Figure 1 (click here) shows approximate contours of the best sensitivity achieved in the survey. We use it to make an empirical estimate of the (in)completeness of the survey. The number of sources detected, as function of highest peak flux density, is corrected by dividing each source by the fraction of the surveyed sky area in which it could have been found. Because the VLA and ATCA data sets have comparable spectral resolutions, we neglect the influence of channel width on the detection probability. The fraction as function of flux density can be found graphically in Fig. 6 (click here). It also shows that all sources with a peak flux density brighter than 530 mJy must have been detected (fraction = 100%, the empirical completeness limit of the survey; or more formally 390 mJy for 99%). If one would confine the area surveyed to the size of the fields used in Lindqvist et al. (1992a, 1997), i.e. 32 squared, the completeness limit would be approximately 65 mJy. In Fig. 7 (click here) we display the number of sources with respect to the highest peak flux density, where we have given preference to the multi-epoch VLA over the one-epoch ATCA data. Because of the variable nature of the OH/IR stars, the distribution is somewhat broadened. By averaging the flux densities in the concatenated monitor data set, the effect should be smaller than in a single epoch observation. However, the exact amount of the broadening is very difficult to calculate, in particular with the mixed observations in this survey.
Figure 6: Sky sensitivity coverage. Shown is the fraction of the surveyed sky area for which we could have detected a source of given flux density. All sources with a peak flux density over 530 mJy must have been detected; for a homogeneous distribution one reads a detection rate of 80% of the 100 mJy sources and only about 20% of the sources around 30 mJy
Figure 7: Number distribution of peak flux densities. The dashed line gives the distribution empirically corrected for incompleteness; it is however an overestimate (see text). For example, the dotted line shows the effect on the completeness correction, when the stars are spatially distributed according to an arbitrarily scaled surface distribution
Strictly, this completeness correction is for a homogeneous number density distribution. As the OH/IR star distribution is concentrated toward Sgr A*, and thus toward the survey center, we expect (and find) the largest number of faint stars in the part of the survey that is most sensitive. Correcting parts of the survey that are less sensitive for these faint stars, just by a linear function based on the sensitivity or geometry of the area, would then assume that pointing the telescopes far away from Sgr A*, the detection probability of an OH/IR star is equal to the detection probability when pointing at Sgr A*. This would imply an overestimation of the completeness correction. We have tried to show this effect in Fig. 7 (click here), by also calculating a completeness correction for an assumed number density proportional to r-2, corresponding to a surface distribution (Lindqvist et al. 1992b). It is evident that the magnitude of the effect depends on the actual concentration of stars; it cannot be extracted directly from our data.
Figure 8: ATCA "maxmap" of the Sgr A complex. Contours are at 0.030, 0.050, 0.070, 0.100, 0.200, 0.500, 1, 2 and 5 Jy. Sgr A* is located in the middle of the right side of the image. The ring-like shape is due to Sgr A East and is not to be confused with the circumnuclear disk
Because of confusion at the location of the Sgr A complex, the survey is in practice less sensitive for detections than the actual instrumental response. As an example of the residual line emission, we show the central part of the "maxmap" of the ATCA survey in Fig. 8 (click here). Note the way the known OH/IR stars stand out with respect to the residuals. Recall that in this region a large continuum emission complex has already been subtracted. Most of the remaining line emission is located as separate spots in a ring-like shape at the inside of the supernova remnant that forms the eastern part of the Sgr A complex (e.g. Ekers et al. 1983; Pedlar et al. 1989). The few absorption measurements done in our data set, have not been explicit about the line-of-sight location of the spots with respect to the absorption and emission lines of the supernova remnant. The ring-like shape cannot be seen in individual or several consecutive frequency channels. The majority of the emission features have a velocity between 25 and , placing them at, or close to either the "" molecular cloud complex behind, or the "streamer" molecular cloud complex G-0.02-0.07 in front of Sgr A* (Zylka et al. 1990). This emission may be indicative for shock fronts, where, on the near and far side of the expanding supernova shell, the shell and the molecular cloud complexes collide. This interpretation would support the recent results of the observation of shock-excited 1720 MHz OH masers by Yusef-Zadeh et al. (1996). In this region, the baselines against which we try to find double peaked OH sources are mostly irregular and make searching difficult. However, because the velocity characteristic is obvious, it is relatively easy to recognise these sources; listing all of them as single - sometimes double - peak detections, on the other hand, has not been the purpose of this survey.
Besides OH emission, the presence of OH molecules also give rise to areas of 1612 MHz absorption, which are not visible in Fig. 8 (click here) because of the "maxmap" procedure followed. We have similar "minmaps" with the detection of 1612 MHz OH absorption. Apart from the supernova remnant and other distinct regions, it is seen most pronounced in the core of the G-0.13-0.08 molecular cloud. The absorption at coincides with the densest NH3 concentration in the GC, that is known to host an ultra-compact HII region, and is very close to two H2O masers (e.g. Güsten & Downes 1983; Okumura et al. 1989). With a projected distance of only 10 parsec from Sgr A*, it is a perfect candidate cloud to investigate present-day star formation in the GC.
Current catalogs on OH/IR stars in the GC can be found in the appendix. When reference data is taken from LWHM, Te Lintel Hekkert et al. (1989; hereafter TLH) and Van Langevelde et al. (1992a), we (re)confirm 87 of the 89 sources that should have been visible in the ATCA data set. In the VLA data set, 67 of the 71 sources are found. An additional 3 sources, that are not (but should have been) seen in the VLA data, are confirmed in our ATCA image cube. We also confirm 4 out of 5 single peak detections of LWHM; 3 of which turn out to have a definite double peaked nature. Unconfirmed from previous surveys in the OH 1612 MHz maser line remain 4 sources: OH359.669-0.019, OH0.204+0.056 plus the single peak OH0.147+0.062 from LWHM, and the double peaked source OH359.897-0.065 from TLH (see the Appendix for two more non-confirmed TLH sources outside this survey).
Summarising the new detections, we find 65 previously unknown double peaked and 3 single peaked OH 1612 MHz masering sources in the VLA and ATCA surveys. We count a total of 52 previously unknown OH/IR stars. Based on a more careful examination of spatial extension and the spectral shapes, we suspect 13 detections to be molecular clouds resembling OH/IR star spectra. We confirm that all three previously known high-velocity OH/IR stars, as well as one newly found high-velocity source, are blue shifted with an absolute line-of-sight velocity exceeding . Because of being detectable in only one survey, we were not able to confirm 15 VLA and 11 ATCA suspected OH/IR stars ourselves. Some of these sources have, however, been detected in the infrared.
OH359.669-0.019: LWHM report a highest flux density of 0.09 Jy, which is, adjusted for the primary beam attenuation, just above our detection limit. We may have been unlucky to observe with the ATCA close to the minimum of its radiation. The source lies outside our VLA survey.
OH0.147+0.062: As opposed to the other single peaks found by LWHM, for which we found a secondary peak in both surveys, this single peak has not been detected here. No counterparts were found in the literature in any other maser line or in the infrared.
OH0.204+0.056: This source lies close to the edge of the survey. LWHM report a highest peak flux density of 0.11 Jy, which makes it only half of our detection threshold. The source is not covered by our ATCA survey.
Generally, the "overlapping" sources - i.e. sources that are in the spatial and velocity domain of both the VLA and ATCA data sets - are detected in each data set. The main cause of "overlapping" sources being detected in only the VLA data is overall sensitivity (i.e. mean flux density over several epochs); the ATCA observations were apparently done when the star was close to its minimum, and the OH maser flux density therefore below the detection limit. Backwards, sources detected only in the ATCA data are either outside the VLA cube, or the ATCA observations were apparently done when the star was close to its maximum, and the OH maser flux density therefore was larger, and above the detection limit, in contrast to the mean flux density. Following we make individual remarks on a small selection of the sources, where we use the abbreviation RR for sources in, or the paper by Rieke & Rieke (1988).
OH359.791-0.081: Single peak, but most likely of stellar origin because it coincides with a long period K-band variable (I.S. Glass, pers. comm.).
OH359.797-0.025: Either a low expansion velocity source, or the red shifted peak of a double peaked source. Not seen in the ATCA data, but the flux and shape justify its entry. Note that although we give a value for the blue shifted velocity, we cannot distinguish between a second peak and a possible effect of the end of the frequency band.
OH359.804+0.201: Listed as single peak in LWHM.
OH359.837+0.052: Listed as single peak in LWHM.
OH359.864+0.056: The only new source for which the absolute stellar velocity exceeds .
OH359.906-0.036: Coincides with, and matches the velocity of source #49 in RR. As for all other cases where we find an infrared counterpart for our OH maser sources, we claim the OH and infrared emission to originate from one (stellar) source.
OH359.936-0.145: Found as single peak by LWHM. Most probably double with the second peak at , but we cannot confirm this; the source is not detected in our ATCA data.
OH359.939-0.034: RR report a source (#32) with a velocity of . Our velocity measurement of is within three times their RMS of . Taking into account the spatial extension of the emission, we are however tempted to attribute the OH emission to molecular clouds on the line-of-sight. If one supposes the velocity of RR is very accurate, one can argue that source #32 can be seen in the OH spectrum with two peaks of mJy, separated by about .
OH359.943-0.055: Matches the velocity of the nearby source GCIRS 19 (Sellgren et al. 1987; source #22 in RR), but we doubt the identification as such. Besides, GCIRS 19 is classified as probably being a supergiant (most recently by Blum et al. 1996), for which one would expect a shell expansion velocity larger than the 6.5 km/s found here.
OH359.947-0.046: Although our absolute positions are not expected to be very accurate (see Sect. 3.1), our VLA position for this source agrees within one arcsecond of the position of GCIRS 5. The velocity measured matches the one given in Krabbe et al. (1991) and Haller et al. (1996). We propose this source to be the OH counterpart for GCIRS 5. Note that the OH maser reveals a low shell expansion velocity, hinting toward a low metallicity, and low mass AGB star.
OH359.954-0.041: The clearest example of interesting spectral structure seen in some (of the stronger) OH/IR stars in the GC. We discuss these peculiar spectra further in Sect. 4.5.
OH359.956-0.050: Found as an H2O maser by Levine et al. (1995) and Yusef-Zadeh & Mehringer (1995). Its nature - a young, massive supergiant or an evolved, intermediate mass AGB star - has been extensively discussed in Sjouwerman & Van Langevelde (1996). The AGB nature of this object, identified with GCIRS24 as its infrared counterpart, has been supported recently by Blum et al. (1996), and was initially motivated by Sellgren et al. (1987).
OH359.957-0.123: A single peak detection, but probably stellar: it has been identified with a long period K-band variable (I.S. Glass, pers. comm.).
OH359.965-0.043: The SIMBAD data base reports this position to be close to the position of IRC-30321. However, the error in the position of IRC-30321 is large (1) and we therefore find it more likely that IRC-30321 coincides with one of the other nearby luminous (previously known) OH/IR stars instead.
OH359.970-0.049: On top of the extended emission we find a double peaked point source. It has most probably been detected by LWHM as their source #69. LWHM apparently mistook a peak of (the molecular cloud) OH359.970-0.047 as their secondary peak velocity, resulting in a relatively high value for the shell expansion velocity. More likely than OH359.970-0.047, this source seems to be related to source #38 in RR. They give a velocity of with an RMS of where we measured .
OH359.971+0.068: Listed as single peak in LWHM.
OH359.980-0.077: Found as an H2O maser by Yusef-Zadeh & Mehringer (1995), but in fact an OH/IR star; see discussion in Sjouwerman & Van Langevelde (1996).
OH359.985-0.042: RR found a source (#51) at , three times their RMS away from our velocity. However, we are confident that the identification of the OH source with source #51 in RR fits.
OH359.990+0.030: This source lies close to IRAS 17423-2855. Te Lintel Hekkert & Chapman (1996) searched IRAS 17423-2855 for OH emission, but did not find it because of a noise level of 440 mJy. The position of OH359.990+0.030 is however more than 20 off from the positions given for several possible near-infrared counterparts for IRAS 17423-2855 in Monetti et al. (1992). Although IRAS 17423-2855 can also be interpreted as an ultra-compact HII region, the association with double peaked OH emission is more convincing for a (post-)AGB object (e.g. Volk & Cohen 1989; Monetti et al. 1994; Te Lintel Hekkert & Chapman 1996).
OH0.001+0.353: A source named OH359.984+0.349 in TLH (#155), observed by Habing et al. (1983) with the 100 m single dish telescope in Effelsberg (Bonn). We confirm the position measured by LWHM; it differs about one arcminute from the position quoted in TLH.
OH0.005+0.360: Because of its spectral shape and because it is not spatially resolved, we suspect this source to be the red shifted peak of a double peaked OH/IR source.
OH0.014-0.046: Close to, and maybe related to the infrared source GCS 6 in Kobayashi et al. (1983).
OH0.053-0.063 and OH0.060-0.018: The former was marked by LWHM as detected by Winnberg et al. (1985); it should have been the latter.
OH0.064-0.308: Located near a compact HII region (#16 in Downes et al. 1979).
The OH/IR stars found for the first time in this survey have less luminous OH masers in their circumstellar shells compared to the previously known OH/IR stars in the GC. In this section we argue that the newly found stars are of similar nature to the ones previously known. This result was used in Sjouwerman & Van Langevelde (1996).
Figures 2 (click here) and 3 (click here) show the spatial and kinematic distribution of the known and previously unknown OH/IR stars detected in this survey. It is clear in Fig. 2 (click here), that the more luminous OH masers, generally the known OH/IR stars, can also be detected further out from the survey center. The fact that there are more stars at positive latitude offsets is an effect of the asymmetry in the survey pointing. Where the survey is sensitive enough, one sees that there is no preferred, and no distinct location for each of the samples. However, we make the observation that the alleged void of known OH/IR stars at small positive longitudes and small positive latitudes, does not comply with the combined sample. Furthermore, a comparison of the location of stars in Fig. 2 (click here) and Fig. 3 (click here) between both samples also suggests a similar distribution in phase-space.
Figure 9: Expansion velocity distribution. The solid line is the distribution of previously known OH/IR stars in our survey, new detections are distributed according to the dotted line. The striking resemblance of both distributions is in large contrast to the expansion velocity distributions of OH/IR stars with a different metallicity, for example in the outer Galaxy and Galactic plane (dashed: Blommaert et al. 1993 plus Blommaert et al. 1994, multiplied by three for display purposes)
Figure 10: OH luminosity distribution. The dashed curve combines the previously known OH/IR stars in our survey and the new detections. The known OH/IR star luminosity distribution is depicted by the dotted line. The solid line outlines the maximum extent of the distribution, i.e.\ corrected for incompleteness according to a homogeneous spatial distribution
To argue further that the samples consist of the same type of stars, we show the distributions of the shell expansion velocity in Fig. 9 (click here). Both samples have shell expansion velocities sharply peaked around . Expansion velocity distributions for many other samples of OH/IR stars can be found in the literature, e.g. Eder et al.\ (1988), Te Lintel Hekkert et al. (1991) and Wood et al. (1992). We want to restrict ourselves by comparing the expansion velocities of the OH/IR star in the GC with samples of the Galactic plane (Blommaert et al. 1993, 1994) and the Galactic bulge (Sevenster et al. 1997).
The distribution of the OH/IR stars found in the Galactic plane is also shown in Fig. 9 (click here). The distribution for the Galactic bulge has its peak around , and is broader than for the GC (see Sevenster et al. 1997). When comparing the expansion velocities of the OH/IR stars in the GC with the ones in the Galactic plane and bulge, we reach two conclusions. First, the expansion velocity distribution in the GC is different from the expansion velocity distribution of any other sample known, which we attribute to generally higher metallicities in the GC (see the discussions in e.g. Wood et al. 1992; Blommaert et al. 1994 and Habing 1996). Second, we note the striking resemblance of the expansion velocity distribution of the known and the previously unknown OH/IR stars in this survey, and argue for generally identical metallicities for both of our samples. Hence, as the shell expansion velocity is a function of metallicity and stellar luminosity, we conclude that the stellar luminosity distribution for both of our samples is identical; as far as we can tell, the central stars are the same.
Because both the known and previously unknown OH/IR stars are intrinsically identical, we can investigate the combined OH luminosity distribution. Figure 10 (click here) shows the result when we assume that all stars are located at a distance of 8 kpc. That the new detections are mainly the low OH luminosity sources can be seen readily from the difference in the total distribution and the known OH/IR star distribution. Also shown is the maximum extent of the distribution when it is corrected for (in)completeness. We calculated this correction by weighing each source by the inverse of its detection probability and assuming a homogeneous number distribution.
Again, this correction is an overestimate. Unfortunately the number of sources with respect to the corrections at the left-hand side of the luminosity distribution is small. In general these are the sources with a low peak flux density for which the survey is not complete. It prevents us to derive a firm conclusion about a possible low luminosity cut-off in the OH maser distribution. We can however state, that the distribution peaks at photons per second within the statistical errors.
In Sect. 4.3 we mentioned non-standard spectral structure seen in OH359.954-0.041 (a category 1 OH variable in vLJGHW). Less clear examples, both variable (cat. 1, 2) and non-, or irregularly variable (cat. 3) sources, are OH359.675+0.070 (cat. 2), OH359.879-0.087 (cat. 1), OH359.938-0.052 (cat. 2), OH0.040-0.056 (cat. 3), OH0.076+0.146 (cat. 1), OH0.083+0.063 (not monitored), OH0.142+0.026 (cat. 1) and OH0.319-0.041 (cat. 3). The masers in OH359.954-0.041 and OH0.040-0.056 are quadruple peaked, the rest are triple peaked. In addition to the "a" spectra in Fig. 5 (click here), Fig. 11 (click here) shows the lower frequency resolution spectra measured in image cube "b" for four of these stars. We have only seen this type of spectra in the concatenated data; spectra with regular additional emission features at roughly either 15 or on the redshifted side of the redshifted peak, or about "outside" both main peaks. Instrumental effects (e.g. the Gibbs phenomenon) and side-lobe features of nearby sources can be excluded.
Figure 11: Spectra from image cube "b" for selected sources. See Sect.\ 4.5
In the case of OH359.954-0.041 we measured a consistent positional offset of 1.34 arcsecond (10000 AU, or 10 times the shell radius that was measured from the phase-lag of the main peaks by vLJGHW) to the southwest, from the main peaks to both outer peaks. The feature "inside" the main peaks is positionally coincident with the main peaks. The outer peaks of OH0.040-0.056 are displaced about 0.6 northeast from the main peaks, again within the errors at mutual excluding positions. Because of the symmetry seen in OH359.954-0.041, and in OH0.040-0.056, and the rather remarkable velocity interval structure in all examples, we tend to conclude that the emission is from (the shell of) the OH/IR star itself; not from another source (an OH/IR binary, or more exoticly, an OH/IR star captured by an OH masering supergiant) seen at the same projected coordinates. Suggestions then range from a double shell, indicating different epochs of interrupted mass-loss as seen for carbon rich AGB stars (Olofsson et al. 1996), creation of other molecules besides CO already early in the mass-loss history, an effect related or similar to the mode switching seen in the H2O maser lines of OH39.7+1.5 (Engels et al. 1997), to bipolar outflow. These stars clearly make excellent test-cases for our understanding of the mass-loss mechanism of evolved stars.
Although the both uniformly sampled VLA and ATCA data sets have a comparable sensitivity, the spectral features are only clearly seen in the concatenated VLA data set. Therefore the features result from either weak emission from many epochs, or strong emission at a limited amount of individual epochs. Making a case for the latter, other examples of similar, one epoch spectra in the literature can be found, however undiscussed, in Te Lintel Hekkert et al. (1991; IRAS15452-5459 and IRAS17253-2824), and in Sevenster et al. (1997; e.g.\ OH353.421-0.894, OH2.186-1.660 and OH5.991+0.252). Furthermore, Eder et al. (1988) mention an identical phenomenon in IRAS18520+0533 (OH38.3+1.9), and suggest it to be mapped. However, OH38.3+1.9 was found unresolved at 1612 MHz with the VLA in CD-array by Lewis et al. (1990), and observations with the European VLBI Network have not been published yet (M. Lindqvist, pers. comm.). Note however, that this OH emission of OH359.954-0.041 and OH0.040-0.056 cannot be detected with VLBI observations, because of severe scattering of the source at decimeter wavelengths (0.6 at 1612 MHz for OH359.954-0.041; Van Langevelde et al. 1992b).