Astron. Astrophys. 340, 351-370 (1998)

3. Results

3.1. Iso-intensity contour maps

A grey scale plot of the HRI field is shown in Fig. 2. HRI sources found by the detection algorithms (see Sects. 2.2 and 3.2) are marked with their HRI number and a ellipse sketches the D₂₅ (see Table 1).

Fig. 2. Grey scale plot of the inner X-ray image seen with the ROSAT HRI. The image was constructed with a pixel size of 2:005 and smoothed with a Gaussian of (FWHM). The center of NGC 3079 is marked with a cross, the D25 ellipse is indicated, and point sources (likelihood ) are enclosed by boxes and numbered (see Table 3). Right ascension and declination are given for J2000.0

Fig. 3 shows a contour map of the PSPC field centered on NGC 3079. The positions of all individual sources detected are marked, with their PSPC number. Note that P20 is detected only in the hard band, and does not appear in the broad-band map. The ellipse sketches the D₂₅.

Fig. 3. Contour plot of the broad band ROSAT PSPC image of the inner of the NGC 3079 field (cf. Sect. 2.2 for how the image was constructed). Contours are 2, 3, 5, 9, 15, 31, 63, 127, and 255 above the background ( cts s-1 arcmin-2, background cts s-1 arcmin-2). ROSAT PSPC detected sources (likelihood ) are plotted as squares with source numbers written alongside (see Table 4). The position of the nucleus of NGC 3079 is marked as a cross, the optical extent is indicated by the ellipse at D25. P16 detected close to the galaxy center, has not been numbered

A smaller portion of the HRI field corresponding to the area of the extended emission seen with the PSPC is presented in Fig. 4. The HRI spatial resolution resolves the emission from the galaxy's plane into a structure at the center and three individual sources, but is not sensitive enough to show the low surface brightness emission at large galactocentric radii.

Fig. 4. Contoured grey scale plot of the central emission region of NGC 3079 for ROSAT HRI. The image is binned to a 1"/pixel and smoothed with a Gaussian function with FWHM=5". Contours are given in units of 0.5 photons accumulated per 4:007 diameter. Contour levels are 3, 5, 9, 15, 31, 63 units. The center of NGC 3079 is marked by a cross, HRI detected point sources by squares

Fig. 10 (see Sect. 4) shows the inner part of the PSPC field superposed onto the optical image of the galaxy. It is immediately apparent that the emission from NGC 3079 is complex and extends both above and below the galaxy's plane. We have also produced maps of the emission in different energy ranges (Fig. 5), namely in the soft, hard1 and hard2 bands defined above. As can be seen by the comparisons of the iso-intensity contour maps, the softer and harder emission show rather different morphologies. The hard2 band emission is aligned with the optical disk of the galaxy, and seems to be rather confined to it, while both the hard1 and the soft images show extensions above and below the plane. This difference cannot be attributed to the different response of the instrument in the different energy bands and indicates the presence of more than one component to the emission of NGC 3079 (see discussion later).

Fig. 5. Contour plots of the central emission region of NGC 3079 for broad (B), soft (S), hard1 (H1), and hard2 (H2) ROSAT PSPC bands. Broad band contours are given as in Fig. 3. Soft band contours are given in units of ( cts s-1 arcmin-2) above the background ( cts s-1 arcmin-2), hard band contours (due to the negligible background in these bands) in units of 1 photon accumulated per diameter. One unit cts s-1 arcmin-2 for the hard bands. Contour levels are 2, 3, 5, 9, 15, 30 units for all contour plots. A cross indicates the center on NGC 3079, squares indicate the positions of the PSPC detected sources in the field, diamonds of HRI detected sources

3.2. HRI/PSPC sources in the field

The source detection procedure yielded 23 sources in the HRI and 34 in the PSPC above the selected likelihood threshold for source existence. These numbers reduce to 20 sources in the inner HRI field (Fig. 2) and 30 sources in the inner of the PSPC image (Fig. 3).

The X-ray properties of the sources are summarized in Tables 3 and 4: source number (col. 1), ROSAT name (col. 2), right ascension and declination (col. 3, 4), error of the source position (col. 5, including the systematic error for the attitude solution), likelihood of existence (col. 6), net counts and error for the 0.1-2.4 keV ROSAT band (col. 7), count rates and error after applying dead time and vignetting corrections (col. 8). For PSPC sources we also list hardness ratios HR1 and HR2 with their relative errors (col. 9, 10). For PSPC sources, positions and maximum likelihood values have been determined from the energy band with the highest detection likelihood, but the count rates refer to the broad band. To distinguish HRI from PSPC sources, a H or a P has been prefixed to the number for HRI or PSPC, respectively. For sources detected in both instruments, the ROSAT names have been derived from the detection with the smaller error radius. These were mainly HRI detections. Only for two sources (H1 and H5) the HRI position errors are bigger than the corresponding PSPC errors due to source variability or big off-axis angle (see Sect. A.1).

Table 3. X-ray properties of the sources detected with the HRI in a diameter field centered on NGC 3079

Table 4. X-ray properties of sources detected with the PSPC in a field centered on NGC 3079

Only one HRI source (H15, at the center of NGC 3079) and two PSPC sources (P16, at the center of NGC 3079, and P28) were flagged as extended by the maximum likelihood detection algorithm.

Besides for the galaxy's center, only one other PSPC source, P19, and 3 HRI sources, H13, H14, and H16, are positioned within the D₂₅ contours of the galaxy. However it is likely that also sources H12, H18, P18, and P21, are related to NGC 3079, and they will be regarded as such in what follows (see Sect. 3.3). Moreover, it is also possible that some of these sources, located in this complex area where extended emission is also seen, are spurious detections picked up by the detection algorithms as a consequence of a bad background model due to the more diffuse component and represent local enhancements. One HRI source (H6) and the corresponding PSPC source P9 are identified with the companion galaxy MCG 9-17-9 (see Sect. 3.5). Properties of other sources outside the D₂₅ ellipse of NGC 3079 are discussed in the appendix.

3.3. The emission from NGC 3079

Complex emission partially filling the D₂₅ ellipse and extending into the halo along the minor axis is detected (see Figs. 2, 3, 4, and 5). On top of this emission, the central nuclear region and five sources (H12, H13, H14, H16/P19, P21) are resolved. Two additional sources (P18, and H18) positioned outside the D₂₅ diameter are within the H I envelope of NGC 3079, and are also probably associated to the emission of NGC 3079. The properties of the nuclear source (H15/P16) are further investigated in Sect. 3.4. The two sources H13 and H14 detected with the HRI are too close to the bright nucleus ( and , respectively) to be resolved by the PSPC. The HRI source detection algorithm did not separate a source at the northern end of the diffuse central emission (distance ), even though the contour map of Fig. 4 suggests a separate peak. Sources P13 and H10 appear to be within the outermost PSPC contour of Fig. 3. While at the present time it is not possible to exclude the possibility that these are unrelated background sources, the evidence of excess emission in this region suggests that maybe these are the peaks of a more extended emission probably connected with the galaxy or with the group. We therefore will discuss these two sources as both truly individual sources and as a more diffuse component.

Table 5 summarizes PSPC and HRI count rates, X-ray fluxes and luminosities of the sources in NGC 3079. For two sources that were detected only in the PSPC, and for 2 detected in the HRI only, a 2 limit to the HRI (PSPC) count rates at the same positions are calculated. These are estimated from circles of radii of FWHM of the source at the off-axis distance (PSPC and HRI, respectively). The fluxes are then corrected by a factor 2 to compensate for the small aperture used to estimate the net counts. No equivalent limit is given for H13 and H14, since they would not be resolved in the PSPC.

Table 5. X-ray parameters of NGC 3079 sources
Notes:
fluxes in units of 10^-14 erg cm^-2 s^-1 for a 5 keV thermal bremsstrahlung spectrum in the 0.1-2.4 keV band, corrected for Galactic absorption
luminosity in units of 10³⁸ erg s^-1 assuming a distance of 17.3 Mpc. Fluxes and luminosities (for which errors are not quoted) of these sources, in particular for the PSPC data, should be taken with caution, since extended diffuse emission is included in the flux of the individual sources due to the relatively large dimension of the instrument point spread function (see text)

While the luminosity of the nuclear source is above erg s^-1 most of the other sources are close to the detection limit and show luminosities in the range erg s^-1 .

The count rates of the individual sources in NGC 3079 are too low to be used to study time variability within each individual PSPC or HRI observation (day time scale). We can however investigate time variability of the sources on one year time scale by comparing the PSPC and HRI observations that were taken 1 year apart. No source variability can be claimed for the sources in NGC 3079.

The HRI upper limit for source P21 however appears to be significantly lower than the PSPC flux. P21 is detected from the algorithm in the hard2 band only, as can also be seen by the maps in Fig. 5. While it is at the moment unclear whether this should be considered a real source, or rather a local enhancement in the diffuse emission that extends to the NE of the galactic plane, it is clear that a variability study is severely hampered by the presence of this latter component, given the widely different spatial resolutions and sensitivity to low surface brightness components of the HRI and the PSPC. Therefore the much lower HRI flux could be in part (totally) due to the different amounts of diffuse component in the detection cell. In fact, when we estimate the background locally, namely from the average surface brightness of the emission at the same radial distance from the center, the net count rate above the extended emission reduces to almost a half, and the flux erg cm^-2 s^-1 , comparable to the upper limit determined from the HRI data.

3.3.1. Radial distribution of the emission

To determine the extent of the emission and the spatial distribution of the detected photons we have produced radial surface brightness plots from the PSPC and from the HRI data, centered at the X-ray central peak (sources P16 or H15). The plots are shown in Fig. 6. The radial distribution of a point source, at the central position, binned as the data and normalized to them in the innermost bin, is also plotted for comparison. In the PSPC data, the point source is simulated separately in energy bins of 0.1 keV which are then normalized to the count rates in the relevant sub-band and co-added. In the HRI data, it is obtained from the analytical formula given in David et al. (1994) for a source on-axis. In both HRI and PSPC data, the overall photon distribution is inconsistent with a point source (however, see later for further analysis of the HRI data).

Fig. 6. Radial distribution of the detected photons, azimuthally averaged in concentric annuli of 15" width. The dashed profiles indicates the radial distribution of a point source and the background level estimated as explained in the text. Point sources detected have not been removed from these profiles

The azimuthally averaged surface brightness distribution of the emission (Fig. 6) extends out to a radius of r (13.5 kpc) in the PSPC data, outside of which the profile becomes constant with radius and consistent with the background map created from the data (see Sect. 2.3). Similar plots in the soft and hard band indicate maximum radii of comparable values. In the HRI, the profile flattens at a radius r (10 kpc). For the HRI, we can therefore determine the field background from a region outside of the galaxy's emission by choosing an annular region around the galaxy of 4 inner and outer radii, respectively. Point sources that lie in the background regions have not been included for background estimates by masking them out with circles of and radii for PSPC and HRI, respectively. Correction for vignetting is negligible at these off-axis angles.

Given the presence of P13 in the PSPC data, and of the apparent connection between this source and the galaxy (see Fig. 10), we have further analyzed the PSPC data by looking at the radial distribution of photons in different directions and in comparison to the expected pure field background, whose shape is represented by the exposure map. In fact, while the background map is constructed from the data and therefore takes partially into account any diffuse emission present in the field, the exposure map should represent the PSPC response to a flat, constant radiation, while taking into account both exposure and vignetting. When properly normalized to the data, the expected field background can therefore be estimated from it (see also Trinchieri et al. 1994, Kim & Fabbiano 1995). Fig. 7 shows the results of the comparison between the spatial profiles in different directions relative to the exposure map. These have been obtained by calculating surface brightness profiles along the major and minor axis using boxes of and , respectively, perpendicular to the axis. Sources P5, P6, P14, P17, P18, P22, P23 and P31 were cut out with a cut radius of FWHM of PSF at 0.3 keV. We normalized according to box area and exposure and corrected for vignetting and dead time. The normalization of the exposure map is determined at to offset from the galaxy.

Fig. 7. Spatial distributions of the surface brightness along the major (above) and minor (below) axis of NGC 3079. ROSAT PSPC counts are integrated in boxes of along the major axis, covering the galaxy disk region, and in boxes of along the minor axis, covering the galaxy halo region. They are centered at the distance given on the X axis relative to galaxy's nucleus. All PSPC sources (except P13, P16, P19, and P21 probably connected to the galaxy) have been cut out with a cut radius of FWHM of PSF at 0.3 keV. Response of a point source at the nucleus of NGC 3079 is given as dotted histogram (normalized to the count rate of the central box)

The profiles along the major and minor axes are clearly more extended than a point source (cf. Fig. 7). The extent along the major axis (2:05 corresponding to 12.5 kpc to both sides of the nucleus) is similar to the optical (about the corrected D₂₅). If the X-ray emission only originated from the galaxy disk, given the galaxy's inclination one would expect a extent to both sides of the nucleus along the minor axis. Fig. 7 instead shows an extension comparable to that along the major axis and possibly more. In the Western direction, excess emission is detected out to (30 kpc), while only a marginal excess is seen at distances greater than 2:05 (12.5 kpc) in the Eastern direction.

We can quantify this excess by noticing that net excess counts are found in the region from - in the western direction, significantly higher than the expected contribution from the single source P13 ( counts, from Table 4) and also much more extended than expected from a single point-like source. While with the present data we cannot exclude that P13 is indeed an individual source, the excess found points towards interpreting it as a local enhancement onto a somewhat irregular emission. No optical counterpart can be seen in the finding charts (see Appendix A.2). The excess in the east is net excess counts. For comparison, we can measure no excess ( and counts) in the N and S directions along the major axis in the same area and at the same radial distance from center.

Asymmetries and irregularities in the emission are also found on smaller scales. Outside of the nuclear area, an almost X-shaped emission (the arm to the SE of the major axis is not as evident as the others) is detected, as indicated rather irregular azimuthal surface brightness distribution outside of the nuclear area shown in Fig. 8. In particular, there are two enhancements relative to neighboring sectors at from the major axis, and clear depressions in the direction almost perpendicular to the major axis ( and in Fig. 8).

Fig. 8. Azimuthal distributions of the surface brightness of NGC 3079. ROSAT PSPC counts are integrated over sectors of 22.5 degree within a radius of 2:05. The central source has been cut out with radius. The sector at is centered at the direction of the major axis and angle is counted from north to the east

To better study the presence of an unresolved core in the data, we have further analyzed the HRI data, and we have produced radial profiles of the net emission in two opposite halves, i.e. East and West. As shown by Fig. 9, where the comparison with the PSF is also shown, there is a suggestion for the presence of an unresolved source embedded in a more extended component, and a significantly steeper decline in the Western half of the plane than the Eastern half. However, the point-source component does not dominate in the inner 20 radius region, so any attempt to study it at the PSPC resolution is hampered by the presence of the extended component.

Fig. 9. Radial profiles of the HRI X-ray surface brightness of NGC 3079 in two opposite halves. The dotted line indicates the profile expected for a point source arbitrarily normalized to the value of the innermost point

3.3.2. Spectral analysis

The morphology of the emission from NGC 3079, coupled with its optical and H I properties, suggests that the X-ray emission comes from three separate regions with presumably very different characteristics. The bright central region ( diameter), resolved in a complex source plus 2 point-like sources by the HRI, is likely to experience a large absorption. Similarly, the emission from the disk, seen edge-on, will be heavily absorbed, with the exception of the very external layers. In the halo region, instead, absorption consistent with the line of sight H I column density from our own galaxy is expected. A possible difference between the two sides of the plane, as the galaxy is not perfectly edge-on, might also be expected. In addition, inter-galaxy gas within the group (see introduction) may add additional absorption.

Given the limited statistics offered by the PSPC data, we have tried to minimize the number of separate regions from which to extract the photons for spectral analysis purposes. We have therefore checked with the aid of the hardness ratios whether different regions showed significantly different spectral characteristics, as could be expected from the considerations above. HR1 and HR2 have been calculated as defined in the previous section for 4 different regions:

a) the central region, defined as a circle of radius, centered at = and = (J2000.0)

b) the disk region, defined as a box of size , positioned along the major axis of the galaxy, shifted to S. The central source region has been masked out in the count extraction (cut radius )

c) the halo region above the plane, defined as a box of size , positioned parallel and adjacent to the E of the disk region

d) the halo region below the plane, defined as a box of size , positioned parallel and adjacent to the W of the disk region

The background was taken from an annulus of and inner and outer radii respectively, centered at the X-ray peak position (source P16).

We found no significant difference between the hardness ratio values for regions c) and d), and we have therefore defined as halo region the combination of c)+d) above. Table 6 summarizes the definitions of source and background regions used to determine hardness ratios and photon energy distributions for spectral fitting. As shown by Table 6, we have also considered the galaxy as a whole, and we have used a local background for the central region, to take into account possible contamination from the disk.

Table 6. Extracted spectra. Source plus background region (and covered area) as well as background region are given

Comparison of the HR1 and HR2 values and also with the theoretical hardness ratios shown in the plots of Fig. 1 clearly indicate that the central region, the disk and the halo occupy different regions of the HR1/HR2 diagram. We have then used simple spectral models to fit the data in the different regions, as indicated in Table 7. Raw spectra have been rebinned to obtain at least the signal to noise level per bin given in col. 1 of Table 7. Rough errors for the fluxes and luminosities in Table 7 are indicated by the statistical errors on the net counts (see Table 6, although additional uncertainties come from the poor knowledge of the spectrum, as can be seen by simply comparing the fluxes derived for different models).

Table 7. Spectral investigations of the extracted spectral files
Notes:
POWL: power law, THBR: thermal bremsstrahlung, THPL: thin thermal plasma
in units of 10²⁰ cm^-2
in units of 10^{- 13} erg cm^-2 s^-1 and 10³⁹ erg s^-1 , respectively, for 0.1-2.4 keV band, corrected for Galactic absorption and calculation of 1 errors
fixed to Galactic foreground for spectral fits

It is apparent that in all cases but the halo region, power law and thermal bremsstrahlung models give a better approximation of the data than the thin plasma model (Raymond-Smith code). Moreover, this latter would prefer a low energy absorption below what is expected from the line-of-sight H I column density, without giving a significant improvement in the best fit value. Power law or thermal bremsstrahlung models give essentially equivalent goodness of fit, and produce spectral models that are a good approximation of the energy distribution of the detected photons. In all cases a significant amount of intrinsic absorption above the line-of-sight value of cm^-2 is suggested, consistent with the idea that the emission comes from within the galaxy, and therefore suffers from the absorption in NGC 3079 itself. The fact that we have obtained very similar results from the disk and central regions is not surprising since, as already remarked above, the extended emission contributes significantly even at small radii, and this, combined with the extremely poor statistical significance of the data, does not allow us to distinguish the presence of a different component (for example from the point source in the nuclear region suggested by the HRI data).

Even though a good fit is already obtained for power law and thermal bremsstrahlung models, for the galaxy as a whole we have also tried to improve on the best fit values in the thin thermal plasma model by assuming two temperatures. This is done mostly to compare ours with published results on similar objects, and it is partly justified by the fact that the requirement of a lower-than-galactic absorbing column in the 1-temperature fit could be suggestive of an additional very soft component. Indeed we find best fit values of and keV, for a the minimum reduced to an acceptable value of 5.3 (for 4 degrees of freedom) and the best fit N_H consistent with the Galactic value.

The halo region cannot be fit by any of the simple models (i.e. the minimum value is larger than 2 in all cases). In spite of the limited significance of our procedure, we have nevertheless tried to fit the data with both a two-component model (i.e. two bremsstrahlung models and a bremsstrahlung and a thin plasma model) and with a thin plasma model with varying abundances. In both cases the minimum value reduces drastically to perfectly acceptable values (1.5 and 2.2) and the best fit values are kT 0.2 and 1 keV (2-T model) and kT 0.5 keV, 5% solar abundance. While it is therefore possible that more sophisticated models might be required for this region, given the limited statistical significance of the data, we cannot discriminate between different scenarios, nor can we be sure that our more sophisticated modeling of the data is correct, since we are left with 1 or 2 degrees of freedom. Since the resulting fluxes that we can derive with the different best fit models are all very similar (see Table 7), we will therefore assume the best fit values from the thin thermal plasma model for counts to flux conversion purposes for the halo emission.

3.4. NGC 3079 nuclear X-ray emission

While in the PSPC the nuclear source can not be resolved (see Fig. 5), a complex source is resolved with the HRI resolution at the galaxy's center (Fig. 4) with an extent of the order of ( kpc). In addition, a connected peak at distance north of the nucleus and a separate peak to the SW (source H13) can be clearly seen.

The radial distribution of the emission, centered on the X-ray peak, and its morphology both indicate that the source is extended and structured and that a possible point source located at the nuclear position could only contribute of the emission in the area. If this source indeed coincides with the active nucleus of the galaxy, its estimated count rate of cts s^-1 would correspond to a luminosity erg s^-1 , calculated assuming a thermal bremsstrahlung model with a low energy absorption equivalent to a column density of cm^-2. The lower limit sign is due to the fact that the absorption in the nucleus is likely to be much higher than the value assumed (which corresponds to an average column density in the disk on NGC 3079, see H I maps (Irwin & Seaquist 1991) and best fit parameters from spectral fits). While a different choice of the spectral model would give very similar values (see Table 7), the low energy absorption adopted influences very strongly the estimate of the intrinsic flux in the ROSAT band. mm-wave estimates of the extinction towards the nucleus of this source indicate that a minimum of H₂ cm^-2 should be expected (Sofue & Irwin 1992). ASCA data (Serlemitsos et al. 1997; Dahlem et al. 1998) in fact suggest the presence of a hard and heavily absorbed component (with absorbing column in excess of cm^-2; however notice that Ptak et al. (1998) give much lower best fit values for absorption to the power law component) in the spectrum of NGC 3079 as a whole (the spatial resolution of ASCA allows only a global measure of the spectral properties of NGC 3079), with a luminosity (2-10 keV) of erg s^-1 . The origin of this emission is not as yet unambiguously interpreted, since it is present both in galaxies with recognized nuclear activity (Low luminosity AGN, LINER) and in starburst galaxies (Serlemitsos et al. 1997; Ptak et al. 1998; Dahlem et al. 1998); therefore, due to the lack of good spatial resolution at high energies, it can either be related to the nuclear activity or to the binary population and starburst phenomenon (or both). Given the high absorption suggested by the data and the presence of nuclear activity, in NGC 3079 this component could come from a very absorbed compact nuclear source. In this case we expect about cts s^-1 in the nucleus with the HRI. Given the large uncertainty (also in absorption) this rate could be consistent with the possible point source contribution from a nuclear source in the HRI image (see above).

Given the strong contamination from the diffuse background, and the small statistics (about 40 cts), we cannot really measure possible time variability in the flux from this source, which would confirm its point-source nature and propose an identification with a super-luminous X-ray binary close to the nucleus or the X-ray detection of the NGC 3079 active nucleus itself.

3.5. NGC 3073 and MCG 9-17-9

We searched for X-ray emission from the companion galaxies of NGC 3079, NGC 3073 and MCG 9-17-9. Only MCG 9-17-9 was detected (H6/P9). The 69 counts detected with the PSPC are insufficient for a detailed spectral fitting, and we could only determine one of the two hardness ratios (see Table 4). Comparison with the plots of Fig. 1 indicates that for these two spectral models the source is probably absorbed above the line-of-sight value, and that its spectrum is relatively hard (i.e. kT 0.5 keV or ).

A weak enhancement is seen at the position of NGC 3073 both in the HRI and in the PSPC, however with a significance far below the threshold for our source catalogs.

Table 8 summarizes PSPC and HRI count rates, X-ray fluxes and luminosity for these sources. To convert count rates to fluxes we assumed a 5 keV thermal bremsstrahlung spectrum (see Table 2); the same distance as for NGC 3079 was assumed for the luminosity determination.

Table 8. X-ray parameters for NGC 3073 and MCG 9-17-9
Notes:
fluxes in units of 10^{- 15} erg cm^-2 s^-1 for a 5 keV thermal bremsstrahlung spectrum in the 0.1-2.4 keV band, corrected for Galactic absorption
luminosity in units of 10³⁸ erg s^-1 assuming a distance of 17.3 Mpc

The comparison of the HRI and PSPC flux of MCG 9-17-9 indicates that the source has not varied between the two observations.

© European Southern Observatory (ESO) 1998

Online publication: November 9, 1998
helpdesk.link@springer.de