1 Introduction

HD 93206 is the brightest member of the open cluster Collinder 228 in the $\eta$ Carinae region; according to Lyngå (1987) the cluster distance is 2600 pc. A variable brightness of HD 93206 was first noted by Morgan et al. (1955). Several radial velocities were published by Feast et al. (1956) and Buscombe & Kennedy (1966). Walker & Marino (1972) determined the type and period of variability and published first photoelectric measurements. The multiple system HD 93206 contains the eclipsing pair QZ Car. Its period is very close to 6 days, depths of minima are 0 $.\!\!^{\rm m}$24 and 0 $.\!\!^{\rm m}$20 in V, maximum brightness is V = 6.22, and spectral type O 9.7 Ib:(n) according to Walborn (1973). Some photoelectric measurements were published by Moffat (1977). Two systems of periodically variable spectral lines were found by Morrison & Conti (1979): the period of the weaker lines corresponds to the eclipsing variable, and is different from that of the stronger lines of about 20 $.\!\!^{\rm d}$73. Therefore it is evident that HD 93206 is composed of two binaries, both single-lined. Recently, the cross-correlation method was applied to nine IUE spectra by Stickland (2000).

Morrison & Conti (1980; hereafter MC) remarked that it would be probably wise to spend ones resources on studying a less complicated binary than QZ Car. Certainly this is true when one tries to add new entries to the table of accurate stellar masses and radii. However, QZ Car is an especially remarkable case. Note, e.g., that there is no other star of such late type with the spectral "(f)'' characteristics among stars listed by Conti et al. (1977). The most similar case is perhaps MY Ser of type O8(f), which is another multiple system, and equally difficult to study.

The system was also studied by Leung et al. (1979; hereafter LMS), who confirmed the spectroscopic results of Morrison & Conti (1979). The mass functions suggest that the eclipsing binary component with so far undetected lines is more massive than its binary companion; in the non-eclipsing pair, which has a longer period, the invisible component is several times less massive than the primary component, and therefore probably has very weak, unobservable lines. LMS solved the light curve by Walker and Marino and found a luminosity ratio close to 3:1 for the two binary components, so supporting the hope that the secondary lines might be discernible in high resolution spectra of high S/N. MC analyzed a larger sample of spectra than LMS and found more reliable elements of both binaries. To avoid confusion, we will denote the two binaries in the following in the same way as MC: the long-period binary is called system A (its visible component A1), and the eclipsing binary is referred to as system B (with the brighter, but less massive component B1, and companion B2).

Figure 1: V photometry phased according to ephemeris (1). Crosses - our data, open circles - Christie, triangles - Moffat, points - Walker & Marino