Raman-scattered O VI Features in the Symbiotic Nova RR Telescopii ¹¹1This paper includes data gathered with the 6.5 meter Magellan Telescopes located at Las Campanas Observatory, Chile.

We obtained high-resolution optical spectra of RR Tel using the MIKE spectrograph (Bernstein et al., 2003). The double echelle spectrograph MIKE covers the wavelength range of 3350-5000 Å (blue) and 4900-9500 Å (red) in its normal configuration. A slit width of $0.7^{\prime\prime}\times 5^{\prime\prime}$ was used, resulting in resolving power $R\sim 27000$ and $\sim 35500$ on the blue and red sides, respectively. To increase the signal-to-noise ratio (S/N), a $2\times 2$ binning was applied in both the spectral and spatial directions. Observations were carried out on 2016 July 30 and 2017 July 26, for total exposure times of 2000 sec and 2400 sec, respectively. With these integration times, the spectra have significant $S/N>100$ for strong lines suitable for profile analysis. ThAr spectra were taken for wavelength calibration.

We reduced the MIKE data using the Carnegie Python Distribution (CarPy) pipeline (Kelson et al., 2000; Kelson, 2003) and NOAO.onedspec package of the Image Reduction and Analysis Facility (IRAF).²²2IRAF is distributed by the National Optical Astronomy Observatory, which is operated by the Association of Universities for Research in Astronomy (AURA) under cooperative agreement with the National Science Foundation. The data are overscan-corrected, bias-subtracted, extracted, wavelength calibrated, and flat-fielded with quartz lamps within the pipeline. Individual exposures were then combined using the IRAF task scombine. Due to the poor weather condition of the night 2016 July 30, flux calibration was not possible. For the observing night 2017 July 26, we observed the spectrophotometric standard star HR 7950 under photometric condition. Flux calibration was performed with the IRAF tasks standard, sensfunc, and calibrate.

The left and right panels of Fig. 2 show the Raman-scattered O VI features at 6825 Å and 7082 Å. To make quantitative profile comparisons of the two Raman O VI features, we transform the observed wavelengths into the O VI Doppler factor $\Delta V_{atomic}$ as shown by the upper horizontal axis. First, we convert the observed wavelength to the vacuum wavelength using the refractive index of air $n_{air}=1.0003$. By measuring the center wavelengths of the strong emission lines (e.g., He I$~{}\lambda$7065 and $\lambda$6678, [N II] $\lambda$6583 and $\lambda$6548, [O III] $\lambda$4959, H$\beta$ and H$\gamma$), we obtain the systemic velocity $\nu_{sys}$ of $-55.5~{}{\rm{km~{}s^{-1}}}$, and $-57.8~{}{\rm{km~{}s^{-1}}}$, respectively, for the 2016 and 2017 data. We correct for the systemic velocity and calculate the corresponding far-UV wavelength of incident O VI photon using the energy conservation principle. Adopting the O VI center wavelengths of $\lambda_{1032}=1031.928$ Å and $\lambda_{1038}=1037.618$ Å, we finally compute the O VI parent Doppler factor $\Delta V_{atomic}$. More detailed information about the transformation between the wavelength and the Doppler factor is available in Heo et al. (2016).

We fit the Raman O VI features with two Gaussian components, from which we derive the peak separation of the Raman O VI 6825 feature $\Delta V=47.5~{}{\rm km\ s^{-1}}$ and $\Delta V=48.7~{}{\rm km\ s^{-1}}$ in the data of 2016 and 2017, respectively. For the Raman O VI 7082 features, the peak separations are measured to be $\Delta V=45.7~{}{\rm km\ s^{-1}}$ and $\Delta V=46.4~{}{\rm km\ s^{-1}}$ in 2016 and 2017, respectively. The Gaussian fitting parameters are presented in Table 3.

We overplot the MIKE data in Fig. 2, whose medium grey lines represent the 2016 data and black lines show the 2017 data. The vertical axes show flux density in units of $10^{-14}~{}{\rm erg~{}cm^{-2}~{}s^{-1}~{}\AA^{-1}}$. Since the 2016 data is not flux calibrated, we multiplied an artificial factor to match the continuum value with the 2017 data. As shown in Fig. 3, the continuum around the bands is virtually flat with a constant value of $2.0\times~{}10^{-14}~{}{\rm erg~{}cm^{-2}~{}s^{-1}~{}\AA^{-1}}$ due to the dust obscuration. The Mira in RR Tel is heavily obscured and so no trace of photospheric absorptions is visible (e.g. Kotnik-Karuza et al., 2006). Dust obscuration in RR Tel is also supported by its optical light curves, which do not show any pulsation of Mira, whereas the Near-IR light curves show a clear periodicity (Gromadzki et al., 2009).

To measure the total line fluxes for the Raman O VI features, we assume that the emission having the Doppler factor $\Delta V_{atomic}$ between $-60~{}{\rm{km~{}s^{-1}}}$ to $100~{}{\rm{km~{}s^{-1}}}$ is a Raman O VI feature. In these regions, the contributions of other spectral lines, including TiO or VO absorption lines from the Mira and telluric lines, are negligible. There is telluric ${\rm O_{2}}$-B band between 6870 - 6950 Å, but not affecting both Raman O VI features (Groppi & Hanson, 1996). Few unknown emission lines with FWHM $<0.5$ Å are also detected at the wavelength range of Raman O VI features. However, their contribution to the total line fluxes are negligible due to the relatively small intensities compared with the Raman features. Therefore we obtain the total line fluxes of the Raman O VI features $F(6825)=30.34\pm 0.11\rm\times 10^{-13}~{}erg~{}cm^{-2}~{}s^{-1}$ and $F(7082)=5.98\pm 0.08\rm\times 10^{-13}~{}erg~{}cm^{-2}~{}s^{-1}$ for the 2017 data.

The Raman line fluxes of RR Tel were studied in previous observations between 1993 and 1996 (e.g., Espey et al., 1995; Crawford et al., 1999; Schmid et al., 1999). A steady decrease was found over the period, and the same behavior is shown for the far-UV O VI 1032 and 1038 emissions. We note that the fluxes of both Raman O VI in the MIKE data have decreased by about 30 % from 1996 October. The measurements are summarized in Table 1. The decrease in the line intensity is consistent with the trend of general slow fading started in 1949 after its outburst (e.g., Nussbaumer & Dumm, 1997; Contini & Formiggini, 1999).

RR Tel was observed on June 14, 2002 with the FUSE satellite as part of the Cycle 3 program C141 ( PI: P.R. Young). The FUSE spectrum of RR Tel was taken for an exposure time of 128 sec through the low-resolution aperture (LWRS: 30 $\times$ 30 arcsec) using TTAG (Photon Address) mode. The SiC2 channel covers a wavelength range of 917 - 1104 Å at a spectral resolution of 12000 - 20000 (Moos et al., 2000; Sahnow et al., 2000). The data were re-processed and archived at MAST with the CalFUSE pipeline version 3.2.1. The extracted data is binned in wavelength with a bin size of 0.013 Å (Dixon et al., 2007). Considering that the observation was made under low-resolution mode, we binned the data by a spectral resolution of 0.362 Å.

In Fig. 4, we present a zoom-in of the FUSE spectrum centered on the two strongest emission lines, O VI 1032 and 1038. The light grey line represents the raw data, whereas the black line shows the binned spectrum. The O VI emissions have line fluxes of $F(1032)=2.01\pm 0.56\times 10^{-10}~{}\rm erg\ cm^{-2}\ s^{-1}$ and $F(1038)=1.18\pm 0.32\times 10^{-10}~{}\rm erg\ cm^{-2}\ s^{-1}$, respectively, yielding the flux ratio of $F(1032)/F(1038)=1.70$. There has been a steady decrease in line fluxes of both O VI emissions from $F(1032)=2.56\times 10^{-10}~{}\rm erg\ cm^{-2}\ s^{-1}$ and $F(1038)=1.54\times 10^{-10}~{}\rm erg\ cm^{-2}\ s^{-1}$ in 1993 (Schmid et al., 1999) - a drop of 22 %. We summarize the observed values of O VI lines and compare them with previous measurements from other research works in Table. 2. We note that the flux ratio $F(1032)/F(1038)$ has remained almost constant at $\sim$ 1.7 throughout this period.

RR Tel was observed with FEROS installed on the MPG/ESO 2.2 m telescope for a total exposure time of 600 sec on November 13, 2003. The spectrum spans the wavelength region 3527 - 9216 Å and has a spectral resolution of $R\sim 48000$ (Kaufer et al., 1999). The data were reduced automatically using the FEROS Data Reduction Software (DRS) pipeline version fern/1.0 and archived at the European Southern Observatory (ESO) database. ⁵⁵5http://archive.eso.org/scienceportal/home The data are overscan-corrected, bias-subtracted, extracted and flat-fielded. The extracted, flat-fielded data are wavelength calibrated and rebinned in wavelength with a bin size of 0.3 Å. The reduced spectra are not flux calibrated. ⁶⁶6http://www.eso.org/rm/api/v1/public/releaseDescriptions/94 In Fig. 2, the light grey lines correspond to the FEROS data. Not being flux calibrated, the FEROS data is multiplied by an artificial factor to compare their profiles with that of the flux calibrated MIKE data.

Following the same method used for the MIKE data, we measure the center of the strong emission lines (He I$~{}\lambda$7065 and $\lambda$6678, [O III] $\lambda$5007 and $\lambda$4959), and obtain the systemic velocity $\nu_{sys}$ of $-59.7~{}\rm{km~{}s^{-1}}$. Through two-component Gaussian modeling of the Raman features, we find that the two peaks of the Raman O VI 6825 are $\lambda_{blue}=6822.75$ Å and $\lambda_{red}=6829.75$ Å. These correspond to a Doppler factor of $V_{blue}=-10.0~{}{\rm km\ s^{-1}}$ and $V_{red}=36.5~{}{\rm km\ s^{-1}}$ resulting in the peak separation $\Delta V=46.5~{}{\rm km\ s^{-1}}$. For the Raman O VI 7082, the peak separation $\Delta V=48.5~{}{\rm km\ s^{-1}}$ is obtained. We summarize the Gaussian fitting parameters for the Raman features in Table 3.

In previous studies, it was found that the Raman O VI features of RR Tel show an asymmetric double-peaked profile, where the red peak is stronger than the blue one (e.g., Allen, 1980; Schmid et al., 1999). Harries & Howarth (1996) investigated the temporal variation of Raman O VI 6825 feature between 1992 and 1994, from which they found a slight decrease in (normalized) line intensity while the overall profiles remained unchanged (see Fig. 8-d in Harries & Howarth (1996)).

By comparing the data obtained in the years of 2003, 2016, and 2017, we find that both Raman O VI profiles show significant variations. In Fig. 2, we overplot the FEROS spectrum taken in 2003 and the MIKE data observed in 2016 and 2017. It is obvious that the asymmetry of both Raman profiles has grown for the last two decades, whereas there is no significant change between 2016 and 2017. To describe the profile asymmetry in a given Raman feature, we use parameter $f_{red}/f_{blue}$, the ratio between the peak values of the two Gaussian components. We find that the asymmetry parameter $f_{red}/f_{blue}$ of Raman O VI 6825 increased $26~{}\%$ from $1.94$ in 2003 to $2.45$ in 2016. This change is even more apparent in the 7082 feature, a $43~{}\%$ increase from $3.94$ in 2003 to $5.65$ in 2016.

First of all, to model the density distribution of H I around the Mira, it is assumed that the stellar wind from the Mira follows a beta law, where the radial wind velocity $v_{r}({\bf r})$ is given by

Here, $R_{*}$ is the launching site of the wind, $v_{\infty}$ is the wind terminal velocity (e.g., Lamers et al., 1999). We choose the parameter $\beta=1$ for simplicity in our Monte Carlo simulations.

The slow stellar wind around the Mira is illuminated by intense far-UV radiation from the hot component, and hence some parts facing the WD are photoionized. Seaquist, Taylor & Button (1984) presented their photoionization calculation to find the ionization front, known as the STB model. In the STB model, the ionization front in the stellar wind region is determined by the balance of photoionization rate by the H-ionizing luminosity from the hot component and recombination rate set by the mass-loss rate.

In the STB geometry, the ionization front is specified by the ionization parameter $X$, which is given by

Here, $a$ is the binary separation, $L_{H}$ is the H-ionizing photon number luminosity, $\alpha_{B}$ is the case B recombination coefficient for H I, and $m_{H}$ is the proton mass. In this work, we adopt the values of $a=56\rm~{}au$, $L_{H}=7\times 10^{47}~{}\rm~{}s^{-1}$ that were used in previous works (Hinkle et al., 2013; González-Riestra et al., 2013).

In view of the fact that the Raman O VI profiles are dependent exclusively on the relative motion between the emitter and scatterer, the double-peak profiles of Raman O VI strongly imply that the far-UV O VI emission lines are formed in regions that move toward the H I region and recede from it. In this work, the far-UV O VI 1032 and 1038 emission regions are identified as a part of the accretion disk formed around the WD.

We, therefore, attribute the red-peak enhanced profiles to asymmetry of the matter distribution in the O VI disk (e.g., Heo & Lee, 2015). The part of the accretion disk on the entering side is moving away from the Mira and hence is characterized by positive Doppler factors for Raman-scattered O VI: we call this region the red emission region (RER) and the opposite side the blue emission region (BER).

It is noteworthy that the profiles of the two Raman O VI features are not identical: specifically, the ratio of the red and blue peak fluxes $f_{red}/f_{blue}$ is 2.4 for the Raman 6825 whereas it is 5.8 for the Raman 7082 as shown in Section. 2.4. The disparate profiles of the two Raman O VI features can be explained by the local variation of the ratio $F(1032)/F(1038)$ in the O VI emission region, which decreases from 2 to 1 as optical depth increases (Heo & Lee, 2015). The RER, assumed to be of high density, is characterized by the flux ratio $F(1032)/F(1038)\sim 1$, whereas the BER is much more sparse, resulting in $F(1032)/F(1038)\sim 2$.

To reproduce the peak separation of the Raman O VI features, the Keplerian velocity is set to $>35~{}{\rm km~{}s^{-1}}$. It corresponds to the physical size of $<0.8$ au when we adopt $M_{WD}=0.65~{}{\rm M_{\odot}}$ as the mass of the WD (González-Riestra et al., 2013). The asymmetry of the matter distribution in the accretion disk is given as a function of the azimuthal angle $\theta$ measured from the line toward the Mira. The best-fit is made using the following density function,

In the right panel, the solid line represents the O VI density profile $n(\theta)/n_{0}$ as a function of $\theta$. The O VI emissivity is dominant in the RER ($0<\theta<\pi$), whose peak is at around $\theta=0.3\pi$.

Considering the disparate profiles of the two Raman O VI, we use the flux ratio $F(1032)/F(1038)$, varying from 1.5 to 2, as a function of $\theta$

which is schematically illustrated in the middle panel and represented by the dotted line in the right panel of Fig. 5.

Adopting the asymmetric O VI accretion disk model supplemented by the locally varying $F(1032)/F(1038)$, Monte Carlo simulations were carried out under various parameters, a mass loss rate ${\dot{M}}$ ranging from $1\times 10^{-6}~{}{\rm M_{\odot}~{}yr^{-1}}$ to $1\times 10^{-5}~{}{\rm M_{\odot}~{}yr^{-1}}$ with a step width of $1\times 10^{-6}~{}{\rm M_{\odot}~{}yr^{-1}}$ and terminal velocities of the Mira wind $v_{\infty}$ with $10~{}{\rm km~{}s^{-1}}$, $15~{}{\rm km~{}s^{-1}}$, $20~{}{\rm km~{}s^{-1}}$ and $25~{}{\rm km~{}s^{-1}}$. These parameter sets give the ionization parameter X from 0.2 to 80.

We find that the double-peak structure of Raman O VI features in RR Tel is diluted if the stellar wind velocity exceeds $v_{\infty}>20~{}{\rm km~{}s^{-1}}$ or the ionization parameter $X<1.25$. With the maximum velocity of $v_{\infty}=20~{}{\rm km~{}s^{-1}}$, we, therefore, constraint ${\dot{M}}<8\times 10^{-6}~{}{\rm M_{\odot}~{}yr^{-1}}$, which gives $X~{}\sim~{}1.25$. The corresponding ionization front is presented in Fig. 1. The best-fit result is obtained with $v_{\infty}~{}\sim~{}20~{}{\rm km~{}s^{-1}}$ and ${\dot{M}}\sim 2\times 10^{-6}~{}{\rm M_{\odot}~{}yr^{-1}}$. In Fig. 6, we present our best-fit profiles for the Raman O VI features. The black lines show the MIKE data, while the light grey lines represent the results of our simulations.

Raman scattering efficiency, $\eta$, is defined as the number ratio $N(FUV)/N(Raman)$ of the incident and Raman-scattered photons. The conversion rate for Raman-scattered O VI features was derived from direct measurement of the O VI far-UV emissions in far-UV spectra and O VI Raman features in optical spectra (Schmid et al., 1999; Birriel et al., 2000). Those authors suggested that $N(6825)/N(1032)\sim 6\%$ and $N(7082)/N(1038)\sim 2\%$, respectively. From our Monte Carlo simulations, we obtained $10.5\%$ for the $\lambda 1032\rightarrow\lambda 6825$ conversion and $4.3\%$ for $\lambda 1038\rightarrow\lambda 7082$ conversion, respectively.

For our best-fit results, we obtain the total line flux $F(6825)=3.0\times 10^{-12}~{}\rm erg\ cm^{-2}\ s^{-1}$ and $F(7082)=0.7\times 10^{-12}~{}\rm erg\ cm^{-2}\ s^{-1}$, respectively. Adopting a distance of $2.6~{}{\rm kpc}$ (González-Riestra et al., 2013; Whitelock, 1988), the corresponding number luminosities of Raman-scattered photons emitted per unit time are $N(6825)=5.6~{}\times~{}10^{45}~{}\rm~{}s^{-1}$ and $N(7082)=1.3~{}\times~{}10^{45}~{}\rm~{}s^{-1}$. Taking the conversion efficiency for O VI, we deduce the far-UV number flux densities $N(1032)=5.3\times 10^{46}~{}\rm s^{-1}$ and $N(1038)=3.1\times 10^{46}~{}\rm s^{-1}$. The corresponding line fluxes are obtained $F(1032)=1.3\times 10^{-9}~{}\rm erg\ cm^{-2}\ s^{-1}$ and $F(1038)=7.3\times 10^{-10}~{}\rm erg\ cm^{-2}\ s^{-1}$, yielding the flux ratio $F(1032)/F(1038)$ of 1.8. This result does not deviate much from the observed value of 1.7, as shown in Section 2.2. In Table 4, we summarize the intrinsic far-UV O VI parameters including, center wavelength, Raman scattering efficiency $\eta$, calculated number flux density $N_{exp}$, expected line flux $F_{exp}$, and the observed flux $F_{obs}$ in FUSE data.