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

2. Observations and data analysis

NGC 253 was observed with the ROSAT HRI and PSPC for 57.7 ks and 22.8 ks, respectively. Both the HRI and PSPC observations are each spread over approximately 3.5 years and consist of 6 HRI and 2 PSPC observation blocks, each subdivided into several observation intervals (OBIs). The date and integration times for the different observation blocks are listed in Table 2.

Table 2. ROSAT observations of NGC 253.

The data reduction was performed with the ESO-MIDAS/EXSAS (ESO-MIDAS 1997, Zimmermann et al. 1997) software package.

2.1. Attitude corrections

Attitude solutions of ROSAT pointings used by the Standard Analysis Software System (SASS, Voges 1992) to produce event files are known to produce residual integral errors of the order of 6" (boresight error) for an observation, and - due to short term fluctuations - systematically broaden the PSF. To improve on the solution we adopted two (subsequent) techniques.

Firstly, for the PSPC observations the positions of point sources during the first and second PSPC observation block were compared, and the two blocks - using different guide star patterns for the attitude solution - were aligned. For the HRI, to improve the intrinsically narrower PSF, all observation intervals were aligned with respect to the first one by computing the centroid for 8 bright point-like X-ray sources visible in each OBI. No offset was seen to exceed 4".

In a second step, the positions of 13 possible optical counterparts (derived from the ROE finding charts, Irwin et al. 1994), coinciding to within with point source HRI error circles, were used to determine the systematical offset of the preliminary attitude solution. The HRI observations were corrected for this offset (translation of 1:009 and 1:001 to the E and N, respectively). The PSPC observations were co-aligned (3:000 and 3:006 to the E and N, respectively, counterclockwise rotation of ) with the HRI solution. All X-ray source positions given in this paper have already been transformed into the sky coordinate system. The systematical error of the final source positions is determined as the residual error of the transformations (2:005).

2.2. Image generation

2.2.1. Images of the HRI observations

To reduce the background due to UV emission and cosmic rays, HRI images were integrated using raw channels 2-8. An 35´ image (0.1-2.4 keV band) was constructed with a bin size of 2:005 and smoothed with a Gaussian filter of 12" FWHM (Fig. 1). The optical extent of the galaxy is indicated by the ellipse. A close-up view of the inner 14´ field of the HRI (Fig. 2) has been formed with a bin size of 1" and smoothed with a filter of 5" FWHM, corresponding to the FWHM of the on-axis HRI PSF. Within a radius of from the center of the HRI, the PSF does not deteriorate significantly. To investigate the detailed structure of the central 2´ region of the galaxy (over-exposed in Figs. 1 and 2), we overlaid X-ray contours over a greyscale plot of the central area of the image used for Fig. 2 (Fig. 3).

Fig. 1. ROSAT HRI image of the NGC 253 pointing for the full HRI field of view. The image has been formed with a binsize of 5" and smoothed with a Gaussian filter of FWHM. Detected X-ray sources (cf. Sect. 2.3.3 and Table 3) are indicated by squares (sources detected with the HRI and PSPC), hexagons (sources only detected with the HRI) or diamonds (sources only detected with the PSPC). The ellipse of NGC 253 has been sketched, the center of the galaxy coincides with the position of the central source. Sources outside the area covered by the disk of NGC 253 have been enumerated in this figure, the others are enumerated in Fig. 2. The right ascension and declination are given for J2000

Fig. 2. ROSAT HRI image for the inner 14´ field of the pointing. The image has been formed with a binsize of 1" and smoothed with a Gaussian filter of . Sources have been marked according to Fig. 1. All sources but the central one (X34) and a bright source 20" to the south of X34 (X33) have been enumerated according to Table 3. The cross () between the sources X25 and X28 marks the position of the SN 1940 E. The positions of the ROSAT sources X17, X21, X33 and X36 coincide with the positions of the Einstein sources E2, E5, E8 and E1, respectively. The Einstein sources E3, E4, E6 and E7 remained undetected during the ROSAT observations, and the positions of these sources have been marked with crosses (+). To make the structure of the over-exposed central region of the galaxy visible, contours (4, 8, 15, 27 and  cts s-1 arcmin- 2) have been over-plotted in the over-exposed area. Fig. 3 shows a close up view of the over-exposed region

Fig. 3. HRI image of the central region of NGC 253. The image gives the inner field of Fig. 2. The contours of the same image have been superposed. Contour levels are at 3, 4, 6, 8, 11, 15, 20, 27, 40, 50 and  cts s-1 arcmin- 2. The circles, each having a radius of , correspond to the extraction radius for the surface brightness profiles (cf. Sect. 3.2 and Fig. 6), and mark the positions of X34 (central source) and X33

2.2.2. Image of the PSPC observations

A 0.1-2.4 keV PSPC image (Fig. 4) was constructed by the superposition of sub-images in 8 standard bands (R1 to R8, cf. Snowden et al. 1994). Each sub-image has been corrected for exposure, deadtime, and vignetting, and the sub-images have been smoothed with a Gaussian filter corresponding to the on-axis PSF of the energy band (FWHM ranging from 52" to 24" for the lowest to the highest energy band). To make the full dynamic range of the PSPC data visible in one image, a greyscale representation is chosen running several times from bright to dark.

Fig. 4. ROSAT PSPC broad band image of the NGC 253 pointing for the same field as presented in Fig. 1. The image has been formed with a binsize of and smoothed as described in Sect. 2.2.2. The greyscale representation runs several times from bright to dark to make the full dynamical range of the data visible. The ellipse of NGC 253 has been sketched, sources have been marked according to Fig. 1 and Table 3

2.3. Constructing a combined HRI/PSPC point source catalog

2.3.1. HRI point source detection

Point sources were searched for in the full HRI field of view with the EXSAS local detect, map detect and maximum likelihood algorithms (Zimmermann et al. 1997), using images of pixel size 5". To reduce the background due to UV emission and cosmic rays only those events detected in HRI raw channels 2-8 were used. Sources with a detection likelihood were accepted. Maximum likelihood values (L) can be converted into probabilities (P) through , thus corresponds to 3.6 Gaussian sigma significance (cf., e.g. Cruddace et. al 1988).

Within a field centered on the nucleus of NGC 253 and extending along the major and along the minor axis, point sources are seen embedded in extended emission structures, not resolved by the HRI. Since the background map in this region does not follow the filamentary diffuse structures, a maximum likelihood algorithm, comparing the local X-ray brightness with the local value of the background map, might pick up extended emission regions as point sources. To avoid this, we only accepted sources detected by the local detect algorithm in regions of enhanced diffuse emission. The local detect algorithm looks for gradients in the image with help of a sliding box method and adapts better to local changes in the background. One exception was within a bright part of the southeastern extension of the central diffuse X-ray emission. Here, we manually excluded a spurious source detected by the local detect algorithm at a position east of the bright source close to the nucleus. Our final HRI source list comprises a total of 49 sources in the HRI field of view.

2.3.2. PSPC point source detection

Source positions and count rates were calculated for the inner PSPC field in the five standard energy bands `broad' (0.11-2.40 keV), `soft' (0.11-0.41 keV), `hard' (0.52-2.01 keV), `hard1' (0.52-0.90 keV), and `hard2' (0.91-2.01 keV). All images used for the source detection have a bin size of 5". As with the HRI, the EXSAS local detect, map detect, and maximum likelihood algorithms were applied to each energy band. Sources with a likelihood 9 () were accepted, and the source lists within each of the different energy bands merged, assuming that detected source positions were identical if their separation is less than 3 times the statistical position error. The final source position was taken from the energy band in which the source was found with the highest likelihood. One PSPC source was located in between two HRI sources (X43 and X46), separated by . The fact that this PSPC source was flagged as extended (extent FWHM) by the detection algorithms strongly suggested that it represents the combined unresolved emission from the two HRI sources, and this source therefore, was removed from the PSPC source list. Later, when merging the HRI and PSPC lists, the PSPC count rates for X43 and X46 were calculated with fixed positions suggested by the HRI.

The energy resolution of the PSPC detector allows the calculation of `X-ray colors' for the X-ray sources, the hardness ratios HR1 and HR2. The ratios are defined on the basis of the net counts in the soft, hard, hard1, and hard2 bands. By definition, HR1 = (hard-soft)/(hard+soft), and HR2 = (hard2-hard1)/(hard2+hard1) (soft here means the counts in the soft band etc.).

2.3.3. Combined HRI/PSPC point source catalog

From the HRI and PSPC catalogs, a combined point source catalog (Table 3) was constructed. Column 1 gives the source numbers, that are used to identify sources in Figs. 1 to 4. The R XJ name (following the naming convention for ROSAT sources according to Zimmermann et al. 1997) is contained in col. 2. Columns 3 and 4 give the source positions (right ascension and declination, equinox 2000.). Sources with spatial separations between the HRI and PSPC positions smaller than the sum of the 90% error radii were taken as identical (the `detector flag' in col. 5 has the entry `B' for both). If a source existed only in the PSPC or HRI source list, an entry `P' (PSPC) or `H' (HRI), is given in col. 5. Sources have the entry confused (`C') in col. 6 if the PSPC hard band image shows extended emission surrounding the source, or the HRI image resolves two point sources separated by a distance smaller than the PSPC hard band PSF. For confused sources, the position information was always taken from the HRI source list. For other `B' sources, the position information with the smaller statistical position error was selected. The position errors, including a 2:005 systematical error, are given in col. 7. The likelihood for source detections is displayed in columns 8 (HRI) and 9 (PSPC).

Table 3. X-ray properties of detected point sources

Columns 10 and 11 give the net counts of the X-ray sources, columns 12 and 13 the count rates, corrected for exposure, deadtime, and vignetting. In the case of HRI non-detections at PSPC source positions, HRI counts were calculated at the position suggested by the PSPC. To do so, HRI counts were extracted with a cut radius of 1.5 times the local FWHM of the HRI PSF at the PSPC source position, and background counts subtracted determined at that position in the EXSAS background maps. For sources with detection likelihoods , upper limits, at a 2 confidence level, are given. PSPC counts for isolated sources were determined using the source detection results. For confused PSPC sources with nearby point sources (separation below the FWHM of the hard band PSF), a `multi source fit' technique (cf. Zimmermann et al. 1997) was used with source positions fixed according to the HRI detections. For PSPC confused sources without nearby point sources, the counts were extracted within a cut radius of the FWHM of the PSPC PSF around the HRI position, and a local background, determined in a concentric ring from to the FWHM around the source position, was subtracted. For all HRI sources, for which no PSPC source with a likelihood was found, upper limits (2), have been calculated.

2.4. Time variability investigations

To study time variability of point sources, counts and count rates were determined for the individual observation blocks listed in Table 2. The rates were calculated with fixed source positions. The `likelihood ratio test' (cf., e.g. Cash 1979, Hogg & Tanis 1983) was used to test for variability. Since the comparison of HRI and PSPC count rates depends on the assumed spectral model (cf. Table 4), the HRI or PSPC observations were analyzed separately.

Table 4. Energy conversion factors for the ROSAT HRI and PSPC for different spectral models ^a Energy conversion factors in units of  erg cm^-2 cts^{- 1 b} Assumed for the conversion from count rates to fluxes for point sources

The `likelihood ratio test' for the six HRI blocks was performed as follows: one starts with the hypothesis of a time constant source. Count rates during the block i are defined by the ratio of the detected counts and the observation time , and these rates should be equal for each block: , in the case of no variability. The likelihood ratio is defined as . If the hypothesis of no variability is true, then due to the counting statistics has a distribution with 5 degrees of freedom.

© European Southern Observatory (ESO) 1999

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