Polarimetry

Polarization measurements are made using a half wave plate to rotate the plane of polarization and a wire grid analyser (grid with 0.4 m spacing on a KRS5 substrate). A CdS waveplate is used for 8 - 13 m observations, and a CdSe waveplate is used at 20 m. The waveplate is mounted outside the cryostat above the dichroic mirror, and is is rotated to four positions in the beam (corresponding to position angle rotations of 45) and an intensity measurement I is made at each position angle . The waveplate is not achromatic and a wavelength dependent correction is applied to the polarization determined. This correction is least at 10 m and increases to shorter and longer wavelengths. The wiregrid analyser is cold and mounted on the CVF wheel. The Stokes linear polarization parameters (Q, U) are calculated from differences in orthogonal polarization planes:
Q = I - I = Q1 - Q2
U = I - I = U1 - U2
I = (I + I + I + I )/2 = (Q1 + Q2 + U1 + U2)/2
The quantities Q1+Q2 and U1+U2 are independent measures of the intensity, and provide a valuable check on data consistency. The circular polarization V is generally negligible.

Figure 4 presents the real-time polarimetry reduction screen showing a field from Sgr A.

 
Figure 4: Image of the real time polarimetry reduction screen, showing a field from Sgr A (Galactic Centre).

The top row of images shows the latest integration in each of the waveplate positions, designated Q1, Q2, U1, U2. Each of these images represents about 1 second of integration. Each of these images is coadded separately, allowing for beamswitching and shift-and-add of images if required. The middle row of images shows the running coadd of Q1, Q2, U1, U2. Each of these images is still an intensity image at one of the waveplate positions. The bottom row of images show the current status of the images in the Stokes parameters Q=Q1-Q2, U=U1-U2, and I = (Q1+Q2+U1+U2)/2, and is the total intensity image. Seeing these images allows one to appreciate the need for high signal-to-noise images, as the Q and U Stokes images are differences of intensity images. The panel in the bottom right corner provides numerical readout from the region around the crosshair in the intensity image. Final images of polarized intensity P = and position angle = arctan(U/Q) are calculated and displayed at the completion of the observing sequence. Should it be necessary to re-reduce the data at a later stage all of the instantaneous Q1,Q2,U1,U2 images are saved to disk. As there are no suitable packages available for dealing with reduction of polarimetric images we have developed a full suite of reduction programs to cope with this most demanding reduction process.

One of the limiting factors in making polarization measurements by this sequential method is that the four intensity values are measured sequentially. This means that small temporal fluctuations, (e.g. seeing, weather and tracking errors) become significant when differences are taken. To avoid this problem, we would like to image the orthogonal polarization planes simultaneously. This would be best done with the infrared counterpart of a Wollaston prism, and we are currently investigating such a device, but unfortunately there seem to be few sufficiently birefringent materials at mid-infrared wavelengths, and there are also difficulties optically mating the two prisms of a Wollaston prism.

The signal to noise ratio required to obtain a polarimetric image depends on the degree of polarization in the source, and the precision required. At mid-infrared wavelengths polarizations are generally low, the maximum ever observed is in BN at 12 %, and polarizations of only 2% are common. To obtain a polarimetric image to 1% polarization means that we need a signal to noise ratio (SNR) of 140 in each of the four images Q1, Q2, U1, U2, so that the Q and U difference images end up with a SNR = 100. This results in a intensity image with a minimum SNR of 280 in the faintest region that polarimetric information is required. We also often measure the polarization to less than 1% resulting in extremely high quality intensity images. The instrumental polarization is measured to be < 1%. Under ``normal'' operating conditions the polarimetric accuracy achieved is limited only by the signal to noise from the source, though systematic effects dominate in marginal operating conditions, limiting the minimum detectable polarization to a few percent.

Calibration of the position angle is effected by two means. An external wire grid polarizer can be placed in the beam to artificially polarize a beam at a known angle and the position angle calibration made from this. Once this ``laboratory calibration'' is made to determine the true polarization position angle in a few ``standard sources'' it is possible and more convenient in an observing environment to obtain position angle calibrations from these astronomical sources. The favoured position angle standard is the BN object in Orion because it is bright and highly polarized, making for quick calibration. The position angle calibration is considered good to about 1.

© Copyright Astronomical Society of Australia 1997