The non-thermal radio emitter HD 93250 resolved by long baseline interferometry
As the brightest O-type X-ray source in the Carina nebula, HD 93250 (O4 III(fc)) is X-ray overluminous for its spectral type and has an unusually hard X-ray spectrum. Two different scenarios have been invoked to explain its X-ray properties: wind-wind interaction and magnetic wind confinement. Yet, HD 93250 shows absolutely constant radial velocities over time scales of years suggesting either a single star, a binary system seen pole-one or a very long period and/or highly eccentric system. Using the ESO Very Large Telescope Interferometer, we resolved HD 93250 as a close pair with similar components. We measured a near-infrared flux ratio of and a separation of arcsec. At the distance of Carina, this corresponds to a projected physical distance of 3.5 A.U. While a quantitative investigation would require a full characterization of the orbit, the binary nature of HD 93250 allows us to qualitatively explain both its X-ray flux and hardness and its non-thermal radio emission in the framework of a colliding wind scenario. We also discuss various observational biases. We show that, due to line-blending of two similar spectral components, HD 93250 could have a period as short as one to several years despite the lack of measurable radial velocity variations.
Massive stars of spectral-type O are typically soft and bright X-ray sources. Their X-ray emission is usually attributed to shocks in the dense region of the winds that grow from the intrinsic instabilities of the line driving mechanism. In the presence of a relatively strong magnetic field, the structure of the wind might be controlled by magnetic force lines and, at least in case of a bipolar field, the wind might be confined towards the equator, producing a typically harder X-ray emission compared to that expected from embedded wind shocks. In binary systems, an additional X-ray emission can also be generated by the supersonic collision of the two winds. One of the most comprehensive surveys of the X-ray properties of massive stars has recently been completed in the framework of the Chandra Carina Complex Project (Townsley et al., 2011), that has provided X-ray spectroscopy of 68 O-type stars in the Carina nebula. Yet not all sources are equally-well understood.
Located at 7.5′ NNW from Eta Carinae in the Trumpler 16 complex, HD 93250 (catalog HD~93250) (O4 III(fc); Walborn et al., 2010) is the brightest X-ray O star in Carina and has puzzled astronomers for over two decades. Assuming that it belongs to the Carina association, HD 93250 is too bright for its spectral type. Its X-ray luminosity, erg s (Gagné et al., 2011), is a factor two to three larger than expected from the canonical relation for O-type stars (Sana et al., 2006; Nazé et al., 2011). Its X-ray emission is variable at the 20-50% level on a time scale of months (Rauw et al., 2009; Gagné et al., 2011) and its X-ray spectrum is extremely hard ( keV, Gagné et al., 2011). HD 93250 is also a known non-thermal (NT) radio emitter (Leitherer et al., 1995).
Most of the puzzling properties of HD 93250 could be explained as part of a binary scenario where the collision between the wind of the two stars produces both an X-ray excess (Luo et al., 1990) and an NT radio emission component in the form of synchrotron radiation in the presence of a population of relativistic electrons (De Becker, 2007). The latter point is especially important in the sense that it provides the opportunity to investigate particle acceleration processes in massive star environments, in addition to evolved objects such as supernova remnants or binaries containing a compact object. For the wind collision scenario to produce enough X-ray and synchrotron radio flux, the winds from both components need to be of similar strength, and the companion is thus expected to be a bright O star. Yet, both high angular resolution observations (Nelan et al., 2004) and RV campaigns (e.g., Rauw et al., 2009) have failed to reveal a binary nature of the system. This has raised the question whether massive single stars could produce NT emission (van Loo et al., 2006). As an alternative scenario, Rauw et al. (2009) suggested that the X-ray over-luminosity could originate from a magnetically confined wind but Gagné et al. (2011) noted that the strong non-thermal emission argued against such a scenario and called for further observations.
As discussed by Sana & Evans (2010), long base-line interferometry is ideally suited to identify binaries in a separation regime that cannot be reached by either spectroscopy, typically limited to semi-major axes up to a few A.U., nor by other high resolution techniques, typically limited to a minimum angular separation of 20 mas, corresponding to 40 A.U. at the distance of the Carina complex.
2 Interferometric measurements
2.1 Observations and data reduction
We observed HD 93250 at the ESO/VLTI (Haguenauer et al., 2010) with the Astronomical Multi-BEam combineR (AMBER, Petrov et al., 2007) on December 27 2010. Observations have been made in configuration A0-K0-G1, providing projected baselines ranging approximately from 50 m to 120 m. We recorded spectrally dispersed fringes between m at a spectral resolving power () of 40.
At VLTI the fast telescope guiding (STRAP) is performed in the V-band. For stars fainter than mag, the quality of the guiding is degraded and the amount of flux injected into the instrument is dramatically reduced. The offered limiting magnitude of AMBER ( mag) has been measured on late-type stars, for which mag typically corresponds to mag. The offered magnitude thus results from the combination of the low flux limit in the observation band with the poor performance of the guiding for stars faint in .
Early-type stars are intrinsically 0.7-0.8 mag brighter in the -band compared to in the -band (Martins & Plez, 2006) although the apparent color depends on the extinction properties. For stars as bright as HD 93250 in the -band ( mag) the telescope guiding performs very well. The fraction of flux injected into AMBER is optimal and the sensitivity of the observations is only limited by the intrinsic flux of the object in the observation band. As a consequence, HD 93250 () could be observed with the 1.8m Auxiliary Telescopes (ATs) despite the fact it is significantly fainter than the offered limiting magnitude.
Calibrated visibility and closure phase values were computed using the latest public version of the amdlib package (v3.1, Tatulli et al., 2007; Chelli et al., 2009) and the yorick interface provided by the Jean-Marie Mariotti Center
The absolutely calibrated visibilities and closure phases are reported in Fig. 1. The target is well resolved in one baseline (K0-A0) and significantly less resolved on the two others (G1-A0 and K0-G1). The closure phase shows a clear departure from 0 (or 180) degrees. As expected, it is impossible to fit the data with a single-disk model as this yields an unrealistic diameter of 3.1 mas, a final and systematic discrepancies in the fit. Altogether, the binary nature of HD 93250 is conclusively revealed by these observations.
We fitted the complete set of data (visibilities and closure phases) with a model composed of two unresolved stars with constant flux ratio over the H and the K bands, a valid assumption for O+O binaries. The separation, the orientation and the flux ratio were the only free parameters of the fit. We then explored the full parameter space to find the best value and we obtained a very good fit with , as shown in Fig. 1. Thanks to the combinations of three baselines of different orientations and of the H and the K bands, the position angle of the binary is unambiguous. By chance, an additional “locking point” for the binary model is provided by the visibility minimum observed in baseline K0-A0.
2.3 Additional observations
HD 93250 has been observed two more times with VLTI/AMBER. On March 24 2011, we used the 8.2-m telescopes (UTs) in configuration UT1-UT3-UT4 but we could not resolve the binary. This is in agreement with the very small angular separation between the components and the smaller baseline provided by the UTs. On March 29 2011, we used the ATs and we resolved HD 93250 again. We measured a binary separation of mas and mas and a flux ratio of , which is fully consistent with our observation of December 2010 reported in Table 1. Given the uncertainties, a rotation on the plane of the sky smaller than about 30° (thus of an orbital cycle) would remain within the error bars. This second observation thus suggests a period larger than 3 yr, a high inclination or a very eccentric system. Alternatively, we might have observed HD 93250 at the same orbital phase (0.5, as the sign of the closure phase is unconstrained in the observation of March 2011).
|Parameter||Dec. 2010||Mar. 2011|
3.1 Optical observations
Rauw et al. (2009) measure RVs from six optical spectra of HD 93250, obtained from 1999 to 2004, and concluded that variations are absent, with a measured 1- dispersion of 1.3 km s. We reanalyzed the Rauw et al. data set, complemented by two additional observations obtained with the Fiberfed Extended Range Optical Spectrograph (FEROS) and available from the ESO archives. The epochs of these supplementary observations are HJD - 2,450,000 = 3365.801 (PI: Bouret) and 3738.828 (PI: Casassus), i.e., respectively, nine months and almost two years after the last Rauw et al. RV measurements. Despite the high signal-to-noise ratio () of the data (typically about 300), we could not detect any RV variation either. The maximum difference between the RVs from the different epochs is 3.6 km s and the 1- dispersion around the mean is 1.1 km s. We also searched for line profile variations in the He ii and Si iv lines, with a negative result. The relative peak-to-peak variations of the amplitude and of the full width at half-maximum (FWHM) of e.g., the He ii 4541 line, are below 5 and 10% respectively, pointing again to an excellent stability of the HD 93250 spectrum over the years.
3.2 Physical constraints
HD 93250 belongs to Tr 16 in the Carina association. Adopting a distance of kpc (Smith, 2006), the angular separation translates to a projected physical distance of 3.5 A.U. The second component appears of a quite similar brightness and is thus likely an O star as well. We estimate the mass-ratio using a mass-luminosity relation given by (Brott et al., 2011). Considering no color difference between the two components of the pair, thus no difference in their bolometric correction, we derive a mass ratio of 0.9.
Assuming a circular orbit and a primary mass of 47 M as appropriate for an O4 III star (Martins et al., 2005a), a 3.5 A.U. semi-major axis corresponds to an orbital period of about 250 d and an orbital velocity of 70 km s. A maximum inclination of 2.5° is required to hide such a signal at the 5- level in the noise associated the RV measurements. Assuming random orientation of binary orbits in space, one expects that such a low inclination is encountered with a 1:10,000 frequency and is thus very rare.
Alternatively, HD 93250 might be a long period, highly eccentric system. We used the method described in Sana et al. (2009) to estimate the likelihood to observe a significant RV shift as a function of the orbital period given the sampling provided by the optical observations and using no constraint on the orbital inclination. The probability that HD 93250 would have displayed a RV shift larger than 5 km s is very close to 1.0 up to periods of d, then drops sharply for longer periods (Fig. 2). As no such a shift is seen, this suggests a minimum orbital period of 30 yr, which would imply an eccentricity as large as 0.9 to match the current separation.
There is however an additional detection bias that results from the blending of two equal-strength lines, which is not taken into account in Sana et al. (2009). Indeed the RV measurement obtained when applying single-Gaussian fitting to blended lines will remain constant (Fig. 3) until evidence for a double-lined profile is seen (typically at RV 100 km s). To estimate this bias, we have generated synthetic profiles of O4 III+O4 III binaries using the non-LTE atmosphere code FASTWIND (Puls et al., 2005), for various RV shifts up to 200 km s and for different rotational broadening. Accounting for the typical of our optical data, we have then fitted a single Gaussian profile. As expected, the measured RVs show no significant variations. For a typical rotational broadening of 100 km s, as appropriate for HD 93250, the measured amplitudes and FWHMs are changing by less that 5 and 10% respectively for a RV difference between the two components of less than 55 and 70 km s (Fig. 4). Using these new values for the detection thresholds, we note that the detection probability already starts to drop for period larger than a few 100s days.
As a conclusion, the fact that RV campaigns have failed to reveal the binary nature of HD 93250 does not necessary imply exceptionally long periods, high eccentricities and/or small inclination angles. It might also result from observational biases affecting binaries with (quasi) identical components.
3.3 Colliding winds
Using the formalism of Usov (1992) adapted to O+O binaries, we estimate the X-ray luminosity that would be produced by the collision of two O4 III winds of equal strength to be of the order of erg s. There is enough uncertainty in the mass-loss rate associated with HD 93250 (see the discussion in Gagné et al., 2011) to accommodate the range of periods and separations discussed in the previous section. It is thus not possible to use the X-ray luminosity to further constrain the orbital properties. Assuming that the wind-wind collision is adiabatic, as expected for wide binaries, the X-ray flux produced in the wind collision should scale with the separation as (Stevens et al., 1992). Given that Rauw et al. (2009) reported a variation of the X-ray flux of the order of 40%, one would thus expect HD 93250 to have a minimum eccentricity of 0.2.
In a scenario where the observed radio synchrotron radiation is produced by colliding winds, the wind-wind interaction region should be located out of the bulk of the stellar wind material. Otherwise, the non-thermal radio photons suffer from significant free-free absorption and cannot escape the radio photosphere. Adopting the mass loss rate and the terminal velocity given by Martins et al. (2005b) (respectively 5.62 10 M yr and 3000 km s), and using equations of Wright & Barlow (1975) and Leitherer et al. (1995), we estimate the radius of the photosphere using the same procedure as in the case of HD 168112 in De Becker et al. (2004). The two frequencies investigated by Leitherer et al. (1995), i.e. 8.64 and 4.80 GHz, yield respectively photosphere radii of 1.4 and 2.2 A.U. As the two components of the system are very similar, the wind-wind interaction region is expected to be located close to mid-separation, i.e. about 1.7-1.8 A.U. according to our results and assuming no significant inclination effect. Interestingly, these values are in good agreement with the detection of HD 93250 at 8.64 GHz, and its non-detection at 4.80 GHz (Leitherer et al., 1995). We emphasize that these considerations are likely to be epoch-dependent. For instance, an increase of the stellar separation in an eccentric orbit could lead to the detection of the synchrotron radiation at 4.80 GHz as well. This scenario can only be verified through a multi-frequency radio monitoring of HD 93250 on time scales covering a significant fraction of the orbital period.
Using the ESO/VLTI, we provided evidence that HD 93250 is an O+O binary system. The binary nature gives a natural explanation of many of the intriguing properties of the object. In particular, it can explain the X-ray over-luminosity and NT radio emission in the framework of a colliding wind scenario. The latter point lends further support to the so-called ‘standard scenario’ where particle acceleration occurs in the wind-wind interaction of a massive binary system. Because we see no change in the orientation of the system in our two VLTI observations separated by three months, we conclude that HD 93250 has a likely orbital period of the order of one to several years. A significant eccentricity is also required to explain the variations observed in the X-ray flux.
We further show that the non-detection of a RV signature based on eight high-resolution spectra covering a seven-year period of time does not require a particularly unfavorable orbital configuration (extremely low inclination, long period and/or high eccentricity). It can also be explained by observational biases resulting from line-blending in nearly equal-mass systems. Further spectroscopic and high-angular observations are require to characterize the orbit of the system, which is needed to quantitatively understand the X-ray and NT radio processes originating from the wind-wind collision.
Together with HD 168112 (catalog HD~168112), HD 93250 was one of the only two known O-type NT radio emitters for which the binary scenario was questioned. Our results provide further evidence that binarity and NT radio emission go hand-in-hand. As was the case for HD 93250, RV campaigns have so far failed to confirm the binary nature of HD 168112. Long base-line interferometry might thus provide a complementary diagnostic tool to test the multiplicity properties of HD 168112 in particular and of single non-thermal radio emitters in general.
- affiliation: Based on observations collected under program IDs 086.D-0586 and 386.D-0198 at the European Organisation for Astronomical Research in the Southern Hemisphere, Chile
- affiliation: UJF-Grenoble 1 / CNRS-INSU, Institut de Planétologie et d’Astrophysique de Grenoble (IPAG) UMR 5274, Grenoble, France
- affiliation: Astronomical Institute, Utrecht University, Princetonplein 5, 3584 CC, Utrecht, The Netherlands
- JMMC software: http://www.jmmc.fr
- Brott, I., de Mink, S. E., Cantiello, M., et al. 2011, A&A, 530, A115+
- Chelli, A., Utrera, O. H., & Duvert, G. 2009, A&A, 502, 705
- De Becker, M. 2007, A&A Rev., 14, 171
- De Becker, M., Rauw, G., Blomme, R., et al. 2004, A&A, 420, 1061
- Gagné, M., Fehon, G., Savoy, M. R., et al. 2011, ApJS, 194, 5
- Haguenauer, P., Alonso, J., Bourget, P., et al. 2010, in Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 7734, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series
- Kraus, S., Weigelt, G., Balega, Y. Y., et al. 2009, A&A, 497, 195
- Leitherer, C., Chapman, J. M., & Koribalski, B. 1995, ApJ, 450, 289
- Luo, D., McCray, R., & Mac Low, M. 1990, ApJ, 362, 267
- Martins, F., & Plez, B. 2006, A&A, 457, 637
- Martins, F., Schaerer, D., & Hillier, D. 2005a, A&A, 436, 1049
- Martins, F., Schaerer, D., Hillier, D. J., et al. 2005b, A&A, 441, 735
- Nazé, Y., Broos, P. S., Oskinova, L., et al. 2011, ApJS, 194, 7
- Nelan, E. P., Walborn, N. R., Wallace, D. J., et al. 2004, AJ, 128, 323
- Petrov, R. G., Malbet, F., Weigelt, G., et al. 2007, A&A, 464, 1
- Puls, J., Urbaneja, M. A., Venero, R., et al. 2005, A&A, 435, 669
- Rauw, G., Nazé, Y., Fernández Lajús, E., et al. 2009, MNRAS, 398, 1582
- Sana, H. & Evans, C. 2010, in IAUS, Vol. 272, Active OB stars, eds. C. Neiner, G. Wade, G. Meynet & G. Peters (Cambridge University Press, 474
- Sana, H., Gosset, E., & Evans, C. J. 2009, MNRAS, 400, 1479
- Sana, H., Rauw, G., Nazé, Y., Gosset, E., & Vreux, J.-M. 2006, MNRAS, 372, 661
- Smith, N. 2006, ApJ, 644, 1151
- Stevens, I. R., Blondin, J. M., & Pollock, A. M. T. 1992, ApJ, 386, 265
- Tatulli, E., Millour, F., Chelli, A., et al. 2007, A&A, 464, 29
- Townsley, L. K., Broos, P. S., Corcoran, M. F., et al. 2011, ApJS, 194, 1
- Usov, V. V. 1992, ApJ, 389, 635
- van Loo, S., Runacres, M. C., & Blomme, R. 2006, A&A, 452, 1011
- Walborn, N. R., Sota, A., Maíz Apellániz, J., et al. 2010, ApJ, 711, L143
- Wright, A. E. & Barlow, M. J. 1975, MNRAS, 170, 41