Multiple variations in the radio light-curve of the colliding wind binary WR 146 (WC6+O): evidence for a third component

The Wolf-Rayet star WR 146 (HM19-3, WC6+O) is the brightest WR star at radi o wavelengths. We have been monitoring this system with the Westerbork Synt hesis Radio Telescope (WSRT) at 1.4 and 5 GHz (21 and 6 cm) since 1989. The time-averaged spectral index alpha(5-1.4 GHz) similar or equal to -0.62 cl early points to a domination by non-thermal radiation, which we associate w ith colliding winds in this binary system. The non-thermal radio flux distr ibution shows a turn-over at low frequency, which we suggest to be due to f ree-free absorption of the synchrotron emission from the colliding wind reg ion by plasma around the system. In the period 1989-1997 the average 1.4-GHz flux density increased from sim ilar to 61 to similar to 73 mJy; in the the period 1989-1999 the average 5- GHz flux density increased from similar to 29 to similar to 37 mJy The ligh t-curves show three different kinds of variations: (i) a slow linear rise i n a time-span of a decade; (ii) a 3.38 yr periodic variation; and, (iii) ra pid non-periodic variations on a time-scale of weeks. We examine whether the slow rise of the flux density could be explained by decreasing free-free absorption in the line-of-sight through the radiophoto sphere of the O component, while moving in an eccentric orbit around the WR component. However, the similarity of the amplitudes (similar to 22% in 10 yr) of the rises at 1.4 and 5 GHz argues against a change in free-free abs orption, expected to be strongly wavelength dependent. This points to an in trinsic flux-density variation, possibly due to modulation of the magnetic field strength resulting from orbital motion in a very-long-period eccentri c binary system. The relation between the flux-density increase and orbital motion is supported by positional measurements of the 5-GHz data. We detect a possible motion of the shock zone relative to one of the contro l sources (Control A) of similar to 0".05 in the 10 yr observing span. At a distance of 1250 pc this motion corresponds to a projected tangential velo city of about 30 kms(-1), which is a plausible orbital velocity for a syste m like WR 146. Superimposed on the 1.4-GHz slow rise, we find a sinusoidal variation with a period P = 3.38 +/- 0.02 yr and a semi-amplitude of 4.3 +/- 0.2 mJy. Adop ting a distance of 1250 pc to the system and a 162 mas WR+O separation, we consider the observed 3.38 yr period too short to be the WR+O binary period by at least two orders of magnitude. We suggest that the periodic variabil ity is caused by a third, low-mass object, modulating the mass flow and/or the magnetic-field of the O component. Unfortunately, our 5-GHz data are fa r too few and not adequately spread over the whole phase to confirm that th ey consistently follow the 3.38 yr period found in the 1.4-GHz data. The erratic 'micro'-variation in the 1.4-GHz light-curve is about 4 sigma o f the typical 0.5 mJy observational uncertainty, on a time-scale of weeks t o months. When irregularities in the mass flow (clumps, inhomogeneities and /or turbulence in the O and/or WR star winds) reach the wind collision regi on, variation in the non-thermal emission can be expected. Such irregularit ies can also affect the free-free line-of-sight absorption at the lowest ob serving frequencies.