The Overall Mechanics

For the example below of the mosaicked image of the galaxy M101, the most simple directory structure is as follows:

/M101/merge
/M101/0104260101/odf
/M101/0104260101/proc
/M101/0164560701/odf
/M101/0164560701/proc
/M101/0212480201/odf
/M101/0212480201/proc

All three observations of M101 are significantly contaminated by SP flaring, 0164560701, however, is a truly pathological case where even after light-curve filtering there remains a considerable amount of residual SP contamination. We use all three observations as an illustrative example. Figure 32 displays the light curves with their accepted time intervals (an output of espfilt) for the three observations.

Figure 32: Light curves of the M101 observations.

The merging processing will take place in the /M101/merge directory. Within the merge directory is an ASCII file listing the paths and prefixes of the individual exposures to be merged which is named mosaic.list:

../0104260101/proc/mos1S001
../0104260101/proc/mos2S006
../0104260101/proc/pnS003
../0164560701/proc/mos1S001
../0164560701/proc/mos2S002
../0164560701/proc/pnS003
../0212480201/proc/mos1S001
../0212480201/proc/mos2S002
../0212480201/proc/pnS003

(The appendices use a slightly different structure that is more consistent with the other reduction scripts.) The following shows the simple set of commands to mosaic the exposures processed for the $0.4-1.3$ keV band to create the image in the top panel of Figure 33:

setenv RAPNT 210.8
setenv DECPNT 54.32

mosaicmerge dirfile=mosaic.list type=1
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=2
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=3
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=4
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=5
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=6
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'
mosaicmerge dirfile=mosaic.list type=7
coordsys=2 crval1=$RAPNT crval2=$DECPNT
pixelsize=4.163999e-2 alpha=1.7 elow=350
ehigh=1100 withcheese=yes cheesemasktype='t'

In the above example, the individual components for an image of the $0.35-1.31$ keV (elow=350 to ehigh=1100) band are cast into a mosaic using the task mosaicmerge. The coord=2 selects the coordinate system to be used (1 would be ecliptic coordinates, 2 would be equatorial coordinates, and 3 would be Galactic coordinates), and crval1=210.8 and crval2=54.32 are the central coordinates of the mosaic projection. Note that in order to reduce problems with typos, I put the coordinates into environment variables, and then reference the environment variables in each call to mosaicmerge.

Figure 33: Merged images of M101 in the $0.4-1.3$ keV band (top) and $2.0-7.2$ keV band (bottom). While the low-energy band looks quite reasonable, the high-energy band shows the effect of uneven residual SP contamination. These images were made with an earlier version of ESAS with $0\farcm03$ pixels.

The pixelsize=0.0416 gives the scale size of the projection in units of arc minutes. A few notes on pixel sizes are appropriate. The “native” pixel size for the ESAS routines cheese, mosspectra/back, and pnspectra/back is $2\farcs5$= $0\farcm041666668$. In the above example, I stuck with that pixel size. The arrays into which the mosaic is cast are $2000\times2000$ pixels in size (currently hardcoded), making an image $1\hbox{$.\!\!^\circ$}38\times1\hbox{$.\!\!^\circ$}38$ which is plenty for M101. Compared to some projects to which ESAS has been applied, M101 is relatively bright, so this pixel size is not ridiculous. Much fainter objects would not have sufficient counts per pixel to have reasonable signal to noise, and so larger pixels would make sense. Thus the limit on the size of the output image is not (generally) an issue. We can, however, imagine projects for which it could be a problem, and are considering what to do about them.

To build a rate image from the components one uses binadaptmerge, which works much like binadapt.

binadaptmerge elowlist=350 ehighlist=1100
withpartbkg=yes withspbkg=true withswcxbkg=true
withmask=true maskfile=mosaic-cheeset.fits
withbinning=false withsmoothing=false fill=0

In this case, the components are combined to create a rate image without binning or smoothing. To bin the output image one sets withbinning=true and binfactor equal to the linear number of pixels to be binned. To adaptively smooth one sets withsmoothing=true and smoothcounts to the number of counts to accumulate for smoothing. One can also control the inclusion of partially exposed region by setting maskthresh to remove regions where the mask is below some value.

Finally, fill controls infilling. If fill is not equal to zero the program will loop through the data, find pixels with no data (a blank pixel), and if the blank pixel has at least three adjacent non-blank pixels the blank pixel is set to the average of all of the adjacent non-blank pixels. The parameter controls how many of these passes will be made. This is purely a cosmetic process to make prettier pictures. However the smoothed images shouldn't be used for scientific analysis anyway.

If the processing of the individual observations had been done in two bands which span the energy range, $0.4-0.75$ keV and $0.75-1.3$ keV, the data can be combined with elowlist='400 750' and ehighlist='750 1300'.

Figure 34: Merged exposure map of the M101 observations in the $0.35-1.25$ keV band.

The mosaics show the effects of both uneven residual SP contamination and significantly different exposure times over the field. The different residual SP contamination is most easily seen at the right-hand edge of the hard band image. Because there is typically many fewer cosmic background events at higher energies, the particle background is relatively much brighter, and the typical hardness of the SP background, the hard band is much more sensitive to residual uncertainties in the SP background determination. Note also that the higher-intensity crescent in the hard band also has a smoother appearance in both bands. This is an artifact of the adaptive filtering algorithm when the exposure time (see Figure 34), and therefore number of counts, is relatively non-uniform.

Figure 35 shows the spectacular MOS mosaic of the Coma Cluster in the $0.8-1.25$ keV band. 28 observations from both MOS instruments are included in the mosaic.

Figure 35: Merged image of the Coma Cluster in the $0.8-1.25$ keV band.