We have used the observations of F-9, taken between 1978 and 1991, at all
wavelengths from the radio to X-rays to derive the nuclear continuum spectral
energy distribution (SED), which will later be used in the photoionization models.
The X-ray, UV, and optical lightcurves are shown in Fig. 1 (click here). All
data were corrected for E(B-V)=0.035 which is derived from the combined
hydrogen column density from our Galaxy and the Magellanic Stream of
(Morini et al. 1986;
Clavel & Santos-Lleó 1990). Using this color excess we applied
Seaton's law (1979) to correct the ultraviolet and the optical. In
the NIR we used the extinction law for this wavelength region from Rieke
& Lebofsky (1985). To evaluate the contribution in the FIR due to cold
interstellar dust from the galaxy disk of F-9, we used the results for normal
early-type spiral galaxies from the IRAS observations derived by De Jong
et al. (1984). The resulting disk flux of 120 and 370 mJy at respectively
60 and 100
, has been subtracted from the IRAS fluxes for F-9 in these
bands. The subtraction of the stellar contribution in the optical and the near
IR will be discussed in Sect. 3.1.1. Unless otherwise specified, we will use
observed wavelength in the following discussion.
Figure: Light curves of the UV, optical and X rays
continua.
a) FES light curve (.
b)
light curve. The full drawn line is the cubic spline
interpolation used to derive the UV fluxes at times when no simultaneous
observations were available.
c)
light curve (Table 2 (click here)).
d)
) light curve (Table 3 (click here)). N.B. At minimum (JD
3000) the object
was too faint to be detected with the FES
The ultraviolet continuum fluxes in Table 1 have been measured in three
`` line-free" windows centered at 1171 (1158-1184), 1400 (1390-1410) and
1910 (1895-1925) Å, and are corrected for E(B-V)=0.035 extinction. When more
than one spectrum was available on the same day, both fluxes and errors have
been averaged. Since is rather noisy -it is measured at the
short wavelength side of the geocoronal emission- and
already has a small contribution of the Balmer continuum and the FeII emission
(Small Blue Bump), both present above
(rest wavelength),
we will use the
) as reference UV continuum. For those cases
where the data in the UV and at other wavelengths are not exactly simultaneous,
we use an interpolation of the UV light curve fitted with a cubic spline using a
smoothing factor of
(reduced) = 3.75 as shown in Fig. 1 (click here)b to the
epoch of the non-UV observation. Table 1 also gives the FES counts
corrected for E(B-V)=0.035. Ground-based optical photometry between 1979-1990
of F-9 has also been reported by Lub & De Ruiter (1992). We have
corrected their B band fluxes at 4298 Å measured with an aperture radius of
for E(B-V)=0.035 (versus
as used by them; De
Ruiter, private communication). These values are given in Table 2 (click here)
together with the corresponding UV fluxes, derived from the interpolated
ultraviolet lightcurve. Both the FES and the B lightcurve are shown
together with
) and X-rays in Fig. 1 (click here).
The FES light curve (Fig. 1 (click here)) is very similar to the UV, but
its amplitude, as is common in Seyfert I galaxies, is considerably
smaller with . The Cross Correlation Function
(CCF; Gaskell & Peterson 1987) of the FES counts with
) peaks at
(Fig. 2 (click here)c),
indicating that there is no delay between the variations at these
wavelengths. As a consequence of this both fluxes are highly
correlated (Fig. 3 (click here)c):
( is in units of
) with r=0.938. Using the
calibration of Imhoff & Wasatonic (1986) one can in an approximate
way convert the FES counts to Johnson V-band fluxes. Using a color index of
for the total central flux of F-9 (Griersmith &
Visvanathan 1979)
we obtain FES
(
) = 0.024
FES(cts), changing relation (1)
into:
Here we have added the estimated error of 6%, associated with the calibration
used (Imhoff & Wasatonic 1986), to the correlation errors. Both
relations (1 and 2) show a non-negligible constant term most likely due to the
stellar contribution from the central bulge of the underlying galaxy. This
contribution dilutes the intrinsic nuclear variation of the AGN itself and
causes the smaller amplitude of the optical variations with respect to the UV.
Similar results are obtained from the optical data of Lub & De Ruiter
(1992) and their light curve is also very similar to the UV one
(Fig. 1 (click here)c) and its amplitude, , is
similar to that of the FES. The CCF of
) with
) peaks at
(Fig. 2 (click here)c), confirming the strict simultaneity at the time
resolution of 96 days of the variations at these wavelengths. Linear
correlation gives (Fig. 3 (click here)d):
r=0.969, using the interpolated light curve. Also here
the non-zero constant term is most likely due to the stellar contribution and is
the main cause for the smaller
amplitude of the
variations compared to the UV. The constant
terms in these relations causes the apparent hardening of the
UV-optical spectral index
(
; e.g.
Table 2 (click here) for
) with increasing UV brightness.
Since we want to use the observed AGN SED later as input spectrum for the
photoionization models at different brightness levels, we need to correct the SED for
these contributions. The constant term in Eqs. (2) and (3) is dominated by the
stellar contribution, however many optical FeII multiplets and the forbidden
lines of
, 4959 as well as
contribute both to
the FES counts and
. To evaluate the Fe II contribution, we
use the results of Wamsteker et al. (1985), deriving a value of
, from high resolution spectra and FeII model fitting. Subtracting this
from the constant in Eq. (2) we obtain (
erg s
cm
Å
(for the V band). Similarly we estimated the FeII contributing to the
, using the same FeII model and the bandpass of this band
(Lub & Pel 1977). This resulted in
.
Subtracting this from the constant in relation (3) gives (
(for the
). In the absence of optical spectra or
images, taken specifically for this purpose, we can compare the results of this
procedure with other independent determinations of the stellar
contribution to
the nuclear flux in those wavelengths were it is important. Griersmith &
Visvanathan (1979) used a De Vaucouleurs law (
distribution)
for the brightness distribution, which they combined with their measured
brightness distribution to calculate the stellar galaxy flux in V for any
aperture. For the FES (taken in V) with an aperture radius of
, we
obtain
, consistent with the above. For the
aperture radius of
, and taking the color index for the
stellar population in the galaxy
(Griersmith & Visvanathan 1979), we obtain (taking
)
, also in good agreement with the above. Since the stellar emission of
the galaxy estimated by Griersmith & Visvanathan (1979) is in
good agreement with those derived by us we will use it in what follows. This
value is approximately half of that derived for
by Lub
& De Ruiter (1992), but they used a disk-like exponential brightness
distribution and may have underestimated the contributions due to FeII
considerably. A disk distribution might not be appropriate for the apertures
used by Lub & de Ruiter (
), because
Griersmith & Visvanathan (1979) suggest that the nuclear bulge
extends out to
. Véron-Cetty et al.\
(1991) find also that a
distribution gives a good
representation of their observations of the galaxy surface brightness
distribution for F-9. With their value of
and
for
the galaxy, we derive the expected contribution to the FES counts to be
and to the
.
Since these independent estimates of the stellar contribution to the optical
measurements of the nuclear brightness are all consistent within the measurement
uncertainties, we can be confident that all constant components in the
observations of the nuclear region of F-9
are properly allowed for in the derivation of the SED of the active galactic
nucleus. Then, using the slopes of the relations (2) and (3), we obtain
and
, equal within the errors. As the second value has
a smaller error, we will use from now on a UV-optical spectral
index
. After correction for E(B-V)=0.035, instead
of E(B-V)=0.06 used
by Wamsteker et al. (1985), this value is similar to the index
obtained from their Table 1,
.
Figure: Variability features of the UV-optical continuum.
a) The power spectrum (Press & Teukolsky 1988 algoritm) of the
variations for . The peak at 4903 days represents the total
temporal range of the data and is not a
real variability time
scale.
b) ACF (Autocorrelation Function; Gaskell & Peterson 1987) for
.
c) CCF for
(solid line), for
(dashed line), FES counts (dot-dashed line) and
(dotted line)
with respect to
. The formal delay between these is
respectively,
,
,
, and
Figure 3: Correlations between the continuum fluxes at different
wavelengths versus the UV flux .
Linear regression lines are shown as dotted lines.
a)
.
b)
.
c) FES counts (
.
d)
.
e)
(1.15
)
1
Table 2: Ultraviolet and optical continuum
Near infrared (NIR) observations of F-9 in J (1.15 ), H (1.58
), K
(2.10
) and L (3.35
) have been reported in Glass
(1986) and in Clavel et al.\
(1989). From these data we only use those coinciding in temporal
coverage with the UV observations and only the data with an aperture radius of
6'' to avoid variations in stellar light contributions in the photometry. All
fluxes have been corrected for E(B-V)=0.035.
Edelson & Malkan (1987) report on four IRAS observations at
12, 25, 60 and 100 made over a period of 14 days, and find no
significant variability in any of these bands. We use here for each band
the mean flux and the rms of the individual measurements. The values for
the mean epoch
, are
,
,
and
for 12, 25, 60 and 100
, respectively. Subtracting the disk dust
contribution at 60 and 100
(120 and
, respectively. See
above), the fluxes for these two bands become
and
. Note the agreement between the 12
IRAS point and
the
measurement (also at 12
) taken
earlier (see Fig. 8 (click here) by Clavel et al. 1989), confirming
that the 12
\
radiation is not variable. The shape of the light curves in J (1.15
), H
(1.58
), K (2.10
) and L (3.35
) are very similar to the UV (see
Clavel et al. 1989), with a considerably smaller amplitude at
,
,
and
,
respectively. Clavel et al.\
(1989) derived delays of respectively
,
,
and
for J, H, K and L,
while we have obtained that the
J band and the UV fluxes are related by (Fig. 3 (click here)e):
with r=0.950, or, converting all units to :
The constant term in relations (4) and (5) is again dominated by the stellar contribution, while also the H, K, and L fluxes are affected, but to a progressively lesser extent.
The stellar emission in the near IR for F-9 with an aperture radius of 6'',
using the surface brightness distribution in V (with an error of 3%) of
Griersmith & Visvanathan (1979) and color indexes of V-K=2.99,
,
and
(Clavel et al.\
1989) is
,
,
and
for J,
H, K and L, respectively. Glass (1986) comments that these
color indexes are consistent with the population found in normal galaxies
(V-K=2.99 is characteristic of giant and supergiant stars of K3-K4
spectral-type (Johnson 1966)). This stellar contribution at J
(
) is similar, within the errors, to the constant term in
Eqs. (4) or (5). Therefore, the behavior in
the J-band can be fully understood in terms of a variable source, which is a
direct extension of the UV-optical power-law, and a constant part due to the
stellar contribution.
Thus there appears to be no evidence for the presence of a steep constant IR power law,
extending into the optical, as suggested by Edelson & Malkan (1986,
1987) and Carleton et al.\
(1987). For the AGN (without the stellar contribution) we obtain an
intrinsic
spectral index of
,
equal to the index obtained between 1400 and 4298 Å, confirming
that the variable J flux is a direct extension of the UV-optical.
Morini et al. (1986) have summarized the F-9 hard X-ray observations
from 1970 to 1984 obtained with the Uhuru, Ariel 5, HEAO (1 and 2) and EXOSAT
satellites. We use here again only data after 1978, for comparison with the UV
continuum. Two GINGA observations on 1989 November 20-21 and in 1990 November
22-25 are included (Makino, private communication). The fluxes
are listed in Table 3 (click here) together with the interpolated UV flux and
the light curves are shown in Fig. 1 (click here)d with the UV and the optical
ones. The GINGA data give hard X-ray spectral indexes of
(
) and
(
), both
with a fixed column density of
and
also show a strong Fe emission line centered at
, with
an equivalent width of
on
and
and
on
.
The only Soft X-ray observation has been taken with the ROSAT satellite at
the same date as the second GINGA observation (
), simultaneous
with one of the UV spectra (in the context of the RIASS program: De
Martino et al. 1991). Walter et al. (1995)
find that for a column density of
two power laws are required. So, he
obtains for the GINGA spectrum a spectral index of
and for
the
flux of 3.04
very similar to Makino's
values, and for the
ROSAT spectrum,
and
,
indicative of a strong soft X-ray excess.
Table 3: Ultraviolet and X-rays continuum
The only radio observation reported for F-9 has been made at 4.7 GHz
(Véron-Cetty et al. 1991) with the Australian Telescope on 24
December, 1990 () and they give a
upper limit to the flux
of 5 mJy, classifying F-9 as a weak radiosource.
The light curve (Fig. 1 (click here)b) shows a wide
range of variability over the
duration of the IUE observations. It shows a striking large and relatively slow flux
decrease since the first observations (1978-August-02,
) through 1984
October 29 (
). After this decline, the source brightened again until
1987 June (
), where it reached
of its maximum
brightness. The maximum amplitude of the whole light curve, expressed as the ratio
of the maximum to the minimum flux (R), is
(in order to
reduce the formal error on
the minimum flux and its error were
taken to be the average of the three
lowest values and their rms).
Pure Fourier techniques can not be used with unequally sampled data. In this case
other algorithms can be applied to allow for the non-uniform
sampling. We applied the algorithm by Press & Teukolsky (1988) to
the UV light curve to obtain the frequency power spectrum. This is shown in
Fig. 2 (click here)a, where different significance levels (0.05, 0.001) for a
random noise hypothesis are also indicated. Small values of this level
indicate a highly significant periodic signal. In the figure three main
maxima can be observed, corresponding to typical variability time scales of
4903, 182 and 56 days, respectively. The strongest peak at 4903 days is
associated with the total duration of the sampling, and does not represent a
real variability time scale. Since the mean interval between two
consecutive observations is 96 days, we can only expect to resolve
variations on longer time scales than this. On the other hand the irregular
nature of the sampling could still allow the detection of faster
variations. This could indicate the existence of the 56 days peak as a real
variability time scale, but since it can not be confirmed from the present
data, we conclude that 182 days represents the dominant typical variability
time scale of F-9 at 1400 Å.
Both and
are strongly correlated with
and the linear correlation analysis
(Figs. 3 (click here)a, b) gives:
with a correlation coefficient r=0.967; and
with r=0.986.
In both cases, the constant term is small compared to the flux and not significantly
different from zero. Most likely the slight deviation from zero of the constant term in
relation (6) is caused by the non-linearity and poorly determined calibrations in
this region of the IUE camera in the window below 1214 Å, while in relation (7) it
is due to the small FeII and Balmer continuum contribution to the flux in the
1910 Å window. The strict proportionality of the fluxes in these windows indicates
that, in spite of the large amplitude, the brightness variation takes place with constant
spectral index between 1171 - 1910 Å. The relations (6) and (7) give spectral
indexes of ,
and
. These two last spectral indexes
(
and
) are somewhat
steeper than the intrinsic UV-optical one (
, Sect. 3.1.2). This
is most likely caused by the contribution of the Balmer continuum which is
correlated with the UV flux (Sect. 4.1) and some FeII
contribution (see Figs. 6 (click here)a-c), resulting in an apparent steepening
of the spectral index. Because the Balmer continuum is a nebular component,
it is more meaningful to use a large wavelength range to calculate the
spectral index linking the UV to the optical, where we have allowed for the
removal of non-AGN associated flux in the spectrum, as done in Sect. 3.1.2.
The proportionality between the three UV fluxes and the similarity of their
maxima and minima in the light curves, suggests the absence of lag between them.
The cross correlation function (CCF; Gaskell & Peterson 1987) of
) and
with
are shown in
Fig. 2 (click here)c and as a reference the auto-correlation function, ACF of
, in Fig. 2 (click here)b. These peak at
and at
, respectively. This absence of a delay in the
variability between the continuum windows, suggest that a coherent emission
mechanism is responsible for the continuous emission at all three
wavelengths or otherwise the variations are isothermal.
The light curve of the integrated X-ray flux between 2 and 10 keV is
clearly different from the UV and optical ones (Fig. 1 (click here)d). The
X-ray light curve decreases from a maximum to a minimum value, after
which the flux raises again to its initial level of 1978. The
amplitude of variations in this time interval is for the decline and
for the increase.
This is contrary to the UV continuum, which shows a much larger
amplitude of
(as calculated from the
interpolated values), in the decrease and an amplitude of
, similar to the X-rays, when brightening. This indicates
that both UV and X-rays vary equally in the increase after the
minimum, but not in the initial decrease from 1978. This suggests that
the good correlation between X-rays and UV breaks down at high levels
of the UV continuum (Fig. 4 (click here)b), as has also been seen in other
AGN such as NGC 4151 (Perola et al. 1986) and NGC 5548
(Clavel et al. 1992). Figure 4 (click here)b shows a good
correlation whereas the UV flux is weak (
), while at UV
flux (
) the X-rays appear to remain constant. For the
intermediate
levels, unfortunately, no data are available. This behavior is obviously
reproduced in the individual UV-X-rays spectral
indexes shown in Table 3 (click here). For the low levels the mean spectral
index is
,
whereas for the high levels it is
,
resulting in a significantly steeper spectrum at the high UV brightness, as
shown in Fig. 4 (click here)c.
Figure: The Balmer Continuum (BaC), X-rays () Flux, and the
UV to X-rays spectral index
as a function of the UV
continuum brightness
.
a) BaC (3837 Å) derived from the data of Lub & De Ruiter (1992); see
Table 2 (click here).
b)
); see Table 3 (click here).
c) UV-X-rays spectral index
In Fig. 5 (click here) we show the intrinsic F-9 continuum SED (observed),
from radio to X-rays, as described above for four different levels of
continuum brightness. To allow direct comparison with the standard
conventions to display SED in the context of different wavelength
domains we show both the frequency spectrum as well as
the energy spectrum
versus
. The first
one allows easy evaluation of the spectral index (slope of the curve)
and the second one is the usual way to represent the SED, where a flat
SED corresponds to equal energy per decade. In absence of
observational support for significant variability in the FIR and radio
domain we have used at these wavelengths the single value determined
above for all UV and X-ray brightness levels of F-9. For the near
IR-hard X-rays range, we use the results of the variability study in
each band with respect to the UV one. We will in what follows refer to
the four levels of UV brightness as it `` high" (
), `` intermediate-high" (
),
`` intermediate-low" (
) and `` low" (
), corresponding to the
interpolated
) for each date. We have chosen these dates because
for three of them there exist values for the
) (see Table
3 (click here)) and for the
, when there are no X-rays data, we
interpolated linearly between the two values nearest to this date,
obtaining
. For each date, the J, H, K and L fluxes have
been taken by interpolating on the light curves (corrected for the stellar
contribution). These observed complete SED's will be applied later, to define the
input continuum spectra for the photoionization models in the detailed evaluation
of the line intensities and the variability characteristics to disentangle the
conditions in the BLR. The flux values used from radio to X-rays, in
units of
(and between round brackets in units of mJy), are given in
Table 4 (click here).
Table 4: Flux values used in the SED
From the combined optical, NIR (J) and ultraviolet variability study we found that
the spectrum in this domain can be represented, after allowing for the galaxy
contribution, by a power law spectrum with a constant spectral index of
(independently of the
brightness). It is
then justified to parametrize the central continuum spectrum with a UV-optical-J
power law of the type
, where
is
determined from the observed
for the different brightness
levels. In the hard X-rays we use a power law with
and the constant value is determined from the observed
). This
spectral index is taken from Morini et al. (1986) and is within the
errors the same as that independently determined from the GINGA observations
(see Sect. 3.1.4). There is only one observation with ROSAT, simultaneous with
GINGA and IUE. Since the wavelength ranges around the soft X-rays are variable,
the soft X-rays are likely to vary also. We can then not use this unique
datum for the four brightness levels defined before (Fig. 5 (click here)).
However, the simultaneity with the hard X-rays and
the UV on
, allows us to characterize at least part of the big blue
bump shape (Fig. 5 (click here)c). In this figure we have used the ROSAT-GINGA
spectral model by
Walter (see Sect. 3.1.4), and the J-UV power law with
fitted to
the
in this date. A strong soft X-ray excess is clearly seen
with respect to the hard X-ray power law, suggesting that the soft X-rays merge
smoothly into the decrease of the big blue bump, which must than have its
maximum at energies less than 0.15 keV (
;
).
Figure 5: Different representations of the SED of the F-9 continuum.
a) versus
from
radio to X-rays for four levels of UV brightness (see Sect. 3.3).
b)
versus
from radio to
X-rays for four levels of UV brightness.
c) SED (
versus
) for the unique
simultaneous
observation IUE-ROSAT-GINGA (
). The extrapolation of the
hard X-rays
(
) into the soft X-rays clearly indicates
a soft X-rays excess. This suggest that the turnover of the big blue bump, which
starts at
, occurs in this range