FIVE MORE MASSIVE BINARIES IN THE CYGNUS OB2 ASSOCIATION

We obtained radial velocities, V_r, for the single-lined system, MT145 via the IRAF cross-correlation task XCSAO in the RVSAO package (Kurtz et al. 1991), using a model stellar atmosphere (Lanz & Hubeny 2003, TLUSTY) of the appropriate effective temperature and gravity as discussed in Paper I. Errors within the XCSAO task are calculated using

$$\begin{equation} \sigma _v = \frac{3w}{8(1+r)}, \end{equation} \tag{ 1 }$$

where w is the FWHM of the correlation peak and r is the ratio of the correlation peak height to the amplitude of antisymmetric noise (Kurtz et al. 1991).

To obtain rough V_r measurements for the double-lined systems, A36, A45, and Schulte 73, we deblended the He i λλ4471, 5876 (A36 and Schulte 73) or Hα (A45 and MT372) lines by fitting simultaneous Gaussian profiles (at fixed Gaussian widths) with the SPLOT routine in IRAF. We repeated this method 10 times while varying the baseline region used to define the continuum each time. We then used the mean of each component's list of Gaussian centers to compute the velocity. Utilizing this method, a number of factors contribute to the radial velocity uncertainty. These factors include, but are not limited to the errors associated with Gaussian profile fitting, wavelength calibrations, emission present in the line cores, and lower signal to noise of the less luminous component's spectral profiles (especially while in a partially or wholly blended state). We estimated the contribution of the Gaussian profile fitting uncertainty by adopting the rms of each component's list of Gaussian centers. Unfortunately, this only produces a lower limit on the uncertainty because the initial Gaussian width estimates also affect the center measurements. For the total 1σ uncertainties, we adopted these rms values combined in quadrature with the typical error in the wavelength calibration (1.4 km s⁻¹ for WIRO and 2 km s⁻¹ for WIYN) and the typical rms observed in the list of velocities obtained from the correlation with the interstellar line template. The total error given for each system reflects a lower limit as we cannot characterize uncertainty contributions from emission present in the line cores (e.g., A36) or the lower S/N of the secondary spectral features (e.g., A45).

We used the method outlined in McSwain (2003) to determine the orbital parameters of each system. We obtained an estimate of the orbital period through an IDL⁸ program written by A. W. Fullerton which makes use of the discrete Fourier transform and CLEAN deconvolution algorithm of Roberts et al. (1987). The strongest peaks in the power spectrum of each star were examined by folding the data at the corresponding period and inspecting the V_r curve visually. Orbital elements were then procured by using the best period as an initial estimate in the nonlinear, least-squares curve fitting program of Morbey & Brosterhus (1974). The best solutions were attained by manually varying the initial guesses of key orbital parameters until a minimum in the rms of the fit was found. Weights for each point were assigned as the inverse of the 1σ V_r error.

While it is not uncommon to quote orbital period errors to one ten thousandth of a day (e.g., Williams et al. 2008; Penny et al. 2008; Linder et al. 2007; McSwain 2007; Hillwig et al. 2006), we performed two additional tests to evaluate the appropriateness of the orbital period uncertainties for each system. In the first test, we explored the confidence range for each orbital parameter by fitting the radial velocity data 1000 times, with the orbital parameters as free variables, while adding Gaussian noise to the data. The noise added to each data point was characterized by the 1σ uncertainty. We compared the resultant rms deviations to the single-fit error estimates for each orbital parameter. The result of this test suggested that the majority of the single-fit parameter uncertainties were reasonable, as the Monte Carlo test routinely produced similar uncertainties. The disadvantage to this test, however, is that we could not examine each fit to see if it was reasonable (i.e., whether the newly folded radial velocities had a reasonable rms). Therefore, with the error in the period strictly in mind, we ran the orbital fitting procedure 50–100 times while incrementing the period by the single-fit error estimate and monitored how the rms changed. The result of this test suggests that we, at most, underestimate the orbital period errors by a factor of 10. With this in mind, the errors for the orbital period throughout this work have been increased by a factor of 10.

The complete list of orbital elements for each binary system appears in Table 2. Listed within the table are the period in days (P), eccentricity of the orbit (e), longitude of periastron in degrees (ω), systemic radial velocity (γ), epoch of periastron (T₀), primary and secondary semiamplitudes (K₁ and K₂), adopted or calculated minimum primary and secondary masses in solar masses (M₁ and M₂), primary and secondary mass functions in solar masses (f(m)₁ and f(m)₂), spectral classifications from this survey (S.C.₁ and S.C.₂), the minimum primary and secondary semimajor axes in solar radii (a₁sin i and a₂sin i), and finally, the rms of the fits (rms₁ and rms₂).

Systemic radial velocities are known to vary depending on which lines are used to measure the radial velocity. This is especially evident with massive stars and evolved stars where the stellar atmosphere velocity gradient may be higher (see Linder et al. 2007 for examples with a few well studied O binaries). To minimize such discrepancies, radial velocities are normally averaged over many lines. In the case of MT145, where we obtain radial velocities by cross-correlation over many lines, this is not an issue. However, in the case of the remaining 4 systems, we only have 1–2 lines (He i λλ 5876, 6678 or Hα) available for radial velocity measurement in the majority of spectra. This places an added uncertainty on the systemic velocity and semiamplitudes measured for A36, A45, Schulte 73, and MT372. In the simplest case, these effects may cause a systematic blueshift to the systemic velocity (such as the case with a noneclipsing system with equal and symmetric stellar winds). Eclipses, uneven stellar winds, and uneven surface temperatures complicate the matter, and these effects can affect the components unequally, producing systemic radial velocities that vary from primary to secondary. We see this effect in two of our systems, A45 and Schulte 73. The component orbital solutions produce two different systemic velocities. For both of these systems, we adopted the mean of the two values as the systemic velocity of the system for Table 2.

We observed MT145 a total of 54 times, including six at Lick, two at Keck, 12 at WIYN, and 34 at WIRO. Excluding two low-S/N spectra from Lick and one from WIRO, a total of 51 spectra were of sufficient S/N to obtain the solution for this O9III star. The strongest signal in the CLEANed power spectrum corresponds to a period of 25.12 days. An alias at 20 days is also present. However, the 20 day period produced a folded V_r curve having a much larger rms than the 25.12 day period. The best-fitting orbital solution and V_r curve for MT145 is shown in Figure 1, corresponding to a period of P = 25.140 ±  0.008 days with an eccentricity of e = 0.291 ±  0.009 and solution rms = 8.9 km s⁻¹. With this period, a semiamplitude of K₁ = 48.9 ± 0.7 km s⁻¹, and an assumed primary mass of M₁ = 22.6 ± 0.5 M_☉ (Martins et al. 2005), we estimate a minimum secondary mass of M₂ = 6.0 ± 0.1 M_☉, corresponding to a mid B star (interpolated from Drilling & Landolt 2000). This yields a mass ratio lower limit of q = M₂/M₁ ≳ 0.27 ± 0.01. To constrain the mass ratio upper limit, we examined the spectra taken while the system was near quadrature for evidence of line doubling or asymmetries originating from the secondary. Neither is present in any of the spectra. Given S/N ratios of up to 150:1 for these spectra, we would expect to see evidence of the secondary star's spectral features at velocity separations of (1 + 1/q)K₁ if the component luminosity ratio is larger than L₂/L₁ ≃ 0.2. This maximum luminosity ratio is determined by combining synthetic spectra at various luminosity ratios and line separations. The absence of secondary spectral features therefore conservatively limits the luminosity ratio to L₂/L₁ ≲ 0.2 and implies a secondary less luminous than a B1V (interpolated from Drilling & Landolt 2000). We thus constrain the mass of the secondary to 6.0 ± 0.1 M_☉<M₂ ≲ 13.8 ± 0.6 M_☉ and the mass ratio to 0.26 ≲ q ≲ 0.63 (where the range in mass ratio incorporates both observational and inclination uncertainties). These limits loosely require i ≳ 31° given the uncertainty in O star masses (Massey et al. 2005). Additionally, the solution produces a minimum primary semimajor axis of a₁sini = 23.2 ± 0.3 R_☉ and a longitude of periastron of ω = 1583 ± 06

Zoom In Zoom Out Reset image size

The ephemerides for MT145 and MT372 are listed in Table 3 and include the date (HJD−2, 400, 000), phase (ϕ), measured radial velocity (V_r), cross-correlation 1σ error (1σ err), and the observed minus calculated velocity (O − C).

A36 is originally introduced as a probable massive member of Cyg OB2 in Comerón et al. (2002) and listed as a possible binary based on the asymmetry in the absorption lines of a high S/N spectrum in Hanson (2003, Figure 1). Otero (2008a) provided a partial photometric solution to this system and listed it as an eclipsing binary of the Algol type with evidence of an O'Connell effect (a light curve having unequal out-of-eclipse maxima; Milone 1968). We provide the first spectroscopic solution to this double-lined system.

We observed A36 a total of 23 times between 2007 October and 2008 August, including five times with Hydra at WIYN and 18 times with the WIRO-Longslit spectrograph. With the 19 higher S/N spectra we obtained a period estimate of 4.680 days from the CLEANed power spectrum with no aliases. From this estimate we obtained the solution for the secondary (rms = 10.0 km s⁻¹) shown in Figure 2 with a period of P = 4.674 ± 0.004 and relatively low eccentricity of e = 0.10 ± 0.01. The period, eccentricity, epoch of periastron, and systemic velocity were applied as fixed parameters to obtain the solution for the primary (rms = 29.3 km s⁻¹). The filled symbols correspond to the primary V_r measurements and the open symbols correspond to the secondary V_r measurements. The combined solutions provide minimum semiamplitudes of K₁ = 169 ± 9 km s⁻¹ and K₂ = 243 ± 2 km s⁻¹ for the primary and secondary, respectively, and produce semimajor axes of a₁sin i = 15.6 ± 0.8 R_☉ and a₂sin i = 22.345 ± 0.003 R_☉. The computed lower limit masses of M₁sin³i = 19.8 ± 0.9 M_☉ and M₂sin³i = 13.8 ± 0.9 M_☉ yield a mass ratio of q = 0.70 ± 0.06. Table 4 provides the ephemerides for A36, A45, and Schulte 73, listing the date (HJD−2,400,000), phase (ϕ), measured radial velocities (V_r1 and V_r2), and observed minus calculated velocities (O₁ − C₁ and O₂ − C₂). Error estimates are shown in parentheses.

Zoom In Zoom Out Reset image size

Based on two of the spectra obtained during the 2008 WIYN run, we tentatively classify the components of A36 as a B0Ib and B0III for primary and secondary respectively. The exposures obtained on 2008 June 10 and 15 show the components at nearly the same phase and in quadrature. Additionally, the light curve shown in Otero (2008a), tells us that the temperatures of the components are similar, and the double-lined nature of the spectra tell us that the luminosities are similar. We can therefore conclude from the 2008 June 10 spectrum shown first from the top in Figure 4 that the spectral types of the components are similar with subtle differences, such as stronger Si ivλλ 4089, 4116 and weaker He i λλ4026, 4121, 4387, 4471 for the primary. Both primary and secondary have no Mg ii λ4481 (confirmed in Figure 1 in Hanson 2003), moderately strong Si iv λλ4089, 4116, strong He i λλ4026, 4471, weak He i λλ4121, 4387, and strong C iii λ4070. After smoothing the WIYN spectra to approximately the resolution of the spectra in the Walborn & Fitzpatrick (1990) digital atlas, we combined a B0Ib (HD164402) spectrum with a B0III (HD48434) spectrum at the appropriate radial velocities to make a comparison. The resulting combined spectrum (second from the top) is also shown in Figure 4 along with the 2001 September 8 spectrum of Schulte 73 (second from the bottom) obtained at WIYN and the associated composite spectrum for Schulte 73 (bottom). Intrinsic He i, He ii, C iii, Si iv, Mg ii, and hydrogen absorption are labeled. The composite spectrum is not an exact match to A36 (note that Si iv 4089 is stronger than that of the primary). However, the mass ratio obtained from Drilling & Landolt (2000) for a B0Ib primary and B0III secondary is q = 0.8 and agrees with our calculated mass ratio of q = 0.70 ± 0.06.

An apparent weakening of the primary's He i lines between phases 0.218 and 0.609, seen in panels 1 and 3 of the time series presented in Figure 3, at least partially explains the large residuals at these epochs and may originate from a Struve–Sahade effect (Linder et al. 2007). This apparent weakening also coincides with an absence of well defined absorption profiles in Hα (panel 2 of Figure 3) between secondary and primary eclipses. These spectral phenomena coupled with the continuous variation of the light curve presented in Otero (2008a) associated with Algol eclipsing systems, an O'Connell effect (which is generally associated with contact or near contact systems; Liu & Yang 2003) and a seemly close separation between components (asini ∼ 28 R_☉), implies that A36 is either a contact or near-contact system. In the very least, it is likely that one or both components are filling their Roche lobes. Because A36 is composed of B stars, one may question whether this is a Be binary. The system does not hold the standard qualifications for a Be binary however, as the mass ratio is much greater than 0.1 and the primary is evolved (Harmanec 2002). Conti & Leep (1974, Figure 3) also show that hot supergiants can have hydrogen emission that very closely resembles the double-peaked hydrogen emission seen in Be stars.

Zoom In Zoom Out Reset image size

Is A36 a member of Cyg OB2? Hanson (2003) places doubt on its status as a member of the cluster based on its disagreement with the 2 Myr isochrone and evolutionary tracks best fit for Cyg OB2 stars. With a systemic velocity offset −36 km s⁻¹ from the mean systemic velocity for cluster stars (−10.3 km s⁻¹; Paper I), A36 is likely either a runaway binary system or a background system masquerading as a member of Cyg OB2. McSwain (2007) found observational evidence for several runaway binary systems, but concluded that they are rare. These systems can be produced by an asymmetric supernova of one of the components (producing an SB1) or by dynamic interactions within a dense cluster. With a stellar density of ρ₀ = 40–150 M_☉ (Knödlseder 2000) it is certainly possible for A36 to have originated this way, but not probable. However, as a background object, A36's systemic velocity inferred from a Galactic rotation curve would place it at a distance of 6.6 kpc, an interarm region where OB stars are unlikely to form. Therefore, without a proper motion to trace its trajectory, or until a complete solution is computed for the photometric and spectroscopic solutions for this star, the status of A36's membership will remain uncertain.

We obtained 24 observations from 2007 October through 2008 September on this double-lined system. Hanson (2003) classified this star a B0.5V, but, as with A36, she also suggests it may be an SB2 based on a high S/N spectrum (see Figure 1 in Hanson 2003). We present the first confirmation of the binary nature of A45 with a spectroscopic solution to this system.

Of the 24 observations, we used 21 to obtain rough V_r measurements with the Hα λ6562.80 line. We did not use the He i λλ5875.75, 6678.15 lines owing to either the weaker He i absorption of the secondary or low S/N of the spectra near He i λ5876. The 18 measurements produced a CLEANed power spectrum with a period estimate of 2.885 days and aliases of 2 days and 1.55 days. The latter is a 1 − ν alias (where ν is the signal frequency per day). Neither alias produced a folded V_r curve with as low an rms as 2.885 days. The best solution for the primary results in a period of P = 2.884 ±  0.001 and eccentricity of e = 0.273 ± 0.002. The period, eccentricity, and epoch of periastron were applied as fixed parameters to obtain the solution for the secondary. The rms of 47 km s⁻¹ indicates there is room for improvement in the secondary's solution and also explains the high uncertainty on the calculated masses, M₁sin³i = 12 ± 1 M_☉ and M₂sin³i = 5.3 ± 0.5 M_☉ (q = 0.6 ± 0.1). With minimum semiamplitudes of K₁ = 126.1 ± 0.3 km s⁻¹ and K₂ = 273 ± 17 km s⁻¹ for the primary and secondary respectively, the solution also produces minimum semimajor axes of a₁sin i = 6.91 ± 0.02 R_☉ and a₂sin i = 14.96 ± 0.01 R_☉ and a systemic velocity of γ = 5 ± 1 km s⁻¹.

We have limited information for classification of the components of A45 owing to the limited spectral coverage of the WIRO-Longslit data sets (∼5300 Å–6700 Å). No attempt to obtain exposures of this system was made during the 2008 WIYN run because of the system's angular separation from the rest of the sample. Examination of the existing spectral coverage and the spectrum in Figure 1 of Hanson (2003) suggests that her classification is more or less accurate in the sense that both components appear to be early B stars. There is no evidence of He ii in either component, He i is moderately strong, Mg ii is weak, and there appears to be no Si iv. An absence of the numerous additional metal lines, such as O ii, C ii, C iii, and N ii seen in evolved B stars suggests that neither component is very evolved and both are likely still dwarves. Additionally, a double-lined system with a mass ratio of q = 0.6 ± 0.1 restricts the secondary component to within one or two temperature classes of the primary component. Therefore, we adopt the classification of B0.5V for the primary and B2V?–B3V? for the secondary component based on the aforementioned line strengths and mass ratio.

Is A45 a member of Cyg OB2? Although A45 has the largest angular separation from the core of Cyg OB2, the systemic velocity agrees with the mean systemic velocity of cluster members and the system does show the same diffuse interstellar band features characteristic of Cyg OB2 stars. Additionally, though the combined luminosity and temperature of the system disagrees slightly with its predicted placement on the H-R diagram presented for Cyg OB2 stars in Figure 12 in Hanson (2003), individual component luminosities and temperatures shift A45's position down on the diagram and into better agreement with other Cyg OB2 members.

Schulte 73 is included in our original sample but excluded from Paper I owing to the less than 3 observations then obtained. Spectra taken during the 2008 June WIYN run, however, revealed the double-lined nature of this star, and re-examination of the 2001 September 8 spectrum confirmed it. The successive 2008 WIRO runs cemented its status as a double-lined spectroscopic binary with a period of ∼17.3 days and a mass ratio nearly equal to unity. The solution computed for Schulte 73 utilized 27 of the 36 spectra obtained. Seven spectra from WIYN in 2008 June were excluded for low S/N owing possibly to a fiber position discrepancy. Two WIRO spectra (2008 July 26 and 27) were excluded because of the extremely blended state of the components.

For the 27 higher S/N spectra, the initial period estimate from the CLEANed power spectrum is ∼16.6 days with an alias at ∼66 days. We dismiss the alias on the basis that progression of the absorption lines from a blended to deblended state is observed over approximately 4–5 nights. The best orbital solutions yield a period of P = 17.28 ± 0.03 days for both components. The exact period, eccentricity, and date of periastron are held constant for the primary in order to facilitate a matching solution with the secondary. Coincidentally, this also provides a better-fitting solution to the primary V_r curve. We attribute the larger error on the period (a magnitude higher than the other three systems) to the smaller number of velocity measurements obtained around apastron (ϕ ≃ 0.3–0.6). The solutions and corresponding V_r curves for Schulte 73 are shown in Figure 6. The filled and open symbols correspond to the primary and secondary V_r measurements, respectively. The errant WIYN measurements in the primary solution (triangles) reflect the lower S/N of the 2008 June data set and the error in the computed period. With an eccentricity of e = 0.169 ± 0.009 and semiamplitudes of K₁ = 117 ± 2 km s⁻¹ and K₂ = 117.9 ± 0.8 km s⁻¹  the solutions produce masses of M₁sin³i = 11.2 ± 0.2 M_☉ and M₂sin³i = 11.1 ± 0.2 M_☉. This results in a mass ratio of q = 0.99 ± 0.02 and our first confirmed "twin" in the cluster. The resulting semimajor axes are a₁sin i = 39.4 ± 0.5 R_☉ and a₂sin i = 39.675 ± 0.004 R_☉ (i.e., a separation of asin i = 79.075 ± 0.5 R_☉), and the calculated systemic velocity is γ = −11 ± 2 km s⁻¹.

Massey & Thompson (1991) classify this star as an O8V. Each of the components show spectral signatures indicative of a later O type star, namely moderate to weak He ii and stronger He i absorption. The primary classification criteria for later O type stars are He ii λ4542:He i λ4387 and He ii λ4200:He i λ4144 (Walborn & Fitzpatrick 1990). The λ4200:He i λ4144 ratio (also sensitive to luminosity class) and a visual comparison with the 2001 September 8 spectrum from WIYN seen second from bottom in Figure 4 agree best with an O8III spectral type for both components. The Schulte 73 spectrum has been smoothed to approximately the resolution of the Walborn & Fitzpatrick (1990) digital atlas. However, without a separation of the spectral components (e.g., tomographic separation), there is uncertainty in this assessment. The components also closely match an O7III and O9III spectral type. The very weak He i λ4144 however, precludes a later temperature class. The O8III (HD36861) and O8III composite spectrum is shown at the bottom of Figure 4. We also tried various combinations of an O8V, O8III and O8II. These did not fit the data as well and left larger residuals when subtracted from the Schulte 73 spectrum.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size