Astron. Astrophys. 342, 101-123 (1999)

3. Results

3.1. HRI and PSPC point sources in the NGC 253 field

In the NGC 253 field (Figs. 1, 2 and 4) 73 X-ray point sources are detected, 23 with both detectors (marked in the figures as squares), and 22 and 28 of which exclusively with the HRI (hexagons) and PSPC (diamonds), respectively.

Assuming a 5 keV thermal Bremsstrahlung spectrum (0.1-2.4 keV band and corrected for Galactic foreground absorption, cf. Table 4), the HRI count rates convert to fluxes between  erg s cm^-2 (source X29) and  erg s cm^-2 (X34). Similarly, PSPC derived fluxes span  erg s cm^-2 (X63) to  erg s cm^-2 (X34). As one can see, the longer observation time for the HRI - about a factor of three with respect to the PSPC observation time - makes up for the lower HRI sensitivity leading to almost identical detection limits for the PSPC and HRI observations.

To allow a crude estimation of the spectral properties of the point sources, their hardness ratios (columns 14 and 15 of Table 3) were calculated. To first order, HR1 traces the absorption and, to a lesser degree, the hardness of the spectrum, while HR2 mainly traces the hardness of the spectrum (cf. e.g. Vogler & Pietsch 1996). The hardness ratios are useful in estimating the absorption and spectral behavior of point sources, for which the low photon statistics do not allow spectral investigations. For non-confused sources the cut diameters for the sub-bands are chosen according to the PSF FWHM for the corresponding off-axis angle and energy sub-band, and the background is subtracted with the help of the background maps. In cases where the error exceeds the counts in one sub-band, only an upper (lower) limit of the hardness ratio is calculated. For that, the counts of the non detected band are chosen to be equal to the upper limit (). Where the errors exceed the counts in both sub-bands, no hardness ratios are calculated. For confused sources, HR1 is not calculated, since problems might arise on account of the large extraction radius used for the PSPC soft band and, to calculate the HR2, counts and errors in the hard1 and hard2 band were deduced with the help of the multi source fit technique.

The light curves of all sources of the catalog (excluding X3 and X37 which are located at the edge of the calculated PSPC images) are presented in Fig. 5. If the source was detected during an observation block with a count rate exceeding its error, filled squares are plotted. Vertical bars represent statistical errors. Upper limits () are indicated with open squares in the case of non-detections. Horizontal lines give the mean count rates during the whole HRI (blocks 1, 4-8) or PSPC (blocks 2, 3) observations. Solid lines represent sources that were detected, dashed lines represent sources for which only upper limits could be calculated. In general, the mean count rates are in good agreement with the results presented in Table 3. The PSPC mean count rates are slightly different from Table 3 for some sources located in diffuse emission regions (e.g. X24 or X27).

Fig. 5a. Light curves of X-ray sources in the NGC 253 field (excluding X3 and X37 which were located at the edge of the investigated field). Where the source was detected during an observation block with a count rate exceeding the error filled squares are used as symbols and the errors are indicated as bars. In the case of errors exceeding the count rates upper limits () are plotted as open squares. The horizontal lines give the mean count rate calculated as explained Sect. 2.4. Solid lines represent sources that were detected according to Table 3, dashed lines represent sources for which only upper limits could be calculated. Stars above the HRI or PSPC light curves mark in which detector the variability was detected. The PSPC count rates have been divided by a factor of 3.2 before they were plotted in the diagram (this is the ratio between the energy conversion factors of the HRI and PSPC for a 5 keV thermal Bremsstrahlung spectrum corrected for Galactic absorption, cf. Table 4). If one detects an isolated point source and the spectral model is right, the light curve should be a straight line

Fig. 5b. (continued)

To further characterize the point sources of the catalog, they were checked for time variability using the likelihood ratio test. The likelihood ratios of the time variability test performed for the HRI and PSPC observations (Sect. 2.4) can be transformed into a probability that the source is time-variable (cf., e.g. Bronstein 1985). In Table 5 we summarize the information on those 24 sources of the total of 73, for which a Gaussian significance for time variability with was found within the HRI or PSPC observation blocks. Besides the significances, maximum count rates and fluxes (cf. Table 4 for the conversion factor and assumed model) are given.

Table 5. Time variability investigations of the different HRI and PSPC observation blocks ^a Gaussian significance for the detection of time variability during the individual HRI (H) or PSPC (P) observation blocks according to Sect. 2.4 ^b Time Variability during the PSPC blocks established. However, the maximum count rate (flux) is in the order of the integral count rate (flux) of the total observation. The problem might be due to diffuse emission surrounding the source as well as due to the very nearby source X42

Time variability for a source might also be established via a comparison of the HRI and PSPC derived fluxes. This method, however, might feign time variability if the wrong spectral model is used for count rate to flux conversion. We calculated the ratio of the energy conversion factors for our assumed model (5 keV thermal Bremsstrahlung), a 0.5 keV thermal Bremsstrahlung spectrum and a thin thermal plasma of 0.3 keV (cf. Table 4). The examples indicate that a wrong spectral model might fake time variability to the order of up to 30 or even more for extreme examples, e.g. a plasma with a temperature below 0.2 keV. Keeping this reservation in mind, time variability is suggested with for four additional sources which were not already picked up by investigating HRI or PSPC blocks, individually: X10, X21, X24, X34. In the case of X10 and X24, the HRI did not pick up a point-like source. The sources X21 and X34 are located within the diffuse emission of the NGC 253 bulge and disk and it cannot be excluded that the PSPC determined counts are affected by subtracting a wrong background.

3.2. Resolving the complex nuclear emission area

Two bright sources, X33 and X34, are detected in the central region of NGC 253, both embedded in a complicated diffuse X-ray emission structure visible in the PSPC and HRI images (Figs. 1 to 4). We tried to disentangle the emission components with the help of the spatial resolution of the HRI and the spectral resolution of the PSPC.

The luminosities as suggested by the HRI count rates are  erg s^-1 and  erg s^-1 for X33 and X34, respectively. The position of the source X34 coincides with the optical center of NGC 253, X33 being located to the south. Together with the sources X33 and X34, the counts within the central emission region encircled by the lowest contour in Fig. 3 amount to HRI counts. We focus in this paper on the point sources and will discuss the diffuse emission components in PEA.

3.2.1. Spatial analysis of the HRI data

The HRI detection algorithm flagged X33 and X34 as extended (FWHM of extent and , respectively). X33 and X34 are, however, embedded in a region of diffuse emission and the detection algorithm may be fooled if this background is not modeled correctly by the background map. To investigate if these sources really are point sources inside diffuse emission, radial surface brightness profiles from 0 to 15" radius were calculated. To visualize the investigated region, circles with radii of 15" around the sources are sketched in Fig. 3. The surface brightness profile of X34 is centered on the intensity maximum correlated with X34. Due to the extent of the source and the slower decay of the intensity towards the east than to the west (cf. Fig. 3), the source position determined by the maximum likelihood algorithm and given in Table 3 is located southeast from this maximum. The width of the individual rings of the profiles is 3". In principal, one could compare the profiles of X33 and X34 to analytical models of the PSF. However, as there was no point source in the HRI field of view, bright enough to allow the correction of the pointing positions on time intervals shorter than the wobble period (400 s), the SASS attitude solution could only be improved on longer time scales (see Sect. 2.1), and one has to expect that the PSF is slightly broadened due to residual artifacts of the satellite wobble movement. We therefore choose to compare the surface brightness profiles of X33 and X34 to the profiles of the bright unconfused point-like sources X21 and X36, which were collected with the same attitude solution as X33 and X34 (Fig. 6). Within the off-axis angles of X21 and X36 (4:06 and 2:02, respectively), the HRI PSF is not expected to deteriorate compared to the central sources.

Fig. 6. Radial surface brightness profiles for the sources X33, X34, X21 and X36, calculated with a radial binwidth of 3" from the HRI observations. The crosses mark the measurements for the individual sources, the length along the x and y axis indicate the binwidth and the error of the brightness, respectively. The open circles show the `experimental' PSF model as obtained from the point-like sources X21 and X36. The PSF predictions for point-like sources at the position of X33 and X34 (cf. Sect 3.2.1) have been marked as open squares

The `experimental' PSF is calculated by averaging the surface brightness profiles of X21 and X36. The maximum of the PSF is normalized to the inner radial bin (). As can be seen in Fig. 6 (circles plotted over the profiles), this PSF model represents a good description of the brightness profiles of X21 and X36. To adapt the model for X33 and X34, the maximum was again normalized to the inner radial bin. The fact that X33 and X34 are sources embedded in a diffuse emission region was taken into account by normalizing the background in the outermost radial bin (). As can be seen in Fig. 6 (squares plotted over the profiles), this model is a good approximation to the profile of X33, though it cannot describe the slow decay of the surface brightness of X34. X33 therefore is likely to be a point source. Integrating the PSF model, 35921 counts ( erg s^-1 cm^{- 2},  erg s^-1) are deduced for X33, a number significantly lower than the number of counts determined by the detection algorithm (52024 counts). That number has been overestimated as X33 was fitted as an extended source and not the whole diffuse background was taken into account in the background map. On the contrary, the X-ray emission of X34 cannot be due to a single point-like source. Within a circle of 9" around X34, 70326 counts were measured after subtraction of the background, whereas the PSF model predicts less than 300 counts for a point source.

3.2.2. Spectral analysis of the PSPC data

To further investigate the structure of the point-like source X33, we made use of the spectral capabilities of the PSPC. A source spectrum was extracted with a cut diameter of . The very small extraction diameter and the varying PSF of the individual channels were corrected for using standard EXSAS procedures. Background was subtracted from a source-free region outside the disk and halo of the galaxy, leading to a raw spectrum of X33 containing 35223 counts. This raw spectrum was binned into energy bands to give a signal to noise ratio . Simple spectral models, a power law (POWL), thermal Bremsstrahlung (THBR) and a thin thermal plasma (THPL) model were fitted. With free absorption, free normalization and free index/temperature, our fits had seven degrees of freedom. Formally, all models achieved values between 0.9 and 1.0. However, as can be seen from the suggested temperatures of the thermal models (THBR:  keV; THPL:  keV), they fall in a range that cannot be properly constrained by ROSAT. The POWL fit (Fig. 7) resulted in an intrinsic absorption (after subtracting the Galactic foreground) of cm^-2, a photon index of and flux of  erg s^-1 cm^{- 2} (0.1-2.4 keV). The errors (1) for the fit parameters were calculated with the help of the error ellipses. The residuals of the fit are very small with the exception of the bin around 0.95 keV. Spectra of X34 (cf. PEA) indicate the presence of a thermal emission component with a temperature around 1 keV. Assuming that this diffuse emission is also contributing at the position of X33, the high residual of the bin around 0.95 keV could be explained. The photon index of X33 suggests a very hard intrinsic spectrum. In comparison, the index of a POWL fit for X34 (same extraction radius chosen as for X33) is and suggests a much softer spectrum. The absorption of X33 ( cm^-2) is clearly less than the absorption of X34 ( cm^-2). The spectrum of X34 together with the spectrum of the underlying diffuse emission is discussed in detail in PEA.

Fig. 7. Results of a power law fit to X33. Top panel: Flux of the observed X-ray emission normalized to the energy, in photons/(), the crosses represent the observed flux (count rates defolded by the spectral model), the solid curve gives the best fit. Bottom panel: residuals of the fit

Combining the PSPC spectrum with the results of the surface brightness profiles obtained from the HRI observations, one derives an improved luminosity for X33 of  erg s^-1, corrected for the Galactic foreground and absorption within NGC 253. This luminosity is significantly lower than the one deduced from our detection catalog and time variability investigations.

3.3. Point sources within the NGC 253 disk

Of the 73 field sources, 32 are located within the ellipse of NGC 253 and attributed to the NGC 253 disk. Some of these sources may be spurious detections caused by diffuse filamentary X-ray emission features which are most clearly seen in the ROSAT PSPC image (Fig. 4), covering the disk and halo of NGC 253. For the HRI, similar problems exist in the central emission region shown in Fig. 3 and in the inner spiral arms of NGC 253 (indeed, one source has already been removed when creating the HRI source catalog). For the PSPC, we exclude from the further discussion those sources in the NGC 253 disk, which are located in regions of diffuse X-ray emission and which have only been detected with the PSPC, namely X50, X51, X52, X57 and X62. A visual check and a comparison with our list of transients verifies that no PSPC bright sources are rejected by this procedure.

All 27 remaining sources have been detected with the HRI, the count rates lying between  cts s^-1 (X26) and  cts s^-1 (X34). At the distance of NGC 253, this converts to luminosities between  erg s^-1 and  erg s^-1. 23 of these sources have been also detected with the PSPC, with count rates from  cts s^-1 (X9) to  cts s^-1 (X34), converting to luminosities between  erg s^-1 and  erg s^-1.

The individual source luminosities are listed in Table 6. For 8 of the 27 NGC 253 sources, either the ROSAT HRI or PSPC data alone proof variability. For the brightest source (X33) statistics were sufficient to search for time variability on shorter time scales. Single observation blocks were analyzed, but no short term time variability could be established. Two further time-variable sources (X21 and X34) are suggested by comparing the HRI and PSPC fluxes. The apparent time variability of X34, however, can be explained as a side effect of the limited resolution of the PSPC hampering the flux determination for this extended source. Therefore, excluding X34, the ROSAT results indicate time variability for 9 of the 27 sources in the disk of NGC 253.

Table 6. Luminosities of sources located within the area covered by the ellipse of NGC 253 ^a Luminosities at the distance of NGC 253 in units of  erg s^-1, cf. Table 5 for assumed model ^b Time Variability during the PSPC blocks established. However, the maximum flux is of the order of the integral flux of the total observation. The problem might be due to diffuse emission surrounding the source as well as due to the very nearby source X42 ^c Probably due to enhancements of diffuse X-ray emission ^d Background quasar

Transients form a special subclass of the time variable sources. For the purpose of this paper a source will be called transient if it remains undetected during at least one observation interval (2 detection limits of the individual intervals  erg s^-1) and shows a monotonic increase/decrease of the peak luminosity, the simplest case being a source only detected during one observation block. This definition of a transient may actually be fulfilled if a source shows an outburst by a factor of a few because of the limited sensitivity of the NGC 253 observations, and in this case the definition is less stringent than for transients in the Galaxy or the Magellanic clouds. From their ROSAT light curves X12, X14 are transients in this restricted sence.

Our investigations establish time variability for all NGC 253 disk sources with luminosity maxima above  erg s^-1, with the only exception of X36, for which no time variability could be found. For fainter sources with no established variability, it remains unclear whether these sources are time-constant or whether the statistics are too low to establish variability.

To further classify the brighter point sources, we have made use of the spectral capabilities of the PSPC. For seven sources (X12, X17, X21, X33, X34, X36 and X40)  PSPC counts are detected. X33 and X34, embedded in the extended nuclear X-ray emission, have already been discussed in Sect. 3.2. Photons for the other sources were extracted with extraction radii of 25", and a local background, determined in a concentric ring from to , is subtracted to reduce contributions from surrounding diffuse emission features. Contributions from other point sources in our catalog to this background are avoided as they are screened out with a cut radius of 25". The very small extraction diameters and the varying PSF of the individual channels were corrected for using standard EXSAS procedures. The counts contained in the raw spectra are listed in Table 7. Due to the PSPC PSF of , we are not able to separate X17 from X18 and X40 from X42. Simple spectral models, a power law (POWL), a thermal Bremsstrahlung (THBR) and a thin thermal plasma (THPL) model were fitted, and for all sources, the formal value of the THBR, THPL and POWL fits were of the same order of magnitude. In Table 7 the results of the THBR fits are listed. All sources are intrinsically absorbed, and the lowest absorption ( cm^-2) is measured for X21, while the fits of all other sources indicate cm^-2. While the THBR fit predicts a temperature around keV for X21 the temperatures of the other sources ( keV) seem to be lower. However, the errors of the fits (1 errors given in Table 7, calculated from error ellipses) are high. With the help of the more precise HRI count rates and using the conversion factors relevant to the PSPC suggested models, we derived two types of source luminosities that are given in Table 7: an "absorbed luminosity" (i.e. calculating the flux for temperature and normalisation as suggested by the fit and using an value that is the fit value minus the Galactic foreground value), is  erg s^-1 for all the sources; an "intrinsic luminosity" (i.e. calculating the flux for temperature and normalisation as suggested by the fit and using an value of zero). Technically, these intrinsic luminosities are higher than the absorbed ones by a factor of 2 for X21 and for all other sources. However, especially the big corrections have to be taken with care as they contain big errors introduced by the relatively low temperatures and high absorption values and the associated uncertainties.

Table 7. Results of thermal Bremsstrahlung fits to X12, X17, X21, X36 and X40 ^a In excess of the Galactic foreground ^b Luminosity corrected for Galactic foreground . The spectral model as received from the PSPC fit was folded with the HRI count rates to obtain the luminosities. For the time variable sources X12, X17 and X40 we used the maximum HRI count rates, for X21 and X36 we used the mean HRI count rate ^c Predicted luminosity corrected for total absorption

3.4. Emission components of the NGC 253 disk

In this section the contribution of the NGC 253 point sources are compared to the total X-ray emission of the NGC 253 disk. One has to be aware that, due to the not completely edge-on orientation of the NGC 253 disk, contributions from hot gas in the lower halo of the galaxy (cf. PEA) will be contained in the integral disk count rate. The measured count rates have been corrected for exposure, deadtime, and vignetting. The background was taken from two source free regions outside the disk and halo of the galaxy, namely a region east of the galaxy encircled by the sources X68, X69 and X72, and a region west of the galaxy encircled by X2, X8 and X20. The PSPC has a very low detector internal background and enables a sensitive calculation of count rates over large areas. The results for the HRI, which has a higher detector internal background, give a higher error and depend more critically on the background regions chosen.

The total count rates within the ellipse of NGC 253, together with the integral point source content for the PSPC and HRI observations are given in Table 8. The PSPC and HRI count rates for the disk are in good agreement (i.e. the PSPC count rate is higher by a factor of , as expected from the energy conversion factors for the different spectral models (cf. Table 4). Excluding the central source X34, which is an extended source (cf. Sect 3.2), the integrated emission of the point sources makes up roughly one quarter of the total emission. With both detectors, an integrated point source luminosity of  erg s^-1 is measured. The PSPC sources X50, X51, X52, X57 and X62, which are probably due to local enhancements of the diffuse emission covering the disk of NGC 253, contribute  erg s^-1.

Table 8. Emission components of the NGC 253 disk ^a Corrected for Galactic foreground absorption ^b The diffuse emission component detected from the disk of NGC 253 cannot be described by a 5 keV thermal Bremsstrahlung model (cf. PEA) ^c The hardness ratios of individual point sources are subject of Sects. 2.3.2, and they have been sketched in Fig. 8.

The central source X34 and the surrounding diffuse emission (cf. the lowest contour level of Fig. 4) contribute (after subtraction of X33 according to Sect. 3.2)  HRI cts s^-1, slightly more than the integral point source contribution ( HRI cts s^-1).

To visualize the spectral behavior of the different NGC 253 emission components, Fig. 8 shows a hardness ratio plot of the total NGC 253 emission (filled hexagon) and the bright point source X33 (square). Five further bright PSPC sources (X12, X17, X21, X36, and X40, all with  PSPC counts), that were also detected with the HRI, have been added to the diagram (marked with diamonds). Sect. 2.3.3 described, how HR2 for the sources (all sources were confused) have been derived. Due to the high FWHM of the soft band PSF, the HR1 values are not given in Table 3. To estimate the HR1 for the purpose of this section, the soft and hard band counts were calculated within an extraction diameter of 1.5 times the PSF of the energy bands, and a local background from 1.5 to FWHM diameter was subtracted. Similar to X33, the sources X12, X17, X21, X36 and X40, are located in the upper right corner of the diagram, and their high HR1 and HR2 values suggest harder spectra and higher absorption than for the total NGC 253 disk emission. The positions of the spurious PSPC sources (X50, X51, X52, X57, and X62) in the hardness ratio digram (left of the hexagon representing the entire NGC 253 disk emission) would indicate lower absorption and softer spectral behavior, giving further support to their identification as spurious detections of diffuse emission from the (outer) disk of NGC 253. Diffuse emission in the disk is investigated further in PEA.

Fig. 8. Hardness ratio plot of point sources in the NGC 253 disk. The square marks the position of X33 in the HR diagram. The filled hexagon marks the hardness ratios of the entire NGC 253 disk emission (cf. Table 7). Crosses indicate the measurements and errors for different sources in the disk of NGC 253. One group (marked with diamonds at the center of the crosses) are bright point sources which were detected with the PSPC and HRI, the second group (left of the hexagon) represents sources only detected with the PSPC and not visible during the ROSAT HRI observations

© European Southern Observatory (ESO) 1999

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