Between 1990 and 1995 we obtained 33 single-dish spectra, from which we display three in Fig. 1 to show the general appearance of the maser during this time. During the bright phase of the pulsation cycle the maser consists of three features, one blue- (labeled B in Fig. 1) and the other two redshifted (labeled G(reen) and R(ed)) with respect to the stellar velocity (20 km s-1), each feature being made up of several individual maser lines. The overall shape of the profile is similar to the double-peaked OH profile, and is therefore reminiscent of radially beamed masers in a spherically symmetric expanding circumstellar shell. In the following we refer to these maser features as radial masers. Close to minimum these masers are extinguished, and from the upper limits ( mJy (3 )) of the VLA observation of August 1991 we can infer that the flux densities vary at least a factor 100-1000 during the pulsation cycle.
The spectra during the bright phases in 1990 and 1993 are strikingly similar. Not only did the three maser features reappear in 1993 with a similar shape but also the individual lines contributing to each feature are present in both epochs. The blueshifted feature B, for example, is a blend of 5 individual maser lines, which have within the measuring errors ( km s-1) the same velocities in 1990 and 1993. This indicates that the radial masers reappeared at exactly the same velocities, after being absent when the star was dim. The velocity range over which H2 O maser emission appears is 5-38 km s-1, almost the same range as displayed by the OH maser (2-38 km s-1 ; Slootmaker et al. 1985), showing that the most redshifted parts of the H2 O maser have the same outflow velocity as the OH maser shell.
During the weak phase in 1991, when the double-peaked maser had disappeared (at phase ) a new maser line at 17.5 km s-1 appeared, reached its maximum at and disappeared at , when the radial masers showed up again. No similar line was observed during the following weak phase in 1995. The new maser is blueshifted by 2.5 km s-1 with respect to the stellar velocity and is either a tangentially beamed maser or originates very close to the star, where the outflow velocity is still small. We believe the first one of these possibilities to be true (see below) and refer to this component in the future as "tangential". Apparently we observed a switch from radial to tangential beaming of the maser, correlated with the variability of the star. In Table 1 we list the integrated fluxes of the blueshifted feature (B) and the tangential line (T) for spectra with sufficient S/N, and their temporal variations are shown in Fig. 2. The striking anticorrelation between the radial and the tangential lines strongly suggests that the two beaming directions are mutually exclusive. None of the spectra, taken close to the phases in which the transition must have taken place (), showed both lines simultaneously.
3.2. Interferometric observations
Visual inspection of the channel maps showed immediately that we were able to resolve some of the maser features spatially. In particular, feature G (cf. Figs. 1 and 3) is clearly separated into two maser groups with a separation of . At 1 kpc this corresponds to a distance of 1.5 cm (100 AU), giving a lower limit for the outer diameter of the H2 O maser shell.
The number of unambiguously identified maser spots in the channel maps is 15 per epoch and their location is shown in Fig. 3. There is a striking similarity between the maps of the three epochs when radial masers were present. Comparison of individual channel maps between the three epochs show that the radial maser lines came from the same positions in 1990 and 1992. Thus, the maser lines reappear at the same velocity and position after being extinguished close to the minimum of the pulsation cycle. We have to conclude that the velocity structure within the H2 O maser zone giving rise to the radial masers is not altered significantly during one pulsation cycle of the star ( years) and that the maser lines reappear as soon as the conditions to pump the maser are restored. This finding is not surprising, because during one period a mass element is able to move only 2 cm outwards, which is small compared to the size of the shell.
The maps of the double-peaked masers are made up, first of all, of maser lines belonging to the blueshifted and the most redshifted feature (labeled B and R in Fig. 1), which are spaced close together and are encircled in Fig. 3. Secondly, there are the maser lines belonging to the G-feature and finally several weak masers located in velocity close to the inner wings of the two maser complexes B and G. The most blue- and redshifted maser lines agree in position on the sky to better than 15 mas, which is strong evidence for radial beaming in an isotropically expanding circumstellar shell. Both maser components define therefore uniquely the line of sight to the star. Averaging over three epochs we find the position of the star to be , (1950) coincident within errors with the position obtained from interferometric observations of the OH maser (Bowers et al. 1981). This position is chosen as origin of the maps shown in Fig. 3.
The VLA observation in August 1991 () showed maser emission only from the 17.5 km s-1 line. Its position is (or 3 cm) away from the star, a surprisingly large value compared to the positional spread of the other maser spots. Unfortunately there is no maser spot in this epoch in common with the others, inhibiting a direct alignment of the maps. We cannot exclude the possibility that the large deviation was caused by phase instabilities during the 1991 August observation, although we consider this to be unlikely given the close coincidence ( mas) of the maps for the other three observing periods. The 17.5 km s-1 line is therefore clearly off the line of sight to the star and hence we conclude that the beam direction is almost tangential. Because of the relatively large distance from the star, one might question the association of the 17.5 km s-1 line with OH 39.7+1.5. However, the striking anti-correlation in variability with the radial masers (Fig. 2) and the observation of a similar "tangential" maser line (then at 20.6 km s-1) during an earlier pulsation cycle (Engels et al. 1986) prove that the association is real.
Fig. 4 shows a plot of relative velocity against projected radial distance from the star for all maser spots identified in the four epochs. For a thin shell with constant expansion velocity all maser spots are expected to fall close to the locus of a single ellipse. This is not evident here, instead a geometrically thick shell with inner and outer radius of 0.6 and 3 cm and inner and outer expansion velocity of 9-10 and 17-18 km s-1 is displayed. The existence of an inner radius is probably due to high gas density close to the star, leading to quenching of the maser emission (Cooke & Elitzur 1985) in the inner part of the circumstellar shell. The outer radius, which is set by the largest separation measured for a radial maser and by the location of the tangential maser line, implies an H2 O maser shell size (diameter) of 400 AU.
The water maser shell for OH 39.7+1.5 lies mostly in the outer part of the acceleration zone of the circumstellar shell, where the velocity gradient is sufficiently small that maser excitation along radial paths occur. As noted in Sect. 3.1 and as is evident from Fig. 4, only the most redshifted H2 O maser velocities reach the terminal expansion velocity given by the OH maser. The B-feature is closely aligned with the R-feature, but has a smaller outflow velocity, implying a position closer to the star. The outer radius, as suggested by Fig. 4, may then give only a lower limit for the maximum size of the water maser shell, because the maser spots determining the outer ellipse in Fig. 4 do not necessarily move with the terminal expansion velocity.
© European Southern Observatory (ESO) 1997
Online publication: June 30, 1998