Birmingham XMM Guide: Imaging Analysis

Create a mosaiced, exposure corrected image
2-D imaging analysis with PSF convolution

Create a mosaiced, exposure corrected image

• A good start is probably to make a combined image of the data from the 3 cameras, and for this, we can use xmmmosaic.csh. First make a directory called mosaic, move to it, and then link to your cleaned events files and your housekeeping (Atthk) file.

  So in mosaic/ I have:

  xun4> ls
  Atthk.ds@             mos2_clean_evt.fits@
  mos1_clean_evt.fits@  pn_clean_evt.fits@
  

• Now we use xmmmosaic.csh which will make a mosaic of exposure map divided images from the three (or a combination of 2) cameras. We need to choose an energy range to work in, and here I chose 300-8000eV though you may want to choose a more specific energy range, in which you think your source will have the best signal to noise. We also choose a binning factor for the images - the pixels of the pn and mos cameras are 4.1" and 1.1" respectively, but the detected events have their positions randomised within these pixels into 'virtual pixels' of size 0.05". This means that a binning factor of 20, for example, will give image pixels of 1" (but remember that the actual detector pixels are larger than this). Note here that the psf images produced by calview have pixels of 1.1", so if we want to use these later in 2-D fitting then we should match that binning here, i.e. a binning factor of 22. Now we run the script:

  xun4> xmmmosaic.csh pn_clean_evt.fits m1_clean_evt.fits m2_clean_evt.fits atthk.dat 22 300 8000 pm1m2bin22_03_8keV_ | & tee pm1m2bin22_03_8keV_log.txt
  
  --------------------XMMMOSAIC.CSH version 1.3---------------------
  
  Input was /exgal6/bjm/scripts/xmmmosaic.csh pn_clean_evt.fits m1_clean_evt.fits m2_clean_evt.fits atthk.dat 22 300 8000 pm1m2bin22_03_8keV_
  
  
  Making PN image, and generating exposure map...
    used XMMEXPMAP.CSH version 1.1 - log written to pm1m2bin22_03_8keV_pn_expmaplog.txt
  
  Making MOS1 image, and generating exposure map...
    used XMMEXPMAP.CSH version 1.1 - log written to pm1m2bin22_03_8keV_m1_expmaplog.txt
  
  Making MOS2 image, and generating exposure map...
    used XMMEXPMAP.CSH version 1.1 - log written to pm1m2bin22_03_8keV_m2_expmaplog.txt
  
  3 data set(s) in use - renormalising exposure maps so their max value when added together is 1
  Mosaicing images...
    mosaic written to pm1m2bin22_03_8keV_mosaic.fits and normalised mosaic (created with exposure maps normalised so their maximum is 1) written to pm1m2bin22_03_8keV_mosaic_norm.fits
  

  This script creates 2 mosaiced images pm1m2bin22_E03_8keV_mosaic.fits, with units of counts per second, and pm1m2bin22_E03_8keV_mosaic_norm.fits which was created with a normalised exposure map, so still has units of counts. This flattens the image, to account for vignetting, but preserves the Poisson statistics, which is useful when making radial profiles and performing 2D fitting

2-D imaging analysis with PSF convolution

• We can also use sherpa to fit a 2-D surface brightness model to an image. This is useful because the source need not be circular, and we can use psf images to convolve with our model. Make a directory called 2drprof in mosaic/ and link from there to the mosaiced image. We will perform the 2-D fitting in Sherpa, and so that the errors on the fit can be estimated correctly, we need to mulltiply the mosaiced image by the average exposure time. This is because the counts per second per pixel are usually very small (~1e-4) which will cause statistical problems in the error estimation in sherpa. The exposure information can be found in the mosaic header, and then we can use fcarith to scale up the image. All further work is then done on this scaled up image.

  We need to take our mosaiced image, and crop it down into a smaller region. I used a square region, trying to avoid chip gaps, and include most of the emission detected above. I used the CIAO tool dmcopy to crop the image,

  xun4> dmcopy "pm1m2bin22_E03_8keV_mosaic.fits[(x,y)=region(pm1m2bin22_box.reg)]" pm1m2bin22_E03_8keV_boxmosaic.fits
  

  I then use calview to generate psf images for the three telescopes. This cannot be done from the command line - you need to use the GUI

  xun4> calview &
  

  Now enter the details for your observation into the calview window, and click on view -> PSF. This will display a psf image in ds9, then go to file -> save as fits image (in ds9) and save the psf image. Repeat for the 3 cameras. Now I use xmmcomb_psf.csh to combine the psfs of the 3 telescopes by adding the fits images created by calview together, after weighting the PN telescopes psf by a factor of 2 because of its extra counts. The resulting image is then normailsed, so the sum is 1.

  xmmcomb_psf.csh pnpsf.fits mos1psf.fits mos2psf.fits 3psf_norm.fits
  
  --------------------XMMCOMB_PSF.CSH version 1---------------------
  
  Input was /exgal1/bjm/scripts/xmmcomb_psf.csh pnpsf.fits mos1psf.fits 
      mos2psf.fits 3psf_norm.fits
  
  Output combined psf image 3psf_norm.fits created with a sum of 1.000000 counts.
  
  ------------------------------------------------------------------------------------
  

  This psf image can then be used as an instrument model in sherpa to convolve a source model through.

• The best way to do the 2d fitting is to write a sherpa script. There are lots of parameters that we can vary, so it's best to impose sensible upper and lower limits, and fit them in a sensible order so as to find a stable minimum. I estimate the background in counts/pixel from a background region of the image, and fit a flat background + 2D beta profile to my cluster emission. The sherpa script I then use is this (and is available from the scripts page):

  xun4> more 2drprof_sherpascript.txt
  data "pm1m2bin22_E03_8keV_boxmosaic.fits" FITSIMAGE
  
  psffromfile[psfinst]
  psfinst.file="3psf_norm.fits"
  psfinst.xsize=50
  psfinst.ysize=50
  fre psfinst.ypos
  fre psfinst.xpos
  instrument=psfinst
  
  const2d[bck]
  bck.c0=1.3327e-5
  bck.c0.max=5e-5
  bck integrate off
  fre bck.c0
  
  source=bck+beta2d[blob1]
  
  blob1 . ampl . max = 0.01
  blob1.ampl.min=0.00001
  blob1.xpos.max=100
  blob1.xpos.min=70
  blob1.ypos.max=100
  blob1.ypos.min=70
  blob1.alpha.max=2.5
  blob1.alpha.min=1.0
  blob1.r0.max=50
  blob1.r0.min=5
  
  blob1 . r0 = 15
  blob1 . alpha = 1.5
  blob1 . xpos = 84
  blob1 . ypos = 84
  blob1 . ampl = 0.001
  
  blob1 integrate off
  
  statistic cash
  
  fre blob1 . r0
  fre blob1 . alpha
  fre blob1 . xpos
  fre blob1 . ypos
  fre blob1 . ellip
  fre blob1 . theta
  show all
  fit
  fre blob1.ampl
  thaw blob1.xpos
  thaw blob1.ypos
  show all
  fit
  fre blob1 . xpos
  fre blob1 . ypos
  thaw blob1.ampl
  show all
  fit
  thaw blob1.r0
  show all
  fit
  thaw blob1.alpha
  show all
  fit
  thaw blob1.ellip
  show all
  fit
  thaw blob1.theta
  show all
  fit
  thaw blob1.xpos
  thaw blob1.ypos
  show all
  fit
  
  save all mysavefile.shp
  write RESIDUALS myresiduals.fits
  write MODEL mymodel.fits
  
  projection blob1.alpha blob1.r0
  
  ------------------------------------------------------------------------------
  

  You'll need to set the values to be sensible for your data, and note that the running time depends strongly on the size of the image, and the psf. Here my image is about 170x170 pixels, and the psf image size is 50x50 pixels. The PSF image is actually larger than this, but I used the psfinst.xsize and psfinst.ysize parameters to crop it down.

  To run this sherpa script, use

  xun4> nice +19 sherpa 2drprof_sherpascript.txt |& tee 2drprof_sherpalog.txt
  

  After you are happy with the fit, you can use the command image fit which displays the data, model, and residuals in ds9, which you can save as fits images. Alternatively, the write commands shown in the script above create these fits images from the sherpa prompt, without requiring ds9.

Updates

• 29/4/02 - Ammended mosaicing section to reflect changes in xmmmosaic.csh, and the use of normalised exposure map corrected images. Added lines to 2D sherpascript to save the residuals and model.