performances of Fresnel rhomb retarders
Given the large wavelength domain of ESPaDOnS, the first natural idea is to use
superachromatic retarders designed along Serkowsky's ideas, like those manufactured by
However, previous experience with them demonstrated that they generate large amplitude
fringing in the intensity and polarisation spectra and thus drastically reduce the
polarisation accuracy of any potential measurements obtained with them.
We therefore decided to use Fresnel rhombs, that were proven to
be much more achromatic and producing almost no fringing patterns in high resolution spectra.
A bunch of 24 single bk7 rhombs (with birefringence smaller than 0.2nm/cm) was
ordered and constructed along detailed specifications, to construct 8 quarter-wave (single) rhombs and
8 half-wave (double) rhombs, with different thicknesses of MgF2 coating to study the effect
on the rhomb retardance. The rhombs are mounted in a specific barrel filled with helium (to avoid
oxydation of the totally reflecting surfaces) and sealed with a soft joint. A dedicated and fully
automated optical bench was also designed and
constructed to measure the rhombs retardance with an accuracy of 0.1deg. The best rhombs were
selected for the polarimeter modules of ESPaDOnS and NARVAL (the copy of ESPaDOnS in construction
for the 2m Bernard Lyot telescope atop Pic du Midi).
The graph on the right shows the retardance curves of 2 different quarter wave rhombs in the series
(full line: first rhomb; other lines: independant measurements of second rhomb, taken over a few
months). It demonstrates that Fresnel rhombs can be designed and constructed so that their retardance
is nominal to better than 0.3% throughout the whole optical domain, while their
optical axis remains stable to better than 0.1deg. This is much better in particular than superachromatic
waveplates, the retardance and optical axis of which vary by about 2% and 4deg respectively in the same
These curves also demonstrate that the retardance measurement is repeatable to better than 0.1deg, and
does not evolve significantly with time, at least on a timescale of a few months. The optimal thickness
of the MgF2 film deposited on the rhombs to achieve such performances is found to be very close to the
theoretical value (of 24nm for our bk7 rhombs).
Amplitude of fringing paterns
We also estimated the amplitude of fringing patterns induced in polarised spectra by Fresnel rhombs
by taking sequences of flat field exposures in different rhomb azimuths, in exactly the same way as one
observer recording polarisation spectra with ESPaDOnS. The graph on the right shows the achieved signal
to noise ratio as a function of wavelength for one such polarisation sequence. The full line depicts
the signal to noise ratio expected from the number of counts on the ccd detector, and the dash-dot line
the signal to noise ratio measured from the check spectrum (derived from spectra recorded in the same
rhomb azimuths and thus free of any fringing patterns from the rhombs). The dashed line, tracing the
signal to noise ratio as measured in the polarisation spectrum, is almost everywhere at the same level
of the 2 others, except in the infrared, where it shows a small drop of about 5% in signal to noise.
This demonstrates that no detectable fringing patterns are observed in the visible domain while a
weak pattern is observed around 850nm with a typical rms relative amplitude of
less than 0.03%; it confirms in particular the superior performance of
Fresnel rhombs for high resolution spectropolarimetry.
© Jean-François Donati, last update May 10 2004