The status of the GI2T has been described in detail in two recent papers
(Mourard
et al. 1994b; Mourard et al. 1994c). The original beam-combiner of the GI2T was
not equipped with polarizing optics (Bosc 1989). For the present study we have used
the simplest form of a polarimeter consisting of two film-sheet linear polarizers.
The directions of these polarizers were accurately set orthogonal in laboratory,
and cross-talk effects calibrated as less than 3
in the continuum centered on
nm. During our observations we introduced the polarizers alternately,
every 10 minutes on average, at the entrance of the GI2T's spectrograph (after beam
recombination, Fig. 1 (click here), also see Fig. 5 of Mourard et al. 1994b). We
sandwiched the recording of fringes in polarized light between series of data in
natural light. This strategy aims at calibrating the effects of instrumental
polarization on the visibility as a function of hour angle.
It is worth emphasizing that in our polarimetric arrangement
the two directions of polarization, noted 
and 
,
correspond to fixed orientations with respect to the coordinate system of the
GI2T's central beam-combiner, i.e. the output pupil geometry. Actually the
celestial field orientations of incoming beams from the north and south telescopes
of the GI2T suffer a differential rotation which affects the fringe visibility
as a function of the Hour Angle. The analytical expression of this degrading effect
has been formulated by Mourard et al. (1994) and atttains at most 5
within
the HA coverage of the present observations (Rousselet-Perraut et al. 1996).
As discussed in Sect. 2.4, our observations are close enough to the meridian
transit so as the visibilities estimated in 
and 
can
be considered as measured parallel to the North-South direction.
Figure 1: Schematic layout of the optical set-up on the GI2T for observing
Cas in linearly polarized light. Sheet-film polarizers 
and
(parallel and perpendicular to the baseline) are introduced alternately
before the
dispersing gratings (G) of GI2T's spectrograph followed by the
camera-mirror (Mc).
The polarized interferograms are recorded on an intensified photon-counting
detector
(CP40)
We have observed 
Cas (HD 5394, B0.5IVe, 
,
= 0h51mn,
= 
) on September 5th and 6th 1994 with two baselines per night. The
dispersed fringes, centered on 660 nm, were recorded as series of data in the three states of
polarizations, i.e. natural, parallel and perpendicular to the baseline. These data
correspond to 20 ms short exposures from the CP40 photon-counting camera for freezing
the atmospheric turbulence (Blazit 1987). All records were obtained at hour angles within
-33 and +96 mn around the meridian transit in order to minimize the GI2T's instrumental
losses of visibility (Rousselet-Perraut et al. 1996). At each baseline on 
Cas,
the records including the three polarizations were carried during 30 mn on average. We
also observed the standard star 
Cep (HD 203280, A7IV-V, 
,
= 21h51mn,
= 
) for HA between -30 to +80 mn for calibrating our observations on
Cas (Tinbergen 1979).
Table 1: Journal of observations of 
Cas and 
Cep with the
GI2T in September 1994. The columns give from left to right: the star, the date, the
universal time (UT), the average Hour Angle (HA) in minutes respective to the meridian
transit, the baseline projected on the sky in meters and the corresponding spatial
frequency in cycles/arcsec for the continuum wavelength of 660 nm.
We reduced the GI2T data following the method that we have developped for analysing
multi-speckle interferograms of the GI2T in dispersed light (Mourard et al. 1994c).
Since both 
Cas and 
Cep are bright objects, the cross-spectral
density analysis of 20 ms short exposures was prefered to the usual spectral density
for avoiding the so-called photon-centroiding-hole of the CP40 (Foy 1988; Blazit
1987).
In practice we cross-correlated fringes of two neighbouring narrow spectral channels
taken in the continuum at 
nm and 
nm. The spectral bandwidth per channel was 
nm in the range
of the correlation limit for long baseline interferometry (Roddier 1981). We
computed the cross-spectral density by Fourier transforming the average cross-correlation
as:
where <> refers to the ensemble average, * to the complex conjugate and
to
the spatial frequency, with B the baseline and 
the wavelength.
Provided 
, we can assume that
(Petrov et al. 1986), thus:
where 
denotes the spatial power spectrum of the source
and 
the interferometric transfer function. We obtained the visibility estimates of our program
stars by normalizing the high frequency energy to the atmospherically unbiased low frequency
energy (Mourard et al. 1994c).
Figure 2: Calibrated visibility points of 
Cep for different states of polarization
(natural light: squares, parallel to the baseline: diamonds, perpendicular to the baseline:
triangles) as a function of hour angle. The typical error per visibility point (
5%)
is given at the left-bottom
After reduction, we corrected raw visibilities at different baselines for the difference
of transmission in the interferometric arms of the GI2T from actual fluxes recorded by
the fine star-guider. We also corrected for the effect of differential field
rotation between the afocal beams of the telescopes using the analytical expressions of
the visibility loss versus the Hour Angle (Mourard et al. 1994a). This should enable us
to detect an eventual variation of visibilities of 
Cas during earth-rotation
related to its intrinsec polarization. Figure 2 (click here) displays the corrected visibilities
of 
Cep versus HA where a slight trend was detected for increasing HA.
This trend has also been noticed for other stars observed with the GI2T (Harmanec et al. 1996)
and is believed to come from systematic instrumental effects of the interferometer.
Using a linear model (Fig. 2 (click here)) we correct for the decrease of visibilities on
Cas from those of 
Cep. The residuals are on the order of
1-2%, which is
well below the atmospheric and photon noise of 5% obtained on the visibilities during this
run. Therefore and in order to push further our analysis we decided to average, when possible,
visibilities of a given polarization for a same baseline regardless of HA. Reduced error bars
enabled us to check for partial resolution of 
Cas in different polarizations for
increasing baselines (Table 2 (click here)).
Table 2: Calibrated visibilities in the continuum light of 
Cas as a function of
projected baseline on the sky (Col. 1) from GI2T observations. Each visibility is obtained
by summing individual measurements at different hour angles. Visibilities are given in %
(
) for
3 different states of polarization: natural (Col. 2), linear parallel to the baseline
(Col. 3) and linear perpendicular to it (Col. 4)