A PROFILE ANALYSIS OF RAMAN-SCATTERED O VI BANDS AT 6825 Å AND 7082 Å IN SANDULEAK'S STAR

Sanduleak's star was discovered in 1977 as an emission line variable object in the Large Magellanic Cloud (Sanduleak 1977). A few years later, Allen (1980) advocated its nature as being that of a symbiotic star, noticing however that the optical continuum appeared weak and featureless, with no clear signature of any late-type giant—which nonetheless is a necessary ingredient at the basis of the symbiotic phenomenon. After almost forty years, the real nature of Sanduleak's star is still a matter of debate: despite no observations unambiguously confirming the presence of a late-type companion, in favor of its binary nature, the bipolar precessing jet was recently discovered by Angeloni et al. (2011). As a matter of fact, bipolar outflows seem to be associated with the binarity of their central source (Sahai et al. 2011). Furthermore, the symbiotic character of Sanduleak's star is supported by both a highly ionized emission line spectrum reminiscent of a dusty (hereafter D-) type symbiotic star (Munari & Zwitter 2002; Belczyński et al. 2000) and by the two intense Raman O vi bands at 6825, 7082 Å, which so far have been convincingly detected only in bona fide symbiotic stars.

These Raman-scattered O vi bands are in fact so unique to the symbiotic phenomenon, that their presence has been used as a sufficient criterion for classifying a star as symbiotic, even in those cases where the cool companion appears to be hiding (Belczyński et al. 2000). These broad spectral bands are known in the astronomical literature at least since the 1940s (Joy & Swings 1945), but their physical origin was explained only much later by Schmid (1989). Schmid proposed that an O vi λ 1032 photon incident on a hydrogen atom in the ground 1s state may be Raman-scattered to become an optical photon with λ = 6825 Å, leaving the scattering hydrogen atom in the excited 2s state; an analogous process for O vi λ 1038 photons to produce the Raman band at 7082 Å.

The resonance doublet O vi λλ 1032 and 1038 arises from ${S}_{1/2}-{P}_{3/\mathrm{2,1}/2}$ transitions. As P_3/2 states have twice as many sublevels than P_1/2 states, under optically thin conditions the O vi λ 1032 flux, F(1032), is expected to be twice as strong as the O vi λ 1038 one, F(1038). However, when the emission region is optically thick, F(1032)/F(1038) ≈ 1, because of thermalization effects (Kang & Lee 2008; Schmid et al. 1999). Since symbiotic nebulae are highly stratified (Allen 1987; Luna & Costa 2013), it is indeed quite common to observe UV resonance doublets with different flux ratios, i.e., emerging from regions with different optical depths (as in CI Cyg, Mikołajewska et al. 2006).

It is also important to emphasize that, while the O vi λλ 1032 and 1038 line profiles are dependent on the observer's line of sight (e.g., in EG And, the O vi lines completely disappear when the white dwarf is at the superior conjunction—Crowley et al. 2008), the Raman band profiles are not. Instead, they mainly reflect the relative kinematics between the H i scattering region and the far-UV emission region. Therefore, the Raman band profiles observed in symbiotic binaries give us the unique perspective of the mass transfer process, as seen from the donor star (e.g., Schmid 1989; Nussbaumer et al. 1989).

Observationally, the Raman O vi bands at 6825 Å and 7082 Å tend to exhibit multiple-peaked profiles, in which the blue peak of Raman 6825 Å is relatively more enhanced than the corresponding one of 7082 Å  (Harries & Howarth 1996; Schmid et al. 1999). Lee & Kang (2007) performed profile analyses of Raman O vi 6825 Å band observed in the two D-type symbiotic stars V1016 Cyg and HM Sge to probe the kinematics of the emission region around the white dwarf. They attributed the observed asymmetric double-peak profiles to the O vi emission from an accretion disk: the peak separation corresponds to a speed of ∼50 km s⁻¹, implying an emitting region of ∼1 AU.

Heo & Lee (2015) further noted that in V1016 Cyg the blue peak of the 7082 Å band is relatively weaker than the 6825 Å counterpart, when the two Raman bands are normalized to an equal red peak strength in the Doppler factor space. They developed a quantitative model in which an accretion stream around the white dwarf is responsible for the double-peak profiles, and where the difference in the profile shape is due to the F(1032)/F(1038) variation in the accretion stream.

Spectropolarimetric data of the same V1016 Cyg by Schild & Schmid (1996) revealed that the red wing of the Raman O vi red peak is polarized in the direction perpendicular to the polarization exhibited by the main part of the bands. This polarization flip is consistent with the emission region made of two components, one moving parallel and the other one perpendicularly (i.e., an outflow-like structure) with respect to the accretion disk plane. Also, Schmid et al. (2000) showed a similar polarization pattern in HM Sge. It is consistent with the presence of a bipolar outflow, which was reported by several observations (Solf 1984; Eyres et al. 2007). Since the multiple-peak structure and the polarization flip in the Raman O vi bands can be intuitively explained by a model with an accretion disk and a bipolar outflow, we then apply and test this theoretical scenario on Sanduleak's star.

In what follows, we present a detailed analysis of the Raman O vi bands in Sanduleak's star. In Section 2, we present archival Far Ultraviolet Spectroscopic Explorer (FUSE) and proprietary Magellan Inamori Kyocera Echelle (MIKE) data with a morphological description of the observed profiles of both the far-UV O vi doublet (Section 2.1) and the optical Raman O vi bands (Section 2.2). Our profile decomposition technique, along with our physical interpretation, which reads these profiles in terms of the relative kinematics between the O vi emission region and the H i scattering region, is fully described in Section 3. In Section 4 we show the results of Monte Carlo simulations, which help to constrain the physical conditions (e.g., column density and spatial extent) of the circumstellar H i region. Discussion and concluding remarks follow in Section 5.

Symbiotic stars are composite interacting binaries: at the origin of their complex photometric and spectroscopic variability lies a series of astrophysical phenomena (accretion processes, stellar pulsations, and orbital modulations, just to quote a few), which act on timescales ranging from minutes to centuries (Munari 2012). Any multiwavelength study aiming at a self-consistent modeling of a symbiotic system should thus take into account all the caveats arising when combining observations taken over different spectral ranges at different epochs.

Specifically with regard to the analysis of Raman-scattering processes, the ideal case would be that of having virtually simultaneous observations of both the far-UV O vi resonance doublet and the optical Raman O vi bands. Birriel et al. (2000) performed near simultaneous far-UV and optical observations to confirm the identification of Raman O VI bands and derive the Raman-scattering efficiencies for nine symbiotic systems (see also Birriel et al. 1998). Even when a high-resolution optical spectrum is at hand, however, far-UV data are as precious as rare. Luckily enough, a search in the Mikulski Archive for Space Telescopes (MAST) returned a publicly available UV spectrum of Sanduleak's star, which we publish here for the first time.

2.1. FUSE Spectrum of O vi λλ 1032, 1038

The FUSE was launched in 1999 and operated until the failure of the pointing system on the satellite, on 2007 October 18. At the conclusion of the mission, the FUSE data archive was moved to MAST, where all the data were reprocessed and archived with the final version of the CalFUSE calibration pipeline software package (v 3.2, Dixon et al. 2007) and made available to the whole community.

Through Program E950, FUSE pointed at Sanduleak's star for a total of ∼22,000 s on 2004 August 14. The science data were collected with the largest apertures on the focal plane assembly (LWRS—30 arcsec²) in TTAG photon collecting mode. They cover the entire wavelength region 905–1187 Å, with a nominal spectral resolution of Δv ≈ 20 km s⁻¹ (Moos et al. 2000; Sahnow et al. 2000), and are placed by the same CalFUSE on a heliocentric velocity scale.

The O vi λ 1032 feature appears among the most intense emission lines present in the far-UV spectrum of Sanduleak's star, second to (in decreasing order) Lyβ, C iii λ 977, and Ly γ. In the present work, we analyze and discuss only the profiles of the O vi λλ 1032 and 1038 resonance doublet (Figure 1), postponing to a forthcoming paper the modeling of the entire FUSE emission line spectrum.

Zoom In Zoom Out Reset image size

Following Nichols & Slavin (2009), in Figure 1 we overplot the two O vi line profiles: the horizontal axis sets the Doppler factor space, whose origin has been chosen at the (respective) line central wavelength in vacuum (Moore 1979):

$$\begin{eqnarray}&&{\lambda }_{1032}=1031.928\ \mathring{\rm A} ,{\lambda }_{1038}=1037.618\ \mathring{\rm A} ,\end{eqnarray} \tag{ 1 }$$

and where O vi λ 1038 has been multiplied by 2 in order to focus on the profile comparison. Based on Figure 5 of Wood et al. (2002) we take the instrumental profile, or line-spread function, as a single Gaussian having the FWHM of 20 km s⁻¹. The convolved profiles are plotted by the thick lines in Figure 1. It is immediately evident that the two profiles coincide with each other, with the remarkable exception of a small excess at a few tens of km s⁻¹ in the O vi λ 1038 line. This excess implies that locally F(1032)/F(1038) is less than 2: as mentioned in Section 1 regarding far-UV resonance doublets in symbiotic systems, this evidence hints toward the presence in Sanduleak's star of an O vi optically thick region, which is approaching the observer with a velocity of ∼20 km s⁻¹. A description of this emitting component within the framework of our scenario is discussed in Section 3.4.

2.2. MIKE Spectrum of Raman O vi Bands

Optical high-resolution spectroscopic observations of Sanduleak's star were carried out on 2010 November 21, using the MIKE spectrograph at the 6.5 m Magellan-Clay telescope, Las Campanas Observatory in Chile. The spectrograph delivers full wavelength coverage from about 4900–10000 Å in its red configuration: with the 0.7 × 5 arcsec slit we used, the spectral resolution was R ∼ 32,000 (Bernstein et al. 2003). The scale of the CCD is about 7.5 pix arcsec⁻¹ (0.13 arcsec pix⁻¹), therefore we binned the detector in the spectral and spatial direction to maximize signal-to-noise ratio. We took a series of 3 × 900 s exposures. Data reduction was performed under IRAF following the standard procedure.

In Figure 2, we display the Sanduleak's star Raman-scattered O vi λλ bands at 6825 and 7082 Å. It is noticeable that the two bands have quite different profiles: while the Raman 6825 Å band exhibits a single-peak profile with a redward extended bump, the Raman 7082 Å band reveals a distinct triple-peak profile. In order to make a quantitative comparison of the two optical profiles, we convert the observed wavelength to the O vi parent Doppler factor as follows: adopting the refractive index of air n_air = 1.0002763 (Ciddor 1996), the air wavelength λ_obs is transformed into the vacuum wavelength λ_vac by

$$\begin{eqnarray}&&{\lambda }_{\mathrm{obs}}{n}_{\mathrm{air}}={\lambda }_{\mathrm{vac}}.\end{eqnarray} \tag{ 2 }$$

Zoom In Zoom Out Reset image size

We then introduce λ_f, the vacuum wavelength corrected for the systemic velocity v_sys, as

$$\begin{eqnarray}&&{\lambda }_{f}=\displaystyle \frac{{\lambda }_{\mathrm{vac}}}{1+\tfrac{{v}_{\mathrm{sys}}}{c}}.\end{eqnarray} \tag{ 3 }$$

Using the energy conservation principle, we find the corresponding far-UV wavelength of incident O vi photon

$$\begin{eqnarray}&&{\lambda }_{\mathrm{UV}}=\displaystyle \frac{{\lambda }_{f}{\lambda }_{\mathrm{Ly}\alpha }}{{\lambda }_{f}+{\lambda }_{\mathrm{Ly}\alpha }},\end{eqnarray} \tag{ 4 }$$

where λ_Lyα = 1215.668 Å.

The O vi central wavelengths of Equation (1) are then used to obtain the Doppler factors ΔV_atomic:

$$\begin{eqnarray}&&\displaystyle \frac{{\rm{\Delta }}{V}_{\mathrm{atomic}}}{c}=\displaystyle \frac{{\lambda }_{\mathrm{UV}}-{\lambda }_{i}}{{\lambda }_{i}},i=1032,1038.\end{eqnarray} \tag{ 5 }$$

The presence in Sanduleak's star of a highly collimated jet (Angeloni et al. 2011) and of Raman O vi bands exhibiting multiple-peak profiles (Figure 2) implies that a plausible model of its O vi emission nebula should at least foresee an accretion disk and a bipolar outflow. Moreover, the F(1032)/F(1038) deviation from 2 discussed at the end of Section 2.1 calls for the inclusion of an additional emitting region characterized by a high optical depth. Our schematic layout of Sanduleak's star is presented in Figure 3.

Zoom In Zoom Out Reset image size

Based on the observed profiles of the Raman-scattered O vi bands presented in Figure 2, we reconstruct in Figure 4 the O vi far-UV resonance doublet by combining five emitting components: three arising from the accretion disk; one from the bipolar outflow; one from the optically thick nebula. For the sake of simplicity, each component is described by a Gaussian function characterized by a central velocity ΔV_atomic, a FWHM Δv, and a peak value f, the latter normalized at the peak value of the 1032 line.

Zoom In Zoom Out Reset image size

Before the detailed discussion of each component with our corresponding physical interpretation, it is worth clarifying once again that the reconstructed O vi λλ 1032 and 1038 Å line profiles are the ones seen by a hypothetical observer at rest with respect to the scattering region (which we assume stationary with respect to the donor star) and can be perfectly (in fact, almost always are) different from the actual observed profiles (Figure 1), which are strongly dependent on the observer point of view.

3.1. Blue Emission Part (BEP) and Red Emission Part (REP) from the Accretion Disk

3.2. Central Emission Part (CEP) in the Accretion Disk

3.3. Bipolar Outflow

3.4. Optically Thick Compact Component

The flux ratio of the two Raman O vi bands F(6825)/F(7082) can be used as a rough proxy to classify a symbiotic star into D-type or S-type. According to Schmid et al. (1999), in fact, D-types tend to show F(6825)/F(7082) ∼ 6, while for S-types F(6825)/F(7082) ∼ 3. The difference in F(6825)/F(7082) seems mainly due to the different H i column density N_{H i} characterizing the Raman-scattering region in the two types of symbiotic stars, where N_{H i} tends to be larger in S-types, because of their smaller binary separation compared to D-type systems. Our MIKE data show F(6825)/F(7082) ∼ 4.5, which is an intermediate value between S- and D-type symbiotic stars.

In this section, we perform Monte Carlo simulations in order to estimate the representative value of N_{H i} in Sanduleak's star by reproducing the observed F(6825)/F(7082). We consider a cylindrical neutral scattering region characterized by a column density N_{H i} measured along the cylinder axis. We place it in front of the giant hemisphere, which is facing the white dwarf. We align the scattering region, assumed stationary with respect to the white dwarf, so that the cylinder axis coincides with the axis connecting the two stars. We vary values of N_{H i} from 1 × 10²² to 5 × 10²³ cm⁻² when the simulated profile for the Raman 7082 Å band becomes stronger than the observed one.

The simulation starts with a generation of an O vi photon (according to the input profiles given in Figure 4), which subsequently enters the scattering region. The O vi photon then moves around through Rayleigh scattering processes and escapes from the scattering region once Raman-scattering occurs. A more detailed description of this family of Monte Carlo simulations can be found in Heo & Lee (2015).

In Figure 5, we show the results of our Monte Carlo simulations of the Raman-scattered O vi λλ 1032, 1038 profiles for various values of N_{H i}, superposed on the 2010 MIKE data. We normalize the simulated spectra with the observed Raman 6825 Å band and look for the best-fitting profile of the observed Raman 7082 Å band. A good fit is obtained for N_{H i} ∼ 1 × 10²³ cm⁻². Our Monte Carlo result shows that F(6825)/F(7082) > 5, appropriate for D-type symbiotic stars, are obtained for N_{H i} < 10²² cm⁻². On the other hand, for N_{H i} > 3 × 10²³ cm⁻² the flux ratio becomes lower than 4, resulting in a poor fit.

Zoom In Zoom Out Reset image size

In this work, we have presented a systematic analysis of the Raman-scattered O vi bands in Sanduleak's star. Following a Gaussian decomposition scheme, we successfully fit the profiles of Raman O vi 6825 and 7082 Å bands by assuming that the O vi nebula can be decomposed into several emitting components, to be identified with the accretion disk, the bipolar outflow, and a further compact, optically thick component. We have also performed Monte Carlo simulations, from which we propose that the neutral region of Sanduleak's star is characterized by N_{H i} ∼ 1 × 10²³ cm⁻².

When remembering the very definition of ΔV_{O vi} and ΔV_atomic, it is suggested to interpret the difference between ΔV_{O vi} and ΔV_atomic (of the order of 14 km s⁻¹) as the terminal wind velocity of the cool component, which so far has escaped any direct detection and is at the origin of the controversial status of Sanduleak's star.

One physical mechanism, which may contribute to the redward spectral shift, can be found in the asymmetric frequency diffusion of resonantly scattered line photons in an expanding medium. In a static medium, a line photon arising from a permitted transition is usually optically thick at its line center. This implies that the escape of a resonance line photon is made through frequency diffusion after a large number of local scatterings. However, in an expanding medium, a line photon shifted blueward in the frequency space will be scattered after it travels to a spot where the resonance condition is satisfied. In this way, the blue part of line radiation is significantly suppressed to form a P Cyg profile. Naively, we expect that locally emergent resonance line radiation in an emission region with an expansion speed of ΔV is shifted redward by the same quantity (e.g., Schmid et al. 1999). In a complex emission region, consisting of a number of emission components, each redward shift is convolved in a complicated way, which unfortunately cannot be disentangled by our present observations.

Spectropolarimetry has the ability to shed much more light on the detailed structure of the O vi emission region in symbiotic stars. Spectropolarimetric observations performed by Harries & Howarth (1996) showed, for example, that Raman-scattered O vi bands are strongly polarized, and that in many cases the red wing is polarized in the direction perpendicular to the polarization direction of the main part.

Interestingly, this behavior is naturally expected in our decomposition scheme in which the band red wing originates from the bipolar outflow, which indeed moves away in the direction perpendicular to the accretion disk. Our profile decomposition thus foresees that the emitting component associated with the outflow will be polarized in the direction perpendicular to the jet axis, whereas the remaining part will be polarized along it. Because the bipolar outflow component is mainly dominant in the red wing, we also expect the degree of linear polarization to be significant there. In a similar way, the blue wing will exhibit a fairly high degree of polarization along the jet axis. However, in the red main part where the REP component and the bipolar outflow component contribute, compensation of the two oppositely polarized components may result in a lower degree of polarization, which is strongly dependent on the relative contribution of each component to the overall profile.

In this context, spectropolarimetric observations represent a crucial validation test of our decomposition scheme and at the same time a unique opportunity to clarify the astrophysical nature of Sanduleak's star.

We are grateful to the anonymous referee for useful comments. We also thank Kwang-il Seon for his help. FUSE data, presented in this paper, were obtained from the Mikulski Archive for Space Telescopes (MAST). STScI is operated by the Association of Universities for Research in Astronomy, Inc., under NASA contract NAS5-26555. Support for MAST for non-HST data is provided by the NASA Office of Space Science via grant NNX09AF08G and by other grants and contracts. This research was supported by the Korea Astronomy and Space Science Institute under the R&D program (Project No. 2015-1-320-18) supervised by the Ministry of Science, ICT and Future Planning.