The X-ray survey in HS 47.5/22 consists of 48 pointings with the ROSAT PSPC
(Pfeffermann et al. 1986). These pointings form a diamond shaped grid
with 
and 
. The observations
were performed between April 1991 and October 1993, and an observation log is
given in Table 4. Figure 1 (click here) gives a schematic view of the field.
Figure 1: Map of the field 47.5/22. Closed contour: the area surveyed with ROSAT,
open contour: 
(Effelsberg observations), +: centres of ROSAT-pointings, 
:
optically selected quasars within the ROSAT area, 
: other optically
selected quasars. Coordinates are J2000.0
Most pointings were observed in more than one interval (OBI). The exposure
times of single OBIs range from 
to 
s.
in Table 4 is the sum of all OBIs at a given position,
the distribution of is presented in Fig. 2 (click here).
Figure 2: Distribution of exposure times of individual pointings. The
vertical line marks the median value
Since the diameter of the field of view (FOV) is 
, the pointings
overlap noticeably, and all but the outer rim of the field is covered by
several observations. Adding up all pointings gives a resulting exposure time
for 73
of the area and more than 
for
the central 
(see Fig. 3 (click here)).
Source detection on the individual pointings was done with the EXSAS software package, version 93JAN (Zimmermann et al. 1993), which uses a sliding window technique. Images were accumulated in the broad (0.1 - 2.4 keV), hard (0.4 - 2.4 keV), and soft (0.1 - 0.4 keV) bands with a pixel size of 15''.
The first step of the source detection took the background from a region
directly surrounding the detection cell, the second took the background from
the cleared and smoothed image. The third step applied a maximum likelihood
analysis to the sources found previously. It computed accurate positions and
their 
errors, vignetting corrected source counts and count rates,
and likelihood of existence, 
(Cruddace et al. 1988).
was used as detection threshold; this value corresponds
to a probability 
that the observed
distribution is a pure background fluctuation, and is equivalent to a signal
to noise 
.
Each band was searched separately to allow the detection of faint sources with
extreme spectra. Detections from different bands were considered identical if
their positions differed less than 
, 
being the larger
position error. The entry with the largest 
defined the position.
After the first pointings were analysed it became clear that some sources had been missed. Most of them were located in the outer part of the FOV. An explanation would be that the detection cell in the first step had been too small, and the sources' PSF leaked into the background extraction area, causing an overestimated background and subsequent non-detection in the next steps. Although possible, this cannot be the only explanation. Tests with varying detection thresholds in the first detection step showed that the differences of the resulting background maps are almost entirely within the noise. So, background contamination is less serious than first assumed, leaving only the possibility of software failure.
Lower detection thresholds resulted in nothing but the "detection'' of random
fluctuations, so each pointing was inspected by eye. Undetected spots
were marked and their positions used as starting point for another run of the
maximum likelihood analysis. Sources above the threshold were added to the
lists, about five per pointing, some of which have noticeably
above the threshold. A repetition of the source detection with the next EXSAS
version (i.e. 94JAN) on several test pointings brought forth all previously
missing sources. This indicates that indeed software problems caused their
non-detection.
The source lists from the various pointings were merged into a single
catalogue. When the positions obtained from different pointings were
compared, the matching criterion was found to be insufficient
because errors of the pointing direction itself could contribute. Therefore,
two sources detected in different pointings are considered identical if
their distance was less than 30''. If the distance was between 30'' and
45'', and at least one source had an off-axis angle 
,
they were considered to be the same as well because of the growing PSF. If at
least one source was found at 
, the distance had to be
larger than 55'' before two detections were considered as two real sources.
Again the entry with the highest determined the position.
Sources found at 
were excluded from the catalogue. The
PSF is too large, and vignetting too effective, to allow the detection of any
but the brightest sources, and these will presumably be blended.
Sources partially shadowed by the window support structure might show even
larger (
) displacements and yet be multiple detections. Since these
cases are very hard to parametrize, the source list was displayed on each
pointing and the dubious detections were checked by eye. The resulting
catalogue, later referred to as "individual catalogue'', contains 489 sources.
in other pointings. In extreme cases, merging these parts might
lead to non-detection.
The average on-axis background count rate is
. With that,
the count rate of a source with given S/N can be computed. A circular
extraction area with radius 1', as is appropriate for on-axis sources,
and an exposure time 
lead to an 
on-axis count rate of
When this source is shifted to larger , the effective exposure
time is reduced due to vignetting. At 
, a reduction of
a factor 0.76 is reached. Simultaneously, the PSF grows, and the extraction
area has to be larger, thus containing more background. An extraction circle
with 
diameter is appropriate at . With
source and background count rates as before, and 
,
one finds 
, clearly not detectable.
If the on-axis and off-axis detections are added together, the resulting
S/N is
If an on-axis value of 
is required, the value is
, which corresponds to the required limit of
, and the source is therefore barely detectable. The
addition of even larger off-axis angles will lead to non-detection.
In the real merged data, more than two pointings with different exposure times
will usually overlap, and various contribute to a certain
position, so that a source is less affected than in the above estimation.
However, regions with will definitely not contribute
to an enhanced sensitivity and are therefore removed in order to obtain a
reasonable point source sensitivity. Even so, sources found close to the
threshold in single pointings may be lost, depending on the actual pointings
and off-axis angles which contribute to that particular position.
EXSAS provides a merging procedure for the relevant files (i.e. photon events table, eventrates table, attitude table, see Zimmermann et al. (1994) for description) which includes re-calculations of all sky coordinates with respect to a new common "detector centre''. This procedure was applied after the outer 20' of each FOV were removed.
Source detection on the merged data was done with the 94JAN version of EXSAS (Zimmermann et al. 1994). Standard procedures were used when possible, but some adjustments were necessary.
The soft and broad bands were the same as before, whereas the hard band had changed to 0.5- 2.0 keV to keep more safely away from the carbon K edge and from the not so well calibrated effective area region above 2 keV. Since the effective area in the now omitted regions is very small, the loss in photon number was unimportant.
The enhanced statistics allowed a further sub-division of the hard band into the 0.5 - 0.9 keV (h1) and 0.9 - 2.0 keV (h2) bands. Since the diffuse X-ray background is most important below 0.5 keV, this enhanced the possibility of finding faint X-ray sources with hard spectra which are invisible in the soft or broad band.
Figure 3: Merged broad band exposure map of the whole field. The contours mark
exposure times of 1 (outermost dotted contour), 5, 8, 13, 20 and 25
(innermost solid contour) ksec, from outside to inside. The axes are in
arcseconds, with 
(J2000.0) as zero
The whole image had to be searched now instead of a circular FOV, and the
PSF size depended no longer on the distance to the detector centre.
Determination of source or background photons had to be weighted by exposure
time since strong gradients occured. For that purpose, merged exposure maps
were created for the five energy bands. Figure 3 (click here) shows the broad band
exposure map of the total field. 
was reached
for 
of the area, and 
for the central
2.3deg2.
Apart from these adjustments, the source detection was performed as described
in the previous section. The automatic count rate calculation was unable to
handle the complex spatial structure of the merged data. Instead, the exposure
time for each source was read directly from the relevant
exposure map. This, together with the number of counts 
from the
maximum likelihood analysis gave the count rate
. Since the exposure maps were already
vignetting corrected, no further correction was necessary.
Each source list had to be checked for spurious detections at the pointing
borders. When the lists from the five bands were merged, a minimum distance
of 
was required. 410 X-ray sources are present in the resulting
catalogue, the "merged list''. 56 have been detected in the h2 band, and
4, 132, 16, and 202, in the h1, hard, soft, and broad band, respectively.
Detection in one band does not necessarily mean that a source is invisible
in the others, on the contrary, most sources show 
in
at least two bands.
in the range
. Due to the complicated PSF structure
in the merged data, a minimum distance of 40'' was required for two sources,
regardless of of the single pointing detection. 320 pairs had
below that limit. The remaining six had
; they were inspected individually to determine
whether they are double detections or indeed different sources.
One pair turned out to consist of two real sources. The merged data entry is a
h2 band detection, in the single pointing visible only as a faint extension
of the other. Four pairs were multiple detections of the same sources. The
large 
were caused by shadowing by the window support
structure.
In the last case, the merged list detection is located between two sources
found in the same individual pointing, separated by 
. Both sources
were of similar magnitude in the detection pointing. In the neighbouring ones,
however, the northern source had disappeared, and its average count rate in
the merged data was low enough to allow a displacement of nearly 2' by the
influence of the southern source. Therefore, the merged data detection was
rejected. Figure 4 (click here) shows the distribution of for
the accepted double detections.
Figure 4: Distribution of position differences of X-ray
sources detected both in individual pointings and the merged data
After the deletion of all double entries, the catalogue contains 574 X-ray
sources with broad band count rates between
and
. Some of the hard band detections were not
detected in the broad band. For these, the hard count rate is kept. 324
sources are found in both detection modes, 165 in individual pointings only
and 85 in the merged data only. 63 of the latter are found in either of the
hard bands with count rates too low to be detected in single pointings.
Of the 165 sources which are detected only in individual pointings, 133 are
faint (
) objects detected
in the central parts of the pointings (i.e. 
). These
are lost when the less sensitive outer parts of the neighbouring pointings are
added. The others are located near the borders between pointings or at the
edge of the entire field. Variability increases the chance of a source being
missed. As was shown by the above example, the average count rate of even
rather bright sources might be too low for detection. These results show the
necessity of searching both the individual pointings and the merged data. Each
mode finds a noticeable amount of sources not detectable in the other.
The final X-ray source catalogue is available as an ascii-file via anonymous
ftp at ftp.hs.uni-hamburg.de in the subdirectory
pub/outgoing/kmout/data. It is also available at the CDS via anonymous ftp
to
cdsarc.u-strasbg.fr (130.79.128.5) or via
http://
cdsweb.u-strasbg.fr/Abstract.html.
The structure of the table is explained in
Appendix.
and , derived with the
formulae in Sect. 2.2.1 (click here) are included in Table 4.
In the merged data, the net exposure time varies between 
at the edges and 
ksec in the central parts. The limit in the outermost
region is determined by the shortest individual pointing which does not
entirely overlap with others (WG 700149 in Table 4). Successive
subfields with increasing minimal net exposure time are then selected, the
average number of background counts 
inside the PSF is measured,
and the number of source counts 
necessary to give
is computed. With an average column density
, and the assumption of a power law
spectrum with energy index 
, the limiting
count rates correspond to fluxes between
and
in the total ROSAT
band. The limits become noticeably fainter for the hard band because the
background is much lower above 0.4keV. The hard count rate and flux limits
are included in Table 1 (click here). It has to be kept in mind, however, that
the flux limits in the table are approximations obtained with the assumption
of one average spectrum.
 | area |
 |  |  |  |
| [sec] | [ ] | ||||
 | 11.5 | 21.0 | 22.3 | 13.0 | 8.0 |
| 2000 | 10.8 | 16.4 | 17.4 | 10.6 | 6.5 |
| 5000 | 8.3 | 8.5 | 9.0 | 4.2 | 1.9 |
| 10000 | 6.0 | 5.4 | 5.7 | 2.6 | 1.2 |
| 15000 | 4.1 | 4.2 | 4.5 | 2.0 | 0.9 |
| 20000 | 2.4 | 3.5 | 3.7 | 1.7 | 0.8 |
| 25000 | 0.3 | 3.2 | 3.4 | 1.5 | 0.7 |
a: values derived
from pointing  1 in Table 4. | |||||
mean limiting fluxes in the broad and hard band, respectively, in
units of 
. Count rates are given in
