Astron. Astrophys. 360, 99-106 (2000)

3. ROSAT observations

The data were collected in two observing cycles. In a 17.991 ksec integration (January 1992) we clearly detected the jet without any image processing (see Fig. 2). As the signal-to-noise ratio (S/N) was not sufficient for detailed studies, 3C 273 was re-observed in December 1994/January 1995 for a total of 68.154 ksec (quoted times as "accepted" by the ROSAT standard reduction analysis).

Fig. 2. Unprocessed image of 3C 273 from the first observing block clearly showing the jet towards the SW.

3.1. Improving the resolution

Images (sky pixel size ) have been produced from the event tables using the EXSAS package provided from the ROSAT group at MPE in Garching. However, the resulting point-spread function turned out to be unsatisfactory. Whereas in the call for proposals the integral point response function of the ROSAT-XTE + HRI was quoted to have a width of 5", the FWHM of the images as produced from the raw data was typically larger than 6". This discrepancy is due to inaccuracies in the aspect solution, which determines the sky coordinates for each photon detected. Obviously in the standard aspect solution the spacecraft wobble with a period of 402 sec is not completely corrected for (see Fig. 3).

Fig. 3. Centroids of the quasar photons for all 10sec intervals of observing block 2.

With a count rate in the HRI of 2.8 cts/sec the signal from the quasar core itself is sufficiently strong to allow a shift-and-add procedure as follows: All integrations are divided into time bins of 10 sec duration and the centre of gravity of the photons from the quasar core are calculated for each interval. Photons were restricted to a raw amplitude between 2 and 8, typical for the quasar. The average accuracy of the centroid positions was in X and in Y-direction. The interval of 10 sec was a compromise between sufficient time resolution and positional accuracy. These offsets from the nominal centre position as a function of time directly correspond to the remnant pointing errors due to the insufficiently corrected wobble motion. They were interpolated in time by splines and the detected position of every photon was corrected as a function of its arrival time.

Fig. 4. Isophote plot of the final image (left). Contours are at 2, 3, ... 8, 16 ... 1024, 1124, ... 1824 counts. Note the jet to the SW and the source to the NE of the quasar. An elongation of the isophotes in the general direction of the jet is evident.

For comparison the same procedure was applied to eight data sets from the ROSAT archive for the white dwarf HZ 43, which is definitively a point source. To derive the radial image profile, each data set was sampled on concentric circles around the quasar respectively white dwarf with radii in steps of . For each circle a constant was fitted to the data giving the azimuthally averaged intensity profile. The "core" of these profiles was decomposed by a least-square-fitting procedure into two Gaussians, neglecting the very extended exponential component discussed in the HRI calibration report (David et al. 1999). The shape of the point-spread-function (PSF) in the raw images changed from observation to observation. But as can be seen from Table 1 the profile of the de-jittered images was constant within the error. As indicated in Table 1 and demonstrated by Fig. 5 and Fig. 6, the resolution could be enhanced to this way. A similar procedure was described by Morse (1994).

Fig. 5. Radial profile of 3C 273 compared with that of the white dwarf HZ 43 (observation #141873) after re-centring of the photons (refer also to Table 1). Not the good coincidence in the inner part with the sum of two Gaussians for both objects. Error bars have been omitted beyond 24 pixels to bring out the difference between white dwarf and quasar halo.

Fig. 6. ROSAT HRI image of 3C 273 from the second observing block before and after re-centring the photons (top). Corresponding images of HZ 43 (#142544) are shown for comparison (bottom). The QSO images are optimised to show the jet, whereas the white dwarf images should best show the circularisation of the PSF.

Table 1. Comparison of the point-spread-function for the white dwarf HZ 43 and 3C 273 in terms of two Gaussian components with widths FWHM1 and FWHM2 (see Fig. 5). Units are sky pixels of .

From this analysis we conclude that there is no discernable difference in the core profile between quasar and white dwarf in the averaged intensity profiles, i.e. the quasar core is unresolved at X-rays. Only in the halo the intensity of the quasar is slightly above the normalized white dwarf profile.

3.2. Fitting the quasar's point-spread-function

The X-ray signal from the jet of 3C 273 is very weak and is located in the wings of the complicated point-spread-function of the bright quasar core (total intensity ratio 400:1). Therefore the X-ray emission from the jet has to be isolated from this underlying point-spread-function background. Using the azimuthally averaged profiles derived above to analyse the point-spread-function is not sufficiently accurate to isolate the jet's weak signal and measure its flux. A better model of the point-spread-function had to be obtained by fitting the signal sampled along concentric circles in four sections by polynomials of order 3, with section boundaries at 30, 120, 210 and 300^o. In this way a satisfactory flat background also interior to the optically visible jet region is achieved (Fig. 7). It should be pointed out that the result of the background modelling does strongly depend on the model parameters. With the current resolution of and the steep intensity gradients towards the quasar core it cannot be completely excluded that the relatively complicated model absorbs a moderately extended component in the inner part of the jet ().

Fig. 7. Background subtracted image of 3C 273. The background underneath the quasar is flat and shows the increased noise due to the intense signal subtracted. The white rectangle indicates the region over which the jet's X-ray emission was summed up. The isophotes clearly show the extent of the X-ray emission all along the jet.

3.3. The X-ray signal from the jet

The count rate of the X-ray emission from the jet was derived from the PSF-subtracted image. In a window encompassing the jet we measured 644 counts. The background in several windows of the same size and at the similar distance to the quasar was measured to be  counts. We therefore deduce a total jet flux of 610 counts in 85068 sec. The same procedure yields 253 counts for the object to the NE of the quasar. To convert this to a flux density we integrated a synchrotron power-law over the effective collecting area of the ROSAT HRI as a function of frequency (effective energy 1.17 keV corresponding to  Hz) including absorption due to a neutral hydrogen column of  cm^-2 (Otterbein 1992) using the cross-sections from Morrison & McCammon (1983) for a solar abundance. To represent synchrotron emission we used spectral indices and -2 and thus obtained an integral flux density of and  nJy for the jet and and  nJy for the object in the NE of the quasar. If we assume a spectral index of (resembling thermal bremsstrahlung) the corresponding values are  nJy for the jet and  nJy for the object in the NE. The error was formally derived from the scatter in the background determination.

Inspection of Fig. 7 suggests that the X-ray emission from the jet is extended in the radial direction all along the visible jet. We therefore rotated the background subtracted image by around the quasar core to sum the jet's signal over the 20 pixel rows encompassing it. The resulting trace along the jet is shown in Fig. 8. This run of the X-ray emission along the jet was analysed in a similar manner as the optical and radio emission (Röser & Meisenheimer 1991): Gaussian profiles with constant width of were placed at the positions of knots A, B, C, D and H ². Their peak was adjusted via a least square fit to represent the trace of X-ray emission along the jet (Table 2).

Fig. 8. Trace along the jet. The thin line represents the background signal due to the quasar, summed over the same area as the jet signal. The thick line is the model fit by 5 Gaussians described in the text.

Table 2. X-ray flux of the jet's components in the ROSAT HRI band (at 1.17 keV corresponding to  Hz).

The strongest X-ray signal originates from the positions of knots A and B, much weaker emission is found further out, possibly out to the hotspot H. Due to the steeply rising quasar background, and to uncertainties in background modelling mentioned above, the onset of the jet's X-ray emission is somewhat uncertain. On the basis of our data no X-ray emission is detected from the jet inwards of knot A.

The X-ray source to the NE of the quasar does not show any optical counterpart on our deep images, neither in the radio nor in the optical. Due to its weakness we cannot say if it is extended or not in the X-rays. Its relation to 3C 273 and its nature remain unknown.

© European Southern Observatory (ESO) 2000

Online publication: July 27, 2000
helpdesk.link@springer.de