The source of the OVI photons is unknown in detail, although it is evidently local to the hot component. The reference model employs a photon source which is static in the binary rest-frame, but if the parent photons are produced in the ionized wind of the cool star then a source which is static in the comoving frame of the wind may be more appropriate. In principle, it may be possible to address this question directly, by studying the orbital velocity dependence of the emission lines and cool-star spectrum, but this will present difficulties in practice. We performed a trial calculation in which the parent photons were assumed to be emitted at zero velocity in the comoving frame of the wind. The results are shown in Fig. 7 (click here).
Figure 7: The polarization spectrum of a model computed with the
source emission at zero velocity in the co-moving frame
(Sect. 6.1 (click here))
This model shows three polarized-intensity peaks, at about +20, +60, and
+110 km s
(in `parent velocity' space), with a PA flip for the
central peak. This model is therefore qualitatively similar to the
reference model, except that essentially all the emission appears
redshifted (because the emission source now sees no part of the wind
approaching it). The well-observed systems described in Paper I nearly
all show a blueshifted peak (cf. Fig. 6 (click here)). The implication
is
that our assumption of a photon source which is (on average) static in
the binary reference frame is more nearly correct for the majority of
real systems than is the assumption of emission from ionized material
outflowing from the red giant. There are, however, possible exceptions;
a blue-shifted polarized component is undetectable for a
few systems (e.g., AS 210, which also shows a constant PA, though not
resolved multiple peaks).
The presence of a blue-shifted feature in the Raman lines also
constrains any possible shift in the parent-line wavelength.
Measurements of emission-line systemic velocities are often difficult to
make, but for a few systems observations suggest a red shift in
high-excitation lines (e.g., Michalitsianos et al. 1988).
Such `intrinsic' shifts in the parent
lines would translate directly into shifts in the Raman-line
wavelengths, amplified by a factor 
. In our models, a Raman-line redshift results from
the expansion of the red-giant wind, and this accounts for the greater
part of the observed displacement from the rest wavelengths; the
observation of blue-shifted components in the Raman lines shows that any
systematic, intrinsic velocity displacement of the OVI lines
must be less than the red-giant wind velocity.
Probably the most important simplifying assumption in the reference model is that there is no region of ionized cool-star wind surrounding the hot component. This assumption does not reflect any fundamental limitation of the numerical code, but was adopted simply because to do otherwise complicates the mass-column calculation (Eq. 11 (click here)) and thereby considerably increases the run time required for a model.
We calculated two models which incorporate
extensive ionized zones. In each case, we assumed that the ionized
region is completely transparent to all photons. Although this
assumption should be relaxed in future work, it is probably not an
unreasonable first approximation; in particular, the electron-scattering
optical depth along any line through the ionized zone is likely to be
, and so it should not have an important polarization signature.
We also assume axial symmetry, which is justified since the recombination
timescale is much shorter than the orbital timescale.
In the first model,
we
approximated the geometry of the ionized zone by
where 
is the distance from the hot component
to a point on the
ionization front (in units of the binary separation) and 
is the angle
between that point and the line of centres.
In this model, the ionization front is concave to the OVI source.
The intensity
and velocity images are
shown in Fig. 8 (click here), and the spectropolarimetry in
Fig. 9 (click here).
Because of the geometry of the ionized zone, there are still extensive
scattering regions `above' and `below' the OVI source which, as
in the reference model, give rise to PA-flipped, redshifted, polarized
intensity at around 6835 Å;
thus many characteristics of this model are similar
to those of the reference model (including the weak, high-velocity redshifted
polarized intensity arising from `fore-and-aft' scattering).
Figure 8: Results for the reference model, with an ionized region
(Sect. 6.2 (click here); equation 34 (click here))
Shown are the intensity, I (scaled logarithmically over
3 decades) and line-of-sight component of the velocity (scaled linearly
over the range -50 to +150 km s
)
In order to produce qualitatively different results within the
framework of a spherically-symmetric outflow model, it is necessary to
incorporate a more extensive ionized zone. To do this, we used
The resulting spectropolarimetry is shown in Fig. 10 (click here). Two
main intensity peaks remain, corresponding to scatterings on either side
of the red giant along (and around) the line of centres. In this model,
however, the ionization front is concave to the red-giant component, so
that the scattering regions `above' and `below' the OVI source
have been removed from the model. The PA is therefore constant across
the line profile.
These results offer the hope that the PA structure in the Raman lines
may prove to be a straightforward, if crude, diagnostic of the extent of
the ionized region. In particular, the presence of a PA `flip' in the
Raman lines of many well-observed symbiotic stars (Paper I)
suggests that the ionized region may not be very extensive in
those systems. In terms of the commonly-adopted parameterization of the
extent of the ionized-hydrogen region
(Seaquist et al. 1984, Nussbaumer & Vogel 1987; 
is the
luminosity in the ionizing continuum), then the model
described by Eq. 34 (click here) has 
, and that described by
eqtn. 35 (click here) has 
. The
results
presented here therefore imply
X values of order unity or less for systems which show a `flip'. However,
Mürset et al. (1991) have estimated X by an independent method for several
systems, including three with well-observed Raman lines reported in Paper I.
All three systems show a `flip', suggesting small X, whereas Mürset et
al. report X of order 10 (SY Mus), unity (RR Tel), and 
(Hen 1092).
The origin of this discrepancy must be resolved before either technique can be
regarded as yielding secure results, but in particular it will be necessary for
the models described here to account correctly for the range of PA structure
observed.
Figure 9: The polarization spectrum of a model computed with an
ionization front which is concave
when viewed from the photon source (Sect. 6.2 (click here),
Eq. 34 (click here))
Figure 10: The polarization spectrum of the model computed with an
ionization front which is convex
when viewed from the photon source (Sect. 6.2 (click here),
Eq. 35 (click here))