Äîêóìåíò âçÿò èç êýøà ïîèñêîâîé ìàøèíû. Àäðåñ îðèãèíàëüíîãî äîêóìåíòà : http://hea-www.harvard.edu/QEDT/Papers/ngc5548.ps Äàòà èçìåíåíèÿ: Thu Jun 1 20:29:22 1995 Äàòà èíäåêñèðîâàíèÿ: Mon Oct 1 23:45:47 2012 Êîäèðîâêà: Ïîèñêîâûå ñëîâà: agn

Testing Unified X­ray/UV Absorber Models with
NGC5548
Smita Mathur, Martin Elvis and Belinda Wilkes
Harvard­Smithsonian Center for Astrophysics
60 Garden St, Cambridge MA 02138
Internet: smita@cfa.harvard.edu
June 1, 1995

Abstract
The bright Seyfert galaxy NGC5548 shows absorption features in its X­ray
and UV spectra. The large amount of optical/UV and X­ray monitoring
data available for this object makes it an ideal candidate to test our model of
a common X­ray/UV absorber. We show that the UV/X­ray absorption is
caused by a highly ionized (2.2! U !2.8) high column density (NH = 3:8 \Theta
10 21 cm \Gamma2 ) gas situated outside the CIV emitting region (2 \Theta 10 16 ! r abs !
2 \Theta 10 18 cm). The gas is outflowing with a mean velocity of 1200\Sigma260 km s \Gamma1
and corresponding kinetic luminosity of ¸ 10 43 ergs \Gamma1 . The gas is more highly
ionized and has a much larger column density than earlier estimates based
on UV data alone. This third example of an X­ray/UV absorber reinforces
our earlier conclusion that outflowing, highly ionized gas is common in the
inner regions of quasars.

1 Introduction
Recently we found that the ionized (``warm'') X­ray absorbers and the associ­
ated UV absorbers were due to the same material in two radio­loud quasars:
an X­ray quiet quasar 3C351 (Mathur et al. 1994) and a red quasar 3C212
(Mathur 1994). In both cases the absorber is situated outside the broad
emission line region (BELR), is outflowing, and is highly ionized. While
this result delineates a new nuclear component in lobe­dominated radio­loud
quasars, it would clearly be much more interesting if the same component
were present in all quasars and active galactic nuclei (AGN) with both X­ray
and UV absorption (Ulrich 1988). In this paper we test this generalization
by applying the same model to the best studied of all AGN, NGC 5548.
NGC 5548, a radio quiet Seyfert 1 galaxy, provides the best test case
for the equivalence of the X­ray and UV absorbers and will yield a highly
constrained determination of the physical properties of the absorber. This
is because NGC 5548 has been extensively studied in reverberation mapping
experiments in the optical and UV (Peterson et al. 1992, Clavel et al. 1991,
Korista et al. 1995). These have led to the accurate determination of the
physical size of the BELR. NGC 5548 has both an X­ray ionized absorber
(Nandra et al. 1991, Fabian et al. 1994a) and an UV absorber (Shull & Sachs
1993), and the reverberation observations place limits on the response time of
the UV absorbers to changes in the UV continuum. Note that this response
time is due to the physical conditions of the absorber only since there is no
light­travel­time delay along the line of sight.
We apply the photoionization modeling method of Mathur et al (1994) to
the X­ray and UV absorbers in NGC 5548 to determine whether consistent
values for abundances of all the observed ions can be obtained. In NGC 5548
the model must meet two extra requirements: it must not lead to a density
for the absorber in conflict with its recombination time; and the distance
of the absorber from the continuum source must not conflict with the well­
determined BELR size.
1

2 Data
2.1 X­ray Data
X­ray observations of NGC 5548 have revealed a complex spectrum. EX­
OSAT showed that the source has a strong soft excess (Branduardi­Raymont
1986, 1989). Ginga showed, moreover that the X­ray spectrum and flux were
variable (Nandra et al. 1991) and that the X­ray variability was correlated
with the ultraviolet variability (Clavel et al. 1992). In addition, complex
absorption was observed by Ginga with a possible Fe­K edge at ¸8 keV,
corresponding to an intermediate ionization stage of iron (¸FeXX) with
Ü F e\GammaK ¸ 0:03, and variable soft X­ray absorption. ROSAT observations
of NGC 5548 (1990 July 16­21) showed an absorption feature in its X­ray
spectrum arising from highly ionized oxygen, demonstrating the presence of
an ionized absorber; E=0:81 \Sigma 0:06 keV, Ü = 0:35 \Sigma 0:13 (Nandra et al. 1993)
Recent observations with ASCA (1993 July 28) confirmed the presence
of an ionized absorber (Fabian et al. 1994a). OVII and OVIII absorption
edges observed at 0.72 keV and 0.86 keV respectively were resolved in the
ASCA spectrum. The optical depths of the edges were Ü OV II = 0:26 +:04
\Gamma:08
and Ü OV III = 0:12 +:07
\Gamma:03 . The combined OVII and OVIII opacity from ASCA
agrees well with that observed with ROSAT. An equivalent hydrogen column
density was NH = 3:8 \Theta 10 21 cm \Gamma2 . However the presence of an Fe­K edge is
not confirmed (Ü F e\GammaK Ÿ 0:1).
The absorption cross­sections of OVII and OVIII are 0:28 \Theta 10 \Gamma18 cm 2 and
0:098 \Theta 10 \Gamma18 cm 2 respectively (CLOUDY version 80.06, Ferland 1991). The
column densities in the two ions are thus NOV II = 0:93 +0:14
\Gamma0:29 \Theta 10 18 cm \Gamma2 and
NOV III = 1:22 +0:72
\Gamma0:30 \Theta 10 18 cm \Gamma2 . Given the total hydrogen column density and
the solar abundance of oxygen relative to hydrogen (= 8:51 \Theta 10 \Gamma4 , Grevesse
& Andres 1989) the fraction of oxygen in the two stages of ionization is
calculated; fOV II = 0:29 +:04
\Gamma:09 and fOV III = 0:38 +:22
\Gamma:10 . The absorber must thus
be highly ionized to have 2/3 of the oxygen in hydrogen­like and helium­like
states.
2.2 Ultraviolet Data: IUE
NGC 5548 has been studied extensively in the ultraviolet (Clavel et al. 1991,
Korista et al. 1995). The UV spectrum shows absorption lines of CIV and
2

NV within the profiles of their broad emission lines (Shull and Sachs 1993;
hereafter SS). The source was found to be highly variable in its continuum
flux and also in the strengths of its emission and absorption lines. The UV
continuum was found to vary by a factor of ¸ 3 while the CIV absorption
line EW varied by a factor of ¸ 2 over a period of about 10 days. There is
no observed lag between the variations in the continuum and the equivalent
width (EW) of the CIV absorption line (\Deltat Ÿ 4 days, SS). Note that any
delay between the continuum and absorption line variability is not due to
light travel time (reverberation) effects, which are not operative along the
line of sight. It is instead due to physical conditions within the absorber,
specifically, the recombination time (see sec. 3.4). SS also infer a blueshift
of 1200\Sigma260 km s \Gamma1 with respect to the systemic velocity of NGC 5548. We
note that the earliest IUE spectra of NGC 5548 are also suggestive of an
absorption feature at this position (Wu, Boggess and Gull 1981, Ulrich and
Boisson 1983). We re­analyzed the 1979 IUE data (SWP05500, SWP05687,
SWP05688, SWP05689, SWP07345, SWP073930) including those reported
by Wu et al. (1981) using current extraction techniques . The CIV absorption
line is present at similar velocity and of comparable strength to the later data
(SS).
No associated Lyff absorption has been reported in the large literature
on NGC 5548. Observationally it is difficult because the Lyff emission line
of this low redshift source lies very close to geo­coronal Lyff. Theoretically,
Lyff absorption was not expected (SS) as the absorbing gas was considered
``fully ionized'' in hydrogen. As a result all the published spectra cut off the
blue wing of Lyff in the AGN emission line in order to remove geo­coronal
Lyff from the spectrum, cutting out also part of the Lyff absorption line.
Because our models predict significant Lyff absorption, we retrieved and
re­analyzed a set of IUE spectra taking care to go to shorter wavelengths. In
Figure 1 we present an IUE spectrum of NGC 5548, without removing the
geo­coronal Lyff. This clearly shows the presence of a Lyff absorption line
at a redshift consistent with that of the other absorption lines.
2.3 Ultraviolet Data: HST
NGC 5548 was observed daily for a total of 39 times by HST in 1993 April/May
to monitor the variability of the source in the ultraviolet (1140 š A! – !2312 š A,
Korista et al. 1995). HST FOS resolution is ¸1š A over this wavelength range
3

(c.f. ¸6š A with IUE). Figures 2 a,b show the emission and absorption line
profiles of Lyman­ff and CIV respectively for the mean FOS spectrum. This
spectrum was constructed by joining the combined G130H and G190H spec­
tra; no scale factors were applied (see sec 2.4, Korista et al.1994). We mea­
sured the EWs of the absorption lines both in the individual HST spectra
and in the mean. The CIV doublet ––1548; 1551 is resolved with EW =
0.32\Sigma0:1 š A and 0.08\Sigma0:03 š A respectively in the mean FOS spectrum (Table
1). The measurement of absorption lines situated within a broad emission
line profile is notoriously difficult, since the real shape of the emission line
is unknown. We estimated the magnitude of our errors by making maximal,
minimal and optimal measurements in each case (see e.g. figure 2a). This
yielded a typical error of about a factor of two in individual spectra and about
30% error in the mean FOS spectrum. The N V doublet is also resolved in
the HST data with EW = 0.1 and 0.04 (\Sigma 30%) in the mean spectrum.
An associated Lyff absorption line, at the same relative blueshift (\Deltaz =
0:002) as the CIV­NV absorption system, is clearly seen with EW 0.5\Sigma 0.3 š A;
a minimum HI column density is thus NHI – 4 \Theta 10 13 cm \Gamma2 . There is another
absorption feature immediately shortward (– obs = 1231:9, \Deltaz = 0:004, EW
¸ 0:6). There are no corresponding CIV or NV absorption features at a
similar relative blueshift and it appears to be unrelated to the associated
absorption system we discuss here. The uncertainty in the Lyff EW is large
due to its proximity both to this nearby absorption feature and to the peak
of the emission line profile.
Note that the CIV EW in the mean HST spectrum is significantly lower
than those reported for the IUE spectra (2 or 8 š A SS) in 1989. The larger of
the two SS EW values was derived by assuming that much of the emission
line peak was absorbed. This is clearly not the case in the higher resolution
HST data. The HST spectra also show that the absorption feature covers
a smaller range in wavelength than was apparent from the IUE data. We
re­measured nine IUE spectra around JD2447580, the time of the maximum
continuum change, using \Delta– ¸10 š A, rather than the 16 š A used by SS. The
exact wavelengths varied due to the inherent uncertainty in the wavelength
calibration for spectra observed with the large IUE aperture. Our EW values
are systematically smaller than SS's small values for the same spectra by a
factor 0.6. We use our revised measurements for the remainder of this paper.
We note that the 1989 IUE CIV EW are still larger that the HST values.
4

This may be due to the measurement difficulties in the low resolution IUE
spectra.
3 Models
In the context of photoionization models we look for an ionized absorber that
satisfies the X­ray constraints. Photoionization models predict the fraction of
atoms in each ionization state, f ion , given the column density (NH ), density
(n) and ionization parameter (U= Q
4úr 2 nH c , where Q is the number of ionizing
photons) of a cloud of gas exposed to a continuum source with a defined
continuum shape. All the photoionization calculations in this paper were
performed using CLOUDY (Ferland 1991).
Mathur et al. (1994) have shown that the particular spectral energy dis­
tribution (SED) of each AGN should be used, rather than a `typical' quasar
SED as this might lead to incompatible physical conditions in a photoion­
ized cloud. The observed radio to X­ray continuum of NGC 5548 is shown in
figure 3 (data from Ward et al. 1987, SS 1993, Korista et al. 1995, Nandra
et al. 1991). The continuum is corrected for Galactic reddening by E(B­
V)=0.033 assuming a fixed conversion of NH=E(B \Gamma V ) = 5:0 \Theta 10 21 cm \Gamma2
mag \Gamma1 and NH (1:65 \Theta 10 20 cm \Gamma2 , Nandra et al. 1993). The shape of the
IR/optical continuum is similar to a typical AGN (Elvis et al. 1994) with
the IR break shifted shortward to 100¯m. The standard radio­quiet AGN
continuum used by Mathews & Ferland (1987) is shown for comparison as a
dashed line. NGC 5548 is much brighter in the X­rays than a typical AGN,
but has a normal X­ray spectral slope, ff E = 0:8 \Sigma 0:2 in the 2 -- 10 keV range
(Nandra et al. 1991). In addition, it has a soft X­ray excess which can be
approximated by a black body spectrum of temperature 150,000 K (dotted
line in Figure 3). The heavy solid line in figure 2 shows the adopted input
continuum. The X­ray column density NH was fixed at the observed ASCA
value of 3:8 \Theta 10 21 cm \Gamma2 (Fabian et al. 1994a).
3.1 The X­ray Absorber
Figure 4 shows the ionization fractions of OVII and OVIII as a function
of U, using the de­reddened continuum for NGC 5548 and assuming solar
abundances (Grevesse & Andres 1989). The number of ionizing photons,
5

Q=2:5h \Gamma2
50 \Theta 10 54 s \Gamma1 where h 50 is the Hubble constant in the units of 50
km s \Gamma1 Mpc \Gamma1 . Note that this value of Q is in fact smaller than that used by
Ferland et al. 1992, Q=4h \Gamma2 \Theta 10 54 s \Gamma1 , which is based on a scaled version
of the standard continuum. This is because most of the ionizing photons
come from the energy range 1--20 Rydberg where the standard continuum
is brighter (figure 3). The observed values are highlighted with thick lines.
Using the constraints on the fractional ionization of OVII and OVIII from the
ASCA data (Ü OV II = 0:29 +0:04
\Gamma0:09 ; Ü OV III = 0:38 +0:22
\Gamma0:09 , Fabian et al 1994a, section
2.1) the allowed range of the ionization parameter is narrow, 2:2 ! U ! 2:8.
(Note that this is a linear scale.) If instead the observed continuum is used,
i.e. with no reddening correction, the inferred range changes to 1:2 ! U ! 1:3
and the dependence of logf ion on U remains qualitatively the same. In the
rest of the paper we quote all the parameters only for the reddening corrected
SED. The values of ionization fraction are independent of the gas density and
so, the density (n) is not constrained by photoionization models alone (see
sec. 3.4). Density n = 10 7 atoms cm \Gamma3 was used in the input to CLOUDY.
We investigated whether the lack of an Fe/K absorption feature observed
by ASCA (section 2.1) is consistent with our model. The ¸ 8 keV Fe­K
absorption edge would most likely be due to an intermediate ionization stage
of iron ¸ FeXX (Nandra et al. 1991), most likely FeXVII, since it dominates
over a wider range of U than other ionization state, since it is neon­like and
so more stable than other iron ions.
We find that, for the best fit model for the oxygen edges, the dominant
stage of iron is indeed FeXVII, log f FeXV II = \Gamma0:77, implying N FeXV II !
2:2 \Theta 10 18 cm \Gamma2 . The ASCA upper limit to an Fe­K edge is Ü ! 0:1 (section
2.1), so N FeXV II = 3 \Theta 10 16 cm \Gamma2 , assuming solar abundance of iron (4:68 \Theta
10 \Gamma5 , Grevesse & Andres 1989). This is consistent with the absence of any
Fe­K X­ray absorption edges. This absorption system, however, falls short
by a factor of ¸ 25 of producing the large observed EW of the Fe­K emission
line even for a uniform spherical shell (model EW ¸7 eV, observed EW =
180 \Sigma60 eV); consistent with the conclusion of Nandra et al. (1991).
3.2 Combined X­ray and UV constraints
We follow the technique developed in Mathur et al. (1994) to look for an
absorption system that satisfies both X­ray and UV constraints. The X­
ray constraints require 2.2!U!2.8, as discussed above. Thus, if the X­ray
6

and UV absorbers are one and the same (as was found to be the case in
3C351 and 3C212) in NGC 5548, then it can be seen from figure 3 that
the ionization fraction of CIV is constrained to be fCIV = 2:0 +0:5
\Gamma0:4 \Theta 10 \Gamma4
(log fCIV = \Gamma3:7 +0:2
\Gamma0:3 ).
In the mean HST spectrum, the CIV doublet ratio is 3.8\Sigma0:2, thus the
CIV absorption lines lie off the linear portion of the curve of growth (see e.g.
Spitzer 1978, Weise et al. 1966). In figure 5a,b,c the curves of growth for
velocity spread parameters b=20, 40, 60, 80 km s \Gamma1 are shown. Figure 4a
shows the observed equivalent width for CIV (Table 2) and the constraint on
log(nf–) derived from the X­ray constraint on fCIV (see above) assuming a
solar abundance for carbon (3.63\Theta10 \Gamma4 , Grevesse & Andres 1989) and oscil­
lator strength f=0.285 (Allen 1973). Similar plots (Figures 5b, c) for NV (at
solar abundance, 3.63\Theta10 \Gamma4 , Grevesse & Andres 1989) and Lyman­ff (using
oscillator strengths of 0.235 and 0.4162 for NV and Lyman­ff respectively,
Allen 1973) lead to the allowed b values given in Table 2. A consistent solu­
tion for all three ions is obtained for b=40km s \Gamma1 , with a small tolerance for
both UV and X­ray constraints to be met.
The measured line widths for the CIV doublet are ¸ 350 km s \Gamma1 (FWHM),
broader that the nominal spectral resolution for this observation and thus
possibly resolved. This would imply that the absorber is dispersed in velocity
space (see also SS). However we note that the resolution of the FOS using
the 4.3'' slit (as in this case) is not well­defined. In particular ¸ 50% of the
light is in very broad wings (Korista, private communication) and thus we
cannot be certain that the line is resolved from these data. As discussed
in section 2.3, the uncertainty in the Lyff EW is large, leading to a large
uncertainty in the HI column density: 13 ! logNHI ! 18 while the model
value ranges from 15:2 to 15:4 (Table 2). An actual HI column density in
the higher end of the observed range would imply that the heavy element
abundance in the X­ray/UV absorber is depleted. Conversely, if it is in the
lower end of the range enhanced metal abundance is implied. A Lyman edge
absorption would be observed if NHI is large (logNHI ? 16:3). The recent
HUT observations ( Spring 1995) will be able to detect such a Lyman edge.
The ASCA observations were made on 1993 July 28, while the HST obser­
vations were made between 1993 April and 1993 May. The close agreement
of U from the ASCA and HST data requires that the UV continuum flux is
similar in both observations. The optical light curve of NGC 5548 (Korista
et al. 1995) shows that the mean continuum level during the HST obser­
7

vations (f – (5100 š A) = 9:14 \Theta 10 \Gamma15 erg s \Gamma1 cm \Gamma2 š A \Gamma1 ) was indeed only 4%
below the continuum at the time of the ASCA observations (¸ 9:4 \Theta 10 \Gamma15
erg s \Gamma1 cm \Gamma2 š A \Gamma1 ).
One additional consistency check is possible. In a highly ionized system
such as this, magnesium is highly ionized (MgVI and higher) leaving no
magnesium in the MgII state (log fMgII ! \Gamma30) and thus MgII absorption
should not occur. This is consistent with the observations (sec. 2.2).
Thus we have a single consistent model that simultaneously, correctly pre­
dicts the fractional ionization of OVII, OVIII, NV, CIV, HI and FeXVII and
the lack of presence of MgII and other low ionization species. We conclude
that the UV and X­ray absorbers in NGC 5548 are one and the same.
3.3 Variability
Our photoionization model of the UV absorber in NGC 5548 must explain
the observed variability of the CIV absorption line (SS) assuming that their
relative measurements are correct (see section 2.3). As can be seen from
figure 5, the absorption line lies slightly off the linear part of the curve of
growth. The slope of the curve of growth in that region is about 0.5. Since
this is a log­log plot a factor of two change in the column density of the CIV
ion would result in a factor of 1.4 change in the EW of the line. In our model,
the variations in N ion are caused by variations in the ionization parameter
of the absorber (figure 4) which is directly proportional to the variations in
the ionizing flux. An increase in U results in a decrease in NCIV (figure 3).
Our measurements of the IUE data show the continuum at 1500 š A varying
from 2.0\Theta10 \Gamma14 to 4.7\Theta10 \Gamma14 erg cm \Gamma2 s \Gamma1 š A \Gamma1 , a factor 2.4, while the CIV
absorption line EW changes from 1.9\Sigma0.3 to 1.4\Sigma0.4 š A, consistent with the
model prediction of a factor 1.5.
During the HST observations the maximum continuum variations were
only a factor of ¸ 1:5 (Figure 6). There is no obvious variability observed in
the CIV absorption line EW. Unfortunately the errors in the CIV EW are
large, due to the inherent measurement difficulty discussed earlier (sec. 2.3)
The two solid lines indicate the variability predicted by our model based on
the observed continuum variations and both zero time lag and a time lag of
two days. The data are certainly consistent with the model (ü 2 = 6:05 for
35 degrees of freedom); however no strong conclusions can be drawn.
8

3.4 Physical Properties of the Absorber
The X­ray/UV absorber in NGC 5548 shows absorption features due to Lyff,
NV, CIV, OVII, OVIII and Fe­XVII. The X­ray and UV constraints can
now be combined to derive the physical properties of the absorber. It is
highly ionized (2:2 ! U ! 2:8) and has a large column density (NH =
3:8 \Theta 10 21 cm \Gamma2 ). The CIV recombination time must be shorter than the
variability response timescale of Ÿ4 days (SS), which sets a lower limit to
the density of the absorber; n ?
¸ 5 \Theta 10 5 t \Gamma1
4 cm \Gamma3 , where t 4 is the lag between
the continuum and absorption line variation in units of 4 days. The thickness
of the absorbing slab is then \Deltar !
¸ 8 \Theta 10 15 cm. It is outflowing with a mean
velocity of v out = 1200 \Sigma 260kms \Gamma1 as inferred by its blueshift with respect
to the systemic velocity of the galaxy (SS) and possibly dispersed in velocity
(FWHM ¸ 300km s \Gamma1 ).
The radial distance of the absorber r abs = (Q=4úUnH c) 1=2 is !
¸ 0.7 -- 0.8
t 1=2
4 pc given the above density constraint. The maximum depth of the CIV
absorption line during the IUE observations is greater than the continuum
level (SS). This implies that the absorber at least partially covers the CIV
emitting region whose size has been accurately determined for NGC 5548 =
7.5\Sigma0:5 lt. days. Combining these two limits gives 8 lt: days ! r abs ! 0:8 pc:,
i.e. 2 \Theta 10 16 ! r abs ! 2 \Theta 10 18 cm.
The lower limit on r abs limits the density to n! 5 \Theta 10 9 cm \Gamma3 . The density
is thus constrained to be 5 \Theta 10 5 t \Gamma1
4 ! n ! 5 \Theta 10 9 cm \Gamma3 . The radial distance
and density are not strongly constrained, nor are the model constraints in
conflict with those from the variability data.
Associated absorption has been observed in ¸ 10% of Seyfert 1 galaxies
(Ulrich 1988) which implies ¸ 10% covering factor of the absorber in all
Seyfert 1s. If we assume this covering factor, f 0:1 , applies to NGC 5548
then the mass of the absorber is M abs ¸ 20f 0:1 M fi . The mass outflow rate
can be calculated, by assuming uniform density: —
M out =M abs v out \Deltar \Gamma1 f 0:1 =10
f 0:1 M fi yr \Gamma1 . This is an upper limit in that the absorber is much thinner than
its radius from the center, so that the time averaged —
M out is likely to be much
smaller. The same rate is a lower limit in that we measure only the velocity
component in our line of sight (see sec. 4.4). This mass outflow rate is quite
a bit larger than the accretion rate of 0.08 M fi yr \Gamma1 needed to power the
L Bol = 5 \Theta 10 44 erg s \Gamma1 continuum source at 10% efficiency. The outflow
would carry a kinetic luminosity of —
M 2
out v 2
out =2 = ¸ 10 43 erg s \Gamma1 about fifty
9

times lower than the radiative luminosity of NGC 5548. A summary of the
properties of the absorber is given in Table 3.
4 Discussion
4.1 Comparison with Previous models
The combination of X­ray and improved UV constraints from HST have led
to derived conditions very different from those reported earlier by SS based
on the IUE data alone. SS necessarily assumed CIV to be the dominant state
of ionization (f CIV = 0.1) leading to U ¸ 10 \Gamma3 . Combining this with the lack
of MgII absorption led to NH ! 10 20:5 cm \Gamma2 . However in our model these
parameters are constrained: 2.2! U !2.6, \Gamma3:4 ! fCIV ! \Gamma3:9. We are thus
led to conclude that the absorber is highly ionized and has a large column
density (NH = 3:8 \Theta 10 21 cm \Gamma2 , Fabian et al 1994a).
4.2 Generalizing UV/X­ray Absorbing Outflows
The physical properties of the long known UV absorber in NGC 5548 are
now well constrained with the identification of a UV/ X­ray absorber. We
can therefore generalize our unification of UV and X­ray absorbing outflows
for lobe­dominated radio­loud quasars to include radio­quiet Seyfert galaxies,
and most likely to all associated absorbers in AGN.
A strong correlation has been observed between the radio properties and
associated absorption line properties of AGN: all Broad Absorption Line
Quasars (BALQSOs) are radio­quiet (Stocke et al. 1992) and all MgII asso­
ciated absorption line quasars are lobe­dominated and radio­loud (Aldcroft
et al. 1994). Associated CIV absorbers are a mixture of lobe­ and core­
dominated radio­loud quasars, and radio quiet quasars. It is of interest to
note that even though NGC 5548 is radio quiet, it is unusual in that its
radio luminosity is dominated by the extended radio flux, which has a larger
physical size than in any other radio quiet Seyfert galaxy (Wilson & Ulves­
tad 1982). In this respect it is similar to a lobe­dominated quasar, providing
a link between lobe­dominated, radio loud AGN with associated absorption
and radio­quiet AGN, perhaps also the BALQSOs. It further suggests that
AGN with UV/X­ray absorbers are edge­on. To have an inclination indicator
10

in radio­quiet AGN would be most valuable, and this possibility should be
studied further.
4.3 Comments on variability in NGC 3227 and MCG­
6­30­15
Variable X­ray spectra and the existence of ionized absorbers have been ob­
served in two more AGN with ASCA: NGC 3227 (Ptak et al. 1994) and
MCG­6­30­15 (Fabian et al. 1994b). In both cases the variations in the prop­
erties of the ionized absorbers were deduced to be incompatible with the
variations seen in the continuum.
In NGC 3227 (Ptak et al. 1994) rapid flux variations were observed with
a time scale of ¸ 10 4 seconds. An OVI absorption edge was detected with
opacity Ü ¸ 0:75. The model fits to the spectra in high and low flux states
showed that (1) for the same value of photon index \Gamma in both high and low
states, NH and U decrease when the luminosity increases; (2) for different
\Gammas, the same NH and U can be fitted in both high and low states. Both of
these models seem unphysical and hence the authors suggested that a warm
absorber model may be too simple.
This, however, may not be the case, since it does not allow for delays due
to the recombination time of the absorber. The best fit Ü in NGC 3227 is
remarkably similar in both high and low states implying that the ionization
structure of the absorber has not changed even though the luminosity has
changed. This implies a recombination time of OVI ? 10 4 seconds. For OVI,
the recombination time, T = 4:3 \Theta 10 4 n \Gamma1
6 sec. (Shull & Van Steenberg, 1982)
where n 6 is the density in the units of 10 6 cm \Gamma3 . This implies a density of
the warm absorber, n! 10 6 cm \Gamma3 . With the model value of U ¸ 0.05 this
places the absorber at r= 10 19 cm, well outside the BELR (which would have
a radius of 7 \Theta 10 15 cm, assuming an L 1=2 scaling from NGC 5548).
It should be noted that the physical parameter which responds to contin­
uum variations is Ü rather than U. Indeed, if Ü remains the same while the
luminosity increases, the fitted value of U, which assumes the system is in
equilibrium, would necessarily decrease.
A similar apparent discrepancy was observed in MCG­6­30­15 (Fabian
et al. 1994b) which has an OVII absorption edge. In this case Ü decreased
significantly over 23 days. The luminosity also decreased over this period,
11

while the fitted value of the ionization parameter increased. Whether the
change in Ü correlates or anti­correlates with the change in luminosity de­
pends upon U (see figure 3). The recombination time of OVII is ¸ 0:28n \Gamma1
6
days (Shull & Van Steenberg, 1982). The data imply that tŸ 23 days and so
n – 10 4 cm \Gamma3 , a very reasonable constraint.
We argue that the warm absorber models in NGC 3227 and MCG­6­30­15
are not unphysical and that the data do not require more complex models.
The information provided by the variability not only helps us understand
this fact, but also provides an important constraint on the density of the
absorber.
The low value of U for NGC 3227 (cf. 2.6 for NGC5548) implies a wide
range of ionization states from object to object for the associated absorbers
in radio­quiet AGN, as in radio­loud.
4.4 Implications for the Nature of the Outflowing
Material
Now that we understand the physical properties of the absorber, we can begin
to ask astrophysical questions.
The lack of differential time delays between the red and blue wings of the
broad line profiles shows that the BELR motions are not primarily radial
(Korista et al. 1995). Yet the absorbing material outside the BELR shows
large outflows, both in NGC5548 and in the other UV absorbers (Ulrich
1988). How is the orbital or random motion at BELR radii converted to an
organized flow? What is the driving force? The continuum radiation pressure
acting on the partially ionized gas (Mathur et al. 1994, Turner et al. 1994)
may play an important role. In a more specific scenario, Glenn, Schmidt &
Foltz (1994) and Aldcroft et al (1995), have independently suggested that in
BALQSOs and radio­loud quasars respectively we are looking almost edge­
on at material blown off a dusty disk (which itself may be the BELR) by
radiation pressure. Since there is some evidence that NGC5548 too is edge­
on the same picture may apply there.
Material escaping vertically from an edge­on disk would then follow a
path that is not purely radial as it is accelerated by the continuum radia­
tion pressure. If the material originates at only a restricted range of radii
in the disk, this could explain a number of otherwise puzzling features. In
12

particular, the limited thickness of the absorber in NGC5548 cannot be un­
derstood in any continuous radial wind model. The X­ray column density
directly rules out significant additional accelerated or decelerated material,
unless it is fully stripped through iron (Nandra et al 1991). Only by making
the wind intermittent, or by allowing for a velocity component in the plane
of the sky (so moving the rest of the flow out of our line of sight) can we
keep the thickness small.
Another consistency check is to confirm that the persistence of the ab­
sorber is consistent with the large observed outflow velocity (1200\Sigma200 km s \Gamma1 ).
In the three years separating the ROSAT and ASCA observations no sig­
nificant change in the OVII/OVIII opacity was detected. (ROSAT Ü =
0:35 \Sigma 0:13 (Nandra et al. 1993); ASCA Ü = 0:38 +0:08
\Gamma0:09 (Fabian et al. 1994a)).
A formal 3oe upper limit for a change in Ü is 0.27 (i.e. ¸80%). The dis­
tance traveled by the absorber in 3 years is 1.2\Theta10 16 cm. In order to limit a
change in U to !80%, this must represent a !34% change in distance from
the ionizing continuum (assuming a constant density system.) This leads to
r abs – 3:5 \Theta 10 16 cm, slightly larger than the minimum distance determined
from the size of the BELR (2\Theta10 16 cm, section 3.4), and so consistent with
our earlier derived parameters for the absorber.
The first IUE observations, some 15 years ago, also showed a CIV absorber
at a similar velocity and depth (section 2.2). These observations provide
more constraints. The outflow cannot be intermittent if it has persisted
for 15 years. Assuming a constant mean velocity for the absorber leads to a
present minimum distance r abs ? 5 \Theta 10 16 cm, and a corresponding maximum
density n! 8 \Theta 10 8 cm \Gamma3 . These are both consistent with the derived absorber
parameters (section 3.4). However, if the apparent width of the absorption
line in the HST data is real (¸ 300 km s \Gamma1 ) and represents the internal
velocity dispersion of the gas over the past 15 years, then it will have led to
a thickening by ¸ 10 16 cm over 15 years, comparable with our estimate of
the present absorber thickness (! 8 \Theta 10 15 cm, section 3.4). In which case
the absorber would have had zero depth when IUE first observed it, which
is unlikely since otherwise it has not changed significantly. This difficulty
disappears if we are looking through a flow with a component in the plane
of the sky, since then we are seeing a steady state flow crossing our path.
This physical picture, derived from observations, is very similar to the wind
models of BALQSOs by Murray et al. (1995).
Testing the model further in NGC5548 could be achieved by measuring
13

the recombination time, and hence the absorber density. This implies a small
sampling interval (! 1 day) accompanied by continuum changes by at least
a factor of two. Monitoring the ionization parameter at a given continuum
flux over a number of years to search for density changes would also test the
model. High spectral resolution observations (e.g. with the HST GHRS)
could examine the velocity structure of the absorber, and perhaps measure
b values of any components. Tests for the edge­on nature of AGN with
UV/X­ray absorbers also need to be pursued.
5 Conclusions
In this paper we have shown that the X­ray and UV absorption in NGC5548
can be explained quantitatively by having the same material produce the
absorption features due to OVII, OVIII (in X­rays), CIV, NV, HI (in the
UV), and not produce observable features due to FeXVII and MgII and other
low ionization lines. This simple model passes further tests by not violating
the size and density limits imposed by the reverberation mapping data. We
conclude that the X­ray and UV absorption does indeed arise from the same
material.
Combining the improved X­ray and UV constraints from ASCA and HST,
together with the reverberation mapping variability constraints leads us to
understand the physical properties of the absorber. The absorber is highly
ionized (2.2! U !2.8), has high column density (NH = 3:8 \Theta 10 21 cm \Gamma2 ),
low density (5 \Theta 10 5 t \Gamma1
4 ! n ! 5 \Theta 10 9 cm \Gamma3 ), and is situated outside the
CIV emitting region (2 \Theta 10 16 ! r abs ! 2 \Theta 10 18 cm). The gas is outflowing
with a mean velocity of 1200\Sigma200 km s \Gamma1 and has a corresponding kinetic
luminosity of ¸ 10 43 ergs \Gamma1 .
We can now generalize our unification of UV and X­ray absorbing outflows
from the lobe dominated radio­loud quasars to include radio quiet Seyfert
galaxies. This may also provide a link to the radio quiet BALQSOs. This
analogy suggests that the UV/X­ray absorbers in radio­quiet AGN may be
viewed close to edge­on, which would be a valuable known parameter if it can
be independently supported. Now that we understand the physical properties
of the absorber, we can begin to ask astrophysical questions, the dynamics
of the outflow and its role in the circumnuclear region of AGNs. A scenario
in which the absorbing material comes off a disk, and is accelerated by the
14

radiation pressure of the continuum source may explain the persistence of
the absorber in spite of its large velocity and thin shell­like geometry.
We thank Brad Peterson and Kirk Korista for valuable discussions and
for early access to the HST spectra. We also thank Jonathan McDowell
for his Tiger software. This work was supported in part by NASA Grant
NAGW5­2201 (LTSA).
15

Table 1: Absorption Line Parameters
Line – obs š A EWš A EWš A
HST a IUE
CIV 1548.2 1572.4 0.30 \Sigma0.10 1.6\Sigma0:3 b
1550.8 1575.4 0.08\Sigma0.03
NV 1238.8 1260.7 0.04\Sigma0.01 \Gamma
1242.8 1262.5 0.10 \Sigma0.03
Lyff 1234.1 0.5\Sigma0:3 \Gamma
a. in mean spectrum.
b. SS mean value scaled by our correction factor, 0.6 (see text).
Table 2: UV and X­ray constraints on b a
Ion log(W – =lambda) log(N–f) b
(from UV) (from X­ray) (km s \Gamma1 )
CIV \Gamma3:61\Sigma0.02 8.9\Gamma9.4 40\Gamma50
NV \Gamma3:92\Sigma0.03 8.5\Gamma9.1 18\Gamma40
HI \Gamma3:8 +0:3
\Gamma0:2 9.9\Gamma10.1 20\Gamma60
a. Ranges are \Sigma1oe.
16

Table 3: Physical conditions of X­ray/UV absorber
Parameter Constraint Derived from:
Neutral hydrogen NH¸ 3:8 \Theta 10 21 cm \Gamma2 X­ray
column density, NH (Fabian et al. 1994)
Ionization Parameter, U 2.2 ! U ! 2.8 Ü(OVII,OVIII) X­ray
(Fabian et al. 1994, Nandra et al. 1993)
Radial Distance, r abs 2 \Theta 10 16 cm ! r abs ! 2 \Theta 10 18 cm LL: BELR covered (Peterson et al. 1993)
UL: n, U
Density, n 5 \Theta 10 5 cm \Gamma3 !
¸ n ! 5 \Theta 10 9 cm \Gamma2 LL: recombination time ! 4 days (SS)
UL: r abs , U
Outflow velocity, v 1200\Sigma260 km s \Gamma1 Line redshift (SS)
Thickness, \Deltar !
¸ 8 \Theta 10 15 cm NH , n
Mass ¸ 20f 0:1 M J n, r abs , \Deltar
(covering factor, f 0:1 = 0.1)
17

References:
Aldcroft, T & Elvis, M. & Bechtold, J. 1995 MNRAS submitted.
Allen, C. W. 1973, Astrophysical Quantities [The Athlone Press]
Arav, H. & Li, Z. 1994, Ap.J. 427, 700
Branduardi­Raymont, G. 1986 in The Physics of Accretion onto Compact
Objects, eds: K. O. Mason, M. G. Watson & N. E. White., p. 407 [Springer­
Verlag; Berlin]
Branduardi­Raymont, G. 1989 in Active Galactic Nuclei eds: D. E. Oster­
brock & J. S. Miller, p. 177 [Kluwer]
Elvis, M., Wilkes, B. J., McDowell, J. C., Green, R. F., Bechtold, J., Will­
ner, S. P., Cutri, R., Oey, M, S., & Polomski, E. 1994 ApJS, 95, 1
Clavel, J. et al. 1991 ApJ 366, 64
Clavel, J. et al. 1992 ApJ 393 , 113
Fabian, A., Nandra, K., Brandt, W., Hayashida, K., Makino, F. & Yamauchi,
M., 1994a in New Horizons in X­ray Astronomy, eds. F. Makino and T.
Ohashi, [Universal Academy Press: Tokyo], p573
Fabian, A. et al. 1994b, PASJ, 46, L59
Ferland, G. F. 1991 ``HAZY'', OSU Astronomy Department Internal Report
Grevesse, N., & Andres, E. 1989 in Cosmic Abundances of Matter, AIP Con­
ference Proceedings, 183, 1. Ed. C. J. Waddington (New York: AIP)
Korista, K. et al. 1995, ApJS, in press (April 15)
Mathews, W. G. & Ferland, G. J. ApJ 323, 456
Mathur, S., Wilkes, B., Elvis M. & Fiore, F. 1994, ApJ, 434, 493
Mathur, S. 1994, ApJL, 431, L75
Murray, N., Chiang, J., Grossman, S. A., & Voit, G. M., 1995, ApJ, in press
Nandra, K. et al. 1991, MNRAS 248, 760
Nandra, K. et al. 1993, MNRAS 260, 504
Peterson, B. M. et al. 1992, ApJ 392, 470
Ptak, A., Yaqoob, T., Serlemitsos, P. J., Mushotzky, R., & Otani, C. 1994,
Ap.JL, 436,
Shull, M. J. & Sachs, E. R. 1993, ApJ 416, 536
Shull, M. J. & Van Steenberg, 1982, ApJS, 48, 95
Spitzer, L. 1978 Physical Processes in the Interstellar Medium, [Wiley­Interscience],
pp 46 -- 57
Stocke, J. T., Morris, S. L., Weymann, R. J. & Foltz, C. B. 1992 Ap. J. 396,
487
18

Turner, T. J., Nandra, K., George, I. M., Fabian, A., & Pounds, K. A. 1994,
ApJ 419, 127
Ulrich, M. H., 1988, MNRAS, 230, 121
Ulrich, M. H. & Boisson, C. 1983, ApJ, 267, 515
Ward, M., Elvis, M., Fabbiano, G., Carleton, N. P., Willner, S. P. & Lawrence,
A. 1987, Ap. J. 315, 74
Wiese, W. L., Smith, M. W., & Glennon, B. M. 1966 Atomic Transition
Probabilities [U.S. Government Printing Office]
Wilson, A. & Ulvestad, J. 1982, ApJ, 263, 576
Wu, C., Boggess, A., & Gull, T. 1981, ApJ, 247, 449
19

Figure Captions:
Figure 1: An IUE spectrum of NGC 5548 (an average of SWP35029,
SWP35284, SWP35461, SWP35636, SWP35849, SWP36018, SWP36369,
SWP36567 & SWP36715) showing the presence of Lyff absorption line.
Figure 2: An HST mean FOS spectrum of NGC 5548 showing (a) Lyff, NV
and (b) CIV absorption troughs. In (a) we also indicate our maximal and
minimal estimates of the absorption line strength (dashed line). The dotted
line indicates the extrapolation to the observed line profile.
Figure 3: Spectral Energy Distribution of NGC 5548: the data are from
Ward et al. (1987), SS (1993) and Nandra et al. (1991). The solid line repre­
sents our adopted SED. The dotted line is 150,000 K black body spectrum.
The dashed line is a `standard' AGN continuum shown for comparison (see
text).
Figure 4: Ionization fractions of OVII, OVIII, CIV, NV and HI as a function
of U. The thick lines mark the observed ranges for OVII and OVIII (ASCA).
N: HST values for CIV and NV; ffi: IUE range. The HST range for HI is
large (see text) represented by the thick curve. The vertical lines define the
best fit model parameter: 2.2 ! U ! 2.8.
Figure 5: The curve of growth for b values from 20 to 80 km s \Gamma1 in steps
of 20 km s \Gamma1 . Horizontal lines are for the observed values of absorption lines
in the mean HST spectrum; (a) CIV (b) NV and (c) Lyff.
Figure 6: Variability in the continuum and CIV absorption line EW during
HST observations. Continuum is in arbitrary units. The dashed horizontal
line is the mean EW. Two solid lines represent the predicted variability for
no time lag and for a time lag of two days.
20

1180 1200 1220 1240 1260 1280
0
NV(Abs.)
NV
21

Wavelength
Flux
1520
1540
1560
1580
1600
1620
5
10
15
20
b
CIV Absorption
22

Wavelength
Flux
1200
1220
1240
1260
5
10
15
20
25
30
35
a
Geocoronal Lya
Lya Absorption
23

24

25

26

10 20 30 40
0
0.2
0.4
0.6
0.8
Day
27