Astron. Astrophys. 358, 910-922 (2000)

4. NGC 6397

NGC 6397 is a nearby cluster, with a collapsed core, in or close to which Cool et al. (1993) detected four X-ray sources (B, C1-3) with a ROSAT HRI observation. Photometry with the Hubble Space Telescope enabled Cool et al. (1995b) to find eight candidate counterparts for these sources, on the basis of high ultraviolet flux or of H emission. The H emission of three stars has been confirmed spectroscopically by Grindlay et al. (1995) who argue that these stars are cataclysmic variables, and responsible for the X-ray emission close to the core.

4.1. Source list and membership

We analyse first the longest observation, obtained in 1995, and use this as a reference for our discussion of the earlier, shorter observations. The standard analysis provides 14 sources, listed in Table 3. Identifications with earlier X-ray sources or optical objects are indicated; 7 sources are new. X 6 has been identified by Cool et al. (1993) as SAO 244944. This star is identical to HD 160177, and is in the Hipparcos Catalogue as HIP 86569. Its position and proper motion are thus very accurately known, and we use it to determine the bore sight correction. This bore sight correction is given in Table 1, and is applied to the X-ray positions; the resulting positions are given in Table 3. The statistical uncertainty in the X-ray position of X 6 is about 0.5"; we therefore estimate that systematic error of the X-ray positions listed in Table 3 is better than 1"; this error should be added in quadrature to the statistical error for each individual source position. The quasar identified by Cool et al. (1993) with X 5 coincides within the error with our position for X 5. However, the active galaxy identified by Cool et al. (1993) with X 2 is 10" from our X-ray position, mainly in right ascension; and we conclude that it is not the X-ray source. The explanation probably lies in the new scale for the size of the HRI pixel that we use (see Sect. 2), which modifies positions of sources at large distance from the center of the HRI image.

Table 3. X-ray sources detected in the NGC 6397 (, kpc, Djorgovski 1993) with the ROSAT HRI, for the standard analysis of the whole field, and separately for two multi-source analyses of the central area. Numbers up to 10 are sources from Johnston et al. (1994), higher numbers are new; cross-identifications with sources discussed by Cool et al. (1993) are listed on the right. All X-ray positions have been corrected for boresight. The positions of the center of the cluster (GC, Djorgovski & Meylan 1993), its core radius and half-mass radius (Trager et al. 1993) and the positions of some optical objects discussed in the text are also listed; epochs are 1992.7 for positions by Cool et al., and 1996.3 for HIP 86569.

The flux detection limit is about outside the blended central region, similar to that obtained for Cen. Analogous to our argument for Cen, we find that all objects detected within are probably cluster members, whereas we expect 1.4 background sources within from the center of of NGC 6397; the sources at therefore may be background sources. We thus cannot decide whether X 12 is a cluster member. Outside the half-mass radius, the sources are more likely to be background or foreground sources. X 5, just outside the half-mass radius, is a quasar (Cool et al. 1993).

4.2. The central sources

In Fig. 4 we show the X-ray contours of the center of NGC 6397 together with the ultraviolet and/or H -emission stars discovered by Cool et al. (1995b). The first models we investigated as fits to the central area of , containing sources X 13/B and X 4/C, are those with successively one, two, three, four and five sources; all with free positions. Using the criterion for significance (see Sect. 2) we find that five sources are required. We refer to the fit with five sources as Model I. The parameters of the five sources of this model are given in Table 3. We do not detect source A of Cool et al. (1993) in the 1995 observation.

Fig. 4. Positions of ultraviolet or H -emission objects in the central area of NGC 6397 (, +, numbered with their ID in Table 2 of Cool et al. 1995b) superposed on X-ray contours of sources X 13/B and X 4/C as observed with the ROSAT HRI in 1995. The candidate counterparts for three X-ray sources suggested by Cool et al. are marked . The X-ray image was smoothed with a 2-d 2" Gaussian. The lower and left axes give pixel numbers for the ROSAT HRI detector, the upper and right axes right ascension and declination with respect to the cluster center. The conversion between pixel and celestial coordinates is accurate to within 1".

Cool et al. (1995b) resolved source X 4/C into three components C1-3. From a list of H emission and/or ultraviolet excess objects (their Table 2), they suggest identifications of ID 1 with C2, ID 2 for C3 and ID 3 for C1. Comparing the positions of the sources in Model I we find that the positions of X 4b and X 4c are compatible with those of ID 3 and ID 1, respectively; we thus identify X 4b with C1 and X 4c with C2. X 4d is a new source. (The offset required to match these positions from Table 3 with those given by Cool et al. (1995b) is slightly larger than our claimed accuracy of ; the remaining difference may be explained by an offset between the Guide Star Catalogue coordinate system and the more accurate Hipparcos coordinate system.) The position of X 4a is not compatible with that of ID 2. The reason for this may be seen in Fig. 4: the two brightest components of source X 4/C have a smaller difference in right ascension than ID 2 and ID 3. If ID 2 is the correct identification for C3, we conclude that X 4a is not identical to C3.

To further investigate this, we note that if the identifications are correct, the distances between the X-ray sources must match the distances between the proposed optical counterparts, which are accurately known from the HST observations. In Model II we fit five sources to the X-ray data of the center of NGC 6397, of which three are forced to be at fixed relative positions, corresponding to the distances between ID 1, ID 2 and ID 3. Model II thus has four fitted parameters less than Model I. The of Model II is 26 higher than that of Model I, i.e. it is a significantly (4-sigma) worse fit. This confirms that X 4a is not ID 2. In Model III we assume that ID 6 of Table 2 in Cool et al. (1995b) rather than ID 2 is the counterpart of C3, and fix the distances between the sources accordingly. This fit has the same as Model II, and thus also is significantly worse than the fit of Model I. Again, the reason for the bad fit is the mismatch in the difference in right ascensions of the two brightest X-ray sources with that between the proposed counterparts: X 4a is not ID 6.

We note that the best position of X 4a is between ID 2 and ID 6, and in Model IV we fit six sources, of which four are forced to be at the relative distances of ID 1-3 and ID 6. Model IV thus has three fitted parameters less than Model I. Its is 6 higher than that of Model I, i.e. it is marginally worse at less than 2-sigma. The parameters of the six sources of this model are also given in Table 3. It is seen that the positions of X 4b, X 4c and X 4d are the same (within the error) in Model IV as in Model I.

Thus, we have two acceptable models. In both models we confirm the possible identifications of ID 3 with C1 (X 4b) and of ID 1 with C2 (X 4c), and we find one new source (X 4d). In Model I the remaining flux is ascribed to one source (X 4a) which is not identical to ID 2. In Model IV the remaining flux is ascribed to two sources, one of which is ID 2/C3 and one is a second new source, X 4e/ID 6. The two acceptable models are illustrated in Fig. 5.

Fig. 5. X-ray contours in the central area of NGC 6397 as observed with the ROSAT HRI in 1995, with the positions of the sources obtained in the best fits. positions found with the best fit for four components in C in which all positions are left free (Model I), positions found with the best fit for five components in C, in which three sources have distances fixed with respect to one other source, at the distances of ID 1, 3, and 6 to ID 2 (Model IV). Other symbols as in Fig. 4. Model I is marginally better at the 2-sigma level.

4.3. The earlier observations

The standard analysis detects X 2, X 5, X 16, X 6 and X 8 in both the 1991 and the 1992 data of NGC 6397, and X 19 in the 1992 data, all at countrates compatible with those of 1995. It also detects sources X 13/B and X 4/C in the 1991 data and in the 1992 data, labelling both as extended. The number of photons in sources B and C is rather small in these short observations. To limit the number of parameters in the fits to the central sources we demand that the distance between the fitted central sources in each model is the same as in the best fit to the 1995 data, but allow the fluxes to be different. The corresponding reductions in the number of fitted parameters for each model are indicated in Table 4.

Table 4. Results of fitting four models to the three data sets of NGC 6397. The table lists the number n of fitted parameters, and the difference with respect to the best model for a given data set. In all fits, the fluxes of all sources are fitted parameters. For the 1995 observation Model I has four sources with free positions, Model IV has six sources of which three have free positions and three have fixed positions relative to ID 3, corresponding to the offsets of ID 1, ID 2, and ID 6 with respect to ID 3. Models II and III are as Model IV, after removing ID 6 and ID 2, respectively. For each Model, the same positions as in the best fit for 1995 are used for the 1991 and 1992 data.

We thus fit four models to each data set. For each year, the best model is set at , and the quality of the other models for that year is determined with respect to this model. The results of our fitting are shown in Table 4. For the 1991 data, the models with five sources are comparable in quality, and the six-source model is not significantly better. For the 1992 data, Model III is marginally better (2 sigma) than Model I and significantly (3 sigma) better than Model II, whereas Model IV is of similar quality.

The fits to the earlier data confirm the conclusions that we draw on the basis of the observation of the long observation of 1995. Model I in which source X 4/C is separated into four components at free positions is acceptable for all three observations. Model II in which X 4/C is separated into four components at fixed relative distances of ID 1-3, is not acceptable for the 1992 data. Model IV in which X 4/C is separated into five components, four of which correspond to ID 1, ID 2, ID 3 and ID 6, also is acceptable for all observations. ID 2 is not required in 1992, and ID 6 is not required in 1991. The latter fact explains why ID 6 is not present in the analysis by Cool et al. (1993) of the 1991 data. These conclusions are confirmed by the countrates that Model IV ascribes to the different sources, listed in Table 5

Table 5. Countrates (counts ksec^-1) assigned to the central sources in Models I and IV in the fits to the observations of 1991 and 1992. Numbers in parentheses indicate the errors in the last digit. For 1995 see Table 3.

To see whether we can confirm the existence of source A of Cool et al. (1993) we have also added the 1991 and 1992 observation (after shifting the 1992 observation by in ; compatible with the shift as determined by Cool et al.). We fit Model I to the added image, and compare it with the fit in which a source is added to model A. We find , which implies that source A is marginally significant at . The position ( with respect to source B) and countrate (0.4 counts/ks) that we find for source A are compatible with those given by Cool et al. (1993).

4.4. Sources not related to the cluster

HIP 86569 is a K1 IV/V star with , , and a parallax of . Hipparcos discovered that this star is a close binary (separation ) of stars with Hipparcos magnitudes and , respectively. At a distance of 60 pc the observed ROSAT HRI countrate converts to an X-ray luminosity in the 0.5-2.5 keV band of erg s^-1 (for assumed 1.4 keV bremsstrahlung with no absorption). This is similar to the X-ray luminosities of single KV stars detected in the ROSAT All Sky Survey, such as HD 17925 (K1V) which has erg s^-1 (Hünsch et al. 1998).

4.5. Discussion

The core of NGC 6397 contains at least four X-ray sources detected with ROSAT, and possibly five. 1 cts ksec^-1 for a source at the distance and with the absorption column of NGC 6397, for an assumed 0.6 keV bremsstrahlung spectrum corresponds to a luminosity in the 0.5-2.5 keV band of erg s^-1. The faintest source we detect, X 4c, is at this level. The brightest source is X 13/B, at a luminosity of about erg s^-1. These luminosities are at the bright end of the luminosity distribution for cataclysmic variables, such as the large sample investigated with ROSAT (Verbunt et al. 1997), as expected for an X-ray selected sample.

Of these sources, X 13 and X 4b have the same flux level in all three observations. Source X 4c is fainter in 1995. The identifications of X 4b and X 4c with cataclysmic variables ID 3 and ID 1 remains probable, as does the argument by Edmonds et al. (1999) that these systems are DQ Her type systems. The distance between X 4b and X 4c in Model I is marginally less than the distance between the proposed counterparts; it is tempting to speculate that this is due to a small X-ray flux contribution of a fourth cataclysmic variable (`CV 4') identified by Cool et al. (1998) and confirmed by Edmonds et al. (1999).

The new source X 4d has been detected in the 1995 observation because of the longer exposure; it may, but need not, be brighter in the 1995 observation than in 1991 and 1992.

Fig. 6. X-ray contours in the central area of NGC 6752 as observed with the ROSAT HRI in the combined image of the 1992-1996 data. The image was smoothed with a 2-d 3" Gaussian. The detected sources are indicated with their numbers in Table 2. The circle gives the half-mass radius of the cluster. The lower and left axes give pixel numbers for the ROSAT HRI detector, the upper and right axes right ascension and declination with respect to the cluster center.

If the remaining flux is assigned to one source X 4a, then this source is not identified, and was brighter in 1995 than in 1992. If the remaining flux is distributed over two sources X 4a and X 4e, the flux of X 4a may still be constant, and X 4a may be identified with the probably DQ Her type cataclysmic variable ID 2. In this case, the flux of X 4e has increased between 1991 and 1995. ID 6 was reported to vary by 1.1 magnitude in five hours by De Marchi & Paresce (1994), but was constant in a ten hour observation by Cool et al. (1998). It is suggested by Edmonds et al. (1999) that ID 6 is a undermassive helium white dwarf, probably in a binary. If it is a single helium white dwarf, it cannot be a variable X-ray source; if it is in a binary with a recycled radio pulsar, it also is unlikely to be a variable X-ray source; if it is in a binary with another white dwarf, then optical and X-ray variability can be due to variable mass transfer from that other white dwarf. However, it is also possible that not ID 6, but a nearby hitherto unidentified star in NGC 6397 is the X-ray source X 4e.

Whether source X 4a alone, or source X 4a and X 4e are present in the core of NGC 6397, and in the latter case whether X 4e is identical to ID 6 requires a better spatial resolution for the X-ray observations than provided by ROSAT.

© European Southern Observatory (ESO) 2000

Online publication: June 20, 2000
helpdesk.link@springer.de