The phase dependence of the photospheric scattering was first examined
by using the reference-model parameters but with no stellar
wind. These models (in common with those reported by Schmid 1992) have
all Raman photons produced at the red-giant surface. The
polarization,
and integrated line intensities,
were measured for models with binary
separations of 
, 5, and 10, at viewing angles ranging
between 0
and 180
. Plots of the line polarizations and
intensities against viewing angle are given in
Fig. 11 (click here).
As expected, the line intensity is a maximum at 
, the
system then being viewed along the line of centres with the
OVI source `in front'.
As the viewing angle increases the scattered
intensity decreases, principally because of the decreasing area of
the scattering surface which is visible. No counts are observed at
viewing angles
as the scattering surface is then totally occulted by the stellar
disk.
The polarization also follows the expected pattern. At
the line polarization is zero, to within the model
errors. As the viewing angle approaches 
the
polarization reaches a maximum, and then returns to zero.
The viewing angle for maximum polarization is noticeably less than
90
for the 
, a characteristic also noted by
Schmid (1992).
This because the
polarization maximum occurs when the bulk of the photons are scattered
through an angle of 
, and when the source is close to the
photosphere this situation occurs at viewing angles of less than
.
Raman-line profiles were
computed
for a range of viewing angles
for the reference model and the ionized-wind models. The
orbital dependence of the integrated
line properties are shown in Fig. 11 (click here),
and the polarization
spectra for the reference model shown in Fig. 12 (click here).
The density structure of the wind in
these
models
means that very little scattering occurs in the photosphere,
and the bulk of the Raman-line flux is produced within one binary
separation of the photon source (see Fig. 4 (click here)).
The phase dependence of the line polarization in the wind models
resembles that of the photospheric-scattering models, with the
polarization maximum occuring near 
, but
the
extent of the Raman-scattering region means that the simple
reflection-like dependence of the Raman line intensity on phase
is lost. In fact, the Raman-line strength and shape is
almost constant as a function of phase, although the polarized
intensity is more variable.
Schmutz et al. (1994) and Harries &
Howarth (1996a) remark on the
near-constancy of the 
6825 intensity profile of SY Mus with
orbital phase, while Harries & Howarth note that this is accompanied
by significant changes in its polarized-intensity profile. The models
shown in Fig. 12 (click here) share these characteristics, and offer a
simple qualitative explanation for them: the extent of the scattering
region
is sufficiently large that the occultation effect of the red giant is
relatively minor, while the changing parent-scatterer-observer angle
directly influences the degree of polarization.
In the preceding sections, the orbital dependence of the lines was
investigated by varying the angle 
. As already stressed, our
models - and any model with `up-down' symmetry - can only
generate position angles of 90
and 0/180
; but, of
course, in practice other (orthogonal pairs of) position angles can be
generated simply by rotating the observer's reference frame about the
line of sight to the system, with 
fixed. We make this point
explicitly so as to avoid any possible misinterpretation of the
discussion in Sects. 7.1 (click here), 7.2 (click here), but also to
emphasize that measurements of polarization position angle provide a
straightforward diagnostic of the relative positions of the binary
components. The importance of this result is that observed changes in
the polarization spectra of symbiotic systems (or other systems in which
the source of polarization is scattering of light from one star in the
atmosphere of the other) provide a useful diagnostic of orbital
parameters, as discussed in more detail by Harries &
Howarth (1996a).