Polarimetry data reduction

Polarimetry observations are similar to map and photometry observations except that they are broken into measurements of a number of integrations (usually 1) for different positions of a half-wave plate. The wave plate normally steps in 22.5 degree increments. The initial data reduction scheme is identical to standard reduction except that the remip task must be run after extinction to remove the instrumental polarisation. For map observations, the intrebin task must then be used to generate an image for each integration (in practice this means an image per wave-plate position). intrebin ensures that the sky rotation angle and the waveplate angle are stored in the FITS headers (using the ANGROT and WPLATE keywords respectively). At this point the images can either be processed by scripts to calculate the Q and U images¹² or use the POLPACK data reduction system (version 2 or higher) which fits a sine wave to each pixel in the input images. The following example uses POLPACK and is similar to the approach used by the ORAC-DR [27] polarimetry recipes [28].

Assuming the output of intrebin is stored in file omc1_reb.sdf (remembering that this file will contain an image per waveplate position named .i1, .i2 etc. There are 16 images in this example), POLPACK must first be told where to find the rotation angle and waveplate position:

% polimp table=$SURF_DIR/polimp.scuba omc1_reb
  16 input images to process...

  Processing 'omc1_reb.I1'
     Setting WPLATE to -2.5
     Setting ANGROT to 70.27795
     Setting IMGID to 'omc1_reb.I1'
     Setting FILTER to '850_-2.5'

<cut intervening information>

  Processing 'omc1_reb.I16'
     Setting WPLATE to 335
     Setting ANGROT to 75.46666
     Setting IMGID to 'omc1_reb.I16'
     Setting FILTER to '850_335'

SURF provides a suitable import table.

The next stage is to generate the I, Q and U images from these individual waveplate images. This can be done by the polcal task directly:

% polcal weights=3 ilevel=2 omc1_reb
   Processing 16 images in single-beam mode...

OUT - Output Stokes cube > omc1_cube

 Iteration: 1...
   Total number of aberrant input pixels rejected: 199

 Iteration: 2...
   Total number of aberrant input pixels rejected: 219

 Iteration: 3...
   Total number of aberrant input pixels rejected: 219

 Iteration: 4...
   Total number of aberrant input pixels rejected: 219

 None of the output pixels failed the test specified by parameter MINFRAC.

In this case, we use WEIGHTS=3 to generate the variance information from the fit since the SCUBA variances are unreliable, although in many cases these variances are not under-estimated. polcal can combine images from separate overlapping fields all in one go if desired. If the intention is to mosaic separate fields within polcal intrebin should be run with the TRIM parameter set to some non-zero value to prevent problems with edge effects during the mosaicing. Also, polcal expects all the images to be referenced to the same pixel origin. This can be achieved using the KAPPA wcsalign command but it is easier to do this in intrebin by making sure that all images are regridded relative to the same RA/Dec centre - the resulting images will all be aligned to the same pixel grid.

Figure: Polarisation E vectors around OMC-1 at 850 microns.

In many cases, a more reliable approach to calculating the IQU cubes with satisfactory variance information is to use the polstack task on a set of 16 images and use that to generate 4 images with associated variances. These variances are determined directly from the data rather than from the fitting. We have found that running polcal on the resulting 4 images (with WEIGHTS=1 since we now wish to use the supplied variance information) and then mosaicking the resultant IQU cubes (e.g. via CCDPACK makemos) to improve signal-to-noise gives the most robust results and provides variance information that agrees with theory.

Once the IQU cube is made polvec can be used to generate the vectors:

% polvec omc1_cube
CAT - Output catalogue > omc1_cat

   2530 vectors written to the output catalogue.

This catalogue can then be binned using polbin, sections selected using catselect(part of CURSA) and plotted using polplot. An example image of OMC-1 can be seen in Fig. .

For more information on POLPACK see SUN/223 [29].

For photometry observations the output from scuphot can be exported to a single-pixel data reduction system, or alternatively, processed as under-sampled images and reduced through POLPACK as described above.