High dispersion spectroscopy of two A supergiant systems in the Small Magellanic Cloud with novel properties
aa r X i v : . [ a s t r o - ph . S R ] M a y Mon. Not. R. Astron. Soc. , 1–32 (2007) Printed 16 July 2018 (MN LaTEX style file v2.2)
High dispersion spectroscopy of two A supergiant systems inthe Small Magellanic Cloud with novel properties
R.E. Mennickent, ⋆ M. A. Smith Universidad de Concepci´on, Departamento de Astronom´ıa, Casilla 160-C, Concepci´on, Chile Department of Physics, Catholic University of America, Washington, DC 20064, USA; Present address: Space Telescope Science Institute, 3700San Martin Dr., Baltimore, MD 21218, USA
ABSTRACT
We present the results of a spectroscopic investigation of two novel variablebright blue stars in the SMC, OGLE004336.91-732637.7 (SMC-SC3) and the peri-odically occulted star OGLE004633.76-731204.3 (SMC-SC4), whose photometricproperties were reported by Mennickent et al. (2010). High-resolution spectra in theoptical and far-UV show that both objects are actually A + B type binaries. Threespectra of SMC-SC4 show radial velocity variations, consistent with the photomet-ric period of 184.26 days found in Mennickent et al. 2010. The optical spectra ofthe metallic lines in both systems show combined absorption and emission compo-nents that imply that they are formed in a flattened envelope. A comparison of theradial velocity variations in SMC-SC4 and the separation of the V and R emis-sion components in the H α emission profile indicate that this envelope, and probablyalso the envelope around SMC-SC3, is a circumbinary disk with a characteristic or-bital radius some three times the radius of the binary system. The optical spectra ofSMC-SC3 and SMC-SC4 show, respectively, He I emission lines and discrete BlueAbsorption Components (“BACs”) in metallic lines. The high excitations of the He Ilines in the SMC-SC3 spectrum and the complicated variations of Fe II emission andabsorption components with orbital phase in the spectrum of SMC-SC4 suggests thatshocks occur between the winds and various static regions of the stars’ co-rotatingbinary-disk complexes. We suggest that BACs arise from wind shocks from the Astar impacting the circumbinary disk and a stream of former wind-e ffl ux from the Bstar accreting onto the A star. The latter picture is broadly similar to mass transferoccurring in the more evolved (but less massive) Algol (B / A + K) systems, exceptthat we envision transfer occurring in the other direction and not through the inner c (cid:13) Mennickent & Smith
Lagrangian point. Accordingly, we dub these objects prototype of a small group ofMagellanic Cloud wind-interacting A + B binaries.
Key words: stars: early-type, stars: evolution, stars: mass-loss, stars: Ae stars:variables-others
Mennickent et al. (2002, hereafter M02) have reported the existence of a number of bright bluestars in the Small Magellanic Cloud the light curves of which exhibit periodic or quasi-periodicvariability in their OGLE (Udalski et al. 1997) I -band light curves. In an initial follow up of the in-vestigation, Mennickent et al. (2006; “M06”) noted that some stars exhibit peculiar spectroscopicproperties and multiple periods. They gave initial estimates of spectral classifications based onlow dispersion spectra, but they noted conflicting properties. In a companion study to this one,Mennickent et al. (2010; hereafter “M10”) have selected two novel Type 3 variables with novelproperties named OGLE004336.91-732637.7 ( ≡ SMC-SC3-63371, MACHO ID 213.15560; here-after SMC-SC3) and OGLE004633.76-731204.3 ( ≡ SMC-SC4-67145, MACHO ID 212.15735.6;hereafter SMC-SC4). These Type 3 stars were chosen according to the sole additional arbitrarycriterion that they are brighter than m v = ff ects from a secondary occur during brief periastron passages. The light curveof SMC-SC4 reveals surprising periodic deep eclipses every 184.26 days, modulated over a su-perperiod of several cycles, and chaotic variability indicative of multiple shorter periods.M06 and M10 have open a variety of questions that can be pursued with high resolution spec-tra, which we undertake to pursue now. We have already remarked as one example that the spectraof SMC-SC3 and SMC-SC4 spectra are unusual in that they show properties of rather conflictingspectral types (strong Na D and strong H α and H β emission components). Therefore, we obtainedthree high-dispersion optical spectra of SMC-SC3 and SMC-SC4 during the period 2002-9. To ⋆ E-mail: [email protected]. Based on observations carried out at ESO telescopes: ESO proposal 69.D-0391(A), backup targets. Mennickent Type-3 variables are SMC Be star candidates with I -band light curves varying periodically or quasi-periodically.c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Table 1.
Summary of observations. The MJD numbers at mid exposure are given. The periods detected in photometric data are also given. Phasesrefer to the ephemeris given in Mennickent et al. (2010). CD refers to cross-disperser.object periods (d) instrument UT-date airmass ∆ λ (Å) grating exptime (s) mjd-obs phase S / N SMC-SC3 238, 15 UVES 2002 / /
17 1.96 3100–8500 CD / /
17 1.91 3100–10400 CD / /
17 1.86 3100–8500 CD / /
17 1.81 3100-10400 CD / /
09 1.43 3390–9410 echelle 500 54413.03654 0.69 25SMC-SC4 MIKE 2007 / /
09 1.42 3390–9410 echelle 500 54413.04421 0.10 40SMC-SC3 Echelle 2009 / /
25 1.44 3940–7490 echelle 2000 55068.24680 0.44 17SMC-SC4 Echelle 2009 / /
25 1.40 3940–7490 echelle 2000 55068.27751 0.66 17
Table 2.
Summary of
FUSE
Observations. Phases refer to the ephemeris given in Mennickent et al. (2010).object UT-date UT-start exptime (s) mjd-obs phaseSMC-SC3 2006 / /
05 04:05:06 2430 54013.18427083 0.01SMC-SC4 2006 / /
30 11:36:26 2736 54069.49946759 0.24 obtain a better understanding of how the spectra appear at short wavelengths we obtained a
FarUltraviolet Spectroscopic Explorer ( FUSE ) spectrum of both stars. These spectra and the
FUSE data form the basis of this investigation. The goal of our study is to elucidate the properties of thecircumstellar environment of these objects and to pave the way to understanding the evolutionarystate of this potential new subclass of Ae stars.
High resolution spectra were obtained on 2002 May 17 with the ESO Ultraviolet-Visual EchelleSpectrograph in dichroic modes at the UT 2 telescope in the ESO Paranal Observatory, Chile. Thethree CCD chips on this instrument allowed sampling the spectral range of 3100–10400 Å. A slitwidth of 1” allowed us to obtain spectra at resolving power ∼
40 000. The spectra were normal-ized to the continuum, and no flux calibration or telluric correction was needed. Additional spectrawere obtained with the
Magellan Inamori Kyocera Echelle (MIKE) spectrograph at the Clay Tele-scope in Las Campanas Observatory, Chile on 2007 November 9. This double echelle spectrographprovided wavelength coverage of 3390 − − − c (cid:13) , 1–32 Mennickent & Smith were obtained at phases 0.23, 0.10, and 0.66, respectively, in the 184.26 day period identified inM10. The zeropoint is taken as the mid-eclipse in the light curves, i.e., at the light minima. ForSMC-SC3 the spectra were taken at phases 0.28, 0.69, and 0.44, according to the 238.1 day periodidentified in M10.In addition, far-UV spectra of SMC-SC3 and SMC-SC4 were obtained through the large sci-ence aperture of the
Far Ultraviolet Spectroscopic Explorer ( FUSE ) in 2006 October-Novemberunder GO Cycle 6 Program F907, as detailed in Table 2. A spectrum of OGLE005100.18-725303.0was obtained on 2006 October 6, as a quasi-standard. The
FUSE spectra cover the continu-ous wavelength range 929-1188 Å over eight MAMA detector segments. The
FUSE spectrumis recorded on two of these segments, thereby insuring the acquisition of two independent andsimultaneous spectra. The spectral resolving power among these varies with wavelength but istypically in the range 15 000-20 000. These spectra were reduced with the CalFUSE version 3.1.8pipeline system, which was similar to version, v3.2, which was used for the final reprocessing ofthe
FUSE archive. Examination of auxiliary files that display the positions of photon events onthe detector in the spatial direction confirm that they originate from an e ff ective point source. Thisfact e ff ectively rules out that the ultraviolet fluxes are contaminated by a nearby comparably brightsource in the science aperture. The optical spectra of the two program stars are complicated, and it is therefore helpful to describetheir general properties before carrying out a quantitative analysis of their spectral features. Thesespectra are redshifted by amounts of roughly 2 Å, which is typical for optical spectra of membersof the SMC.
In many respects the optical spectra are similar to the spectrum of a typical A supergiant. Thisstatement is demonstrated in Figure 1, which shows the high-level Balmer lines and the blue-greenUVES spectra of SMC-SC3 and SMC-SC4 and of the Galactic A5 II star HD 74252 from theUVES atlas (Bagnulo et al. 2003). The cores of the hydrogen and metallic lines in the spectraof the program stars are influenced by two competing e ff ects. On one hand, the strengths of themetallic lines are weakened in our objects due to their low metallicities. This e ff ect is largely com-pensated by the presence of a substantial disk contribution, the details of which we present below. c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Figure 1.
The optical UVES spectrum of the target stars and the Galactic A5II standard HD 74252 taken from the UVES atlas covering (left panel)the high level Balmer lines and (right panel) the metallic lines in the region 4900-5450 Å. The standard and SMC-SC3 have been o ff set in fluxfor convenience, and SMC-SC3 and SMC-SC4 have been blueshifted to the rest frame. The general appearance of the metallic lines shown forthe two SC targets is similar to that of the standard’s lines. The Fe II lines 4923 Å 5018 Å 5176 Å and 5316 Å, are discussed in the text. Carefulinspection shows that the wings of these particular lines show emission in SMC-SC3. The cores of both hydrogen and metallic lines are stronger in SMC-SC3 than in SMC-SC4. De-tailed inspection shows that most of these lines arise from once-ionized Fe-like ions and of mixedstages of light metals (Mg I, Si II, Ti II, Ca I), and these will be discussed in § § α emission. In contrast to the merelyfaint Beals Type I P Cygni emission, e.g., in the Bagnulo et al. (2003) spectrum of the A6 Ia stan-dard HD 97534 the H α profiles are not only strong but double peaked with nearly equal “V” and“R” emission components - see Figure 3. The strengths of these components are shown in Table3. These H α profiles also resemble those of Be stars at the high end of their distribution of H α strengths, although they are not quite extraordinary among Be stars. These attributes indicate thepresence of a flattened, Keplerian disk. In contrast to the SMC-SC4 profiles, the H α emission ofSMC-SC3 decreased over time, suggesting that these changes were monotonic over time and notdue to stellar or binary activity.The optical spectrum of SMC-SC3 is likewise well populated with sharp lines of light and Fe-group elements. The Fe II lines, those arising from levels at 2.9 eV are particularly prominent. Allof these are characteristic of the spectrum of a Be star disk seen at high inclination. In a few casesdiscussed below weak symmetric emissions are superposed on some of the strongest Fe II lines c (cid:13) , 1–32 Mennickent & Smith
Figure 2.
A comparison of the Na I D doublet lines in our UVES spectra, smoothed over 4 points. These systems are placed together to showthe relatively symmetric profiles for SMC-SC3 and the great variations for the profiles of SMC-SC4. In the latter star the blue / red wings appeardepressed in the 2002 and 2009 spectra, respectively, but the MIKE 2007 spectra shows that these shadings are caused by blendings of BACsubcomponents of widely varying relative strengths. The zeropoints of the velocity system are referred to the rest frames of the D2 (5896 Å), and inso doing the various radial velocities of the primary star have been removed. The counts for the SMC-SC4 spectra have been displaced downwardby 150 units for convenience. The emission features in the LCO spectra of SMC-SC4 are telluric D line emission. no r m a li z ed f l u x + c on s t an t wavelength (A) SMC-SC3 2001SMC-SC3 2007SMC-SC4 2001SMC-SC4 2009SMC-SC3 2009SMC-SC4 2007 Figure 3.
The UVES (2002), MIKE (2007) and LCO (2009) spectra depicting the H α profile for both the program stars. Fluxes are normalized toa unit continuum level (with an o ff set of -25 for SMC-SC4), and heliocentric corrections have been applied. in the optical spectrum. However, unlike the spectra of B[e] stars, lines arising from metastablelevels are not present, nor are [Fe] emission lines. Since these are a defining traits of B[e] stars,this star cannot be a B[e] (or “A[e]”) star.In general the fitting of disk components to the metallic lines can be performed to computerough column densities by assuming a local disk temperature. The spectrum of SMC-SC4 di ff ersfrom the case of SMC-SC3, and for that matter from the spectrum of nearly all other known A Table 3.
Equivalent widths ( EW ), maximum intensity relative to the continuum ( I / I c ) and full width at half maximum ( FWHM ) for the H α emission line in 2002, 2007, and 2009. Object - EW (Å) I / I c FWHM (Å)SMC-SC3 112-104-72 24.1-22.7-17.1 4.39-4.46-4.41SMC-SC4 28-26-33 6.3-5.7-7.1 5.16-5.25-5.02 c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties − from what we will call the main (or red)sharp component, the velocities of which adhere closely to those of the hydrogen line cores inthe spectrum. Among the metallic lines the BACs are generally of comparable strength and oftenstronger than the main components. They tend to be barely visible or absent in weak resonancelines, e.g., in the region 3820-3850 Å. In the strong Fe II lines and the Na I D doublet a secondaryBAC is also present at approximately -100 km s − relative to the main component. Section 3.7.1will be devoted to a discussion of these novel features. / K features
Despite the comingling of photospheric and disk features in many lines of the optical spectrum, aspectral type for our stars can be determined from the Ca H or K lines (with correction for metal-licity) and the wings of the hydrogen lines. The high level Balmer and Paschen lines are consistentwith a middle A-type spectral type. However, these features can vary significantly among spectraof A-type standards separated in type or luminosity class. As already mentioned, the emission inthe lower members also interferes with a comparison to profiles from spectral standards or synthe-sis models. Consequently, we found the Ca II K line to be a more reliable indicator of the positionsof the stars in the HR Diagram. The wings of this line are sensitive to electron pressure and thus lu-minosity class. The hydrogen lines and Fe-line metal lines already guide the spectral classificationto the middle A range.Figure 4 exhibits matches we have found for the K lines of SMC-SC3 and SMC-SC4 with thespectral standards HD 97534 (A6 Ia) and HD 74272 (A5 II) from the UVES atlas, again from theUVES atlas. We have likewise fit the spectra with SYNSPEC models for T e ff = =
3. The explicit assumption we made in our fittings was that there is no measurable contribu-tion at this wavelength from another continuum source such as a secondary star. In general, thefits are very good, except that the models predict a narrower core than is observed in SMC-SC3or even the core of the Ia standard (Fig 4a). In addition, the blue core of both the SMC-04 andclass two standard shows an apparent “emission” feature” that the model atmospheres (with nochromosphere) do not produce (Fig. 4b). Otherwise, we estimate the precision to be ± ± K in T e ff and ± . c (cid:13) , 1–32 Mennickent & Smith
Figure 4.
The spectral region surrounding the Ca II K line in our two program stars from our MIKE (2007) spectra. Spectra of the comparison starsHD 97534 and HD 74272 are represented by dashed lines. Fits from SYNSPEC models using Kurucz atmospheres are given by the dotted lines.
Figure 5.
The far UV spectrum covering 1000-1070 Å for the B3 II quasi-standard of NGC 330-30 and one of our program stars, smoothed over8 points. The spectrum of the standard is o ff set vertically for clarity. Several molecular H and atomic interstellar (“ISM”) lines are noted. Weidentify photospheric lines by coded letters. These codes have the following meaning: a) Fe III-Cr III 1007Å, b) C II 1010.4 Å, c) S III 1015.5 Å, d)Fe III 1018.4 Å, e) Fe III 1021.7 Å, f) C II 1036-1037 Å, g) Fe III 1039.9 Å, h) Cr III 1042.9 Å, i) Fe III 1045.2 Å, j) Mn III 1055.5 Å, k) Fe III1058.8 Å, k) Fe III 1060.7 Å, l) Si IV 1066 6 Å, and m) S IV-Mn III 1073 Å. The dominant Lyman β line is also indicated. spectrum including the Ca K and hydrogen lines to be due to middle A supergiants, specificallytypes near A6 I and A5 II, respectively. In addition we judge that we should just be able to discerna 6% (3 magnitudes) contamination to the K line from flux of a contributing secondary. Because the program
FUSE spectra were obtained through the instrument’s Large Science Aper-ture, we were able to estimate the far-UV flux in addition to surveying the spectral line strengths.The spectral lines are consistent with an early to mid B type spectrum and inconsistent with pre-dictions for an A star. We concluded from this that the far-UV flux is emitted by a secondary starthat is too faint to show visible lines in the optical range. Our initial reconnaissance of the hydro-gen Lyman and the metallic lines suggested that the spectrum is representative of a star of type c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties -14 -14 -14 F l u x SMC03 (in red)
B2B3B4 -14 -14 -14 -14 F l u x SMC04 (in red)
B2B3B4
Figure 6.
FUSE ) spectra for the region surrounding the C III 1176 Å feature in our two program stars, represented by the red line, and in threeGalactic stars representing spectral types B2, B3, and B4 near the main sequence. For the panel a) these stars are HD 37367, HD 45057, andHD 201836. For panel b) we have substituted the spectrum of the B3 star NGC 330-B30 for the spectrum of HD 45057. Both B3 star spectra givemuch the same comparison. The dotted line is the SYNSPEC spectral simulation for a T e ff =
18 000K, log g = / h] = -0.7 model. B0 to B5. To compare our far-UV spectra with those of other metal-poor early B stars fainter thansupergiants, we canvassed the entire
FUSE archive. We found only four SMC stars within a factorof ten of the same far-UV brightness and with O9-B5 spectral types quoted in the literature (M02,Evans et al. 2006, Blair et al. 2009). One of these stars is in the SMC cluster NGC 330 and islisted by Evans et al. as B30. The second star, HV 1620, is a well known O9 eclipsing binary. Twostars are SMC OGLE survey stars that we selected in 2006 for
FUSE observations. An analysisof these OGLE stars has not been published and we note specifically that the spectral type givenin the
FUSE archives as “B2 V” for OGLE005745.25-723532 was not based on a spectrum andis not to be trusted. The optical spectrum of OGLE005100.18-725303 was discussed and given inM06. Table 4 gives the coordinates, m v , and spectral types where possible for these stars. Of thesefour stars with known far-UV fluxes only the flux of HV 1620 is comparable to the fluxes of ourprogram stars. However, this is a late O star and is clearly not a good match to the far-UV spectraof our program stars. The other three are some 8-10 times brighter, suggesting that they have eitherearlier types, higher luminosity classes or both. The Ly β lines and other metallic features in thespectra of the OGLE stars and HV 1620 are weaker than those of our program stars, suggestingthat their spectra lie in the O9-B1 range.In Figure 5 we show a comparison of the spectra over the range 1000-1070 Å for each ofthe two program stars and NGC 330-B30. Although these spectra are o ff set for clarity, overplot-ting them against one another shows that the wings of the dominant Ly β line are a close match, c (cid:13) , 1–32 Mennickent & Smith
Table 4.
Candidate early-type B stars with FUSE spectraObject RA Dec m v Sp. TypeNGC330-B30 00 56-09.4 -72 27 58.9 14.22 B3 IIIHV 1620 00 54 38.6 -72 30 04.2 14.08 O9 V + O9 IIIOGLE005100.15-725303 00 58 00.1 -72 53 03.9 13.56 B1 II-IIIeOGLE005745.25-723532 00 57 45.2 -72 35 32.0 13.82 – suggesting that this star is close to B3 in type. We depict also identifications for 13 available pho-tospheric lines, as taken from the Far-UV Spectral Atlas of B near the Main Sequence (Smith2010). A comparison of the metallic line ratios of the Si IV and S III relative to the Fe III andCr III lines, and in turn to the C II 1037 Å line also suggests nearly equal spectral types for B30and both program objects. In addition, this is also true of the possible luminosity class diagnosticSi III 1113 Å / Si IV 1066 Å (Pellerin et al. 2002, Smith 2010). The value of this ratio is ∼ > B and V magnitudes of about 17. This is consistent with the stars’ positions on the upper edge ofthe main sequence, or approximately luminosity class III and a log g of 3.5 to 4. Finally, we pointout that the far-UV flux of SMC-SC3 is some 0.57 magnitudes brighter than SMC-SC4, which isprobably due to a combination of brighter luminosity and / or slightly earlier type.With new luminosity class estimates determined, we repeated the spectral synthesis on the C IIIaggregate at 1176 Å. According to the Pellerin et al. (2002) and Smith (2010) spectral atlases ofB stars, this feature undergoes a maximum for B0-B2 V spectra. Spectra of stars just hotter andcooler than this maximum show telltale secondary lines in the wings of this aggregate that allowone to distinguish between an O9 and a middle-B spectral type. We found in our spectra thatdespite the stars’ low metallicities these features already have strengths close to the maximumexhibited in Galactic spectral standards. In addition, owing to the decrease in Stark broadening,the C III components begin to resolve even for moderate rotations. However, the components arenot resolved in our spectra. These considerations suggest that the spectral types are in the rangeB0-B4. This is in accord with the comparison of the various indicators just discussed. In Figure 6we exhibit spectra in the region of the C III aggregate compared to B2, B3, and B4 luminosity classV standards. To provide some redundancy we show in Fig. 6b the spectrum of NGC330-B30 as acomparison to SMC-SC4 instead of the Galactic B3 star (Ski ff c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties
11C III aggregate for a representative photospheric model (T e ff = = = -0.7).The fit is also good.If we regard NGC 330-B30 and OGLE005100.18-725303.9 as secondary spectral standards fora low-metallicity early-type B star, the similarity of their profiles with the profiles of our programstars indicate that the profiles of the latter are consistent with isolated stars, although there is ahint of weakening in the C III feature of SMC-SC3. This inference is confirmed by the agreementof the synthesized C III profile with the SMC-SC4 feature, with again a hint of weakening of aseveral percent in the case of SMC-SC3. The depths of these lines suggest that the contribution ofthe A primary can be no more than a few percent. We consider it highly unlikely that our targetsare interlopers happening to be visible in the FUSE aperture. Rather, it appears that the B starsare secondaries with bright giant or supergiant A companions and that they are representativesof binaries caught at similar evolutionary stages. Our discussion of radial velocity variations tofollow confirms this inference for SMC-SC4. In this picture the B and A components should haveapproximately the same stellar mass. In addition, in either of the two wavelength regions wherethe spectrum of one or the other binary component dominates, we cannot detect the flux from theother.According to the model atmospheres we used, a B3 V star’s flux (assumed to be T e ff =
18 000 Kand log g =
4) at a wavelength of 3933 Å should be five times the flux of an A5 (8 500 K) of thesame radius. Recall that we noted above that we should be able to barely discern a hypothetical6% contamination to the K line from the B star. This means that the A star has to have a radius ofsome five times the B star’s for the B star not to dilute the K line flux. This is consistent with thestellar radii of middle A supergiants provided that the B star has a radius of ∼ ⊙ . This gives aradius of ∼ ⊙ for the A supergiant. This is consistent with values in the literature (e.g., Verdugoet al. 1999) and the mass estimated from evolutionary tracks of ≈ ⊙ in M10, and for simplicitywe will take 9 M ⊙ for the secondary mass too. Based on our models, even with its larger radius theA primary should not contribute to the flux at 1176 Å. We conclude from this analysis that eachcomponent of these binaries overwhelms the flux of the other in its dominant wavelength regime.The primary conflicting feature with this overall description, as already noted, is the strong H α emissions in both optical spectra, which is usually an attribute of a Be rather than an Ae star. c (cid:13) , 1–32 Mennickent & Smith
Before conducting a quantitative analysis of the spectral lines in the disk, it was necessary to firstidentify them and then determine their radial velocities. To obtain radial velocities from the opticalspectra, we correlated measured wavelengths of identified lines with their theoretical wavelengths(Kurucz 1993). The results are summarized in Table 5. For SMC-SC4 we give the mean of the“main” (red) component (which is taken as the velocity of the A primary), as well as the ve-locity of the blue component. To verify the wavelength systems from our Th-He-Ar comparisonspectra, we utilized several telluric lines from the oxygen B-band (absorption) and the GalacticISM component in the Na I D lines. The internal errors, measured by weighting the internal rms ’sfor individual ions, do not exceed ± − , and so we quote ± − . SMC-SC3 shows aweighted average velocity of 105, 106 and 108 km / s in the 2002, 2007 and 2009 spectra, i.e., itshows virtually no radial velocity (RV) variability. The spectra of SMC-SC4 disclose mean RVsof of + +
108 and +
161 km / s at these epochs. Although the RVs are clearly variable, we willexercise considerable caution in interpreting these di ff erences below.We noticed moderately strong V and R emission components in the lower members of boththe Balmer and Paschen series. The relative strengths of these features decrease to invisibility atH ζ . These lines are flanked by absorption wings. Typically the blue peak is stronger, and they havea deep central absorption. As Table 6 shows, the central absorption cores are found blueward ofthe centroid of the V and R emission peaks, typically by 4-8 km s − . This table displays smalldi ff erential shifts in these cores as one progresses up the Balmer sequence from H β and a smallerpeak separation at lower order Balmer lines. The Balmer emission decrement is steep, meaningthat emissions quickly drop o ff from H α toward the intermediate Balmer lines. All the above aresignatures of quasi-Keplerian optically thin H α -emitting disks.In most metallic lines of SMC-SC4, remarkably, we also found sharp blue absorption compo-nents (discussed later as “BACs”). These have a blue shift with respect to the primary absorptionline components. These shifts average -48, -50 and -36 km s − for the 2002, 2007 and 2009 spectra,respectively. We noticed that in 2009 the BAC is stronger than the main component for the Na Dand Fe II group lines. c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Table 5.
Summary of heliocentric radial velocities (in km / s). For SMC-SC4 we give the velocity of the main component and / or the associatedBAC. The number of lines included in the averages is also listed. Errors reflect the rms of the RVs per line within an ion. For H I we considercentral absorptions in the higher level Balmer lines. The mean RVs exclude He I lines.Ion SMC-SC3 SMC-SC3 SMC-SC3 SMC-SC4 SMC-SC4 SMC-SC42002 2007 2009 2002 2007 2009CaI - 108 ±
12 (2) - - - -CaII - 106 ± ± ± ± ±
11 (10) 105 ± ±
10 (3) 55 ±
11 (6);101 ±
10 (3) 125 ± ± ± ±
12 (27) 116 ±
10 (4) 52 ± ± ± ± ±
10 (22) 106 ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ± ±
24 (3) - - 48 ±
16 (4); 94 ± ± ± ± ± ± ±
27 (7) 116 ± ± ±
10 (5); 119 ±
13 (2) 169 ± ±
13 (2) - - 60 ± ± ±
26 (7) 107 ± ±
12 (7) 164 ±
10 (5)TiII 106 ± ± ± ± ± ± ±
16 (9) 129 ± ±
15 (12)mean 105 ± ± ± ±
2; 100 ± ±
7; 106 ± ±
5; 166 ± Table 6.
Heliocentric radial velocities (in km / s) for H I emission line components and shift of the central absorption relative to the centroid of thered and violet emission peaks for years 2002 / / ∆ λ ) in higher order lines.Emission line star blue-peak central-abs red-peak shift ∆ λ (km / s)H δ SMC-SC3 - / / NP - / /
106 - / / NP - / -2 / - - / / -H γ SMC-SC3 - / /
21 - / /
92 - / /
190 - / -6 / -14 - / / β SMC-SC3 24 / /
27 95 / /
94 159 / /
160 -4 / -7 / / / α SMC-SC3 55 / /
41 93 / /
91 153 / /
142 -11 / -8 / -1 98 / / δ SMC-SC4 - / NP / NP - / /
140 - / NP / NP - / - / - - / - / -H γ SMC-SC4 - / NP /
48 - / /
110 - / NP / NP - / - / - - / - / -H β SMC-SC4 46 / NP / - 105 / /
101 183 / /
185 -16 / - / - 137 / - / -H α SMC-SC4 55 / /
51 114 / /
106 171 / /
170 -1 / -4 / -4 116 / / We used the Far-UV Spectral Atlas of B Stars Near the Main Sequence (Smith 2010) to identifya number of lines in our
FUSE spectra to measure radial velocities of the B stars at the timeof our observations. The results for SMC-SC3 and SMC-SC4 are + ±
12 km s − and + ±
15 km s − , respectively. FUSE spectra are prone to several systematic sources of errors in theirwavelength calibrations (Dixon et al. 2007), and thus the zeropoint di ff erences may easily be aslarge as ±
20 km s − . Nonetheless, we noticed that the measured wavelengths of the same lines forthe same spectra closely coincided in nearly all cases. On the basis of the spatial positions of thetwo dimensional spectra, we believe the apertures were placed consistently over the stellar imagesand that the radial velocity di ff erence between the B star spectra is no more than 10 km s − . c (cid:13) , 1–32 Mennickent & Smith
Much of the analysis in this paper relies on spectral line synthesis, so we first describe the methodsused for this analysis. We have adopted the results of the fitting of the wings of the Ca II K-lineand assumed stellar parameters for SMC-SC3 and SMC-SC4, namely log g = × solar (e.g., Dolphin et al. 2001, Mighell et al. 1998, Dufton et al. 2005). The spectra were analyzedusing the synspec and circus (disk) spectral synthesis programs (Hubeny, Lanz, & Je ff ery 1994,Hubeny & Heap 1996).Using circus , we were able to estimate rotational broadenings, V sin i , from the absorptionlines in both the FUSE and optical spectra of our two program stars. For SMC-SC3 the A primaryand B secondary have broadenings of 20 ± − and 50 ±
10 km s − . For SMC-SC4 the corre-sponding rotational broadenings for the A and B binary components are 28 ± − and 75 ± − .For circus analysis one must assume an input temperature for a putative intervening disk struc-ture, T disk , and several other parameters. The circus program permits as many as three separateabsorbing or emitting regions to be specified, each with its own areal coverage, local temperature,and radial velocity. We utilized this feature in some cases to specify two distinct temperature re-gions, since the line identifications indicated that the lines we modeled arise in regions having atemperature range of 5-8 kK. At these temperatures the fraction of the ions considered plateaus,and thus the errors in this parameter are not large. The strengths of disk components of the Na Dand K I profiles can be most easily modeled by temperatures of 5 kK or less, consistent with thepicture from the H α emission profiles of the disks extending out to at least several stellar radii.Nonetheless, our models show that the column densities required to fit these components decreasemonotonically with temperature. Thus these values are not well constrained. The strengths andwidths of the Fe II and Si II lines suggest a microturbulence ξ of about 10 km s − , and we assumedthe disk to be approximately Keplerian. For lines having both photospheric and circumstellar com-ponents, these components exhibit no net shift. Therefore, the latter structure appears to be in aKeplerian orbit. We have also assumed that the foreground disk segment fully covers the star.However, if our lines of sight toward di ff erent portions of the star sample di ff erent disk conditionsalong the way, our column densities will be underestimated.These assumptions required the matching of the computed equivalent widths to the observa-tions using a one or two parameter fit in disk and column densities. The columns needed to fit the c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties emission profiles in our modelswith temperatures for a medium with a larger projected area than the star’s - that is, as if they areformed in lines of sight that do not intersect the star disk. We will express the column densities inexamples below by using the unit “stellar area.”Our modeling represents an oversimplification of the true and unknown disk conditions be-cause the disk geometry and velocity fields are poorly known. In addition, the compututation ofline strengths requires a common estimated excitation and ionization temperature. Nonetheless,for exploratory purposes we believe that our models provide insight into the thermal and velocitydi ff erentiation within the disks. Generally, departures from LTE in the line transitions will resultin an underpopulation of the higher levels relative to lower and resonance levels, and therefore anincrease in the absorbing column density for lines arising from excited levels. We have computedour absorption column densities in this scattering approximation in our CIRCUS simulations. In view of the implied complex nature of the structures surrounding both program stars, we willdiscuss quantitative fits to several line profiles mainly to get an idea of the column densities, radialvelocities, and turbulences in the disk gas. Except for a model fitting to the excited He I line andresonance lines (which form in cold media which we cannot specify accurately), we will limit ourquantitative analysis to the optical metal lines arising from atomic levels of a few eV. Because ofthe uncertainties involved, we discourage the reader from not interpreting our numerical resultstoo literally.Nearly all the features in the optical spectra of SMC-SC3 are characteristic of absorptions inan extensive and flattened, di ff erentiated disk or envelope. The high-level Balmer and Paschenlines can be resolved out to H30 (Fig. 1a) and P24, but any disk contribution to them cannot bewell determined because these features are common in A supergiants spectra. Our SYNSPECsimulations of these high atomic levels lead to estimates for the characteristic density in the lineformation in the atmosphere of 1-3 × cm − .In our initial modeling we discovered that the disk temperature and therefore the column den-sity through the disk along the line of sight cannot be constrained to a single temperature, partic- c (cid:13) , 1–32 Mennickent & Smith ularly for metallic absorption lines in the blue / near-UV region. For example, given assumed diskgas temperatures values of 6 kK and 5 kK, we are able to generate good CIRCUS fits to weaklines in the 3800-3850 Å region in our spectra with column densities of about 3 × cm − or7 × cm − , respectively. These values carry uncertainties of at least a factor of three and assumefull disk coverage of the star. We were able to get a better handle on these parameters by findinga region of the spectrum at 5250-5280 Å that contains both Fe I and Fe II intermediate strengthlines. Our modeling for the region of 5250-5280 Å is depicted in Figure 7. The ionization tem-perature in the disk is already well constrained by the Fe I / Fe II ratio to be ≈ disk = cm − , consistentwith our less refined analyses of the lines in the far blue.A salient attribute of the strongest Fe II lines of SMC-SC3, (which arise from a common mul-tiplet having 2.9 eV) is that they exhibit broad absorption components and even broader emissions.Certainly, the emission components and even much of the absorption does not appear to be pho-tospheric. For example, among the strongest three lines, 5169 Å, 5018 Å, and 4923 Å the strongerthe line, the more pronounced the absorption and overlying emission wings, a point which we nowelaborate.Among the strong disk Fe II lines in this spectrum, we chose to fit the 5018 Å line, which isdepicted in Figure 8. Our fit for this feature was achieved by confirming the column density of adisk from the emission component of a few Fe II lines. These lines are 4923 Å, 5018 Å, 5169 Å,5316 Å. We estimated from their relative emission strengths that that they are mildly opticallythick, τ L ∼
3. For the optical spectrum of SMC-SC3 this corresponds to a column density of3 × cm − and a temperature of 8 kK. This is the same temperature and nearly the same columndensity as we found for the lines in Fig. 7. This agreement allowed us to fit the absorption featuresof these Fe II lines straightforwardly. The full profile was fit in CIRCUS with a two temperature-component simulation. We fit the emission using the same column density and temperature as justfound for the absorption component. Next we needed to broaden the simulated emission featureby a gaussian macroturbulence of 70 km s − . The amplitude of the emission required adding a freeparameter, the emitting gas area, which we found to be 11 stellar areas in this case. We note here c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Figure 7.
The fit of yellow Fe-like lines for SMC-SC3 (upper) and SMC-SC4 (lower) of our UVES spectra to a two-temperature (8 kK, 6 kK)model. For the warmer component the column densities were 5 × cm − and 2 × cm − for SMC-SC3 and SMC-SC4, respectively. Thecolumns for the cool component were 3 × cm − and 1 × cm − . The weak feature appearing at redshifted wavelength 5258 Å, not present inour simulation, is likely to be an Fe I line the published oscillator strength of which is inadequate. The Fe I 5255 Å line arises from a 10 eV level.In the SMC-SC4 this feature exhibits weak redshifted emission. Figure 8.
A fit of the Fe II 5018 Å line (UVES spectrum) for SMC-SC3. This profile was fit by two independent models, one with a temperatureof 8 kK (see text for other details). The second component was fit to the emitting wings with the same gas temperature and by assuming an area of11 stellar areas and a gaussian turbulence function of 70 km s − . Wavelengths are in the observed system for the 2002 epoch.c (cid:13) , 1–32 Mennickent & Smith that we actually have no independent handle on the volumetric density of the matter in which theFe II lines are formed. Using the column lengths of ∼ ∗ , characteristic of the derived emittingareas ( ∼
11 stellar areas), we can estimate characteristic densities of 3-10 × cm − . Values of thisorder of magnitude should not result in visible forbidden features, and indeed they are not seen.The Si II and O I triplet in the SMC-SC3 spectrum are formed in plasma with temperaturesin the range 6-9 kK as well. Although their cores are sharp, and consistent with being formed in amedium distant from the star, their strong wings suggest a second component formed closer to thestar. The Na I D and K I multiplets in this spectrum have the same combination of sharp core andextended wings. This is also true for several of the Fe II lines and the Si I 6347 Å, 6371 Å doublet,which have high optical depths and therefore characteristic mean formation sites in the outer diskwhere the temperature and column length are low. For the Na I and K I lines we could fit the linecores with a column above the photosphere of 5 × cm − for T disk = × cm − for 5 kK.We address next the remarkable emissions in the He I 5876 Å, 6678 Å, and 7265 Å lines, allof which arise from 21 eV levels. We show the stronger of these lines, 5876 Å and 7265 Å, inFigure 9, corrected for their mean redshift of 32 km s − . These lines are the only ones in the spectraof both stars that di ff er markedly from the other line system(s). The observed 5876Å / ffi ciencies of formation in media of 15 to 23 kK,with corresponding emission regions encompassing of 8-12 to 3-4.5 stellar areas and column den-sities of no more than 1-5 × cm − for the two respective gas temperatures. If this heated gasundergoes continuous cooling down to 8 kK, it is likely, given the small volume of formation im-plied by the He I emission widths, that the resulting emission generated in the Fe II lines would behidden by the much broader Fe II emission component formed over the larger disk volume. Much of what we found for SMC-SC3 is true also for SMC-SC4, and we have already pointedout that the inferred temperature distribution for the disk around SMC-SC4 is indistinguishablefrom the SMC-SC3 disk. For the most part the column densities in our fits to the main (unshifted)metallic absorption line features run a factor of two to three times lower. For example, for a 6 kK c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Figure 9.
The regions of the He I 5876 Å and 7065 Å weak emission lines in the spectrum of SMC-SC3 in both UVES (2002) and LCO (2009)spectra. Here the velocity zeropoint is taken as the mean, -32 km s − , noted in Table 5 and is -74 km s − from the average velocity of the featuresin the optical spectrum. Note the near absence of velocity shifts between the two. The fluxes are 4-point smoothing of the raw data. model the 3800-3850 Å lines could be fit with column densities of about 7 × cm − and theFe II lines in the 5250-5280 Å region by some 2 × cm − . Most of the strong lines in this spectrum have profiles di ff erent from those of SMC-SC3. The mostnoticeable di ff erence is the presence of one and sometimes two discrete BACs in the SMC-SC4spectrum. These “BACs” are so named because their morphology is reminiscent of the “DACs” inresonance lines of radiation-driven winds of hot stars. A primary BAC generally occurs at about-50 km s − to the blue of a redder or “main” component, which we so name because the latter’svelocity are coincident with the RVs of the high level Balmer lines. The double entries in SMC-SC4 columns of Table 5 indicate the commonness of single BACs throughout the optical spectrum.The yellow optical region is rich in intermediate strength Fe II lines with χ ≈ × cm − , or several times that of themain (static) components. This is largely because these components are mildly optically thick.The pattern of two BACs is common among the strongest of these 3 eV Fe II lines. Figs. 11aand 11b exhibit double component pattern in the 5018 Å and 4923 Å lines, respectively. In bothfigures a “secondary” BAC shifted by about -100 km s − relative to the main (red) componentis conspicuous. The BAC strengths for the Fe II lines can change over time, and Fig. 11a showschanges for the 4923 Å line. As expected, the strengths of BACs among various lines arising fromthe same ion and comparable excitation levels increase with line strength.BACs are relatively weak in lines arising from near-ground state atomic levels in low ioniza- c (cid:13) , 1–32 Mennickent & Smith tion species such as Mg I, Fe I, and Ti II. They are particularly noticeable in lines arising frommoderately excited states of abundant metallic ions like Fe II, Mg I, Si II and even the O I 7771-5 Å triplet (9 eV) - see Figure 10. The excited lines in the N I 8703-48 multiplet ( χ =
10 eV) exhibita similar behavior as the O I triplet. Many of the strongest lines, including the high order Balmerlines, and the Na I D and K I 7699 Å doublets at first seem to show a shading toward one wing orthe other, depending on the epoch. However, as the 2007 MIKE spectrum makes clear in Fig. 2,this tapering is due to the blending of two BACs. In this spectrum the “tapering” is fully resolvedinto two sharp components.
In addition to BACs, absorptions and emission components appear often in the extreme wings ofthe Fe II lines. However, their behavior is in some respects stranger than the BACs. Both panels ofFig. 11 show that the flux in one of the wings can change from absorption to weak emission amongmembers of the same multiplet in the same spectrum. The di ff erence in the fluxes in the red wingsof the 4923 Å and 5018 Å lines is particularly surprising because their di ff erence in log g f is onlyabout 0.1 dex. These di ff erences can be seen even more dramatically in Figure 12a, again in thesame spectrum. Note in the blue wing the strongest line in a multiplet, 5169 Å, shows absorption.As one proceeds to the weaker members in the series, e.g., 5316 Å, the wing exhibits the strongest emission.The increased activity with time for the “weak” 5316 Å line is exhibited in Figure 12b. Thisfigure also shows weak absorption in the red wing, i.e., the opposite wing in which emission isseen. Other lines show this same behavior (e.g., Fig. 11b).As one proceeds to higher excitation lines the secondary BACs disappear, and often so dothe main components. Figure 13 shows vestigial red wing emissions of the excited Si II 6347 Å,6371 Å doublet similar to those just exhibited in the Fe II 5316 Å profile in Fig. 12b.In Figure 12a we point out the curious di ff erence between the BAC and emission strengths ofthe same 2009 spectrum: whereas the strengths of BACs of lines in the same multiplet or super-multiplet increase as expected with oscillator strength, the strengths of the emissions go in reverseorder, that is, the strongest line ( λ absorption and the weakest line showsthe strongest emission. This is depicted in Fig. 12a, and the reversal can be seen in spectra fromother epochs as well. This behavior sets the far wing features apart phenomenologically from theBACs. c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Figure 10.
Comparison of the O I triplet in the two program stars (UVES spectra). These lines, located at 7771.94 Å, 7774.17 Å, and7775.39 AA, arise from a level at 9 .1 eV. Note the broadening and the sharp, blueshifted BAC components in the SMC-SC4 spectrum.
Figure 11.
Panel (a):
Overplotting of the Fe II 4923 Å and 5018 Å lines from the UVES spectra of SMC-SC4, exhibiting two
Discrete BlueAbsorption Components (“BACs”). The 4923 Å line was redshifted by 94.7 Å. The red wing absorption (5018 Å line) and emission (4923 Å line)discussed in the text are present in this figure. MIKE (2007), and LCO (2009) spectra of the Fe II 4923 Å profile. The 2007 and 2009 spectra havebeen shifted to force the major absorptions to coincide. Red wing absorption and blue wing emission are present in the 2009 spectrum.
Figure 12. (a) Overplotting of Fe II 5169 AA, 5018 Å, 4923 Å, and 5316 Å (in order of decreasing strength) in the LCO (2009) spectrum. Eachof the other profiles have been coshifted to the wavelength centroid of 5018 Å. Note the increasing flux in the blue wing at 5019-5020 Å) withdecreasing line strength. (b) The Fe II 5316 Å line of SMC-SC4, exhibiting two
BACs; the MIKE (2007) and LCO (2009) spectra have been shiftedin wavelength to force the major absorptions to coincide. The emission activity at 5316.5 Å and 5318.5 Å is discussed in the text.c (cid:13) , 1–32 Mennickent & Smith
Figure 13.
A fit to the Si II 6347 Å and 6371 Å doublet (solid line) of the UVES (2002) spectrum, emphasizing the BACs and the filled in maincomponents for the SMC-SC4, such as also found in other strong Fe II lines and the excited Fe II 5255 Å line (e.g., Figs. 7, 11 & 12a). The observedspectrum is depicted for SMC-SC3 in order to guide the eye to the filled in red wing emission in the SMC-SC4 spectrum. The solid line fit wasattained with a model having a disk temperature of 8 kK, a column of 3 × cm − , and a projected surface of 1 stellar area. The 6371 Å profileshave been shifted by -14.9 Å with respect to the 6347 Å for convenience. We end this section by reporting that we encountered di ffi culties in attempting to fit the Fe IIline emissions strengths of Fig. 12 with our CIRCUS models, even though it was easy to fit asimilar emission in the Si II profiles shown in Fig. 13. The source of the problem is the combinedconstraints a ff ecting the formation of the Fe II lines. As one increases the gas temperature in themodels the Fe II emissions increase up to a point. However, above 9 kK the flux emitted by agiven gas volume decreases rapidly as the ionization of iron shifts to Fe ++ . Above this tempera-ture, regardless of the column density (the lines are optically thick already), the emission requiresincreases in the emitting volume and therefore the area. In practice, an emitting area of at least 10stellar areas is required – substantially higher values are needed for gas temperatures greater than10 kK or less than 8 kK. These simulations bring out a contradiction according to our assumptions: whereas the blue wing absorption in the strong λ ff erent temperatures. Clearly the forma-tion of various components of the Fe II lines in this spectrum occurs in separate regions havingdiverse properties. We will return to this point in the next section.
The primary attributes that initially drew our attention to the program stars were their multipleperiods in the OGLE I -band light curves. However, it was the complex nature of their optical c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties α and other low members of the Balmer series, whichis uncharacteristic of an A star, even one of high luminosity. The acquisition of far-UV spectra hasclarified that these objects are A + B luminous binaries, and three optical spectra have confirmedradial velocity variations for SMC-SC4. The current investigation discloses no new clues to theorigin of the 15 day period of SMC-SC3 or of the “chaotic” photometric activity for SMC-SC4discussed in M10. However, we can remark further on spectroscopic activity pertinent to the 184day “eclipse” period.In § ff erent epochs tofaithfully represent the motion of the A star around its orbit. We also note that in the only blueBalmer line covered in all three observations, H δ , the wings suggest that the RV variations arelarger than 100 km s − . For this reason we do not trust these few measurements to estimate binaryorbit attributes such as eccentricity and inclination reliably. Nonetheless, reasonable to infer thatthe RV variations are due to the same 184 day eclipse period discussed by M10, even though theeclipses are almost certainly not due to the secondary star. We will comment on further on thisgeometry in § ⊙ and a 184 day binary period,one finds a semimajor axis for the binary systm of 1.66 AU. Considering next the disk, if oneassume Keplerian orbits and adopts a measured separation between the V and R emission com-ponents in the H α profile of 110 km s − , the characteristic orbital radius for the disk becomes 4.9sin i AU. Furthermore, the well resolved nature of the two emission peaks implies once again thatthe sin i factor is not far from unity. Then assuming further only that the disk and orbital planesare at least roughly coincident, the ratio of the size of the disk to distance between the stars isabout a factor of 2-3 (4.9 sin i / cir-cumbinary (CB) disk around the SMC-SC4 system. Confirmation of this picture comes from thefact that the V / R emission ratio of the H α profile of SMC-SC4 changes from the 2002 to the 2007epoch even at near-like binary phases. We would expect the details of the emission profile to bedi ff erent if the emission came from the environment of one of the binary components. We use thedescriptor “circumbinary” hereafter to describe the structure causing the disk emission and mostof the exophotospheric absorptions in its spectrum. We have not made the case for RV variability c (cid:13) , 1–32 Mennickent & Smith for SMC-SC3, but given the discovery of its B secondary this seems to be a matter of discoverygiven our poor phase sampling.As for SMC-SC3, if one again takes the sum of masses as 18M ⊙ and a period of 238 days, onederives a semimajor axis of 5.4 AU of the system. This is larger than the scale of the SMC-SC4 sys-tem, but then the H α emission is some four times stronger for SMC-SC3 (Table 3), which impliesa more extensive disk outside this radius.Altogether, the high resolution spectra have revealed a number of novel properties that aredi ff erent for each star and yet reminiscent and of one another. the two stars, and the overarchingquestion is how the evolutionary state of the binaries and the remarkably slow rotations of the com-ponent stars might be responsible. These novel properties include emission, the unusual excitationsimplied by the emissions in He I (SMC-SC3 only), Si II, and Fe II lines, and for SMC-SC4 thepresence of multiple components, including “BACs” in the strong resonance and intermediatestrength lines, and emission components in permitted Fe II and Si II lines,Before evaluating the spectra of of our program stars further, we should state our assumptionsabout the dynamics of the binary-wind system that will facilitate consideration of the simplestpossible picture that at the same time is consistent with the complex features we see. Becauseof their presume relatively high space density in the region of the SMC first investigated, weassume that both program stars are immediately post-main sequence. We have also assumed thatM ≈ M , and this is based on the absence of any evidence of extensive mass transfer, anomalousabundances, or rotational spin up. The stellar radii we determined particularly from the surfacegravities inferred from the Ca II and C III lines suggests that the stars’ radii are much smaller thantheir Roche lobes. This leads to the simplifying picture that, except for wind-wind or wind-CB diskinteractions, the wind kinematics and geometries are not di ff er radically from those of single stars.Of course, the fate of wind e ffl ux stopped by collisions with the other wind or a disk, as suggestedby shock features, suggests a more complex scenario. We assume that the residual momentum ofthe B star wind forces some of this stalled matter to the A primary, perhaps in a relatively narrowneck defined by the inner Lagrangian point. If any matter also returns to the B star it would not beobserved in our optical spectra. The H α emissions in both program stars are strong but not at the high end of the range amongclassical and pre-main sequence Be stars. The presence of H α emission has long been used to c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties α line is invariably generated byrecombination in dense circum-stellar / binary environments of hot stars. Yet the extended wings ofthese features hint at the importance of electron scattering, and this implies that the disk extendsto low densities and large radii.The presence of V , R emissions in strong Fe II lines of SMC-SC3 is a characteristic in a num-ber of well-known classical Be stars, These spectroscopic signatures are formed by scattering indense, high optical-depth disks as well. An important characteristic of these emission profiles isthat they remain almost symmetric in our observations, suggesting a stationary disk. However, asmall persistent excess in the R emission may hint at a few km s − expansion of this structure. Sep-arate from these considerations, the superposition of broad absorption profiles of strong Fe II (e.g.,Fig. 8) and other excited lines like the Si II doublet suggests the presence of a separate absorbingsource close to the A star component of SMC-SC3.A clue to understanding the energetics of the disk of SMC-SC3 is the presence of He I linesfrom the primary’s photospheric and circumbinary lines. As with the hydrogen emissions, theselines cannot be excited by the photospheric radiation field. A ready alternative is the shocks pro-vided between wind-wind collisions between the stars. In the absence of profiles of the UV res-onance lines, this is still a speculative idea. Nonetheless, the fact that the velocities of the He Ilines are so di ff erent from all the absorption line velocities indicates that they are formed in adi ff erent physical region that need not even be confined to the disk plane. From our fittings of thestrengths of He I ( § × cm − is typical. Adopting a value of N e ∼ cm − , typical of CS disks of Be and Ae stars, thecolumn length implies a radial extent of 3 × km ( ∼
2% of the A star’s radius) for the formationregion; if the temperature this extent could be much less. This is consistent with shock formationand inconsistent with mechanisms that excite helium atoms over a large volume, e.g., irradiationby X-rays. c (cid:13) , 1–32 Mennickent & Smith
The kinematics of the disk surrounding the SMC-SC4 system are quite di ff erent from the SMC-SC3.For example, the metallic lines do not show a symmetric two ( V , R ) component emission that sug-gests the presence of a stationary disk nor that the optical depths of the circumstellar or circumbi-nary absorptions have as high optical depths. In contrast, there are the blue and occasionally redabsorption components, as well as emissions in the wings of the main line whose positions seemto be determined by the orbital phase.Although ubiquitous, novel, and likely present at all orbital phases, the BAC features are notunique to the SMC-SC4 system. In a little noted discovery, Heydari-Malayeri (1990) reported thatmany of the metallic absorption lines in the optical spectrum of the SMC supergiant N82 consistof double components, the dominant member of which is blueshifted by about -46 km s − . This isalmost the same value found for SMC-SC4. The Balmer lines also show double-peaked emissionsintermediate in strength between those we find SMC-SC3 and SMC-SC4. In addition, Heydari-Malayeri found that lines of di ff erent ions have various red to blue emission component ratios. Thiscaused the RVs of members of ions exhibiting these components to have a large RV scatter. Thisauthor classified N82 as a “sgB[e]” star largely on the basis of the presence of infrared emissionand the appearance of [Fe] emission lines. However, an important supporting justification for theauthor’s classifying N82 as Be rather than Ae was the star’s strong Balmer line emissions. Becausewe now know that this is not necessarily a sine qua non criterion for Be classification, the argumentthat this object is a B star is weakened. The presence of [Fe] emission in the spectrum of N82 butnot in our program star spectra suggests that N82’s attributes are phenomenologically distinct.However, its spectrum is similar in including BAC components in metallic lines and substantialemission in the lower Balmer lines.Overall, the striking attributes of the BAC-like components in the SMC-SC4 spectrum are: a) their ubiquity among many types of lines, b) their consistent blueshifts of -50 to -32 km s − in atleast three orbital phases, and c) their concentration mainly among lines arising from atomic levelsof a few eV. Points a and b argue that the BACs appear the same around the orbital cycle and areperhaps caused by impacts of an axisymmetric outflow.We also found, depending on the epoch, that absorptions in either the far blue or red wingsof some Fe II lines show an extraordinary sensitivity to excitation and line strength. The flux inemission features exhibits a reversed dependence on the line’s oscillator strength, with the weakerlines showing the greater emission. The sensitivity of emissions to line strength fits in with the c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties km in the case of the 3 eV Fe II lines, in the formation length where the emissions of (weaker)lines are formed. The sensitivity to line strength and excitational information also hints that theenergy input into the putative shocks has a small “bandwidth,” that is, if the outflows were lessor more energetic then some other range of excited states would be collisionally populated. Thisis consistent with the wind energetics because winds have a characteristic velocity when theyimpact stationary matter at a fixed distance. According to these ideas, one expects the excitationsof atoms exhibiting emission to depend on the wind velocities, and thus to some degree also theorbital separation between the stars. Perhaps these conditions are “just right” for SMC-SC3 toexcite lines at ≈ From the dissimilar nature of the optical and far-UV spectrum (and the RV variations for SMC-SC4),we have been obliged to adopt a binary model involving nearly two equal mass A + B components.The strong double-peaked emission in the H α line against the continuum of an A bright giant orsupergiant implies the presence of a flattened Keplerian disk.The amplitude of the radial velocity variations compared to the almost stationary Balmer emis-sion cores shows that at least in SMC-SC4 the H α -emitting structure is a circumbinary structure.We cannot rule out the possibility of RV variations in SMC-SC3 on the basis of their apparent con-sistency in just three observations. Indeed, the requirements to produce blueshifted He I line emis-sion suggest unusual energetics in a region that has a di ff erent radial velocity than the A star. Theorbital separation of the binary components, estimated above for SMC-SC4 to be ≈ i AU,is about - of the characteristic CB disk radius. The absence of forbidden emission or metastableabsorption lines argues that most of the disk does not have a low density. The optical / IR colorsimply that reddening from cool dust is negligible.We envision that each of the binary component stars has a radiative wind, with the A star’swind being weaker, slower, and perhaps denser near its surface. These winds will interact alongan annulus centered on the line connecting the two stars. The shock-heating in this wind willshow a maximum intensity at velocities in between those of the stars, though not necessarily at thebarycenter. This would explain the displacement of emission of the He I lines for SMC-SC3 and For ballpark numbers winds of early B stars may have ˙M ∼ − –10 − M ⊙ yr − and a terminal velocity of 1000-2000 km s − . Typical values forwinds of A supergiants are thought to be ∼ < − M ⊙ yr − and 300 km s − (e.g., Verdugo 2002).c (cid:13) , 1–32 Mennickent & Smith
Fe II lines for SMC-SC4. The wind continuously replenishes the CB disk, but we have no directestimate of the relative amounts of disk or post-shock matter returning to the two stars or ofit escaping through the outer edge of the CB disk. The symmetric emissions in the Fe II andother lines in the SMC-SC3 spectrum have a separate origin, and probably originate in the quasi-stationary CB disk.Apart from the 1-2 BAC pattern in the spectrum of SMC-SC4, far wing absorptions appear inthe opposite wing from far wing emissions ( § ff erent sight lines to the A star and to flowing CS streams in the sky plane.This complicated behavior is reminiscent of red- and blueshifted components in time-resolvedultraviolet spectra in Algol systems. Peters & Polidan (1984, 1998) have noted the existence offlows associated with “High Temperature Accretion Regions” caused by streaming of wind e ffl uxfrom the cool giant star as it expands and transits through the binary’s inner Lagrangian point. Thisstream collides with a disk around the receiving hot secondary. This is the reverse of the presentcontext, in which the receiver is the cooler star. In the present case we envision that the collisionpoints occur behind the orbiting primary and generate shocks. These are manifested in spectraas Doppler shifted emission components. Disk matter cools and eventually falls into the cool (A-type) star, rather than only through the inner Lagrangian point. The result is a complex array ofrelatively high velocity blue- and redshifted emission and absorption components that change theirmorphology around the orbital cycle. The high positive velocities compared to the A star in the2009 spectrum (Fig. 11b, 12a) indeed suggests that matter falls toward the A star as seen from itstrailing side, as Peters & Polidan found for the Algols. One aspect of this picture is that it predictsthat the shock geometry can be complex and not necessarily repeatable from the cycle to cycle.Even at the same time the volumes producing emission for lines of di ff erent ions are not the same.It is tempting to speculate that the infalling matter inferred from our 2009 spectra of SMC-SC4 isassociated with the observed eclipses. Actually, any such picture must be complicated because itis the “wrong” (2007) spectrum which nearly coincides with the egress from a continuum lighteclipse. This would suggest that the stream spirals nearly a complete circuit (some 65-90%) aroundthe A star until it settles onto the star (the 2007 spectrum was obtained beyond the midpoint andduring the egress of a photometric eclipse; see Table 1 here and Figure 6 in M10). Spectroscopi-cally, this may be the cause of the second blueshifted absorption component of the strongest lines(e.g., dashed line in Fig. 2, solid line in Fig. 11b). At this binary phase the blue component mightactually be a better measure of the A star’s velocity in orbit. The 2009 spectrum was taken at the c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties Even as luminous variables in the SMC, our two program stars exhibit a extraordinary set ofphotometric and spectroscopic properties. These include a multiperiodic photometric variability,a ≈ B3 type spectrum in the FUV, an optical A-type spectrum with Balmer emission lines andmetallic lines from an intervening disk. For SMC-SC4 the spectral anomalies include the presenceof one and sometimes two Discrete Blue Absorption Components (BACs) in metallic lines and RVvariations. We have remarked that the optical spectrum of the SMC supergiant N82 in the SMCexhibits BACs and strong H α emission. However, unlike N82, neither of our stars shows forbiddenlines. Because UV spectra are not available for N82, nor an optical RV campaign mounted, itssuggested binary nature has not been tested.We argue that our program stars reflect a comparatively brief stage in the life of intermediatemass ∼ ⊙ binary components. With separations of a few AU between the components, thewinds of the two stars can interact and produce a large array of spectroscopic phenomenology.This phenomenology can include emissions in lines of intermediate and / or highly excited ions(e.g., He I) and time-variable pattern of sharp absorptions, in addition to the BACs. It is an openquestion why as many as two objects have such similarly unusual properties from a sample of onlybright eight Mennickent Type 3 variables. In any case we may dub these objects prototypes of apresumably small group of Magellanic Cloud wind-interacting A + B binaries.We speculate that the emission components originate from the interaction of that componentof the A and B star winds in the zone near the line that intersects the centers of the two stars. Thedouble peaked emissions in the Balmer lines disclose an emitting Keplerian disc, which from thepeak separation has a larger extent than the binary separation. We have suggested, in part becauseof the He I line emissions in the SMC-SC3 spectrum, that wind-wind interactions cause localizedheating and ionizing photons capable of exciting the observed level of H α model.The Fe II line emission strengths are anticorrelated with the line’s oscillator strengths. We haveattributed this extraordinary circumstance to their formation in a geometrically thin but optically c (cid:13) , 1–32 Mennickent & Smith thick and two layered column, where emission is formed in the deeper hotter shock and the ab-sorption is just outside this region (though in the foreground of the observer’s sight line). We havealso speculated that wind-wind interactions in the zone between the stars collisionally excite He Iline emission and also produce Balmer emission in the quasi-Keplerian circumbinary disc.We have surmised that the primary BAC (absorption) feature is caused by the violent impact ofthe expanding A star wind into this CB disk. However, although we do not have enough observa-tions to characterize the time-dependent secondary components well, it appears that the strengthsof the emission components in either one of the far wings of the 3 eV Fe II lines are physicallyassociated with absorptions in the opposite far wing.We have speculated that a rich line profile variability, including the so-called secondary BACsand emission components, is a consequence of a matter stream originating from the wind of the Bstar and settling on the A star. This is a reverse Algol scenario in which matter is not constrainedto flow only through the inner Lagrangian point. Perhaps this settling processes is responsible forthe eclipses in the continuum light curve of SMC-SC4.A number of observing programs can be suggested to test these ideas and conceivably o ff ernew alternatives: • A high-resolution monitoring of these objects should be undertaken to construct a radialvelocity curve and determine the periods, separations, and inclinations of these systems. Thesesame data can be used to trace the velocities of the various components in the primaries’ spectra.For SMC-SC4 the features of interest are the BACs and the pairs of absorptions and emissions inthe wings of the Fe II lines. For SMC-SC3 it is important to determine whether the velocity of theHe I lines is constant or shows a small amplitude variation around the system’s velocity. • UV observations of the prominent resonance lines are needed to characterize the wind strength,velocity, and its behavior through the orbital cycle. We expect that these features are diagnosticsof the B star’s wind. It may also be possible to trace velocity changes in signatures of the A star’swind, such as an arguably Ca II K emission feature (Fig. 4b). In general, we expect the profiles ofthe UV resonance lines to be highly complex and variable. • Any understanding of these particular objects rests on the expansion and adequate definitionof this class, which in term entails finding more members. So far our small sample is drawn from agroup of luminous optical / IR variables in the SMC. One prong of this search would be to find starswhose spectra exhibit BACs. To date we have obtained moderate resolution spectra of a number ofluminous B stars in the OGLE surveys of both Clouds. We find that two LMC stars have properties c (cid:13) , 1–32 igh dispersion spectroscopy of two A supergiant systems in the SMC with novel properties = α emission profile intensities, strong absorptions amonghigh level lines, and variable (possibly regular) light curves. Although the resolution of thesespectra is su ffi cient to show at least strong forbidden emission lines, none are visible. If many ofthese stars are members of the group exemplified by SMC-SC3 and SMC-SC4, it will indicatethat their properties are not isolated to a particular evolutionary age and / or angular momentumstate. This is a possibility that we cannot yet dismiss. REM acknowledges support by Fondecyt grant 1070705, the Chilean Center for AstrophysicsFONDAP 15010003 and from the BASAL Centro de Astrof´ısica y Tecnologias Afines (CATA)PFB–06 / REFERENCES
Bagnulo S., Jehin E., Ledoux C., Cabanac R., Melo C., Gilmozzi R., The ESO Paranal ScienceOperations Team, 2003, Msngr, 114, 10; http: // / santiago / uvespop / Blair, W. P., Oliveira, C., et al. 2009, PASP, 121, 634Dixon, W. V., Sahnow, D. J., Kruk, J. W., et al. 2007, PASP, 119, 527Dolphin A. E., Walker A. R., Hodge P. W., Mateo M., Olszewski E. W., Schommer R. A., Suntze ff N. B., 2001, ApJ, 562, 303Dufton P. L., Ryans R. S. I., Trundle C., Lennon D. J., Hubeny I., Lanz T., Allende Prieto C.,2005, A&A, 434, 1125Evans, C. J., Lennon D. J., Smartt, S. J., Trundle, C. 2006, A&A, 456, 623EHeydari-Malayeri, M. 1990, A & A, 234, 233Hubeny, I & Heap, S. R. 1996, ApJ,, 470, 1144 c (cid:13) , 1–32 Mennickent & Smith
Hubeny, I., Lanz, T., & Je ff ery, C. S. 1994, Newsl. Anal. Astron., 20, 30Jaschek, M., Jaschek, C., & Andrillat, Y. 1988, A&AS, 72, 505Kurucz, R. L. 1993, Kurucz CD-Rom ff , B. A. 2009 VizieR on line catalog: “Catalog of Stellar Spectral Classifications”Smith, M. A. 2010, ApJS, 186, 175Soszynski, I. 2007, ApJ, 660, 1486Soszynski, I. 2010, priv. comm.Udalski A., Kubiak M., Szyma´nski M., 1997, AcA, 47, 319Verdugo, E., Talavera, A., & Gomez de Castro, A. I. 1999, A&A, 346, 819Verdugo, E., Gomez de Castro, A., Ferro-Fontan, C., et al. 2002, ASPC, 259, 574Verdugo, E., Henrichs, H. F., Talavera, A., et al. 2005, ASPC, 337, 324 c (cid:13)000