8. Speculations in the absence of firm answers
The analyses presented above indicate several (in some cases mutually exclusive) ways to possible speculative interpretations. Since they can only be tested by new dedicated observations, I find useful to mention them explicitly.
8.1. Alternative models of CMa
The result displayed in Fig. 10 is potentially quite exciting since it may indicate that the semi-amplitude of the 1 372 RV curve is largest around the phases of the periastron passage of the putative 34 675 binary system. This hypothesis offers an explanation of the 1 372 period as the tidal pulsational mode.
Note also that
2 34 675-1 + 1 371906-1 = 1 3453
The 1 345 period is reminiscent of the periods found from the multiperiodic analyses of the RV data. This can be interpreted in at least two different ways. One possibility is that the medium-term light changes represent a beat period of the two short periods. This seems to contradict the simulations with artificial data, however. Another is to assume that the true physical periods are twice longer, i.e. 2 74 and 2 69 and represent in fact the synodic and sidereal periods of rotation of the primary star in the 34 675 binary system. For a high eccentricity of the 34 675 period, 2 7 could be the proper value for the spin-orbit synchronization at periastron.
It will be important to decide whether the 1 372 period undergoes a slow cyclic change or a periodic change with the 34 675 period. If the former is confirmed, such a variation could - in principle - be explained in terms of the light-time effect in a wide binary. Already in 1981 I suggested (Harmanec 1982) that the recurrent shell phases observed for Be stars like BU Tau (HD 23862), V832 Cyg (59 Cyg) or Cas (HD 5394) could be causally related to the fact that these Be stars might be long-periodic binaries. Duplicity of BU Tau was indeed discovered by Gies et al. (1990) during a lunar occultation of the star and the authors explicitly suggested that the shell episodes of BU Tau are driven by tidal interactions near periastron. Duplicity of V832 Cyg was discovered by McAlister et al. (1984) by means of speckle interferometry but no relation to the spectral variations of the star has been demonstrated as yet. There is a suspicion that the major emission episodes of o And (HD 217675) occur with a period of its closer speckle-interferometric companion (see Harmanec et al. 1987 and references therein). One speckle-interferometric observation of CMa was obtained in the spring of 1996 (JD about 2450300) and no binary component at distances from 0:00035 to 1:005 was found (Hartkopf 1997, priv.com., Mason et al. 1997). I estimate that for the Hipparcos parallax of 0:0000353 and orbital periods from 2700 to 8100 d, the angular projection of the semimajor axis of the putative binary would be 0:00028 - 0:00060. Clearly, the binary interpretation of the long-term changes of CMa is not ruled out and continuing interferometric observations are desirable.
8.2. What causes the changes with the 1 372 period?
The true nature of the periodic 1 372 (or 2 744) changes remains unexplained. One seeks an explanation for the large-amplitude RV variations of the absorption-line cores, accompanied by lower-amplitude RV changes of the Balmer emission lines, occurring in antiphase to the absorption RV curve (Baade 1982a), and by virtually no variability of either brightness or the RV of the outer wings of the absorption line. All quantitative attempts to explain these changes by a low-mode of photospheric pulsations, corotating structures or orbital motion in a close binary led to serious problems but it is conceivable that a more sophisticated version of one of these models will succeed.
A natural explanation of the fact that the RV of the Balmer emission lines of CMa varies in exact antiphase to that of the absorption lines would be to assume that CMa is a close binary with an orbital period of 1 372. Taken at face value, the semi-amplitude of the emission RV changes (of 5.1 km s-1 only) leads to rather extreme assumptions about the nature of the binary components and the binary model does not appear tenable. However, before the binary origin of the 1 372 period is definitively ruled out, one should apply some disentangling technique, like Hadrava's (1995b) KOREL program, to spetral observations from a limited period of time (to eliminate the effects of varying emission strength). If two binary components with similar relatively broad-lined spectra were present, then the observed RV amplitudes would certainly be affected by both, the mutual blending of the two sets of lines and by the secularly varying strength of the emission. Note that for the binary system of two B stars V436 Per (1 Per) Harmanec et al. (1997) found that the direct RV measurements gave a RV curve with an amplitude of some 10 km s-1 only while the disentangling revealed that the true RV curves for both binary components have amplitudes of about 100 km s-1.
© European Southern Observatory (ESO) 1998
Online publication: May 15, 1998