The Complex X-Ray Absorbers of NGC 3516 Observed by BEPPOSAX

In this paper we present the analysis of two broadband (0.1-150 keV) BeppoSAX observations of the Seyfert 1 galaxy NGC 3516. The two observations were taken 4 months apart, on 1996 November 8 and 1997 March 12. We report a dramatic change in the degree of obscuration of the central source between the two observations and propose, as possible explanations, transient absorption by either a stationary-state cloud of cold gas crossing the line of sight or a varying-state, initially neutral and dense amount of expanding gas with decreasing density and therefore decreasing opacity. We also report the detection of a second highly ionized absorber/emitter, which causes deep Fe XVII-XXII K edges at ~7.8 keV to appear in both of the BeppoSAX spectra of NGC 3516 and possibly produces the soft X-ray continuum emission in the 1 keV blend of Fe L recombination lines detected during the epoch of heavy nuclear obscuration.


INTRODUCTION
NGC 3516 is a bright ergs (F 2h10 keV \ 7.8 ] 10~11 cm~2 s~1 ; Reynolds 1997) nearby (z \ 0.009, D \ 54 Mpc, for km s~1 Mpc~1) Seyfert 1 galaxy, which is H 0 \ 50 known to host complex and variable systems of both mildly and highly ionized absorbers in the UV and in the X-ray. In the UV band IUE spectra of NGC 3516 taken between 1978 June and 1989 October (Ulrich & Boisson 1983 ;Voit, Shull, & Begelman 1987 ;Walter et al. 1990 ;Kolman et al. 1993) show the presence of at least two systems of C IV resonant-absorption lines ; one is broad (D2000 km s~1) and variable ; the other is narrow (D100 km s~1) and constant. Both are produced by ionized and outÑowing gas that obscures both the continuum and at least part of the broad emission lines (BEL). Four years later the broad absorption line component was no longer detected in the spectra of NGC 3516 taken with the IUE (from 1993 ; Koratkar et al. 1996) or later with the Hopkins Ultraviolet Telescope (in 1995 ; Kriss et al. 1996a) and the Goddard High Resolution Spectrograph (GHRS) Crenshaw, Moran, & Mushotzky 1998 ;Goad et al. 1999, hereafter GEA99), suggesting either a change in the covering factor or in the ionization structure of the gas responsible for that feature or an increase in the emissivity of the C IV BEL. The complex X-ray absorber was observed by ROSAT (in 1992 ;Mathur, Wilkes, & Aldcroft 1997 ;hereafter MWA97) and ASCA (Kriss et al. 1996b ;Reynolds 1997 ;George et al. 1998) through the detection of two strong absorption features at the energies of the O VII and O VIII K edges (0.74 and 0.87 keV, respectively). The optical depths measured in the ROSAT and ASCA spectra using phenomenological twoedge models are similar to one another (q O VII D 0.6, q O VIII D 0.3).
Neutral absorption has not been unambiguously found in this source. While the 1979 Einstein observation (Kruper, Urry, & Canizares 1990) did not detect absorption by neutral gas, a subsequent low energy resolution EXOSAT observation in 1985 showed the presence of cold gas, with cm~2 a †ecting the low-energy spectrum (Ghosh N H D 1022& Soundararajaperumal 1991. A 1989 Ginga observation (simultaneous to IUE ; Kolman et al. 1993) showed a lowenergy cuto † of the intrinsic power law, indicating heavy absorption of the nuclear continuum along the line of sight. However, because of the limited energy coverage of Ginga (2È20 keV bandpass), it was not possible to establish the ionization state of that absorber, which was consistent with either a high column of neutral gas cm~2) (N H Cold \ 4 ] 1022 or a lower column of more highly ionized gas. In this paper we present data from two broadband BeppoSAX (Boella et al. 1997a) observations of NGC 3516. Preliminary results were presented by Stirpe et al. (1998). The paper is organized as follows : In°°2 and 3 we present our analysis of the data, while in°4 we discuss our main results. Finally, we present our conclusions in°5.

DATA REDUCTION AND ANALYSIS
NGC 3516 was detected with a good signal-to-noise ratio in the three main narrow-Ðeld instruments (NFI) on BeppoSAX, spanning 0.1È120 keV. In Table 1 dates and exposure times of the two BeppoSAX observations of NGC 3516 are presented, along with the source count rates measured by the Low-Energy Concentrator Spectrometer (LECS ; Parmar et al. 1997), Medium-Energy Concentrator Spectrometer (MECS ; Boella et al. 1997b), and Phoswich Detection System (PDS ; Frontera et al. 1997). Data from these three instruments were screened following standard criteria, as detailed in Fiore, Guainazzi, & Grandi (1999). In particular, PDS data were screened using Ðxed rise-time thresholds. Data from the three MECS units were merged to improve the statistics. The scientiÐc products for the 283 analysis are extracted from the equalized event Ðles, using both the standard FTOOLS package (Version 4.2) and the new release (Version 1.1) of the CIAO (Chandra Interactive Analysis of Observation) software (M. Elvis et al. 2001, in preparation ;Siemiginowska & Fruscione 20008). LECS and MECS source counts were extracted from circular regions of the 6A radius centered on the source. The same extraction regions were used to extract background counts from the long-exposure "" blank Ðelds ÏÏ provided by the BeppoSAX Science Data Center.9 PDS source and background counts were collected during the on-and o †-source modes of the instrument, respectively. For all the instruments, spectra were binned following the instrumental resolutions and allowing a maximum of three channels per resolution element.
2.1. Model-Independent Evidence for L ong-T erm Spectral V ariability NGC 3516 underwent moderate (factor of 1.5È2) 0.1È10 keV Ñux variations during both BeppoSAX observations (on a timescale of hours) ; however, the ratios of the soft (0.1È2 keV)-to-hard (2È10 keV) light curves were consistent with being constant, indicating no signiÐcant spectral variability across these energy bands on a 1 day timescale (the duration of each observation).
A dramatically di †erent behavior was observed on the longer 4 month timescale between the two observations, during which the source experienced strong spectral variability. The 0.1È2 keV count rate increased by a factor of D4, while the corresponding increase in the 2È10 keV band was only 1.9 (Table 1). Figure 1 shows the model-independent ratio between the raw LECS (0.1È2 keV), MECS (1.5È10 keV), and PDS (13È100 keV) spectra for the two BeppoSAX observations (1996 November and1997 March). The amplitude of nonlinear e †ects in the responses of the NFIs (potentially a †ecting the results of such an analysis) is, at all energies, much smaller than the amplitude of the features visible in this ratio (Boella et al. 1997b ;Frontera et al. 1997 ;Parmar et al. 1997).
In the ratio in Figure 1 the 1È5 keV energy band is a smoothly increasing function of energy. It then Ñattens at keV, around a value of D0.6. In the PDS band the E Z 5 ratio is D0.8. This implies a steepening of the ratio between 10 and 13 keV. A sharp emission feature around 1 keV is visible superimposed on these smooth changes. All this strongly suggests that some, and perhaps all, of the several components that make up the 0.1È100 keV spectrum of NGC 3516 varied independently of each other from one observation to the other. In particular, the smooth and monotonic rising of the ratio up to D5 keV, along with its Ñattening above this energy and up to 10 keV, suggests a drastic decrease in the degree of obscuration of the central X-ray source between the two observations rather than an intrinsic variation of the slope of the primary X-ray continuum. In this case the asymptotic value of D0.6 around which the ratio Ñattens above D5 keV would be the intensity increase factor of the intrinsic X-ray power law between the two observations. The higher PDS count ratio of D0.8 may suggest a higher relative intensity of a Compton-reÑection component in the 1997 observation. There is no clear indication of a similar e †ect at the energy of the Fe K line. The emission feature at D1 keV may be the signature of iron L emission, visible only during the 1996 observation, and therefore either external to the nuclear environment and detectable by the BeppoSAX LECS only when the soft nuclear continuum was heavily obscured or internal to the nuclear environment but only partially obscured by the amount of neutral (or mildly ionized) gas obscuring the X-ray source in 1996. Intrinsic variation of the intensity of the Fe L complex in the hot plasma is, of course, a further possibility. We use this scenario to guide the modeling and Ðtting of the spectra in°°3 and 4.

SPECTRAL MODELING AND FITTING
We have performed global modeling and Ðtting of the 0.1È150 keV spectra of both BeppoSAX observations of NGC 3516 using the XSPEC, Version 10.0 package. In Table 2 we summarize the results of this analysis. Errors are quoted at the 90% conÐdence level for one interesting parameter (i.e., *s2 \ 2.71 ; Avni 1976). The cosmology NOTE.ÈBoth models include a cuto † power law plus Galactic absorption, a Gaussian line plus cold reÑection, and a highly ionized absorber. 96BF also includes a partial covering absorber (C sou D 0.95) and a thermal component. 97BF instead includes a "" warm absorber.ÏÏ See°3 for details. a Errors are calculated Ðxing the slope at the best-Ðt value. Vol. 544 used is km s~1 Mpc~1 and For all the H 0 \ 50 q 0 \ 0.5. absorption (neutral and/or ionized) models that we use in this paper, we adopt solar abundances as estimated by Grevesse & Anders (1989) and Grevesse & Noels (1993). Photoionization models for Ðtting purposes include photoelectric absorption but not resonant absorption and gas emission and have been built following Nicastro et al. (2000). Finally, we use pure Compton-reÑection models built by running iteratively the routine PEXRAV (Magdziarz & Zdziarski 1995) for the inclination angle of 30¡ (consistent with the best-Ðt interval measured ; see°3.1.3) and a range of values of the slope ! of the incident continuum and the e-folding energy of the high-energy exponential cuto †. For Ðtting E c purposes we always combine the Compton-reÑection model with a cuto † power law, linking the two ! and values to E c the same values and leaving the two normalizations free to vary independently.
We Some are common to both spectra : (1) a linelike feature at D6.4 keV, (2) a deÐcit of counts at D8 keV, and (3) a systematic deviation across the PDS band. However, at low energy keV) the shapes of the residuals are quite (E [ 4 di †erent from one spectrum to the other. Our best-Ðtting parameterizations of N96 and M97 are thus quite di †erent and are summarized in Table 2. The two best-Ðtting models (hereafter 96BF and 97BF, respectively) include several common components, which make up our "" base model ÏÏ : (1) a power law with a highenergy cuto †, (2) a neutral absorber (to account for Galactic absorption along the line of sight), (3) a hot photoionized absorber (HA ; see Table 2) parameterized by a column density and an ionization parameter U (to account for N H the 8 keV deÐcit), (4) a Gaussian emission line (to account for Ka iron Ñuorescence), and (5)  components that are not held in common : (6) a neutral absorber only partially obscuring the direct view of the primary X-ray power law, (7) soft emission by an optically thin plasma for the N96 spectrum, and (8) a second, more mildly ionized absorber for the M97 spectrum.
3.1. T he "" Base Model ÏÏ Our estimates for the parameter values of our base model were consistent with each other between the 96BF and 97BF parameterizations (see Table 2). In the following discussion we describe and discuss the individual components of this model, referring to the numerical results obtained for the M97 spectrum, the one with higher statistics.
3.1.1. T he Intrinsic Power L aw and the High-Energy Cuto † The 0.1È100 keV primary continuum of NGC 3516 is well described by a simple power law F(E) \ AE~! [F(E) in photons s~1 cm~2 keV~1], with ( Table 2). ! \ 2.04~0 .07 0.06 This component varied in Ñux by D70% between the two observations [i.e., A(N96)/A(M97) \ 0.6] without changing in shape. We then conclude that little, if any, of the dramatic spectral variability experienced by the source can be due to variation of the primary continuum, either in shape or intensity. We could only estimate a lower limit (E [ 450 keV) on the e-folding energy of the high-energy cuto † and were not able to measure any signiÐcant change of this parameter between the two observations.

Iron L ine and Compton ReÑection
Both the iron-line and the Compton-reÑection components were required by our data at high signiÐcance. Removing each of these components, one at time, from the 97BF model and reÐtting the data gives, respectively (with *l indicating the number of free parameters eliminated by the 97BF model), *s2 \ [25 (*l \ 3) and *s2 \ [41 (*l \ 1).
The best-Ðt energy of the iron emission line (E \ 6.41^0.15 keV) is consistent with that of Fe IÈXXII Ka transitions. When modeled with a simple Gaussian (as in our adopted best-Ðtting parameterization 97BF), the line is consistent with being narrow, with a 90% upper limit on its width of 0.27 keV. We note that the energy resolution of the BeppoSAX-MECS is only D500 eV at 6.4 keV. Furthermore, an accurate determination of the exact shape of the iron-line proÐle is hampered by the presence in the data of the D8 keV edgelike absorption feature. However, we have checked the consistency of our data with more complex scenarios, replacing the Gaussian emission line in the 97BF model with a relativistically broadened and distorted emission line from the accretion disk (the model DISKLINE in XSPEC ; Fabian et al. 1989), as suggested by previous ASCA observations (Nandra et al. 1997b(Nandra et al. , 1999. We Ðxed the internal radius on the accretion disk at 6 gravitational radii, the energy of the emission-line centroid at 6.33 keV (the best-Ðt energy in 97BF), and the spectral index of the emissivity law at [2 and reÐtted the data. The result was inconclusive. We obtained the same s2/l \ 86/95 as in 97BF. The outer radius on the disk was only barely constrained and larger than 11.5 gravitational radii. All the other model parameters were consistent within the errors in the 97BF measurements. We also checked the extreme hypothesis that all of the negative residuals visible in Figure  2 at D8 keV could be explained in terms of a very distorted blue proÐle of the iron line, eliminating the HA component and reÐtting the data. This yielded a worse s2 than 97BF (*s2 \ [18 for *l \ 3) and accounted only partially for the absorption feature at D8 keV. We then concluded that a relativistically broadened and distorted emission-line proÐle with no Fe XVIIÈXXII K absorption is not a necessary component for the description of our data.
Thanks to the broad band of BeppoSAX, we were able to detect a bump in the 10È30 keV data (see Figs. 2a and 2b). We model this bump as being due to Compton reÑection of the primary power law o † a slab of almost neutral matter reradiating the isotropic incident X-ray continuum. During both the 1996 and 1997 observations we measured relative fractions of reÑection consistent with each other (Table 2) and fully consistent with the value of 1 corresponding to a solid angle of 0.5 as seen by a central isotropic source of primary radiation. Their ratio, is consistent with 0.8~0 .2 0.7, the value derived in°2.1 from our model independent analysis. If the same matter were also producing the Ka Ñuorescence iron line (but see also°2.1), we would expect the equivalent width of this line to be D100È150 eV (e.g., Mushostzky, Done, & Pounds 1993), consistent with the measured value (see Table 2).

T he "" Hot ÏÏ Absorber
Both the N96 and M97 spectra of NGC 3516 show a clear absorption feature around 8 keV (Fig. 2). A neutral reÑection component alone cannot explain the edge feature at D8 keV, even allowing the inclination disk parameter to vary in a wide range (from 0¡ to 60¡). We accounted for this feature by including a hot photoionized absorber (HA) in 96BF and 97BF, obscuring the line of sight. The same model was used for NGC 3516 to Ðt the absorption feature detected at the same energy (E D 8 keV) by Ginga (Kolman et al. 1993). We stress again that the model used to Ðt the data includes neither gas emission nor resonant absorption. Removing this component from both 96BF and 97BF and reÐtting the data gives, respectively, *s2 \ [13 and *s2 \ [29 for *l \ 2.
The best-Ðtting parameters U and do not vary (within N H the errors) between the two observations (Table 2), and their values indicate that a high column density [N H HA \ (3.2~0 .6 0.5) cm~2, very similar to the value found by Kolman ] 1023 D 3 ] 1023 cm~2] of highly ionized gas, is (U \ 224~8 3 131) absorbing the primary continuum of NGC 3516 in both data sets. In such an absorber elements lighter than oxygen and neon are fully ionized and therefore do not imprint any visible features on the emerging spectrum. However, iron is still not fully ionized and contributes strongly to the opacity of the absorber at the energies of the photoelectric K and L absorption edges of Fe XVIIÈXXII. Absorption by Fe XVIIÈ XXII K imprints deep absorption edges at D7.8È8.0 keV on the emerging spectrum, while L absorption by the same species a †ects the whole 1È5 keV band, gently Ñattening a low-resolution spectrum rather than producing visible sharp and resolved single absorption features on it (unless Fe L resonant-absorption lines dominate on recombination lines).10 Fitting low-resolution data with a phenomenological single-edge model, accounting for the D8 keV absorption feature, would result in a wrong estimate of the slope of 10 The relative contribution of Fe L resonant-absorption lines, compared to emission lines, depends critically on the geometry and dynamics : if the gas has a high microturbulence velocity and a lower or comparable bulk velocity, then Fe L resonant absorption dominates and imprints a visible "" 1 keV ÏÏ feature on low-resolution spectra (Nicastro, Fiore, & Matt, 1999a). See also°4.2. the intrinsic 0.1È100 keV power law, which would appear to be Ñatter by *! \ 0.28^0.13. Ionized, and possibly relativistically smeared, edgelike features at energies greater than 7.1 keV have often been detected in the hard/low state of Galactic black hole candidates, where the accretion disk is supposed to be hotter than that of active galactic nuclei (AGNs) by a factor of D30 (Ross, Fabian, Brandt 1996 ;Done & 1999). In this contest to further test the Z 0 ycki uniqueness of our interpretation, we applied an ionized reÑection model, which possibly also can account for an ionized edge feature to our data. The *s2 between a model with an ionized reÑection (model PEXRIV in XSPEC ; Magdziarz & Zdziarski 1995) and the best Ðt (BF) is *s2/ *l \ 25/1 and *s2/*l \ 23/2 if the iron abundance parameter is left free to vary. The variation of the inclination angle of the ionized reÑector (h \ 30¡, 45¡, and 60¡) does not inÑuence the s2 test. Moreover, the expected ionized iron line at 6.7 keV, consistent with the ionization degree of the edge, is not detected even though a ionized line could have been missed because of the low-resolution instrument.
The preferred interpretation is thus in terms of a highionization, high column density absorber, which can simultaneously account for both the D8 keV edge and the overall continuum shape at medium energies, and, most important, it provides the statistically better description of the available data.
3.2. Departure from the "" Base Model ÏÏ As pointed out in°3, the base model does not provide a fully satisfactory description of either the N96 or the M97 spectra. Additional components are present in our best-Ðtting parameterizations 96BF and 97BF. We describe them in the following sections.

T he "" Cold ÏÏ Absorber and the Soft T hermal Emission in N96
To model the 0.1È5 keV residuals of Fig. 2b, we added to our base model a neutral absorber of column density N H Cold, which only covers a fraction of the nuclear X-ray C sou source. Fitting the N96 data with this model produces a signiÐcative improvement in the s2 (*s2/*l \ 73/2) and gives cm~2 and N H Cold \ 2.31~0 .15 0.33 ] 1022 0.7 \ C sou \ 1. (We note that when using an ionized absorber model, the ionization parameter is consistently zero while the column density of the warm gas is indicating absorption by DN H Cold, neutral gas as the most likely explanation.) Nevertheless, the residuals continue to show a clear linelike, unresolved feature around 1 keV (Fig. 3, upper panel).
At this energy recombination L lines from Fe XVIIÈXXIV are expected to strongly contribute to the emission from either a collisionally ionized plasma (CIP) with a temperature of D1È2 keV or from a photoionized plasma (PIP) with a high-ionization degree (U D 200) and column density cm~2). Only for Ðtting purposes we use here a (N H D 1023 CIP model, while a more accurate discussion on the physics of this component and on its possible identiÐcation is deferred to°4.2. We added a CIP component (the model MEKAL in XSPEC ; Mewe, Gronenschild, & van den Oord 1985) to our model and reÐtted the data. The Ðt is again signiÐcantly improved by the addition of this component (*s2 \ 13 ; for two additional parameters, see model 96BF), and the residuals are now Ñat over the entire 0.1È100 keV band. The best-Ðt temperature of this collisional plasma is keV. The covering factor of the neutral gas kT \ 1.78~0 .39 0.91 is now larger than in the previous Ðt, indicat-(C sou ing that only a small or null fraction of the nuclear radiation escapes absorption.

T he "" W arm ÏÏ Absorber in M97
Our base model did not provide a satisfactory description for the M97 spectrum either. Several features remained at keV as shown by the residuals in Figure 3b (bottom E [ 2 panel). These strongly suggest the presence of a typical "" warm ÏÏ absorber obscuring the line of sight and imprinting deep O VII and O VIII K edges at 0.74 and 0.87 keV (see Reynolds 1997 ;George et al. 1998). For this source, the signature of a complex "" warm ÏÏ absorber had been previously detected by ROSAT -PSPC and ASCA (Kriss et al. 1996b ;MWA97). We then added a photoionized absorber component to our base model and reÐtted the M97 data. The residuals are now Ñat over the entire BeppoSAX band, and the reduced s2 is 0.90 (Table 2, 97BF model).

T he V ariable Absorber
The 1996 November spectrum of NGC 3516 is heavily absorbed by almost neutral gas that covers most of the primary continuum source (here we choose the C sou D 0.95, best-Ðt value obtained by Ðtting the soft emission with the CIP component ; see°3.2.1) and whose equivalent hydrogen column density (as derived by a Ðt with a neutral absorber) is as large as cm~2 (exceeding N H Cold \ 2.3 ] 1022 the Galactic value by almost 2 orders of magnitude). This value would be even larger if the gas were mildly ionized (O IIÈIII dominating). There is no evidence for variation of the degree of obscuration of this gas during the observation (D20 hr), and this component is not required by the 1997 March data (4 months later),11 and the UV continuum of this source in an HST -FOS observation taken D18 days after the 1996 November BeppoSAX observation (GEA99) shows no marked extinction by dust (assuming a Galactic gas-to-dust ratio, we should expect mag for A v D 10 N H \ cm~2 ; Gorenstein 1975 constraints on the geometry and the physical state of this component and to speculate on di †erent possible scenarios.

T ransient Cold Absorption by a Broad Emission L ine Cloud ?
The typical timescale of variability of the X-ray continuum (e.g., Nandra et al. 1997a) gives an upper limit on the linear size of the emitting region D D c*t D 1014 cm. In order for a homogeneous cloud of absorbing gas to cover almost the entire primary emitting region, either the cloud is at a distance R close to the central source, comparable to D, and its linear sizes l are comparable and/or larger than D or R is much larger than D, and the source appears pointlike as seen by the cloud. Photoionization arguments strongly support the second picture. In fact, assuming R D D gives an ionization parameter where U [ 2500n 10 1, is the electron density of the gas in units of 1010 cm~3. n 10 We used the ionizing luminosity of NGC 3516, derived assuming the 2200 Ñux and from George et al. (1998) A a ox and X-ray Ñuxes and spectral shape from our data. Such an ionization parameter would be associated to almost neutral gas distributed between O I and O III) (U [ 0.01Èoxygen only for implausibly high electron densities n 10 [ 2.5 ] 105.12 More plausible electron densities of n 10 \ 0.1È 100, as those estimated for the broad emission line clouds (BELCs) in AGNs (Blanford et al. 1991 ;Krolik et al. 1991 ;Ferland et al. 1992), would result in very highly ionized gas, which is inconsistent with our data. We then assume that R ? D.
Reverberation mapping studies of the BELs in NGC 3516 (Wanders et al. 1993) have estimated a distance R BELC of the BELCs from the source of ionizing continuum of D11 lt-days (D3 ] 1016 cm), supporting the possibility that the gas obscuring the X-ray continuum of NGC 3516 in 1996 is associated with a BELC. At this distance, and assuming the ionization parameter mean 10 BELC \ 0.1È100, sures and the gas is only mildly U(R BELC ) D 0.0003È0.03, ionized and therefore compatible with the nature of the absorption seen in the X-ray band in 1996.
All this suggests that a BELC crossing our line of sight could have caused the obscuration of the X-ray primary source in 1996. As a consistency check we can calculate the time needed for a BELC to cross our line of sight. t cross Assuming as transverse velocity and the measured Hb v BELC FWHM of 4000 km s~1 (Wanders et al. 1993), we obtain (where l is the linear size t cross \ l/v^2500N H22 Cold n 10 1 v 4000 1 of the cloud). For the measured we have then N H Cold, t cross D s. Since we know that must be longer than 20 6000n 10 1 t cross hr (the duration of the 1996 observation) and shorter than D4 months (the time elapsed between the two BeppoSAX observations), the above equation constrains the electron density of the gas to the range 6 ] 10~4 cm~3 [ n 10 [ 0.1. We stress again that if the gas were mildly ionized (as is the case for the BELCs), then we would be underestimating the equivalent obtained by Ðtting the data with a neutral N H Cold absorber model, and therefore the interval of densities given above would shift toward higher values. This range of densities contains the estimated BELC value. A much narrower interval could in principle be set using the time elapsed between the 1996 November observation and the last 12 We note that this is one of the highest densities one may expect to Ðnd in the nuclear environment of a Seyfert 1, being of the order of the density of the matter in a standard Shakura-Sunyaev (1973) accretion disk around a black hole of 107È108 accreting at 0.1È0.01 times the critical M _ , rate, and in proximity of the last stable orbit.
HST -FOS observation of GEA99, showing no Mg II resonant absorption nor extinction in the UV continuum in *t \ 18 days. This would give a lower limit on the density of n 10 Z 0.01. In the framework of the proposed scenario, a BELC crossing the line of sight, we estimated the probability of such an event based on an estimate of the Ðlling factor of the  (Osterbrock 1989) and cm~3 n e D 1010 (Blanford et al. 1991). We then obtain V eff D 6.7 ] 1044 cm3. The probability for a cloud to Ðnd itself along a particular line of sight at any radius is then given by P D This low probability supports the V eff /V tot D 8.1 ] 10~6. picture of an event that occurs very rarely, as indeed observed in AGN (Kruper et al. 1990).

A V ariable State Absorber ?
Another intriguing possibility is that the ionization state of the absorber varied between the two BeppoSAX observations, becoming more ionized and therefore transparent to the soft X-ray radiation. In this scenario we may speculate that the "" warm ÏÏ absorber seen in the 1997 March spectrum of NGC 3516 is the same gas that was obscuring the line of sight 4 months earlier but with a very di †erent ionization degree. The same picture was drawn for this source by MWA97, who proposed a model for a variable warm absorber in NGC 3516 to reconcile the UV and the X-ray spectra within a unique scenario (see also°4.2). In the MWA97 model the presence and subsequent disappearance of a broad C IV absorption component at 1549 was con-A sistent with a variable state X-ray absorber, from mildly ionized during the 1989 Ginga observation to highly ionized in subsequent ROSAT -PSPC and ASCA observations.
In the framework of the scenario above described, we derived some dynamical properties of this variable state absorber. At a given ionizing Ñux the ionization degree of a photoionized absorber can vary because of variations of the distance of the gas from the ionizing source or of the electron density in the gas (or both). In the Ðrst of these two extreme hypotheses we assume that the absorber has a constant density in the range (typical of n 10 \ 10~4 to 10~1 warm absorbers in Seyfert 1 galaxies ( MWA97 ; Nicastro et al. , 2000. The gas is almost neutral during the 1996 November observation, which gives an upper limit on its ionization parameter of above this value, and for U [ 0.01 ; the ionizing continuum shape of NGC 3516, the opacity of the gas at low X-ray energies drops drastically. This gives a range of possible distances from the central source of R Z 2 cm (the larger the density the smaller the ] 1017È5 ] 1018 lower limit on R). If the gas were moving toward the central source with a radial velocity keeping its density conv rad , stant, it should have required less than months to t [ 4 cover the distance *R \ R/10 needed for its ionization parameter to increase by a factor of D100 (to reach the values observed during the 1997 November observation). The estimated radial velocity of the gas is thus v rad D This limit would become even higher if the *R/t Z 0.06c. cloud became denser while collapsing toward the central source. We note that the most direct implication of this extreme hypothesis is that UV and high-resolution X-ray spectra of Seyfert 1 galaxies, during such relatively shorts events, should show highly redshifted resonant-absorption lines with varying strength. Such phenomenon have never been observed so far.
Alternatively, the change in the ionization state of the absorber might be a result of a change in density (as in MWA97). To change from an almost neutral state (U [ 0.01) to a warm state (U D 5.37) would require a drop in density by a factor of For example, the (U P n H 1) Z500. density of a BELC would have to change from a few times 1010 cm~3 to about 2 ] 107 cm~3. A column density of 1022 cm~2 and a density of 1010 cm~3 imply a thickness of 1012 cm for the cold absorber. Similarly, the thickness of the warm cloud would be about 5 ] 1014 cm. Such an expansion has to occur on a timescale of months. If the cloud [4 is expanding adiabatically, its velocity dispersion would then have to be about 500 km s~1. Such a velocity dispersion is observed in associated absorption lines in AGN, so the derived value is completely reasonable. We conclude that a scenario based on a variable state absorber is plausible (see°4.3 for further details). Unfortunately, no other X-ray observation was performed between the two BeppoSAX pointings to further constrain the variability timescale in the soft X-ray energy band.
Both cases discussed above were based on the assumption of no changes of the intensity of the photoionizing continuum. Variations of the ionizing Ñux could have contributed to the observed changes of the ionization degree of the gas. If, between the two observations, the continuum increased by a factor of D100 and then recovered the level measured during the 1997 observation, then the gas could have been almost instantaneously ionized by this dramatic event but not yet had enough time to recombine to its equilibrium neutral state. Alternatively, the Ñux level seen during the 1996 observation could have been preceded by a long-duration event of very low intensity that caused the gas to recombine to an almost neutral state. The gas during the 1996 observation would then be underionized with respect to the observed ionizing continuum level. Both cases would require low electron densities of the absorber (n e [ 105 cm~3 ; see  for details).

A Hot Ionized Absorber/Emitter Gas ?
Two apparently unrelated spectral features are clearly visible in the N96 spectrum of NGC 3516 : (1) soft emission (with a hump at D1 keV) not obscured (or only partly obscured) by the amount of neutral gas covering the primary X-ray source and (2) the signature of heavy absorption by highly ionized gas at D8 keV. In our spectral analysis of N96 we modeled these two features as being due to two distinct components : a collisionally ionized emitting plasma and a photoionized absorber, respectively. Extended emission by the host galaxy surrounding the AGN has frequently been observed in Seyfert galaxies, e.g., Elvis et al. (1990), Wilson et al. (1992), and Weaver et al. (1995). However, in the particular case of NGC 3516 the extended component was probably unresolved by the ROSAT -HRI (Morse et al. 1995). Wilson et al. (1996) found that in their sample (which includes NGC 3516) the 0.1È2.5 keV intrinsic luminosity of the extended components ranges between 1040 and 1042 ergs s~1. In the present data we measure ergs s~1, a value only mar-L 0.1h2.5 keV \ 5 ] 1042 ginally consistent with the observed range, even allowing for intrinsic variability. Based on the correlations found between far infrared (FIR) and 2 keV luminosity of bright spiral galaxies (Fabbiano, Gioia, & Trinchieri 1988), we also compared the 2 keV luminosity ergs s~1 L 2 keV \ 9 ] 1041 keV~1 measured with the mean value for the 11 L 2 keV objects of the Fabbiano et al. (1988) sample with L FIR \ 1042.8È1043.2 ergs s~1. This range includes the value of 1043 ergs s~1 obtained for NGC 3516 (Bonatto & Pastoriza 1997). We found ergs s~1 , again much L 2 keV Gal^1039È1040 lower than measured in our data. We then conclude that the soft X-ray emission visible in 1996 cannot be due entirely to an extended component in the host galaxy.
An intriguing alternative is that both the D8 keV absorption edges visible in N96 and M97 and the soft emission in N96 could be self-consistently accounted for by a unique photoionized absorber/emitter in the nuclear environment of NGC 3516. If photoionized, this gas has a temperature of T^2 ] 106 K, and iron (the lightest abundant element not yet fully ionized) is smoothly distributed between Fe XVII and Fe XXVI. Because of the high temperature, the emissivity of this gas is large and contributes to the overall spectrum in an amount that depends linearly on its covering factor around the central source. In particular, a strong contribution by Fe XVIIÈXXIV L recombination lines is expected at D1 keV, where a clear emission feature is present in the N96 spectrum. Models for photoionized gas accounting for gas emission, other than photoelectric and resonant absorption, can then be used to constrain the geometrical conÐguration of the observed absorber/emitter.
In the following discussion we try to test this hypothesis quantitatively. The basic assumptions of this hypothesis are that (1) the neutral absorber in 1996 covered almost entirely as from the Ðt with the CIP component ; see (C sou [ 95%,°3 .2.1) the nuclear source of primary X-rays but left uncovered a much larger fraction of the extended (1ÈC env ) nuclear environment (i.e., the warm and hot scattering media), which is seen through scattering of the primary continuum ; (2) during the 1997 observation both the primary nuclear continuum and the reprocessed components are seen directly ; (3) no other component can signiÐcantly contribute to the line emission.
We proceeded to build (following Nicastro et al. 2000) two series of photoionization models, which include gas emission and resonant absorption, for the two ionized absorbers observed in NGC 3516. We used column densities and ionization parameters as measured Ðtting the N96 and M97 spectra with models accounting for photoelectric absorption only (Table 2). We built these models for a number of values of the covering factors and as f warm f hot seen by the central source, varying from 0 to 1, and for particular values of outÑowing and microturbulence (v out ) (p) velocities of the absorbers/emitters. In particular, we used km s~1, k m s1 (observed in v warm out \ 500 p warm \ 1100 the UV for the broad C IV system of NGC 3516 ; Kolman et al. 1993), and km s~1 and km s~1. For v hot out \ 500 p hot \ 200 the hot absorber, the adopted ratio maximizes the v hot out/p hot contribution of line emission compared to resonant absorption (Nicastro et al. 2000). The particular choice of the absolute values of v and p separately, however, is not critical. The warm component was not included in the models for the N96 spectrum, since we did not see it in the data (which can be either because the component was not present or because it was completely obscured by the amount of almost neutral gas covering the line of sight during that observation). We also neglected the emission contribution from this gas. However, given the temperature and ionization degree measured in 1997 (in the photoionization hypothesis), emission of the warm component is expected to be very weak and should inÑuence mostly the 0.5È0.6 keV portion of the spectra where weak O VII and O VIII emission lines are expected. This contribution would hardly be detectable with BeppoSAX (Nicastro et al. 2000). The hot component instead has been included in both models. We held constant between the two observations. For the f hot cold absorber in 1996, we used the best-Ðt column density and from Table 2. C sou We did not Ðt these models directly to the N96 and M97 spectra but proceeded as follows : We varied the values of the three parameters and to let (1) the 0.1È2 f hot , f warm , C env keV Ñux predicted by models for N96 match the best-Ðtting 96BF 0.1È2 keV Ñux and (2) the 0.1È10 keV ratio between N96 and M97 models match the observed raw data ratio N96/M97. In this particular scenario we could put relatively strong constraints on the covering factors and f hot but not on which we then Ðxed to the value of 0.5. C env f warm , We found that and lie in the ranges 0.3È0.4 and f hot C env 0.15È0.25, respectively. Figures 4 and 5 show two models for the N96 and M97 spectra of NGC 3516 (Figs. 4a and 4b, upper and bottom panels, respectively) and their ratio (Fig. 5) in the energy band 0.1È10 keV for and f hot \ 0.35 The ratio between the raw 0.1È10 keV N96 and C env \ 0.2. M97 data is also plotted in Figure 5c. The agreement between the models and the data is quite good. In particular, we note that the 1 keV emission feature present in the data is well reproduced by Fe L emission of the hot photoionized components, which was much more visible in 1996 because of the heavy obscuration of the primary X-rays by the cold absorber. We then conclude that a highly ionized component is present in the nuclear environment of NGC 3516 and that its emission is visible only when our view of the primary X-rays is at least partially covered by a large amount of neutral absorber.
This component could well be present in all Seyfert galaxies but has eluded detection so far. If this component is spherically distributed around the central source and if the  covering factor and column density measured in f hot N H hot NGC 3516 are representative of the entire class, then we would expect that absorption K edges due to highly ionized iron should be observed in almost 30%È40% of all known Seyfert galaxies. Furthermore, in all heavily obscured sources (i.e., Seyfert 2 galaxies or objects with transient absorption, like NGC 3516) we should detect emission by this hot gas at energies below 1È2 keV. Soft X-ray emission indeed is observed frequently in many bright and wellstudied Seyfert 2 galaxies (Turner et al. 1997 ;Comastri et al. 1998). High-resolution spectrometry, like that available with Chandra and/or XMM, will test this speculation unambiguously by allowing the detection of narrow emission and/or absorption lines even in objects with a low degree of nuclear obscuration with high accuracy, as already studied in NGC 5548 and NGC 3783 (Kaastra et al. 2000 ;Kaspi et al. 2000).

T he Complex X-Ray/UV Connection
Following MWA97, we investigate whether the X-ray absorbers in NGC3516 will produce any associated UV absorption lines. If the cold absorber is mildly ionized, the maximum degree of ionization would be U D 0.01 ; as seen in°4.1.2, a higher value, with the ionizing continuum shape of NGC 3516, would imply a drastic drop of opacity of the gas at low X-ray energies. In such an absorber the dominant ions would be O IÈIII, Mg IIÈIII, and C IIÈIV. For a total column density of a few times 1022 cm~2, the predicted column densities of Mg II and C IV are 2.5 ] 1017 and 4.7 ] 1017 cm~2, respectively. For the warm absorber, the C IV column density would be 5.4 ] 1014, but Mg is too highly ionized to have observable Mg II. For the cold absorber, the Mg II and C IV column densities are very large, and even for a velocity dispersion parameter, b, of 100 km s~1, the resonance absorption lines would have large equivalent widths (EW Z 3 A ).
In the warm phase the C IV column density is less than that in the cold absorber. However, the predicted EW of the C IV j1549 absorption line would still be detectable (EW Z 1.0 for b \ 100 km s~1). The absorption line EWs would A be even larger for a larger b. Therefore, for reasonable parameters, the X-ray absorbers in NGC 3516 will contribute to the absorption in the UV. How broad the lines would be will again depend on the b parameter. The UV observations closest in time to our X-ray observations (*t \ 18 days) are presented by GEA99 (°4.1). With the HST -FOS intrinsic Mg II absorption was not detected. This constrains the cold X-ray absorber (if still along the line of sight during the UV observation) to be almost completely neutral. GEA99 found C IV jj1549, 1551 absorption doublets with FWHM 651^16 km s~1 and 405^9 km s~1, respectively, and EWs 2.87^0.09 and 1.47^0.04 respectively. The A A , lines are not resolved with the FOS and are likely to be saturated. GEA99 quote a C IV column density in the range between 6.3 ] 1014 and 1015 cm~2, very similar to our predicted value for the warm absorber. Earlier observations with GHRS have resolved the C IV absorption line into four distinct components, two of which are from the host galaxy of NGC 3516. The two nuclear components have column densities consistent with the GEA99 values. We conclude that the nuclear C IV absorption line (possibly component 1 of Crenshaw et al. 1998) is likely to arise in the warm absorber observed by BeppoSAX. At the least the warm absorber makes a substantial contribution to the absorption seen in the UV (based on C IV strength). The hot X-ray absorber is too ionized to contribute to the UV absorption lines, since carbon, oxygen, and magnesium are fully ionized.
X-ray and UV observations show that the system of X-ray absorbers in NGC 3516 is clearly complex and highly structured with multiple ionization stages just as in the UV absorption lines (°1). This complexity is enhanced by variability on many di †erent timescales, from days to years. It is possible that on the very short timescales of a day the variability of absorption is governed by the photoionizationrecombination timescales, while on the longer timescales from months to years the dynamical time may play the dominant role, via either bulk motion or adiabatic expansion. It is conceivable that the di †erent ionization stages trace di †erent phases in the evolution of an absorber. A neutral absorber may expand as it outÑows, become a warm absorber, and eventually become hot and later completely transparent in the X-rays, or the "" components ÏÏ may be part of a quasi-static structure in an outÑowing wind (Murray & Chiang 1995 ;Elvis 2000). Again high-resolution Chandra and XMM observations of this source will be crucial to disentangle the many X-ray components and clearly establish (or reject) a link with the UV components.

CONCLUSIONS
We presented two BeppoSAX observations of NGC 3516, performed in 1996 November and1997 March. Our main Ðndings are the following : 1. The source underwent strong spectral variability between the two BeppoSAX observations, taken 4 months apart, entirely because of the drastically di †erent degree of obscuration of the central source. During the Ðrst observation the nuclear X-ray continuum was absorbed by an equivalent column of cold hydrogen of 2.3 ] 1022 cm~2, while in 1997 November the absorption was fully compatible with that from our Galaxy. This is the Ðrst time that such a change in the degree of obscuration of a Seyfert 1 has been clearly observed. We investigated several possibilities and propose as possible explanations either a BELC crossing the line of sight or a varying-state absorber.
2. We discovered the presence of a very highly ionized absorber/emitter in the nuclear environment of NGC 3516. This gas is visible in absorption (and with a stable physical state) in both BeppoSAX observations.Thanks to the large fraction of primary X-rays absorbed by neutral gas in 1996, we also were able to detect emission from this gas and to give an estimate of its covering factor f hot \ 0.3È0.4.