The strong correlations obtained between the intensity of the different broad components and the UV flux, confirms that the main excitation mechanism of the gas is photo-ionization by a central source. To obtain better insight in the physical conditions and the spatial relations of these parts of the BLR which appear to constitute separate regions, we have made separate photo-ionization calculations for each of the regions to compare with the results described in the previous section.
We used the photo-ionization code CLOUDY (Ferland 1991: CLOUDY
80.06). The initial parameters needed by CLOUDY are: the shape and
intensity of the incident continuum (
), the chemical
composition, the distribution of the gas (radial extension and the run of
density with radius) and the covering factor. We have assumed solar
composition, an open (small radiative covering factor) spherical geometry and
constant density. For the ionizing continuum we have used the observed
continuum derived in Sect. 3.3 at four levels of brightness
(Fig. 5 (click here)), corresponding to the 
of 23.27 (high),
10.04 (high-intermediate), 3.78 (low-intermediate) and 1.31 
(low), respectively. To connect the frequency gap between the Radio and the
FIR range, linear
interpolation has been used. To account for the big blue bump, the
optical-UV continuum has been parametrized with a power law
,
fitting 
to the observed 
for each level. The cut-off
energy is a critical factor in the input spectrum. Clavel et al.\
(1990) and Binette et al. (1989) suggested 0.83 and 0.73 ryd,
respectively, based on photoionization models for Fairall-9. The simultaneous
observation with IUE, ROSAT and GINGA (Fig. 5 (click here)c) shows however that
such a cut-off frequency does not correspond with the observed
instantaneous spectrum of Fairall-9. Although it might be possible that the
spectrum incident on the BLR is different from that which we observe, at
this stage we have no convincing arguments to apply a spectrum different
from the observed one. Then, a considerably slower drop off is indicated and
a cut-off at 3.5 ryd as shown in Fig. 17 (click here) was applied. In the
X-rays and higher frequencies we used a power law similar to the UV-optical
one, with a spectral index 
and
an arbitrary cut-off at 0.1 MeV (consistent with the 
ray observations of
other Seyfert 1 nuclei with CGRO). The constant was
defined by the observed integrated flux at 
. To bring all in an
absolute scale we used 
and
all fluxes have been corrected for redshift with z=0.0461. We have
calculated models with a range of total hydrogen density of the gas, n,
between 
and 
and for the total hydrogen
column density, 
, a range between 
. The inner radius of the gas cloud, r, was chosen over a range from 50 to
600 light-days, consistent with the delays obtained for the BLR line
components with respect to the UV continuum derived in Sect. 4.3.
With these initial conditions many models have been calculated, obtaining
from them the line luminosity for a geometric covering factor of 1
(
). Comparing the results of the models for this value of
with the observed luminosities, the real covering factor can
be obtained. The ionization parameter, U, is determined from the incident
continuum luminosity, the density and the distance assumed for each model.
So, comparing the relative intensities of the lines obtained from the models
with the observed ones, the best fitting model can be found.
Figure 17: Fit to the simultaneous observation
IUE-ROSAT-GINGA. : Observation
(between the UV continuum and the J band 
has been used).
: Fit with a cut-off energy
of 3.5 rydberg used as input for the
photoionization models. : SED with a 0.63 rydberg cut-off energy. It
is clear from these simultaneous observations that
the 
cut-off energy proposed
by Binette et al. (1989) and Clavel & Santos-Lleó
(1990) is incompatible with the data
To compare the models and the observations we have used the observed line
ratios for the components for 
/
, 
/
and 
/
, imposing that
one single model fits all three line ratios, at all brightness levels of the
continuum. Since the results for all models with 
and
are the same, we only consider the first. The
conclusions for each broad component are:
The three line ratios are well described by models with the following
characteristics distance (light-days); 
; 
:
;
;
;
;
.
The results for these best fitting models are shown in Fig. 18 (click here).
Since 
is nearly independent of the distance, also
models with inner radius up to 600 light-days could be allowed too, but at
these distances the other two ratios are too large at some continuum
levels, restricting the distance range to 
. We
see that all models are consistent with the observed values of
and 
, but underestimate 
, at bright continuum levels, since they converge to 1. The mean ratio
of 
, which increases with distance, for distances of
50, 100, 150, 200 and 250 light-days is, respectively, 12.6, 17.0, 21.7, 26.1
and 36.1. The ratio 
for the central component
estimated from the data of Wamsteker et al. (1985) is 
(Sect. 4.3.3), corresponding to models with radius of 50 and 100 light-days.
Figure 18: Line ratios for the central component and the corresponding
photoionization models.
The different models are drawn with several types of lines over
the observed ratios: distance (light-days) - 
(
(
(- - -),
(
- 
-),
(
), 
(
- 
) and
(--)
We show in Fig. 21 (click here) a comparison between the spectrum derived from
the models and the individual line components with the observed mean
spectrum. The mean -emission lines only- spectrum shown in Fig. 21 (click here)
has been corrected by z=0.0461 and corresponds to an observed 
of 
and to a 
central component
luminosity of 
(
. The comparison in
Fig. 21 (click here) is thus for an high-intermediate level of the continuum,
with an observed value 
. For this level
the models give U=0.003, 0.009, 0.039, 0.022 and 0.014 and corresponding
covering factors of 12%, 5%, 4%, 4% and 3% for distances of 50, 100,
150, 200 and 250 light-days, respectively. These model spectra are similar,
except perhaps the 50 light-days model, which produces a weak CII
, which is not observed. The 
and 
intensities are
somewhat higher than observed. In Fig. 21 (click here)a we show the mean
spectrum together with the model of the central component for 100
light-days, 
and 
model. The photoionization models with CLOUDY confine the region where the
central component is produced to a distance from the ionizing source of 
, with a column density of 
, an hydrogen density of 
and a covering
factor of 
, with U between 0.003 and 0.039 (Table 14 (click here)).
The distance derived from the models is thus consistent with the lower end of
the range derived from the delay of this component with respect to the UV
continuum, 
.
13
Table 14: Resulting parameters for the broad components
The acceptable models for the observed ratios of this component are similar
in their parameters range to the central component ones. They are characterized by (as
above in Sect. 5.1.1):
; 
;
; 
and 
.
Larger distances are excluded because they
result in a too high 
ratio. The results for these models are
shown with the observed ratios in Fig. 19 (click here). Similarly to the central
component, all models reproduce the observed 
/
and 
/
ratios
very well, but underestimate 
/
. The mean spectrum has in the blue component
a Ly
luminosity of
. For the comparable high-intermediate
level of the continuum, the models have U=0.035, 0.089, 0.039, 0.022 and
0.014 and predict a covering factor of 5%, 2.9%, 2.5%, 2.3% and
2.1% for 50, 100, 150, 200 and 250 light-days, respectively. Compared
with the mean spectrum, the spectra of the five models are very similar, and
we show the 100 light-days, 
and
model for the blue component with the observed
mean spectrum (Fig. 21 (click here)b).
Figure 19: Line ratios for the blue component and the corresponding
photoionization models.
The different models are drawn with several types of lines over the observed
ratios: distance (light-days) - 
(
(
(- -
-), 
(
- 
-),
(
), 
(
- 
) and
(--)
In summary, the models indicate that the gas producing the blue component
of Ly
, CIV, SiIV and MgII is situated at a distance of the ionizing
source of 
, with a column density of 
, an hydrogen density of 
and a covering
factor of 5 - 2%, where U is confined between 0.089 and 0.014. This range of
parameters is similar to the central component one (Table 14 (click here)). The
distance derived from the photoionization model for the blue component is
consistent with the mean delay of this component respect to the UV
continuum (
).
The models, which fit best to the observed ratios for this component are characterized
by (as above in Sect. 5.1.1):
; 
;
; 
and 
.
Also here larger distances are excluded because they result in an overestimate
the 
ratio. These models are shown in Fig. 20 (click here)
with the observed ratios. All models fit perfectly to 
/SiIV and
/
, but 
/
is always underestimated.
Figure 20: Lines ratio for the red component and the corresponding
photoionization
models.
The different models are drawn with several types of lines over the observed
ratios: distance (light-days) - 
(
(
(- - -),
(
- 
-),
(
), 
(
- 
) and
(--)
For this component we have been able to isolate also the 
/
ratio. Its
observed value is low (between 2 and 8) which can only be explained with
models that, at similar distances, have an order of magnitude lower hydrogen
density and a very small column density (
). This might
indicate that the BLR is composed of optically thin clouds besides of the
optically thick ones considered above, which might also explain better the
ratio, as have been recently suggested by Shields et al.\
(1995). However, even though the previous models (optically thick) seem
to underestimate the 
line strength, we will
use them since at the required low column density no consistent solution was
feasible for all other line ratios.
The mean spectrum has a red component Ly
luminosity of 
. For the high-intermediate level of the
continuum, the models have U=0.035, 0.0089, 0.039, 0.022 and 0.014 and
a covering factor of 6%, 4%, 3.3%, 3.0% and 2.9% for 50,
100, 150, 200 and 250 light-days, respectively. Comparison with the mean
spectrum shows that the model spectra are very similar, and in
Fig. 21 (click here)c the model shown is for 100 light-days, 
and 
. In summary, the
photoionization model calculations indicate that the gas producing the red
component of 
, 
, 
and 
is situated at a distance of the
ionizing source of 
, with 
, 
and a covering factor of 
with U between 0.009 and 0.039
(Table 14 (click here)).
The models alone do not
allow to distinguish the gas producing the red component from that
producing the blue component, because the same parameter range is
applicable for both (and very similar to those for the central component as well). On
the other
hand the delay of the red component of 
with respect to
the continuum falls outside the range suggested by the models, in contrast to
the results found for the central and blue component. One possible
explanation for this is that the gas emitting the red component is situated
along the line of sight to the observer or that the regions producing
both the red and the blue component are at the same distance falling in
toward the central source (see Sect. 7 for a more detailed explanation).
The difficulty associated with the 
ratio for all models
can be explained if: (a) The carbon abundance is less than the solar one (see the
next section); or (b) The continuum shape in 
) between 13.6
and 48 eV (
and 16, respectively) is not as steep as the
exponential cut-off used for the input continuum spectrum
(Fig. 17 (click here)), so that the fraction of ionizing photon density to
which CIV responds, decreases with respect to the photon density which Ly
responds, keeping in mind that the cut-off frequencies suggested by
Clavel & Santos-Lleó (1990) and Binette et al.\
(1989) are inconsistent with the observed soft X-rays excess
(Fig. 17 (click here)); or (c) There is not a smooth connection between the UV
and the soft X-rays, with an absorption in the continuum between 
, resulting in an increase in the 
ratio.
Such absorption can be present if optically thin material, transparent at
and absorbing at 
, is present between the
ionizing source and the BLR (Ferland et al. 1990); or (d) It is
necessary to introduce in the models two populations of clouds: optically
thin and optically thick ones, as it has been shown by Shields et al.\
(1995) to be possible for the specific case of F-9.
Figure 21: Observed mean spectrum and the obtained spectra from the
photoionization models for each component.
a) The observed mean spectrum and the model with 100 light-days,
and 
for the central
component.
b) The observed mean spectrum and the model with
100 light-days, 
and
for the blue component.
c) Equal to the central
component for the red one.
d) The sum of each model for the different
components over the observed mean spectrum. Take into account that the narrow
component has not been fitted