The Swift XRT (Burrows et al 2005, SSRv, 120, 165)
is a sensitive, broad-band (0.2 - 10 keV) X-ray imager
with an effective area of about 125 cm**2 at 1.5 keV.
The 600 x 600 pixel CCD at the focus provides
a 23.6' x 23.6' field of view with a pixel scale
of 2.36". The point spread function
is 18" (HPD) at 1.5 keV.
These XRT surveys represent the data from the first 12.5 years of Swift X-ray observations.
They include all data taken in photon counting mode. A total of just over 8% of the sky
has some non-zero exposure. The fraction of sky exposed as a function of the exposure is given
in the following table:
Exposure>0 10 30 100 300 1000 3000
1000 30000 100000300000
Coverage
8.42 8.37 8.29 7.67 7.29 5.68
3.40 1.26 0.35 0.044 0.00118
The individual exposure and counts maps have been combined
into a Hierarchical Progressive Survey (HiPS) where the data are stored in tiles
in the HEALPix projection at a number of different resulutions. The highest resolution
pixels (HEALPix order 17) have a size of roughly 1.6". Data are also stored at lower
resolutions at factors of 1/2, 1/4, 1/8, 1/16, and 1/32, and in an all sky image with a resolution
1/256 of the higest resolution. An intensity map has been created as the ratio
of the counts and exposure maps.
These surveys combine the basic count and exposure maps provided as standard products
in the Swift XRT archive in obsid/xrt/products/*xpc_(sk|ex).img.gz. The surveys were created as follows:
All of the exposure maps available in the archive in mid-May 2017 were combined
using the CDS-developed Hipsgen tool. This includes 129,063 observations for which
both count and exposure files were found in PC mode. Three exposures where there was
a counts map but no exposure map were ignored. A few exposure files had more than
one exposure extension. 1,082 files had two extensions and 1 file had 3 extensions.
The 1084 HDUs in extensions were extracted as separate files and included in the total exposure.
The value of 0 was given to the Hipsgen software as the null value for the FITS files.
This caused the CDS software to treat such pixels as missing rather than 0 exposure.
The counts data was extracted from the counts maps for each observation using SkyView
developed software. For any pixel in which a count was recorded, the corresponding
exposure file was checked and if there was any exposure (in any of the associated
extensions), then the count was retained. If there was no exposure in any of the extensions
of the corresponding exposure file, the counts in the pixel were omitted. Once a count
was accepted, the overlap between the counts map pixel and the pixels of the corresponding
HiPS tile (or tiles) was computed. Each count was then assigned entirely to a single
pixel in the HiPS tile randomly but with the destination pixel probabilities weighted by area of
the overlap. Thus if several pixels were found in a given counts map pixel they
might be assigned to different pixels in the output image. The HiPS pixels (~1.6") used were
of substantially higher resolution than the XRT resolution of 18" and somewhat higher than
the counts map resolution of 2.36".
A total of 183,750,428 photons were extracted from the counts maps while 15,226 were rejected
as being from pixels with 0 exposure. There were 501 pixels which required special
treatment as straddling the boundaries of the HEALPix projection.
The resulting counts tiles were then clipped using the exposure tiles that had been
previously generated. Basically this transferred the coverage of the exposure tiles
to the counts tiles. Any counts pixel where the corresponding exposure pixel was a NaN
was changed to a NaN to indicate that there was no coverage in this region.
During the clipping process 137,730 HiPS level 8 were clipped (of 786,432 over the entire sky). There were
12,236 tiles for which there was some exposure but no counts found. During the clipping process
2 photons were found on pixels where there was no corresponding exposure in the exposure tiles.
This can happen when the pixel assignment process noted above shifts a photon just outside the
exposed region but should be -- as it was -- rare. These photons were deleted.
After creating the clipped level 8 counts maps, level 7 to 3 tiles and an all sky map
where generated by averaging pixels 2x2 to decrease each level.
When adding the four pixels in the level N map together
only pixels whose value was not NaN were considered.
Finally an intensity map was created by dividing the counts tiles by the exposure tiles.
To eliminate gross fluctuations due to rare counts in regions with very low exposure, only
regions with exposure > 1 second were retained. A total of 30 photons were deleted due to
this criterion.
Note that while any sampler may in principle be used with these data, the Spline sampler may
give unexpected results. The spline computation propogates NaNs thought the image and
means that even occasional NaNs can corrupt the
output image completely. NaNs are very common in this dataset.
Also, if the region straddles a boundary in the HEALPix projection,
the size of the requested input region is likely to exceed memory limits since the HiPS
data are considered a single very large image. Provenance: Data generated from public images at HEASARC archive. This is a service of NASA HEASARC.