Astron. Astrophys. 339, 897-903 (1998)

3. Results

3.1. Detections

With integration times of 30 to 60 minutes on 10 targets, 2 sources were detected. Source OH359.971-0.119 was detected in December 1995 at a SNR of 6.5 and a velocity of -10.50.9 . An attempt to re-detect OH359.971-0.119 in January 1996 was unsuccessful. It was detected in the simultaneously correlated VLA data, and should therefore be detectable with the VLBA. We are convinced that the detection from December 1995 is real, as the maser can be seen consistently in two phase-referencing blocks that are more than two hours apart. In the Jan 96 experiment we detected OH359.810-0.070 at a SNR of 7.6 and a velocity of -34.10.9 . We show the images for the channel with the highest peak flux in Fig. 3 and Fig. 4. Positions and velocities (Table 3) were measured from the image with the AIPS task IMFIT.

Fig. 3. Detection of OH359.971-0.119 at -10.5 . Contour levels are -2, 2, 4 and 6 times the RMS noise level

Fig. 4. Detection of OH359.810-0.070 at -43.1 . Contour levels are -2, 2, 4 and 6 times the RMS noise level. Note that some of the flux is scattered into two more components The main component is also detected in an adjacent channel

Table 3. VLBA maser positions referenced to Sgr A*

The detection of OH359.971-0.119 in Fig. 3, was obtained with a phase connection to Sgr A* over 4:06. The phase connection for OH359.810-0.070, at a distance of 8:02, is not as convincing as for OH359.971-0.119. Only from a detection in an adjacent channel we have a consistent position for the peak identified in Fig. 4 and quoted in Table 3. Phase referencing over 8:02 should produce similar errors to that over 4:06, but better understanding of this distance dependence is needed.

3.2. Non-Detections

In the first experiment (VLBA test of May 95, without the VLA), no maser sources were detected. In the subsequent experiments, only 2 detections were obtained from 11 attempts on 10 sources. However, 7 out of 8 sources were simultaneously detected in the correlated output of the phased VLA. The VLA detections and VLA correlator setup can be found in Table 4, where we have assumed a rather low flux density of 6 Jy for J1733-130 to estimate a lower limit on the maser source flux densities. Most maser sources therefore have a flux density sufficient to detect them in the VLBI measurements as well. We conclude that the variability of the SiO maser does not explain the low VLBI detection rate.

Table 4. "Phased" VLA maser observations and detections

Positional errors from the OH maser surveys can be as large as one arcsecond; a possible reason for not detecting a maser. We mainly used OH maser positions from Lindqvist et al. (1992a), which are consistent at the level with our own SiO maser positions (Sjouwerman et al. 1998). However, Lindqvist et al. (1992a) do not give a position for Sgr A*, leaving the possibility that the positional error with respect to Sgr A* is larger than . The source would then fall outside our field of view.

However, we believe that the detection rate is mostly limited by difficulties in making a phase connection. This is partly due to atmospheric instabilities. In combination with low signal to noise on Sgr A*, it is likely that not all phase slopes have been removed in our 20 second fringe-fit interval. Recall that we observed with cycle times (40 seconds) longer than the coherence time (about 20 seconds); shorter cycles are currently impossible because Sgr A* is difficult to detect already. We conclude that it is important to observe in the best available conditions, when the atmospheric water vapour contribution is low, and possibly with larger continuum bandwidth.

3.3. Positional accuracy

It is important to make estimates of the accuracy of the measured positions. The first component in the positional uncertainty is simply the noise in the observations of the target sources. Although this determination is also affected by residual effects in the phase connection, we simply take the formal error of the fit in the map for this. This will yield a conservative estimate.

One also has to account for the uncertainty in the phase-reference calibration scheme. Ideally the residual visibility phase of the calibrator source on all baselines should be zero and without any phase slopes. Phase deviation from zero will result in a distortion of the image of the target, possibly scattering the flux of the target over multiple images. However, Fig. 2 (bottom frame) shows that the average on Sgr A* is indeed zero, albeit with a large scatter. It suffices to estimate the RMS phase error , due to noise on the reference and estimate the uncertainty on the derived position. The estimates for these errors in position can be found in Table 3.

An additional problem can originate from the VLA data. The VLA may well be a dominant factor in our results, because of its sensitivity and the short baselines it provides. Therefore it is important to check our assumption that the atmosphere over the VLA sub-arrays is identical. The maximum magnitude of the phase error due to phase instabilities on a VLA baseline can be estimated from the VLA data.

At the start of the phased VLA observations, we observed J1733-130 for 8 minutes in "auto-phasing" mode and more than 5 minutes in "extended-phasing" mode (Fig. 5). In "auto-phasing" mode, the antenna phases are contineously monitored for deviations from the source model and the derived corrections are fed back into the system to form the optimal phased-array response. This method can only be applied if the source is strong and compact enough with a well determined position. For the maser sources we had to rely on extrapolating the corrections from a calibrator ("extended-phasing" mode). Fig. 5 shows the phases are generally zero for the "auto-phasing" mode data on J1733-130. However, during the "extended-phasing" mode (after t = 24 minutes in Fig. 5), the phase starts to drift due to uncorrected changes in the atmosphere. Nevertheless, the deviations stay well within a range of 50 degrees, sufficient to coherently average the VLA antennas.

Fig. 5. Residual phase for all baselines in VLA-Ref in December 1995: first the auto-phasing mode, then from t = 24 minutes the extended-phasing mode. Note that the offsets are not entirely random for short time intervals ( 1 minute)

Fig. 6 shows the measured phase error for the J1733-130 Dec 95 observations. The maximum length T before the effect will be calibrated was 200 seconds. We see that the "auto-phasing" scans have a phase error of about . Next, the average phase error in sub-array VLA-Mas is about 50% larger than for VLA-Ref in the "extended-phasing" mode. The larger extent of the VLA-Mas sub-array produces larger phase errors, because the phases and necessary corrections for the outer antennas change fastest. We take this RMS phase error as an estimate of the maximum , total VLA phase error and estimate the positional error for the case that the position would only be derived from VLA baselines in Table 3.

Fig. 6. Average phase error over the VLA baselines in December 1995 over different time intervals T on the calibrator source J1733-130. The star symbols (lower line) are averages over all the "auto-phasing" scans in VLA-Ref. The filled circles are determined from the "extended-phasing" scans for VLA-Ref; the open circles are determined from the 'extended-phasing' scans in the larger sub-array VLA-Mas. VLA-Mas, which contains the maser source observations, is the relevant sub-array. The averages for the 'auto-phasing' scans for VLA-Mas lie just below the 'extended-phasing' VLA-Ref scans (filled circles) and are therefore omitted for clarity

The positions given in Table 3 are positions of the source relative to the adopted position of Sgr A* (Table 1). Any error in the absolute position for Sgr A*, will affect the positions for the masers in Table 3. For example proper motion of Sgr A* has been inferred and is consistent with the motion of the Sun in the Galaxy (Backer & Sramek, 1982). However, for measuring proper motions with respect to Sgr A*, our primary goal, precise absolute positions are not important.

Another source of systematic error can be the processing model. We have relied on the accuracy of the VLBA correlator model to register positional information. The experiment described here offers no independent means to check this, but Reid & Menten (pers. comm.) find in a project at similar frequencies, a systematic effect of 0.3 mas introduced by inaccuracies in the troposphere model. This positional uncertainty is added to our list of errors.

Finally an intrinsic error in the stellar position remains because the SiO maser surrounds the stellar atmosphere at roughly 5 AU. This effect can be estimated to be on the order of 1 mas or less for 43 GHz, still much larger than the errors calculated so far.

© European Southern Observatory (ESO) 1998

Online publication: October 22, 1998
helpdesk.link@springer.de