当前位置:首页 >> >>

XMM-NEWTON High Resolution Spectroscopy of NGC 5548

Astronomy & Astrophysics manuscript no. H3936 (DOI: will be inserted by hand later)

February 2, 2008

XMM-NEWTON High Resolution Spectroscopy of NGC 5548
K.C. Steenbrugge1 , J.S. Kaastra1 , C. P. de Vries1 , R. Edelson2
1 2

SRON National Institute for Space Research, 3584 CA Utrecht, The Netherlands Department of Physics and Astronomy; University of California, Los Angeles; Los Angeles, CA 90095-1562, USA

arXiv:astro-ph/0302493v1 24 Feb 2003

Received 21 August 2002 / Accepted 14 February 2003 Abstract. We analyze a 137 ks exposure X-ray spectrum of the Seyfert 1 galaxy NGC 5548 obtained with the XMM-Newton Re?ection Grating Spectrometer. Due to the long exposure time, the spectrum is of higher statistical quality than the previous observations of this AGN. Therefore, we detect for the ?rst time in NGC 5548 inner-shell transitions from O III to O VI ions, and the Unresolved Transition Array of M-shell iron. The warm absorber found from this X-ray observation spans three orders of magnitude in ionization parameter. We detect O III, which is as lowly ionized as the warm absorber detected in the UV band, to Fe XXIV. For O VI the column density determined from our X-ray data is an order of magnitude larger than the column density measured in previous UV observations. We conclude that there is substantially more low ionized material than previously deduced from UV observations. However, only a few percent of the warm absorber detected in the X-rays is lowly ionized. A 99.9% signi?cant increase in the derived absorbing column density with higher ionization states is observed. The out?ow velocities determined from the X-ray absorption lines are consistent with those deduced from the UV lines, evidence, together with the detection of O VI, that the X-ray and UV warm absorber are different manifestations of the same phenomenon. From a simple mass conservation argument, we indicate that our data set is consistent with an out?ow with small opening angle formed due to instabilities in the accretion disk. Possible due to uncertainties in the radiative transport mechanism, an apparent deviant iron to oxygen abundance is detected. No strong relativistically broadened emission lines of O VIII, N VII and C VI were detected. Key words. Galaxies: active – Galaxies: Seyfert – X-rays: galaxies – Techniques: spectroscopic – Galaxies: individual: NGC 5548

1. Introduction
The immediate environment of active galactic nuclei (AGN) is poorly understood. The black hole is fed from an accretion disk, which becomes visible through high energy radiation. Possibly powered by radiation pressure, an out?owing wind is formed from the accretion disk. In some AGN, the wind is visible through its continuum and line absorption of the radiation from the nucleus. These absorption lines were already detected in the UV part of the spectrum (Mathur, Elvis & Wilkes 1995), and have now been con?rmed in the X-ray spectra (Kaastra et al. 2000). NGC 5548, a Seyfert 1 galaxy, has been well studied in the X-rays due to its relative X-ray brightness, and the fact that it is a nearby active galactic nucleus with a redshift of z=0.01676 (Crenshaw & Kraemer 1999). This Seyfert galaxy was studied in detail with ROSAT (Done et al. 1995). They detected an O VII and O VIII K-shell absorption edge, and concluded that there was a warm absorber present. Therefore, NGC 5548 is an ideal Seyfert 1 galaxy to study with the high resolution spectroscopy instruments on-board of XMM-Newton and Chandra.
Send offprint requests to: K.C. Steenbrugge Correspondence to: K.C.Steenbrugge@sron.nl

Table 1 list the observations made with these instruments. In this paper we discuss a 137 ks exposure time Re?ection Grating Spectrometer (RGS) spectrum from the guest observation program of XMM-Newton.

2. Observations and data reduction
Here we present data taken with the RGS instrument (den Herder et al. 2001). The EPIC data (the continuum spectrum, and the Fe Kα line) are presented separately (Pounds et al. 2002). The time variability of the combined XMM-Newton and RXTE data sets will be presented by Markowitz et al. (2002). Shortly after launch two of the eighteen CCD chains failed. ? For RGS 1 this causes a data gap in the 10 ? 14 A range (where most Fe-L and Ne absorption lines are), for RGS 2 the gap is ? between 20 ? 24 A (where the O VII triplet is). The data were reduced using the developer’s version of the XMM-Newton data analysis package SAS version 5.3, which comprises a correct calibration of the effective area and the instrumental O-edge. The latter is important as the 1s-2p absorption lines of O II and O III are cosmologically redshifted toward the instrumental O I edge. In this paper we analyze the ?rst and second order spectra obtained from both RGS 1 and 2. Although the second order


K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON

Table 1. High resolution X-ray observations of NGC 5548. LETGS: Low Energy Transmission Grating in combination with the High Resolution Camera; HETGS: High Energy Transmission Grating in combination with the ACIS camera both onboard Chandra; RGS: Re?ection Grating Spectrometer onboard XMM-Newton. The exposure time is given in in ks.
Date 1999 Dec. 11 2000 Feb. 5 2000 Dec. 24 2001 Jul. 09–12 Instr. LETGS HETGS RGS RGS Expos. 86.4 83 28 137 References-comments Kaastra et al. 2000, 2002 Yaqoob et al. 2001, Kaastra et al. 2002 noisy spectrum low ?ux state present work

forbidden line can be determined in second order. However, comparing ?rst and second order results for the warm absorber, we ?nd a smaller out?ow velocity in the second order data set. The difference in wavelength between ?rst and second order, for the strongest line in the second order, O VIII Lyβ (both orders were ?tted with the xabs model, see Sect. 3) is ?0.019 ? ± 0.005 A. Finally, in the ?rst order data set there is a weak Galactic O VII recombination line, which has a rest wavelength ? of 21.6 A. This absorption line also shows an apparent veloc? ity shift of ?230 km s?1 (?0.017 A), although no blueshift should be expected. To correct for this wavelength calibration ? problem we applied a constant wavelength shift of ?0.017 A to the model for the ?rst order spectra. If not corrected, this wavelength shift would result in a velocity error between 130 km s?1 for the longest wavelengths to 620 km s?1 for the shortest wavelengths. The residual error in the wavelength scale should ? be within the 8 mA uncertainty of the calibration. All velocities quoted in this paper are corrected for this wavelength shift.

3. Modeling the warm absorber
In NGC 5548 there is an intricate pattern of densely spaced absorption lines as well as the possible presence of weak relativistic emission lines (Kaastra et al. 2002a). Blending, mostly of iron lines, is important in many wavelength ranges, making equivalent width (EW) estimations rather complicated. On the other hand, in the X-ray band many of the observed ions have more than one detected absorption line, signi?cantly improving the column density and velocity out?ow measurements. This is also important in order to disentangle possible partial covering factors and column densities of the ions. The possibility of relativistic emission lines complicates the analysis of the continuum spectral shape, which is already complicated to derive due to the presence of many edges and absorption lines. We consider here two approaches for ?tting the warm absorber. In the ?rst approach all absorption lines and edges of each individual ion are calculated, using a separate value for the column density for each ion. The transmission of all ions is then combined to calculate the total transmission of the warm absorber. Out?ow velocity and velocity broadening are coupled for the different ions. This is the slab model in SPEX (Kaastra et al. 2002b). In the second approach the column densities of the ions are coupled through a photoionization model. In this case only the total column density, ionization parameter, elemental abundances and the out?ow and broadening are free parameters. This is the xabs model. For a full description of both methods see Kaastra et al. (2002b). The results obtained with the xabs model are more dependent on theory, because the ionization parameter determines the column density ratios of the ions. An advantage of the xabs model is that different ionization components, each with their own velocity structure can be combined. This is harder to accomplish with slab components, in particular if the velocity components are barely separable with the current spectral resolution. This is the case in NGC 5548, and would lead to strongly correlated parameters in the spectral ?tting procedure. One further difference between the models is that ions which

Fig. 1. The ?uxed spectrum of NGC 5548, RGS 1 and 2 are added together and no error bars are plotted for clarity. The ? typical 1 σ error on the data is 1 photon m?2 s?1 A?1 . Some of the strongest features are identi?ed.

spectrum has a lower count rate and only a wavelength cover? age from 5 to 19 A, it does have a higher wavelength resolution than the ?rst order. Throughout the paper we have ?tted RGS 1 and 2 simultaneously, but for clarity of representation we have added the RGS 1 and RGS 2 spectra in the ?gures. All the spectral analysis was done with the SPEX package (Kaastra et al. 2002b). Fig. 1 shows the ?uxed spectrum of NGC 5548, indicating the Unresolved Transition Array (UTA) of inner-shell iron absorption lines and some other prominent features.

2.1. Wavelength scale accuracy
In general the RGS wavelength scale can be reconstructed to ? an accuracy of 8 mA (den Herder et al. 2001). However, in our case larger uncertainties exist, due to errors in the recorded satellite attitude and right ascension in the data ?les available. From the analysis of the ?rst order data we ?nd that the O VII ? forbidden line is blueshifted by 285 km s?1 (?0.021 A). The O VII forbidden line is not blueshifted in the earlier LETGS data of NGC 5548 (Kaastra et al. 2002a). No blueshift for the

K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON


Table 2. The parameter values quoted are those obtained in the xabs model ?tting and differ from the best ?t continuum without absorption. The continuum model was used as a basis for adding the warm absorber, either with a slab component (see Table 3) or with xabs components (see Table 4). The quoted χν 2 is for the best ?t continuum parameters without absorption, and thus not the parameters quoted in the table, which we derived including absorption.
χν 2 equals 1.96 for 2441 degrees of freedom norma (1.57 ± 0.03) × 1052 ph s?1 keV?1 Γ 1.77 ± 0.02 mbb: normb (6.5 ± 0.7) × 1032 m1/2 T (97 ± 6) eV a at 1 keV. b norm=emitting area times square root of electron density. pl:

have a rather small cross-section are automatically included in xabs, but since these are more dif?cult to discern individually, they are not usually included in the slab modeling.

4. Data analysis 4.1. The continuum
The continuum model consists of a power-law spectrum with Galactic absorption, and a modi?ed black body (see Kaastra & Barr 1989 for more details). A second modi?ed black body with a high temperature is required to describe the EPIC data, but is not detected in the limited RGS band (Pounds et al. 2002). This indicates that RGS data alone cannot constrain the continuum accurately. Table 2 gives the continuum parameters used for the analysis of the RGS data. The temperature (T ) of the modi?ed black body was allowed to vary as well as the normalizations of the power-law and the modi?ed black body. However, the photon index (Γ), which is very sensitive to small changes in the model, as well as the Galactic H I column density (1.65 × 1024 m?2 , Nandra et al. 1993) were kept ?xed throughout the ?tting process. The obtained parameters are within calibration uncertainties consistent with those obtained from simultaneous EPIC and BeppoSAX observations.

4.2. Spectral ?t using the slab model
As a ?rst approach to the analysis of the warm absorber, we modeled it by adding a slab component to the continuum model. We re?tted the data, leaving the same parameters free as above for the continuum, as well as the overall out?ow velocity, and all the relevant ion column densities. Finally, after obtaining this overall ?t we also let the photon index vary, and now obtain a reasonable result. The results of this ?t were then used to ?ne-tune the column density and out?ow velocity of each ion which has a detectable absorption line or edge in the RGS wavelength range. To do so we decoupled subsequently for each ion its out?ow velocity from the out?ow velocity of all the other ions in the model. This method was followed to minimize the effects from blending. The signi?cance of the different

ionic columns can be derived from the best ?t column densities and their error bars. The result of this ?tting is summarized in Table 3. This Table also lists the ionization state for the maximum column density of the particular ion. Note however, that each ion exists for a range of ionization states, and thus the quoted ionization is only an indication. Adding the slab component signi?cantly improves the χν 2 from 1.96 to 1.26 (with 2404 degrees of freedom). For the ions with only an upper limit for their column densities, absorption is only detected through continuum absorption (Ar XIV, Fe V and Fe VII) or a single/few weak lines (O IV, Si X, Si XI, Fe I, Fe XII and Fe XIV). As a result the column density as well as the velocity of these ions are poorly determined. The other ions with poorly determined velocity shifts, but with solid column densities, are those for which the column density is determined from an edge (C IV, N V and Si IX), for which the absorption lines are strongly blended (Fe XII), or because only a few weak absorption lines are detected. We ?nd a very well constrained column density for C IV and N V, although both these ions have in the RGS band only ? ? an absorption edge, at 36 A and 25 A (rest wavelength) respectively. As the depth of the edge correlates with the assumed underlying continuum, these column densities are rather more uncertain than the ?t results indicate. To ascertain how these column densities change the overall continuum shape we re?tted the data, but now forcing a column density for C IV and N V of zero. For the new ?t, which gives a similarly good χν 2 , the normalization of the modi?ed black body increased by an order of magnitude, while the temperature of the same component decreased by an order of magnitude. As a result of this change in the continuum model, the column densities found for O III and Ne IX also changed signi?cantly. In this respect the xabs modeling is more reliable, as the column densities determined for ions that are only detected through an edge, are based upon the determined ionization state. The same method for modeling the warm absorber was used for ?tting the second order data. The second order data for RGS 1 and 2 were not ?tted together with the ?rst order data, as the systematic shift in the ?rst order spectra is corrected for in the applied model and not in the data itself. However, both orders are in excellent agreement, see Fig. 2. The average out?ow velocity found, including only those ions with well determined velocities in both orders, is ?350 ± 250 km s?1 for the ?rst order data set, in comparison with ?420 ± 340 km s?1 for the second order data set. For further analysis we focus on the ?rst order data, because the better statistics outweigh the better spectral resolution of the second order data.

4.3. Spectral ?t using the xabs model
Using the same continuum model as in the previous section, we replaced the slab model by the xabs model for the warm absorber. First we ?tted only the ionization parameter (ξ), the hydrogen column density (N H ) and the out?ow velocity (v), in addition to the continuum parameters that we leave free throughout the ?tting procedure. The abundances were kept to


K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON

Table 3. Best ?t values for the logarithms of the column densities in m?2 and the out?ow velocity, using the slab component. The logarithm of the ionization parameter for which the ion has the largest column density as determined from XSTAR is given in column 4.
log NH (m?2 ) 21.4 ± 0.1 20.4 ± 0.2 21.1 ± 0.1 20.6 ± 0.3 20.1 ± 0.1 20.5 ± 0.1 20.0 ± 0.6 < 20.2 20.7 ± 0.3 20.5 ± 0.1 21.57 ± 0.05 22.24 ± 0.03 21.82 ± 0.06 21.5 ± 0.2 20.5 ± 0.4 20.6 ± 0.6 21.4 ± 0.1 20.5 ± 0.5 20.8 ± 0.5 21.3 ± 0.1 < 20.4 < 20.1 < 20.5 20.3 ± 0.2 20.3 ± 0.1 20.1 ± 0.2 v (km s?1 ) ? ?490 ± 200 ?510 ± 70 ? ?470 ± 130 ?390 ± 150 ? ? +430 ?250 ± 160 ?450 ± 90 ?310 ± 70 ?30 ± 200 +10 ± 230 ? ? ?560 ± 250 ? ?970 ± 990 ? ? ? ?400 ± 380 ?820 ± 420 ?410 ± 80 ?640 ± 200 log ξ (10?9 W m) ? 0.6 0.2 1.2 0.0 0.7 1.5 ? 1.8 ? 0.7 0.0 0.5 1.2 1.8 1.7 2.3 1.9 2.4 1.5 2.1 2.6 0.9 1.3 1.7 2.1 1.1 1.7 2.1 log NH (m?2 ) 20.1 ± 0.3 19.4 ± 0.4 < 19.4 20.1 ± 0.3 < 19.8 19.9 ± 0.3 < 19.7 19.8 ± 0.4 < 19.7 20.1 ± 0.1 20.1 ± 0.1 19.9 ± 0.1 20.1 ± 0.1 < 19.7 19.4 ± 0.6 < 19.5 < 19.7 19.7 ± 0.3 20.1 ± 0.1 20.1 ± 0.1 20.49 ± 0.09 20.4 ± 0.2 20.5 ± 0.2 20.6 ± 0.2 20.4 ± 0.4 20.7 ± 0.5 v (km s?1 ) +90 ± 280 ?630 ± 210 ? ?470 ± 400 ? ? ? ? ? ?340 ± 450 +130 ± 220 ?670 ± 430 ?430 ± 400 ? ? ? ? ?500 ± 450 ?480 ± 170 ?420 ± 240 ?350 ± 230 ?960 ± 710 ?65 ± 550 ? ? ? log ξ (10?9 W m) 2.4 1.8 2.1 2.3 < ? 4.0 ? 3.0 ? 1.3 ? 0.9 ? 0.3 0.1 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.6 2.1 2.3 2.5 2.8 3.0 3.2 3.3 3.5

a b


See text for details. Possible error in the wavelength.

Table 4. The best ?t results for the ?rst order data xabs ?t. Further details are given in the text.
Component A B C log ξ (10?9 W m) 2.69±0.04 1.98±0.06 0.40±0.03 log N H (m?2 ) 25.68±0.05 25.52±0.06 24.15±0.01 out?ow v (km s?1 ) ?311±60 ?440±100 ?290±70 abundancesa : C <4.9 0.6±0.2 1.8±0.6 N 2.2±1.0 <0.3 1.7±0.5 Ne 2.5±1.4 0.1±0.7 b Na <8.3 3.9±3.9 <350 Mg 1.0±0.8 <0.6 <5.8 Si 2.2±2.2 <0.2 <3.9 S 0.6±0.7 0.5±0.2 b Ar 2.2±2.2 0.4±0.5 <5.4 Fe 0.70±0.06 0.11±0.03 7.4±0.6 a The abundances are relative to O, which is kept at solar value. b Frozen to solar value, see text for explanation.

produces no X-ray lines, we normalized the abundance relative to oxygen, which was frozen to solar abundance. Finally we added two more xabs components to the model, both with lower ionization parameters, signi?cantly improving the spectral ?t. Adding the ?rst xabs component improved χν 2 from 1.97 to 1.66. The addition of the second xabs component, which was the one with the lowest ionization parameter, further improved χν 2 to a value of 1.46. However, for this xabs component we found two abundances which have unphysical values. Namely, Ne and S are overabundant by a factor of 43 and 106, respectively. From detailed ?tting, it is clear that both these abundances are mainly determined from the absorption edges in the spectrum, and are thus rather uncertain. Fitting the model again, but ?xing the abundance of Ne and S in the lowest ionization component to solar, we ?nd a similar good ?t, namely with χν 2 of 1.48. The addition of the last xabs component gives a ?nal χν 2 = 1.22. Table 4 lists the best ?t parameters, with the xabs components labeled to be consistent with the components found by Kaastra et al. (2002a). The signi?cance, according to F-tests, is at least 100, 99.8 and 99.9 percent for components A, B and C, respectively. Only in the last step we left the photon

the solar values (Anders & Grevesse 1989). Later, the abundances were left as free parameters. However, as hydrogen

K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON


Table 6. Comparison of the O VII forbidden line from three different observations.
Instrument HETGS LETGS RGS v (km s?1 ) ?70 ± 100 0±50 ?110 ± 80 σv (km s?1 ) <310 <330 <490 ?ux (ph m?2 s?1 ) 0.8±0.2 0.8±0.3 0.9±0.1

Fig. 2. Best xabs ?t to the ?rst (top) and second order (bottom) data. First order data are rebinned by a factor of 2, second order by a factor of 3. For the ?rst order data a CCD gap related ? feature can be seen at 13 A and a dead pixel is seen short-ward ? ? of 16.5 A. Between 12.5 and 14 A only RGS 2 gives results, hence the larger error bars. For second order data a dead pixel ? is seen at 14.05 A and a CCD gap related feature can be seen ? at 14.45 A. Table 5. Values determined for the forbidden line of O VII.
? (?170 ± 22) mA (0.9 ± 0.1) ph m?2 s?1 ? < 36 mA < 490 km s?1 ? λa (22.096 ± 0.005) A vb (?110 ± 80) km s?1 a Wavelengths are given in the rest frame of NGC 5548. b Velocity shift from comparing rest and observed wavelengths. O VIIf: EW ?ux FWHM

Fig. 3. Detail of RGS 1 showing the O VII triplet as well as absorption lines for O VI and O V. The xabs model was used for ?tting the warm absorber. The Galactic O VII absorption line is indicated. are formed at the same distance from the ionizing source and under the same conditions, a blueshift of 620 km s?1 is inconsistent with the lack of blueshift for the forbidden line. Because the forbidden line is strong, models predict that even tenuous gasses should produce a weak but detectable inter-combination line. One possible explanation is that the inter-combination line ? blends with an O VI absorption line at 21.79 A. Another emission feature for oxygen is the radiative recom? bination continuum (RRC) of O VII between 16.9 and 17 A (see Fig. 4). As there is severe blending with iron absorption lines, we approximated the RRC simply with a delta function. We ?nd a 2 σ detection for this RRC and the narrowness of the feature (unresolved by the instrument) indicates a low temperature. The ?ux is 0.06 ± 0.03 ph m?2 s?1 which is consistent with the 0.06 ± 0.10 ph m?2 s?1 value as determined from the LETGS spectrum (Kaastra et al. 2002a). The presence of inner-shell absorption lines of oxygen together with the higher ionized O VII and O VIII absorption lines allow for an accurate ionization determination, independent of elemental abundances. The ionization parameter determined for oxygen spans three orders of magnitude from log ξ = ?1.8 for O III to log ξ = +1.8 for O VIII. The deepest absorption line in the spectrum is the O VIII Lyα line. For O VIII we also detect the Lyβ, Lyγ and Lyδ lines and some hints for the Ly? line in absorption (see Fig. 2). The multitude and depth of these absorption lines allow thus for an accurate column density and out?ow velocity determination (see Table 3). These higher order lines are important for the

index free, but this did not improve the χν 2 . The continuum parameters for this model correspond within 1 σ with the continuum parameters as found from the slab ?t.

4.4. Emission and absorption by oxygen
Fig. 3 shows the spectrum around the O VII triplet. In Table 5 we list the parameters of the O VII forbidden line in emission. For the O VII forbidden line Table 6 gives the ?uxes, out?ow velocity as well as an upper limit to the width of the line, as measured from the HETGS and LETGS spectra (data taken from Kaastra et al. 2002a) and our data set. Note that this line has the same measured ?ux over a period of 1.5 years, despite large variations in the continuum. In all three observations the line is unresolved. The above indicate that the O VII forbidden line is formed at least 0.46 pc from the ionizing source. Interestingly, the inter-combination line is not present in ? the data. An excess is observed at 21.74 A rest wavelength, ? shifted by 0.06 A from the rest wavelength of the intercombination line. The relative wavelength accuracy of RGS is ? 2 mA. Assuming that the inter-combination and forbidden lines


K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON

detection of variability in the warm absorber, as they have an optical depth close to unity. However, we did not detect significant variability in the measured column densities as compared to the previous LETGS observation, which has rather large error bars. Also within this observation no variability in the warm absorber was detected. Fig. 3 shows that our model overestimates the O VII resonance line. A probable explanation is that the emission component of this line partly re?lls the absorption line due to the imperfect spectral resolution. In the LETGS data (Kaastra et al. 2002a) this line has a clear P Cygni pro?le, and also in the RGS there is excess emission at wavelengths just longer than the resonance line. Both models for the warm absorber don’t include possible P Cygni pro?les. The excess emission just long-ward ? of 22.25 A is noise, as the width of the feature is narrower than the point spread function of the instrument. The Galactic absorption by O VII, detected with 1.5 σ signi?cance is also indicated in Fig. 3. Shortward of the forbidden of O VII there is a deep O VI ab? sorption line, and 0.25 A longward of the forbidden line there is a deep O V absorption line. Both are inner-shell absorption lines, indicating the importance of these transitions in AGN. For O V two absorption lines are detected. However, the line ? at 22.33 A is redshifted by about 700 km s?1 relative to the ? weaker line at 19.92 A, and the other oxygen absorption lines. To check the possible uncertainty on the wavelength for the strongest O V line, the O V wavelength and oscillator strength were also calculated using the Cowan code (1981) (Raassen, private communication). This gives a rest wavelength of 22.38 ? ? A, closer to the observed rest wavelength of 22.39 A, and indicates that uncertainties in the rest wavelength can be of order ? 0.05 A; i.e. equivalent to the instrumental FWHM.

Fig. 4. Detail of the ?rst order data showing the UTA between ? 16 and 17.5 A. The thick line is the best ?t through the data, the thin line indicates the same model if we set the iron abundance for the lowest ionization component to 0. A dead pixel occurs ? ? at 16.4 A. The gaps in the model at 17.05 and 17.15 A are due to missing data. Only the deepest features are labeled.

4.5. Absorption by iron
Iron is the best indicator for the ionization structure of the warm absorber, as it has absorption lines in the RGS band from neutral Fe I to the highly ionized Fe XXIV. For NGC 5548 we detect absorption from Fe VI to Fe XXIV (see Table 3). The lowest ionization component is represented by the UTA (Sako et al. 2001) formed from inner-shell transitions of Fe VI through Fe XI. These ions represent the lower ionized iron ions. Fig. 4 ? gives a detailed view of the spectrum between 16 and 17.5 A, where the UTA is located. Note that although there are some individual absorption features visible, the dominant effect of the UTA is the depression of the continuum due to an unresolved blend of transitions with small cross-sections. The UTA clearly shows the importance of these inner-shell transitions in determining the continuum correctly. The moderately ionized component is represented by Fe XIII through Fe XIX, where we only ?nd an upper limit for Fe XII and Fe XIV. The highest ionized component is represented by absorption by Fe XIX through Fe XXIV. For Fe XIX through Fe XXII we detect a multitude of absorption lines between 8 (lowest wavelength that we included in the ?t) and 15 ? A. For Fe XXIII and Fe XXIV only two absorption lines are detected, resulting in uncertain velocity determinations.

It is thus clear that all three ionization components as ?tted with xabs are amply con?rmed by iron absorption. The question arises whether a continuous ionization structure would be a better ?t to the data. From Fe VI through Fe XXIV we detect all iron ions with the exception of Fe VII, Fe XII and Fe XIV. Between Fe XII and Fe XV all column densities are low. Forcing Fe XII to Fe XVI to have a similar column densities as Fe XI or Fe XVII worsens the ?t, increasing χν 2 by 0.24. This indicates that for log ξ between 1 and 1.4 (ξ between 10 and 25) (in 10?9 W m) the column densities are lower than for the other ionization states. However, no clear depression in column densities is seen for the transition between the middle and highly ionized component. A more continuous ionization structure than presented by the xabs model for the warm absorber is not excluded by the data.

4.6. Absorption by carbon and nitrogen
For carbon we detect a strong C VI Lyα line (see Fig. 5), the Lyβ line (see Fig. 7), Lyγ and hints of Lyδ lines (see Fig. 6). These higher order lines are important in the correct determination of the column density as well as any possible time variability, as their optical depth is around unity. C VI and N VII are produced mainly in the middle ionization component, while C V and N VI are produced in the middle and lowest ionization component. Finally, C IV and N V are produced only in the lowest ionization component (see Table 3 for an indication of the ionization state). The C VI Lyα line is rather deep ? and at a relatively long rest wavelength of 33.736 A, resulting in a very good velocity shift determination. The velocity component found from the X-rays is consistent with component 3 from the UV, which has an out?ow velocity of ?540 km s?1 instead of ?510 km s?1 . There is a hint in our data that the line has substructure, consistent with the earlier LETGS results.

K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON


Fig. 5. Detail of ?rst order RGS 1 and 2 data, indicating the C VI Lyα absorption line as well as some C V, S XII, Ar IX, Ar X and Ar XI absorption lines. A dead pixel is seen at 34 and ? 35.8 A.

Fig. 7. Detail of ?rst order RGS 1 and 2 data, indicating an N VI absorption line as well as the C VI Lyβ line. Further absorption by S XII, S XIV and Ar XIII are labeled. Dead pixels are seen at ? ? 28.1 A and 30.3 A. A CCD gap related feature is seen at 28.9 ? A. Table 7. Values for the relativistically broadened emission lines in NGC 5548, MCG?6-30-15 and Mrk 766 (Sako et al. 2001). LETGS measures are taken from Kaastra et al. (2002a).
RGS LETGS MCG?6-30-15 Mrk 766 ia 46±35 46(?3,+8) 38.5±0.4 34.3±0.74 qb 3.9±5 3.9±0.6 4.49±0.15 3.66±0.22 rin c 100±30 <2.6 3.21±1.2 <2.3 rout c 400±90 >10 100(?48,+95) 81(?20,+73) C VId <0.8 <0.7 2.0±0.2 2.3±0.2 N VIId <0.03 1.1±0.4 4.4±0.2 5.8±0.3 O VIIId 0.2±0.1 0.6±0.2 13.1±0.6 11.4±0.7 a The inclination derived for the AGN from the relativistic lines, in degrees. b The emissivity slope. c In GM/c2 . d ? EW measured in A.

Fig. 6. Detail of ?rst order RGS 1 and 2 data, indicating the N VII Lyα absorption line as well as the C VI Lyγ and Lyδ lines. Absorption of S XIII, Ar XIII, and Ar XIV are labeled. Dead pix? ? ? ? ? els are seen at 23.4 A, 24.3 A, 25.7 A, 26.2 A, 26.6 A, and 27.6 ? A.

4.7. Relativistically broadened emission lines
After ?nding a model for the warm absorber that is as complete as possible, we added relativistically broadened emission lines of O VIII, N VII, and C VI Lyα to the model. Such lines were observed in MCG?6-30-15 and Mrk 766 (BranduardiRaymont et al. 2001, Sako et al. 2002). In NGC 5548 Kaastra et al. (2002a) found evidence for weak broadened emission lines of O VIII Lyα and N VII Lyα. However, no indication for a relativistically broadened C VI Lyα line was detected. We applied the best ?t parameters found from the LETGS model to the RGS data but found no signi?cant improvement for any of the relativistically broadened emission lines. Leaving the parameters for the Laor line pro?le (Laor 1991) free, we ?nd that a broad Gaussian would better

describe the line/excess. However, the improvement in ?t is still small for all three emission lines. The results are given in Table 7, together with the best ?t values for the LETGS data of NGC 5548 (Kaastra et al. 2002a), and the RGS data of MCG?6-30-15 and Mrk 766 (Branduardi-Raymont et al. 2001, Sako et al. 2002). Note, that the values found for the EW between our RGS results and the LETGS results are consistent within 3 σ, except for the inner radius. These relativistic lines should be time variable, considering that they are formed in the accretion disk close to the black hole. Possibly, time variability explains the differences between the previous LETGS results and ours. Given their weakness, in the further analysis and in all the plots we did not include these broadened emission lines. Fig. 8 gives the wavelength band where the relativistic O VIII Lyα line should be situated, and a small excess is seen between ? 18 and 18.4 A.


K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON

Fig. 8. Best xabs ?t to the ?rst order data, detailing the wavelength band around the position of the relativistic O VIII Lyα emission line. Note that in this plot no relativistic line is ?tted. ? ? ? ? ? Dead pixels are seen at 17.4 A, 18.3 A, 18.7 A, 18.8 A, 19 A, ? and 19.95 A. A CCD gap related feature can be seen between ? 20.7 and 20.8 A.

Fig. 9. The velocity shift as given in Table 3 versus the the ionization parameter. The vertical line represents the average out?ow velocity.

5. Discussion 5.1. Out?ow velocity versus ionization
No evidence is found for a correlation between the out?ow velocity and the ionization parameter (see Fig. 9). Rather the out?ow velocity, for ?rst order, is consistent with ?380 km s?1 and ?350 km s?1 for the slab and xabs model respectively over the three orders of magnitude sampled in ionization scale. This is consistent with the UV data (Crenshaw & Kraemer 1999), and the results for the C VI Lyα line in the LETGS data (Kaastra et al. 2002a). Kaastra et al. (2002a) found a trend of decreasing out?ow velocity for higher ionized iron and oxygen ions, although all out?ow velocities were consistent with an out?ow velocity of ?340 km s?1 . This result is not reproduced here. The out?ow velocities observed in both data sets are consistent with the ?ve UV velocity components, suggesting a similar origin of both absorbers.

Fig. 10. The derived absorbing column density per ion as derived from the column densities given in Table 3. There is an increase of the derived column density with higher ionization. similar shift, the derived abundances and the derived absorbing column density are rather insensitive to the input SED. In general, the derived absorbing column density increases for higher ionization parameters. The correlation coef?cient, for all data points, excluding upper limits, is 0.71, and the related signi?cance is 0.999997; for iron these values are 0.82 and 0.9982, respectively. The derived absorbing column density from the three component xabs model is very similar to the absorbing column density as derived from the slab model, and a similar increase for higher ionization is observed. In general the abundances found are consistent with solar. An exception is iron, for which both models have lowly ionized iron overabundant compared to lowly ionized oxygen, while a reverse trend occurs for highly ionized iron and oxygen. As carbon and nitrogen are consistent with oxygen, the effect is probably due to uncertainties for the lowly ionized iron ions. A similar overabundance for iron in the lowest ioniza-

5.2. Hydrogen column density versus ionization
The derived absorbing column density per ion, as determined from the slab modeling, versus the ionization parameter is plotted in Fig. 10. In the Figure we use the ionization parameter corresponding to the state with maximum column density. In reality, however, the ion is produced over a range of ionization states. Fig. 10 thus assumes a continuous ionization distribution, and is dependent on the spectral energy distribution (SED) in our models with XSTAR. We tested a model with and without a UV bump, and found a maximum displacement in log ξ of 0.3 for very highly ionized iron, and a maximum change in the derived absorbing column density of 0.1 on the log scale used. However, as most ions at a particular ionization state have a

K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON


tion state was noted by Blustin et al. (2002) for NGC 3783. In the LETGS observation of NGC 5548 (Kaastra et al. 2002a) found a similar abundance pattern for iron, throughout the ionization range. The explanations listed by Kaastra et al. (2002a) for this discrepancy have been tested using this data set. First, a calibration effect is unlikely as RGS and LETGS give consistent results. Secondly, we determined the column density for lowly ionized oxygen ions from our own X-ray spectrum; and thus we do not need to rely on non-simultaneous UV measurements. Spectral variability is thus not a possible explanation for the apparent overabundance of lowly ionized iron. Thirdly, we tested the sensitivity of the model to different values for the line broadening, ensuring that saturation due to narrow line widths cannot be responsible for the overabundance of lowly ionized iron. Reducing the width of the lowly ionized absorption lines, has the effect that iron saturates ?rst, worsening the overabundance already found. The maximum width of the lowly ionized lines is set by the width of the inner-shell oxygen lines, which are unresolved. Possible explanations for the iron overabundance are an uncertainty in the wavelength of the M-shell iron absorption lines; or if the optical depth, τ ? 1, an uncertainty in the radiative transfer models used. Another possibility is the omission of certain processes like re-emission and Bowen ?uorescence in the radiative transfer models used. Finally, the iron abundance could be non-solar, although the physical process causing such an overabundance of iron in Seyfert galaxies is unknown. Nonsolar abundances have been detected in IRAS 13349+2438 (Sako et al. 2001) and NGC 1068 (Kinkhabwala et al. 2002).

In the case of a Schwarzschild black hole, as considered here, η = 0.057. For our average out?ow velocity of 380 km s?1 the upper limits of the opening angle vary between 3.8 × 10?7 sr for log ξ = ?1.8 and 0.024 sr for log ξ = 3, with log ξ in 10?9 W m. For an average ionization state of log ξ = 1.7 in 10?9 W m, we ?nd an upper limit to the opening angle of 0.0019 sr. The upper limits to the opening angle indicate that the out?ow is mostly in a narrow stream, where the densest part of the stream is the narrowest and lowest ionized. For higher out?ow velocities or a higher mass conversion ef?ciency, i.e. in the case of a Kerr black hole, the stream is even narrower. From Fig. 10 we determine that the derived absorbing column density scales with the ionization parameter, NH ? ξ α . α ranges from 0.25 to 0.5 depending on whether one uses the NH obtained for the iron or oxygen ions at the lower ionization states. We de?ne NH ≡ nd, where n is the density and d is the thickness of the out?ow in the line of sight. Equating NH to ξ α , where ξ = L/nr2 , one can determine a relation between the density, thickness and distance. If one further assumes that for a ?xed distance (r) from the ionizing source, there is a range in densities, resulting in a range in ionization states, one ?nds that the density is related to the thickness n ? d?5/4 or n ? d?3/2 , respectively. From both simple models we conclude that the out?ow occurs in streams with a very small opening angle for the dense and lowly ionized gas, and with a larger opening angle for the less dense and higher ionized gas.

5.4. Comparison of UV and X-ray detection of O VI
? O VI has absorption lines in both the UV (λλ1032, 1038 A) ? ? ? and X-ray band (22.01 A, 19.34 A and 21.79 A). Our X-ray column density of O VI of 1020.5 m?2 is nearly an order of magnitude larger than the UV column density (1019.62 m?2 ) of Brotherton et al. (2002). However, the out?ow velocities in the UV and the X-rays are comparable. The difference in derived absorbing column density indicates that the UV data are severely saturated, leading to an underestimate in the column density. Arav et al. (2002) compare the column density of O VI as observed in this data set and an earlier FUSE UV measurement, and conclude that the column densities could be equal if velocity dependent covering is considered. However, simultaneous observations are necessary to exclude the possibility of column density variability (see Arav et al. 2002 for more details). Generally, in the X-rays, we ?nd signi?cantly more low ionization gas than previously deduced from UV data. The discovery of O III to O V absorption lines, in addition to O VI, gives us for the ?rst time an estimate for the column densities of these ions, which are inaccessible in the UV.

5.3. Out?ow geometry
The geometry and physics of the warm absorber as observed in some AGN is poorly understood. One of the outstanding questions is whether this out?ow is spherical, in which case the ionization of the gas is mainly dependent on the distance. Alternatively, the out?ow could be in localized streams formed due to instabilities in the accretion disk. The ionization parameter is then mainly dependent on the density. As distance and density cause a degeneracy in the determination of the ionization parameter, the location of the out?owing X-ray and UV warm absorber is uncertain. Constraining the geometry of this out?ow will help solve these outstanding problems. ˙ From mass conservation (Mloss = 4 πr2 nmp v?), the ionization parameter (ξ = L/nr2 ), the out?ow velocity (v), and the bolometric luminosity (L), one can determine the mass loss ˙ rate, Mloss , as a function of the opening angle (?) of the out?owing wind. This is summarized in eq. (1), where mp is the proton mass. Assuming that the mass loss rate of the wind cannot be higher than the mass accretion rate onto a black hole and that the system is stationary, one can equate both and obtain an upper limit for the opening angle. Eq. (2) gives the mass ˙ accretion rate, Macc , with a mass conversion ef?ciency of η. 4πLmp v? ˙ Mloss = ξ L ˙ Macc = 2 c η (1) (2)

6. Summary
We have presented here the highest signal to noise high resolution X-ray spectrum of NGC 5548 obtained yet. The spectrum shows a rich structure in narrow and broad spectral features. We detected a very weak O VII RRC, consistent with a low temperature. Inner-shell oxygen lines, together with the higher order absorption lines from O VIII Lyα and C VI Lyα have an optical depth near unity, important in the detection of spectral vari-


K. C. Steenbrugge et al.: Spectroscopy of NGC 5548 with XMM-NEWTON Kaastra, J. S. & Barr, P., 1989, A&A, 226, 59 Kaastra, J. S., Mewe, R., Liedahl, D. A., Komossa, S., Brinkman, A. C., 2000, A&A, 354, L83 Kaastra, J. S., Steenbrugge, K. C., Raassen, A. J. J., et al., 2002a, A&A, 386, 427 Kaastra, J. S., Mewe, R. & Raassen, A. J. J., 2002b, Proceedings Symposium ’New Visions of the X-ray Universe in the XMMNewton and Chandra Era’ Kaspi, S., Brandt, W. N., Geoerg, I., et al., 2002, ApJ, 574, 643 Kinkhabwala, A., Sako, M., Behar, E., et al., 2002, ApJ, 575, 732 Laor, A., 1991, ApJ, 376, 90 Markowitz, A., Edelson, R., Vaughan, S., et al., 2002, submitted to ApJ Mathur, S., Elvis, M. & Wilkes, B., 1995, ApJ 452, 230 Nandra, K., Fabian, A. C., George, I. M., et al., 1993, MNRAS 260, 504 Pounds, K., Reeves, J. N., Page, K. L., et al., 2002, accepted by MNRAS Sako, M., Kahn, S. M., Behar, E., et al., 2001, A&A, 365, 168 Sako, M., Kahn, S. M., Branduardi-Raymont, G., et al., submitted to A&A Yaqoob, T., George, I. M., Nandra, K. et al., 2001, ApJ, 546, 759

ability of the warm absorber, although no spectral variability was detected in this data set or from comparison with the earlier LETGS spectrum. Uncertainties in rest wavelengths for the inner-shell ions complicate the study of these absorption lines, certainly for the lowly ionized ions. Possibly this or uncertainties and the omission of certain processes in the radiation transfer models lead to an deviant iron to oxygen abundance ratio. The detected warm absorber spans three orders of magnitude in ionization, from ions as lowly ionized as those measured in the UV band, to nearly fully ionized iron. Only a few percent of the X-ray warm absorber column density is lowly ionized, represented by oxygen and iron inner-shell transition lines. The out?ow velocities determined from the X-rays are all consistent with those measured in higher resolution UV spectra, and no correlation between the ionization state and velocity was detected. The similarity in out?ow velocity and the detection of O VI give strong evidence that the UV and X-ray warm absorber are different manifestations of the same out?owing wind phenomenon. However, we detect an order of magnitude more O VI in the X-rays than previous UV measurements. In general, we detect substantially more low ionized gas, than is deduced to be present from UV observations. Probably the UV band underestimates the true column densities due to saturation and velocity dependent covering factor. From simple arguments, we found that the out?ow detected is consistent with a narrow stream, with smallest opening angles for the densest and lowest ionized gas.

This work is based on observations obtained with XMMNewton, an ESA science mission with instruments and contributions directly funded by ESA Member States and the USA (NASA). SRON National Institute for Space Research is supported ?nancially by NWO, the Netherlands Organization for Scienti?c Research. RE acknowledges support from the NASA XMM-Newton grant NAG5-10032. We thank Ehud Behar (Columbia University), Ton Raassen (SRON) for helpful discussions on the uncertainties of inner-shell absorption line wavelengths. References
Anders, E. & Grevesse, N., 1989, Geochim. Cosmochim. Acta 53, 197 Arav, N., Kaastra, J. S., Steenbrugge, K. C. et al., 2002, submitted to ApJ Blustin, A. J., Branduardi-Raymont, G., Behar, E., et al., 2002, A&A, 392, 453 Branduardi-Raymont, G., Sako, M., Kahn, S. M., et al., 2001, A&A, 365, L140 Brotherton, M. S., Green, R. F., Kriss, G. A. et al., 2002, ApJ, 565, 800 Cowan, R. D., 1981, The theory of atomic structure and spectra, Univ. of California Press, Berkeley Crenshaw, D. M. & Kraemer, S. B., 1999, ApJ, 521, 572 Done, C., Pounds, K. A., Nandra, K. & Fabian, A. C., 1995, MNRAS, 275, 417 den Herder, J. W., Brinkman, A. C., Kahn, S. M., et al., 2001, A&A, 365, L7


All rights reserved Powered by 甜梦文库 9512.net

copyright ©right 2010-2021。