Background

Maser activity in radio recombination line (RRL) emission was first recognized in 1988 in data obtained with the IRAM 30 m telescope [4]. One of the latest additions to the family of astrophysical masers, this new phenomenon has been detected so far only in the circumstellar shell of the optically inconspicuous Be star MWC349. Investigations over a wavelength range of 1 to 3 mm, showed that the maser activity begins near 100 GHz. The maser line flux FL grows rapidly with increasing frequency, reaching an easily detectable antenna temperature of ~2 K at H30a (transition between principal quantum numbers n = 31 ® 30) near 232 GHz.

The RRL maser profile includes two ~14 kms-1 wide spikes superposed on a weak and broad pedestal feature (Fig.1). Interferometry with OVRO at 1 mm [6] has shown that the two spikes, dubbed `blue' and `red', are separated by 80 a.u. in a direction coincident with the orientation of the neutral circumstellar disk inferred previously [1, 3, 13]. It is therefore thought that the maser is located along symmetrical tangential lines of sight where, on a rotating disk, the velocity shear is a minimum [2, 10]. If the velocity separation of the blue and red spikes, ~50 kms-1, truly reflects the rotation velocity of a Keplerian disk, a central (stellar) mass of 30 M¤ is implied.

An important property of a maser is whether it is saturated or not. For the RRL maser, evidence from small but systematic variations of the width of the spikes and from the variabilty of their flux suggests that the degree of saturation increases systematically with frequency, showing signs of saturated behavior beginning near 1 mm.

Why ever higher frequencies?

The systematic increase of maser line flux from 100 to 250 GHz suggested that the maser might be even stronger at submillimeter wavelengths and thus be easily detectable in the 0.85 mm atmospheric window. Theoretical considerations also suggested that the principal requirements of a maser, inversion of the level populations and sufficiently long coherent gain paths, are satisfied even at much higher frequencies. Walmsley's calculation [12] of the NLTE departure coefficients showed that the line absorption coefficient, kL, stays negative at submm wavelengths for a large range of electron densities ne. His work indicates that with decreasing quantum numbers n the (negative) kL peaks at ever higher ne. At H30a, for example, regions of ne = 107 cm-3 would be seen masing, whereas at H21a regions of ~108 cm-3 would contribute most. With increasing frequency, the RRL maser therefore probes denser regions of the sources, presumably located progressively closer to the center. We know that a range of densities up to values of at least 1010 cm-3 exists in the circumstellar disk [3].

Figure 1. Spectra of the H21a transition at 662.4 Ghz (top) and of the H26a transition (bottom), both obtained with the JCMT on March 9th, 1994 within the space of 4 hours. For the H21a line, a new SIS receiver built by the MPE Group [8] was used. The sky opacity at zenith was ~0.75, integration time was 1.1 hours.

Another even more intriguing aspect of Walmsley's calculations is that the (negative) kL maxima increase with frequency. An ever shorter path s is needed for the higher frequency transitions to achieve unit gain. While for H41a (near the onset of the maser activity) s is of the order of 200 a.u., which is conspicuously close to the inferred outer disk diameter [13], s is only ~30 a.u. at H30a. This suggests that the maser gain is increasing with frequency, and the question arises where does all this stop?

Apparently not with the H26a transition. We have observed the 353.6 GHz transition several times with both the JCMT and the 30 m telescope [9] and it follows the predicted pattern of increasing line flux and saturation very nicely. We were therefore compelled to try an even higher frequency transition, and we chose H21a at 662.4 GHz, which is located in a frequency range where the atmospheric opacity is still not prohibitively large. The JCMT is the telescope best suited for such a study and a good receiver is available thanks to the long-standing cooperation between the Max-Planck-Institut für Extraterrestrische Physik and the Joint Astronomy Centre.

Getting data at the edge of feasibility requires, among other things, patience. In our case, more than 4 years have passed from the first proposal until successful observation (and publication [11]). We were rewarded with a high--quality detection of H21a (Fig. 1), constituting the highest frequency recombination line observed with radio techniques, and at the same time the highest frequency astrophysical maser of any kind.

Why did it take us so long? Poor high frequency weather accounts for most of the delay, and we think that success was possible at the end due to a combination of several factors: decent weather, an excellent receiver, an efficient telescope at 662 GHz, the presence of the receiver builders on the mountain, and flexible scheduling of the telescope. A general summary of this winter's high frequency observations on the JCMT is given elsewhere in this issue by Andy Harris and Linda Tacconi.

New results

The most striking finding from the H21a spectrum is that the maser flux is still increasing. Fig. 2 shows that the increase is well described above 250 GHz by a power law spectral index a = 3.4. Some of this growth is simply due to the line width. Since it is nearly constant in velocity, its frequency width grows linearly with n, and the line flux density grows then as a ~2.4. This value is very similar to the spectral index of the line absorption coefficient, as would be expected for saturated growth.

The velocity separation of the blue and red spikes of the H21a transition is found to be 50.6 ±0.2 kms-1, indistinguishable from the H26a value. This result came as a big surprise, since we expected the H21a maser to originate much closer to the center than the lower-frequency transition, where a Keplerian disk would rotate much faster. This dramatic velocity increase is not seen. The explanation of this confusing finding could be that the disk is braked inside the H30a emission radius, or even stranger that the density structure in the inner disk is rather peculiar. An interferometric measurement of the angular separation of the H26a spikes would help to find the answer.

Measuring wind speeds

Despite the overall similarity of the H21a and H26a spectra there are some interesting differences which are evident in the H26a/H21a ratio spectrum in Fig. 3. The relative increase in line strength is least at the radial velocities of the maser spikes, while the ratio is much higher between the spikes. This testifies again in favour of saturated growth of the spikes, but it also means that the low velocity gas, likely originating mostly on the disk near the central line of sight, is also subject to maser amplification.

Figure 2. Line fluxes (time averages when several observations are available) of maser spikes. Transitions n+1 -> n in the range n = 41 ... 21 are shown.

The two most prominent peaks in Fig. 3 are at extreme velocities, much beyond those thought to occur within the disk. We attribute these peaks to stimulated (weak maser) emission of gas in the strong ionized wind. Conditions for amplification in the nearly spherically expanding wind are optimal along the central line of sight in front and to the rear of the star. Their mean velocity, vc = 8.5 kms-1, is in good agreement with the velocity centroid of the maser spikes, and lends support to the interpretation of vc as the stellar velocity.

Furthermore, in this interpretation the velocity separation of the peaks would then measure twice the expansion velocity (i.e. wind speed) of the gas emitting at submm wavelengths. The wind speed inferred from this novel method is somewhat smaller than the conical value of 50 kms-1, but not greatly so. Whether this means that at submm wavelengths we start to see into the acceleration region of the wind, awaits further analysis.

Is MWC349 unique?

Given the highly controversial physical nature of MWC349, the only known source of a RRL maser, we (and others) have spent considerable effort in searching for a second maser in a large variety of astronomical objects, so far in vain. Probably the most telling result is the absence of masers in Wolf-Rayet stellar winds [5]. Given their strong ionized winds, optical depths should be high enough (and inversion is ubiquitous) for masers to occur. We interpret our failure to mean that Wolf-Rayet stars do not have the disk structure, or velocity coherence, required for RRL maser emission.

We have monitored the maser in MWC349 at many transitions over the previous 5 years. We find that the basic double-peaked pattern is remarkably stable. In particular we find only minute variations of the spike velocity separation, in contrast with the considerable independent intensity changes of the spikes. This is not the behavior expected from a disk created by violent mass loss, typical of the final stages of a star's life. The pattern relates more naturally to a steady and massive disk, such as those built up by accretion. The high central mass derived from the disk rotation and the high ionizing luminosity required by the presence of an ionized wind and the RRL maser then suggest a scenario in which a massive ZAMS star is observed during the fleeting period in its evolution when it is no longer obscured by its nascent cloud, yet its high luminosity and powerful wind have not yet succeeded to fully destroy its accretion disk.

Figure 3. Ratio of H21a and H26a spectra smoothed to a resolution of 2.1 km/s and plotted on a logarithmic scale. The velocities labelled B and R correspond to the blue and red maser spikes. The average velocity of the ratio maxima labelled front and rear is indicated by the vc arrow.

Acknowledgements: We are deeply appreciative of the expert assistance given us by the JCMT telescope operators during these observations, and of course, without the truly excellent receiver built by our colleagues at the MPE in Garching these observations could not have been successful. We thank Ian Robson, Director of the JCMT, and Reinhard Genzel of the MPE for arranging this continuing collaboration.

[1] Cohen, M., Bieging, J.H., Dreher, J.W., Welch, W.J. 1985, ApJ, 292, 249

[2] Gordon, M. 1992, ApJ, 387, 701

[3] Hamann, F., Simon, M. 1986, ApJ, 311, 909

[4] Martín-Pintado, J., Bachiller, R., Thum, C. & Walmsley, C.M. 1989, A&A, 215, L13 (paper I)

[5] Matthews, H.E., et al. , in preparation

[6] Planesas, P., Martín-Pintado, J., Serabyn, E. 1992, ApJ Letters, 386, L23

[7] Ponomarev, V.O., Smith, H.A. & Strelnitski, V.S. 1994, Ap.J. 424, 976

[8] Schuster, K.F., Harris, A.I., Gundlach, K.-H. 1993, Int. J. IR and MM waves 14, 1867

[9] Thum, C., Matthews, H.E., Martín-Pintado, J., Serabyn, E., Planesas, P., Bachiller, R. 1994, A&A, 283, 582

[10] Thum, C., Martín-Pintado, J., Bachiller, R. 1992, A&A, 256, 507 (TMB)

[11] Thum, C., Matthews, H.E., Harris, A.I., Tacconi, L. Schuster, K.J., and J. Martín-Pintado, A&A, in press

[12] Walmsley, C.M 1990, A&AS, 82, 201

[13] White, R.L., Becker, R.H. 1985, ApJ, 297, 677

H.E. Matthews, JAC

C. Thum, IRAM, France