NARVAL
Polarisation analysis


Polarimetric design

The polarimeter design is the same as that of ESPaDOnS (see corresponding ESPaDOnS page). It consists of one fixed quarter-wave retarder sandwitched between 2 rotating half-wave retarders and coupled to a Wollaston beamsplitter, all working at infinite conjugate ratio (thanks to 2 triplets performing the required f/25 to f/3.6 focal ratio transformation required by optical fibers). The polarimeter yields continuum subtracted linear or circular polarisation spectra of the stellar light (when used in polarimetric mode).

Fresnel rhombs

Rather than using cristalline plates, NARVAL, like ESPaDOnS, is analysing the radiation polarisation using Fresnel rhombs coated with MgF2 to make them almost perfectly achromatic throughout the whole wavelength domain. The curve on the right shows, as an example, the retardance of the quarter-wave rhomb of NARVAL (measurement: dashes; fit: full line), whose retardance is nominal to within 0.4° troughout most of the spectral range (and there is obviously still room for improvements). This is far better than what usually achieved with the so-called 'superachromatic' cristalline waveplates (designed after the Serkowsky/Pancharatnam design), typically achieving retardation deviating from nominal by as much as ±2°.
The second advantage of rhombs is that the optical axis is constant with wavelength, as opposed to the 'superachromatic' cristalline waveplates whose optical axis varies by as much as a few °.
Last but not least, Fresnel rhombs produce undetectable fringing in the resulting high-resolution spectra (down to a rms precision of about 0.05%), as opposed to superachromatic plates known to induce fringing patterns with an amplitude of as much as a few %, ie usually one magnitude or more larger than the signal we aim at detecting.

At the moment, the half-wave rhombs of NARVAL are not as good as the quarter-wave rhomb as a likely result of a coating problem. This problem will be fixed soon.

Crosstalk between polarisation states

Birefringence stress on the entrance triplet, eg caused by cold weather contracting its metal barrel and squeezing the inner optics, can produce crosstalk between polarisation states. Prior experience with ESPaDOnS showed that this crosstalk can reach as much as 20% in the worst cases.
This crosstalk is small on NARVAL, of order 1% at most (about the detection threshold).

As an example, a portion of the spectrum of the chemically peculiar star 53 Cam, well known for hosting a very strong magnetic field, is shown on the right, as seen by NARVAL in all polarisation states (from bottom to top, intensity spectrum or Stokes I in black, circular polarisation spectrum or Stokes V in red, and linear polarisation spectra or Stokes Q and U in green and blue - note the different expansion factors used for the different Stokes parameters). Linear polarisation signatures are detected in some lines, but their shapes and aplitudes do not correlate with that of circular polarisation signatures, as would expected if circular polarisation was significantly leaking into linear polarisations. In particular, the lines at 501.57 and 502.17 nm, usually displaying strong circular polarisation signatures but very little linear polarisation signatures, show indeed no detectable Stokes Q and U signatures in the spectra secured with NARVAL.


© Jean-François Donati, last update 2007 Jan 20