High - resolution optical and infrared spectroscopic observations of Cir X - 1 Helen

We present new optical and infrared (IR) observations of Cir X-1 taken near apastron. Both sets of spectra show asymmetric emission lines. Archival optical observations show that an asymmetric Ha emission line has been in evidence for the past 20 years, although the shape of the line has changed significantly. We present an eccentric (e , 0:7–0:9† low-mass binary model, where the system consists of a neutron star orbiting around a (sub)giant companion star of 3–5 M(. We suggest that the broad components of the emission lines arise in a highvelocity, optically thick flow near the neutron star, while the narrow components of the optical and the IR lines arise near the companion star and a heated ejecta shell surrounding the binary respectively. In this model, the velocity of the narrow component reflects the space velocity of the binary; the implied radial velocity (1430 km s after correcting for Galactic rotation) is the highest velocity known for an X-ray binary. Key words: binaries: spectroscopic – stars: individual: Cir X-1 – X-rays: stars. 1 I N T RO D U C T I O N Cir X-1 is one of the most puzzling X-ray binaries known. Like the peculiar systems SS 433 and Cyg X-3, it cannot easily be classified into any of the major categories of X-ray binaries. Indeed, there is even doubt as to whether it is a high-mass X-ray binary (HMXB) or a low-mass X-ray binary (LMXB). Since its discovery in the early 1970s, Cir X-1 has been studied intensively at X-ray wavelengths. The X-ray properties of Cir X-1 were found to differ dramatically each time it was observed (see, e.g., Kaluzienski et al. 1976; Tennant 1988; Tsunemi et al. 1989; Shirey et al. 1996). Periodic modulation of the X-ray flux was found at a period of 16.6 d (Kaluzienski et al. 1976). A radio counterpart was detected (Clark, Parkinson & Caswell 1975), and was found to flare at the same period as the X-ray modulation (Haynes et al. 1978). These flares were initially detected at peak flux levels of .1 Jy; since the 1970s, the flux of the source has decreased dramatically, and it has only occasionally been detected above 50 mJy (Stewart et al. 1991). This radio source is located 25 arcmin from the centre of the supernova remnant G 321.920.3, and is apparently connected to the remnant by a radio nebula (Haynes et al. 1986). Stewart et al. (1993) have imaged arcmin-scale collimated structures within the surrounding nebula, suggesting an outflow from the X-ray binary. Fender et al. (1998) have imaged an arcsec-scale asymmetric jet aligned with these larger structures, raising the possibility that the outflow from the system is relativistic. Recently, Case & Bhattacharya (1998) have revised the estimated distance to G 321.920.3 (and hence to Cir X-1, assuming that they are associated) to 5.5 kpc, which is substantially smaller than the original suggested distance to Cir X-1 of 10 kpc (Goss & Mebold 1977). The discovery of Type I X-ray bursts (Tennant, Fabian & Shafer 1986b) suggests that the compact object is probably a weakly magnetized neutron star. The close association of Cir X-1 with the supernova remnant suggests that the system may be a young (,105 yr old) runaway system from a supernova explosion (Stewart et al. 1993). The optical counterpart to Cir X-1 was identified as a highly reddened star with strong Ha emission (Whelan et al. 1977). This object was later shown to consist of three stars within a radius of 1.5 arcsec, the southernmost of which is the true counterpart (Moneti 1992; Duncan, Stewart & Haynes 1993). The long orbital period and periodic X-ray activity suggested a high-mass system in an eccentric orbit (Murdin et al. 1980); however, the variability of the optical emission, the faintness of the optical counterpart, and several of its X-ray characteristics suggest that the companion is a low-mass star. The lack of spectroscopic studies in the optical band means that most of the fundamental orbital parameters of Cir X-1 have not been determined. Moreover, Cir X-1 shows very different properties q 1999 RAS E-mail: [email protected] 416 H. M. Johnston, R. Fender and K. Wu from time to time. This make it difficult to construct a coherent picture for the system from observations of different wavelengths at different epochs. Here we present new spectroscopic observations of Cir X-1, and use these, together with analysis of archival observations of the system, to suggest a more coherent model for the system. 2 O B S E RVAT I O N S A N D DATA R E D U C T I O N 2.1 New optical observations Cir X-1 was observed on 1997 June 4 using the 3.9-m AngloAustralian Telescope (AAT). The mean orbital phase of the observation was 0.51, calculated according the ephemeris of Stewart et al. (1991). The RGO Spectrograph was used in combination with the TEK 1k CCD in the 82-cm camera and a grating of 270 groove mm in first order, resulting in a dispersion of ,1:08 A pixel21 over a wavelength range 6060–7165 Å. The spatial scale was 0.25 arcsec pixel; the spectral resolution, measured from the arc lines, was 5.4 Å. A 1.5-arcsec-wide slit was used, oriented north–south so that both Cir X-1 and star 2 of Moneti (1992) were in the slit. The atmospheric seeing was about 1 arcsec. Five 1800-s integrations were taken, interspersed with CuAr arc-lamp exposures, before cloud prevented the acquisition of any more data. The bias and pixel-to-pixel gain variations were removed from each exposure using standard procedures in iraf. Cosmic rays were removed using the method of Croke (1995) to compare adjacent frames. Because of the presence of the nearby confusing star (Moneti’s star 2), special care needed to be taken to measure the flux from our object. At every position along the dispersion direction, we fit two Gaussian profiles, with fixed widths …FWHM ˆ 5:6 pixels† and separation (6 pixels), to the skysubtracted frames. The amplitude of these Gaussians was used as the estimate of the flux from Cir X-1 and star 2 at each wavelength. We then determined the wavelength calibration using the CuAr arc lamp exposures. We fitted a low-order polynomial to the arc line wavelengths as a function of pixel number: the rms scatter of the fits was ,1=4 of a pixel. A rough flux-calibration was performed by comparing with the spectrum of the observed flux standard LTT 4364 although, since the night was nonphotometric, this flux calibration should be considered only approximate. 2.2 Infrared observations K-band spectroscopy of Cir X-1 was obtained using the Cryogenic Array Spectrometer/Imager (CASPIR) on the ANU 2.3-m telescope at Siding Spring Observatory on the night of 1997 June 20. The K grism was used with the SBRC 256 256 InSb array, giving a dispersion of 21.5 Å pixel over a wavelength range of 1.94–2.49mm. The spatial scale was 0.5 arcsec pixel. A 5-arcsec slit was used, oriented east–west: note that this means that Moneti’s stars 2 and 3 both contributed light in our spectrum. The telescope was nodded by ^12 arcsec along the slit to provide sky frames at the same position as the object. Argon lamp spectra were taken to perform wavelength calibration, and two nearby bright stars (BS 5699 and 5712) were observed in order to remove atmospheric spectral features and perform flux calibration. Standard data reduction procedures were followed, using the local caspir package running in iraf. Bias and dark frames were used to linearize all frames, the sky background was subtracted from the object frames, and pixel-to-pixel variations were corrected. The chip distortion was corrected in order to align the dispersion and spatial directions along rows and columns of the chip; the sky background was then subtracted and spectra extracted. A low-order polynomial was fitted to the argon lines, and these calibrations applied to the object spectra. Flux calibration was achieved by dividing the observed spectra by the spectrum of a nearby mid-G-type star and then multiplying by a model for the absolute flux distribution of the calibrator. Residual terrestrial atmospheric features were then corrected using an early-type star.