the last century, which would be consistent with the presence of a thicker, more extended disk resulting from a higher mass transfer rate. Finally, we note that the mass loss into the circumbinary disk removes angular momentum from the central binary, consistent with the observation that the orbital period is decreasing (Šimon, 1999). Thus, contrary to the predictions of conservative mass transfer models, the components of RY Scuti are continuing to be drawn together even after the mass ratio reversal has occurred. We may therefore expect that an episode of more intense interaction and mass transfer and mass loss lies in the immediate future for RY Scuti, and continuing observations will be vital in documenting this key stage in the evolution of massive binaries.

We thank Daryl Willmarth and the staffs of KPNO and CTIO for their support in making these observations possible. This material is based upon work supported by the National Science Foundation under Grants No. AST-0506573 and AST-0606861. Institutional support has been provided from the GSU College of Arts and Sciences and from the Research Program Enhancement fund of the Board of Regents of the University System of Georgia, administered through the GSU Office of the Vice President for Research. RDG is supported by NASA, the NSF, and the US Air Force. We gratefully acknowledge all this support.

## References

• Ak et al. (2007) Ak, H., et al. 2007, A&A, 463, 233
• Antokhina & Cherepashchuk (1988) Antokhina, E. A., & Cherepashchuk, A. M. 1988, Sov. Astr. Lett., 14, 105
• Antokhina & Kumsiashvili (1999) Antokhina, E. A., & Kumsiashvili, M. I. 1999, Soviet Ast., 25, 662
• Bagnuolo et al. (1994) Bagnuolo, W. G., Jr., Gies, D. R., Hahula, M. E., Wiemker, R., & Wiggs, M. S. 1994, ApJ, 423, 446
• Cowley & Hutchings (1976) Cowley, A. P., & Hutchings, J. B. 1976, PASP, 88, 456
• Djurasĕvić et al. (2001) Djurasĕvić, G., Eshankulova, M., & Erkapić, S. 2001, A&A, 374, 638
• Eggleton (1983) Eggleton, P. P. 1983, ApJ, 268, 368
• Friend & Castor (1982) Friend, D. B., & Castor, J. I. 1982, ApJ, 261, 293
• Gaposchkin (1937) Gaposchkin, S. 1937, Harvard Annals, 105, 509
• Gehrz et al. (2001) Gehrz, R. D., Smith, N., Jones, B., Puetter, R., & Yahil, A. 2001, ApJ, 559, 395
• Gehrz et al. (1995) Gehrz, R. D., et al. 1995, ApJ, 439, 417
• Gies & Bolton (1986) Gies, D. R., & Bolton, C. T. 1986, ApJ, 304, 371
• Gies et al. (2003) Gies, D. R., et al. 2003, ApJ, 583, 424
• Giuricin & Mardirossian (1981) Giuricin, G., & Mardirossian, F. 1981, A&A, 101, 138
• Golovatyi & Skul’skii (1992) Golovatyi, V. V., & Skul’skii, M. Yu. 1992, Soviet Ast., 36, 550
• Gray (1992) Gray, D. F. 1992, The Observation and Analysis of Stellar Photospheres (2nd ed.; Cambridge: Cambridge Univ. Press)
• Harmanec (1987) Harmanec, P. 1987, in IAU Colloq. 92, Physics of Be Stars, ed. A. Slettebak & T. P. Snow (Cambridge: Cambridge Univ. Press), 339
• Harmanec (2002) Harmanec, P. 2002, Astronomische Nachrichten, 323, 87
• Harmanec et al. (1996) Harmanec, P., et al. 1996, A&A, 312, 879
• Herrero et al. (1995) Herrero, A., Kudritzki, R. P., Gabler, R., Vilchez, J. M., & Gabler, A. 1995, A&A, 297, 556
• Hilditch (2001) Hilditch, R. W. 2001, An Introduction to Close Binary Stars (Cambridge: Cambridge Univ. Press)
• King & Jameson (1979) King, A. R., & Jameson, R. F. 1979, A&A, 71, 326
• Kreiner (2004) Kreiner, J. M. 2004, Acta Astron., 54, 207
• Lanz & Hubeny (2003) Lanz, T., & Hubeny, I. 2003, ApJS, 146, 417
• Lubow & Shu (1975) Lubow, S. H., & Shu, F. H. 1975, ApJ, 198, 383
• Lucy (2005) Lucy, L. B. 2005, A&A, 439,663
• Lucy & Sweeney (1971) Lucy, L. B., & Sweeney, M. A. 1971, AJ, 76, 544
• Martins et al. (2005) Martins, F., Schaerer, D., & Hillier, D. J. 2005, A&A, 436, 1049
• McErlean et al. (1999) McErlean, N. D., Lennon, D. J., & Dufton, P. L. 1999, A&A, 349, 553
• Milano et al. (1981) Milano, L., Vittone, A., Ciatti, F., Mammano, A., Margoni, R., & Strazzulla, G. 1981, A&A, 100, 59
• Morbey & Brosterhus (1974) Morbey, C., & Brosterhus, E. B. 1974, PASP, 86, 455
• Morel et al. (2004) Morel, T., Marchenko, S. V., Pati, A. K., Kuppuswamy, K., Carini, M. T., Wood, E., & Zimmerman, R. 2004, MNRAS, 351, 552
• Morgan et al. (1955) Morgan, W. W., Code, A. D., & Whitford, A. E. 1955, ApJS, 2, 41
• Nazarenko & Glazunova (2003) Nazarenko, V. V., & Glazunova, L. V. 2003, Astr. Rep., 47, 1013
• Nazarenko & Glazunova (2006a) Nazarenko, V. V., & Glazunova, L. V. 2006a, Astr. Rep., 50, 369
• Nazarenko & Glazunova (2006b) Nazarenko, V. V., & Glazunova, L. V. 2006b, Astr. Rep, 50, 380
• North et al. (2007) North, J. R., Tuthill, P. G., Tango, W. J., & Davis, J. 2007, MNRAS, 377, 415
• Petrovic et al. (2005) Petrovic, J., Langer, N., & van der Hucht, K. A. 2005, A&A, 435, 1013
• Popper (1943) Popper, D. M. 1943, ApJ, 97, 394
• Ryans et al. (2002) Ryans, R. S. I., Dufton, P. L., Rolleston, W. R. J., Lennon, D. J., Keenan, F. P., Smoker, J. V., & Lambert, D. L. 2002, MNRAS, 336, 577
• Sahade et al. (2002) Sahade, J., West, R. M., & Skul’sky, M. Yu. 2002, Rev. Mexicana Astron. Astrofis., 38, 259
• Shafter et al. (1986) Shafter, A. W., Szkody, P., & Thorstensen, J. R. 1986, ApJ, 308, 765
• Šimon (1999) Šimon, V. 1999, A&AS, 134, 1
• Skul’skii (1992) Skul’skii, M. Yu. 1992, Soviet Ast., 36, 411
• Skul’skii & West (1993) Skul’skii, M. Yu., & West, R. M. 1993, Astr. Rep., 37, 594
• Smith et al. (2001) Smith, N., Gehrz, R. D., & Goss, W. M. 2001, AJ, 122, 2700
• Smith et al. (1999) Smith, N., Gehrz, R. D., Humphreys, R. M., Davidson, K., Jones, T. J., & Krautter, J. 1999, AJ, 118, 960
• Smith et al. (2002) Smith, N., Gehrz, R. D., Stahl, O., Balick, B., & Kaufer, A. 2002, ApJ, 578, 464
• Tarasov (2000) Tarasov, A. E. 2000, in ASP Conf. Ser. 214, The Be Phenomenon in Early-Type Stars, ed. M. A. Smith, H. F. Henrichs, & J. Fabregat (San Francisco: ASP), 644
• Umana et al. (2000) Umana, G., Maxted, P. F. L., Trigilio, C., Fender, R. P., Leone, F., & Yerli, S. K. 2000, A&A, 358, 229
• Valdes et al. (2004) Valdes, F., Gupta, R., Rose, J. A., Singh, H. P., & Bell, D. J. 2004, ApJS, 152, 251
• Wade & Rucinski (1985) Wade, R. A., & Rucinski, S. M. 1985, A&AS, 60, 471
• Walborn (1976) Walborn, N. R. 1976, ApJ, 205, 419
• Walborn (1982) Walborn, N. R. 1982, AJ, 87, 1300
• Walborn & Fitzpatrick (1990) Walborn, N. R., & Fitzpatrick, E. L. 1990, PASP, 102, 379
You are adding the first comment!
How to quickly get a good reply:
• Give credit where it’s due by listing out the positive aspects of a paper before getting into which changes should be made.
• Be specific in your critique, and provide supporting evidence with appropriate references to substantiate general statements.
• Your comment should inspire ideas to flow and help the author improves the paper.

The better we are at sharing our knowledge with each other, the faster we move forward.
The feedback must be of minimum 40 characters and the title a minimum of 5 characters