A HIGH RESOLUTION VIEW OF THE WARM ABSORBER IN THE QUASAR MR 2251-178

XMM-Newton observed MR 2251-178 three times from 2011 November 11–17 over three consecutive satellite orbits. Each observation was approximately 130 ks in length, with the details of the three observations listed in Table 1. First order dispersed spectra were obtained with the RGS (den Herder et al. 2001) and were reduced using the rgsproc script as part of the XMM-Newton SAS software v11.0. The spectra from each of the orbits were found to be consistent with each other, with the only variation being due to a 10% change in the count rate of the source over the three observations. Therefore, spectra and response files for each RGS were combined to produce a single spectrum with a total net exposure of 389.1 ks. There were no periods of strong background flares during the observations; the background rate in each RGS was only 7%–8% of the total source rate. Prior to spectral analysis, channels due to bad pixels on the RGS CCDs were ignored, as well as the two malfunctioning CCDs for RGS 1 and RGS 2, respectively.

The net background-subtracted count rates were 0.496 ± 0.001 s⁻¹ and 0.535 ± 0.001 s⁻¹ for RGS 1 and RGS 2, respectively, yielding a total of over 4 × 10⁵ counts for the two RGS spectra together. Spectra were binned into Δλ = 0.02 Å bins, which over samples the RGS spectral resolution by a factor of times four compared with the FWHM resolution. Due to the high count rate statistics, χ² minimization was employed in the subsequent spectral fitting. An additional ±3% systematic error was added in quadrature to each combined RGS spectrum in order to allow for systematic differences between the two grating spectra. A constant multiplicative offset was subsequently allowed between RGS 1 and RGS 2 in all the spectral fits, which was found to be within ±3%. The data were fit over the 0.33–2.0 keV energy range in the observed frame.

HETG on board Chandra (Weisskopf et al. 2000; Canizares et al. 2005) also observed MR 2251-178 from 2011 26 September to 2 October, approximately 40 days before the XMM-Newton observations. As per the XMM-Newton observations, the Chandra observations occurred over three consecutive orbits, with the last sequence somewhat shorter than the first two—see Table 1 for details. Spectra were extracted with the ciao package v4.3. Only the first order dispersed spectra were considered for both the Medium Energy Grating (MEG) and the High Energy Grating (HEG) and the ±1 orders for each grating were subsequently combined for each sequence. No significant spectral variability was observed between the three sequences and the spectra were consistent, with only modest ∼10% variations in source flux. Therefore, the spectra were combined from all three sequences to yield a single first order spectrum for each the MEG and the HEG, yielding respective net source count rates of 0.485 ± 0.001 s⁻¹ and 0.245 ± 0.001 s⁻¹, respectively, for a total exposure time of 392.9 ks. Thus, the total counts obtained exceeded 1.9 × 10⁵ and 9.5 × 10⁴ counts for the MEG and the HEG, respectively. Note that the background contribution to the count rate was negligible.

The resulting 2011 source spectra were subsequently binned to Δλ = 0.02 Å and Δλ = 0.01 Å bins for the MEG and the HEG observations, respectively, which samples their respective FWHM spectral resolutions. The MEG and HEG spectra were analyzed over the energy ranges of 0.5–5.0 keV and 1.0–9.0 keV, respectively. The C-statistic was employed in the subsequent spectral fits to the HETG; although the overall count rate is high, the total counts per bin dropped below N = 20 in some bins toward the lower energy (longer wavelength) end of each grating spectrum. In the case of χ² minimization, this would lead to the continuum level being somewhat underestimated at soft X-ray energies.

An archived Chandra HETG observation of MR 2251-178 was also taken on 2002 September 11, with a total net exposure of 146.3 ks. First order spectra for the MEG and the HEG were re-extracted as above, yielding count rates of 0.317 ± 0.001 s⁻¹ and 0.164 ± 0.001 s⁻¹, respectively. Thus, the 2011 observation was approximately 50% higher in count rate or flux than the earlier 2002 observation and therefore the 2002 dataset provides a lower flux comparison spectrum. The data were binned and analyzed over the same energy ranges as per the 2011 observation and the C-statistic was employed in all subsequent spectral fits.

Initially, we concentrated on the 2011 RGS and HETG observations. All parameters are given in the rest frame of the quasar at z = 0.06398, unless otherwise stated, and spectral parameters are quoted in energy units (thus, 1 keV is equivalent to 12.3984 Å). In all the fits, a Galactic absorption of hydrogen column density of N_H = 2.4 × 10²⁰ cm⁻² (Kalberla et al. 2005) was adopted and modeled with the "Tuebingen–Boulder" absorption model (tbabs in xspec) using the cross-sections and abundances of Wilms et al. (2000). Figure 1 shows the overall 2011 fluxed RGS spectrum of MR 2251-178, plotted against a power law with Γ = 2 in the soft X-ray band and in the quasar rest frame at z = 0.06398. The spectrum shows several clear signatures of a warm absorber and emitter. A deep absorption trough is present between 0.7 and 0.8 keV that is most likely identified as an UTA, due to 2p → 3d transitions from lower ionization M-shell iron (i.e., Fe less ionized than Fe xvii; Behar et al. 2001). The iron M-shell UTA has been commonly observed in high-resolution grating spectra of many AGNs (McKernan et al. 2007), e.g., IRAS 13349+2438 (Sako et al. 2001), NGC 3783 (Kaspi et al. 2000, 2001; Krongold et al. 2003), NGC 5548 (Kaastra et al. 2002; Andrade-Velázquez et al. 2010), Mrk 509 (Pounds et al. 2001; Yaqoob et al. 2003; Smith et al. 2007), NGC 7469 (Blustin et al. 2007), Mrk 841 (Longinotti et al. 2010), IC 4239A (Steenbrugge et al. 2005b), NGC 3516 (Holczer & Behar 2012), Ark 564 (Papadakis et al. 2007), MCG-6-30-15 (Lee et al. 2001; Turner et al. 2004), NGC 4051 (Pounds et al. 2004a), Mrk 279 (Costantini et al. 2007), I Zw1 (Gallo et al. 2004), 1H 0419-577 (Pounds et al. 2004b), and PG 1114+445 (Ashton et al. 2004).

Zoom In Zoom Out Reset image size

Several narrow absorption lines appear to be present between 0.85 and 1.0 keV, likely due to K-shell 1s → 2p lines of neon as well as higher ionization L-shell (2p → 3d) lines of iron (i.e., Fe xvii–xxii). A broad absorption trough appears to be present near 1.3 keV in the rest frame, close to the expected K-shell lines of Mg, the origin of which is discussed in Section 4. Strong and resolved line emission is also especially prominent in the RGS 1 spectrum between 0.56 and 0.58 keV, at the expected energy of the O vii triplet.

For comparison, the fluxed 2011 HETG spectrum of MR 2251-178 is shown in Figure 2. The spectrum is plotted against a power law with a photon index Γ = 1.6 for comparison purposes only; as is discussed later in Section 4.2, the likely intrinsic photon index of the source is perhaps much steeper (Γ ≳ 2) once all the layers of absorption in MR 2251-178 have been accounted for. Although the power law provides a good representation of the HETG spectrum above 3 keV, the data/model ratio residuals show pronounced curvature due to the presence of the known warm absorber in this AGN. Indeed, fitting a single power law (modified by Galactic absorption only) provided a very poor representation of the whole HETG spectrum fitted from 0.5–9.0 keV, with a very hard photon index of Γ = 1.33 ± 0.02 and an unacceptable fit statistic of C = 3806.4 for 2360 degrees of freedom (dof). Note that a multiplicative cross-normalization constant was included between the MEG and HEG spectra; the HEG normalization was found to be slightly lower (0.97 ± 0.01) than the MEG normalization (which was normalized to 1.00).

Zoom In Zoom Out Reset image size

In order to investigate and identify the atomic lines present in the HETG spectra, a more complex continuum shape was adopted to better account for the clear spectral curvature. A power-law continuum was adopted and modified by a neutral partial covering absorber (the pcfabs model in xspec). While this simple partial coverer is not meant to provide a physical description of the spectrum, its advantage is that it provides a better parameterization of the spectral curvature, while not imparting any discrete atomic lines on the spectrum, and thus provides a reference continuum from which lines can be identified. A similar approach was also taken by Gofford et al. (2011) to provide an initial parameterization of the broadband Suzaku spectrum of MR 2251-178. In addition to the partial covering absorption, a phenomenological absorption edge component was initially included to account for the pronounced spectral drop above 0.7 keV, due to a possible combination of the Fe M-shell UTA and the O vii edge. Again, this was not meant to provide a physical fit to the spectrum. The edge energy was E = 730.6 ± 2.1 eV with an optical depth of τ = 0.36 ± 0.05. The partial coverer had a column density N_H = (2.9 ± 0.2) × 10²² cm⁻² and a covering fraction of 0.35 ± 0.03, while the photon index was Γ = 1.69 ± 0.03. The overall fit statistic was much improved compared with the power-law only case, with C = 3047 for 2356 dof.

Figure 3 (MEG) and Figure 4 (HEG) show the residuals against the neutral partial covering model in the soft X-ray band below 2 keV. A wealth of absorption lines are clearly present in the HETG spectrum against the continuum model over the 0.7–2.0 keV energy range (or 6–18 Å). In order to parameterize the lines, successive narrow Gaussian absorption lines were included in the continuum model; an individual line was deemed to be statistically significant if its addition to the model resulted in an improvement of the fit statistic of ΔC > 9.2, corresponding to 99% significance for two interesting parameters. The width of the absorption lines was initially assumed to be less than the instrumental resolution. The parameters of all 31 of the statistically significant absorption lines detected in the soft X-ray HETG spectrum are shown in Table 2. A narrow structure is also clearly present in the UTA region around 0.73–0.76 keV; these are parameterized by two lines in Table 2. This structure may be due to iron in the ionization states Fe vii–x, based on a comparison with the blends of transitions noted in Behar et al. (2001).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Table 2. Soft X-Ray Absorption Lines in the 2011 Chandra HETG Spectrum

IDa	Eatomicb	Equasarc	EWc	ΔCd
O viii	653.5 [18.972]	$654.4^{+0.1}_{-0.3}$ [18.946]	−0.8 ± 0.3	9.8
Fe vii–viii 2p − 3d	...	$733.3^{+0.5}_{-0.9}$ [16.908]	−1.6 ± 0.5	20.4
Fe ix–x 2p − 3d	...	750.4 ± 1.2 [16.522]	−1.5 ± 0.6	18.0
Ne v	871.4 [14.228], 873.7 [14.191]	873.3 ± 0.5 [14.197]	−1.2 ± 0.4	15.6
Ne vi	885.0 [14.010], 883.3 [14.036]	884.5 ± 0.5 [14.017]	−1.7 ± 0.4	27.3
Ne vii	898.2 [13.804]	897.7 ± 0.4 [13.811]	−1.6 ± 0.5	29.9
Ne viii	909.2 [13.637]	908.6 ± 0.3 [13.646]	−1.0 ± 0.3	12.1
Fe xix 2p − 3d	918.0 [13.506]	$918.6^{+0.2}_{-0.4}$ [13.497]	−1.0 ± 0.4	15.2
Ne ix	922.0 [13.447]	922.6 ± 0.4 [13.439]	−1.5 ± 0.5	25.7
Fe xx 2p − 3d	967.3 [12.818]	966.4 ± 0.7 [12.829]	−1.3 ± 0.5	17.2
Fe xx 2p − 3d	987.8 [12.552]	986.4 ± 0.8 [12.569]	−1.2 ± 0.4	14.7
Fe xxi 2p − 3d	1000.9 [12.387]	1010.1 ± 0.8 [12.274]	−0.9 ± 0.4	9.8
Ne x	1021.5 [12.137], 1022.0 [12.132]	1022.8 ± 0.4 [12.122]	−1.8 ± 0.3	56.6
Fe xxii 2p − 3d	1053.6 [11.768]	1052.9 ± 1.0 [11.766]	−1.0 ± 0.4	13.6
Ne ix 1s − 3p	1073.8 [11.546]	1074.1 ± 0.7 [11.543]	−1.0 ± 0.3	18.5
Ne ix 1s − 4p	1127.1 [11.000]	1128.6 ± 0.6 [10.986]	−1.2 ± 0.3	36.3
Ne ix 1s − 6p	1165.0 [10.642]	1165.5 ± 0.7 [10.638]	−0.7 ± 0.3	11.2
Ne x 1s − 3p	1210.9 [10.239]	1211.7 ± 0.5 [10.232]	−0.6 ± 0.3	11.6
Mg vi	1276.8 [9.711]	1276.9 ± 0.9 [9.710]	−0.7 ± 0.3	12.6
Mg vii	1291.6 [9.599]	1294.5 ± 1.3 [9.578]	−0.7 ± 0.3	10.8
Mg viii	1306.4 [9.491], (1304.2 [9.507])	1306.4 ± 0.7 [9.491]	−1.4 ± 0.3	41.8
Mg ix	1323.1 [9.371]	1322.6 ± 1.1 [9.374]	−0.6 ± 0.3	9.7
Mg xi	1353.1 [9.163]	1352.7 ± 0.8 [9.166]	−0.8 ± 0.3	18.2
Mg xii	1472.6 [8.419], 1471.7 [8.425]	1473.0 ± 0.5 [8.417]	−1.2 ± 0.2	47.3
Mg xi 1s − 3p	1579.3 [7.851]	1581.3 ± 0.9 [7.841]	−0.6 ± 0.3	12.6
Si viii	1772.8 [6.994]	1772.6 ± 0.6 [6.994]	−1.3 ± 0.3	38.8
Si ix	1792.2 [6.918], (1788.2 [6.933])	$1791.9^{+0.9}_{-1.2}$ [6.919]	−0.9 ± 0.3	19.2
Si x	1810.3 [6.849], 1807.3 [6.860]	$1809.6^{+1.4}_{-2.2}$ [6.851]	−0.8 ± 0.3	15.5
Si xi	1830.6 [6.773]	1830.2 ± 1.1 [6.774]	−0.7 ± 0.3	14.1
Si xiii	1866.4 [6.643]	1866.0 ± 0.9 [6.644]	−1.0 ± 0.3	23.5
Si xiv	2006.1 [6.180], (2004.8 [6.184])	2007.0 ± 1.2 [6.178]	−0.8 ± 0.4	12.3

Notes. ^aLine identification. Lines correspond to the 1s − 2p transition unless stated otherwise. ^bKnown atomic line energy in eV. Values are from www.nist.gov, Behar et al. (2001) for Fe M-shell UTA, and Behar & Netzer (2002) for inner shell Ne, Mg, and Si. The corresponding wavelength in Å is given within brackets. ^cMeasured line energy and EW in the quasar rest frame, units eV. The corresponding mean wavelength value in Å is given within brackets. ^dImprovement in the C-statistic upon adding the line to the model.

Download table as:  ASCIITypeset image

Further low ionization gas appears to be present in the form of a multitude of inner K-shell lines of Ne, Mg, and Si. These are 1s → 2p absorption lines whereby the L-shell is partially occupied, i.e., due to charge states corresponding to Li, Be, B, C, N, and O-like ions, etc. We refer the reader to Behar & Netzer (2002) for a compilation of these inner-shell lines; we adopt the known energies (wavelengths) of these lines from this paper in Table 2. Indeed, such lines have been detected in other high signal-to-noise grating spectra of Seyfert 1 AGNs, such as NGC 3783 (Kaspi et al. 2002; Blustin et al. 2002), NGC 4151 (Kraemer et al. 2005), Mrk 509 (Kaastra et al. 2011a), NGC 3516 (Holczer & Behar 2012), NGC 4051 (Lobban et al. 2011), and NGC 5548 (Steenbrugge et al. 2005a). In the MR 2251-178 HETG spectrum, absorption lines due to Ne v–viii (i.e., C-like through Li-like ions) are detected from 0.87–0.91 keV (13.6–14.3 Å) in the rest frame (Figure 3). Similarly, 1s → 2p inner-shell lines from Mg are detected due to Mg vi–ix (N-like through Be-like ions) from 1.26–1.33 keV (9.3–9.8 Å). Likewise, inner-shell absorption is also detected from Si, from Si viii–xi (N-like through Be-like) around 1.8 keV (6.5–7.0 Å). The inner-shell absorption is also independently detected in the HEG (Figure 4), as well as the MEG (Figure 3) spectra. Thus, the detection of the strong Fe M-shell UTA, plus the inner-shell absorption due to Ne, Mg, and Si suggests the imprint of a significant amount of absorption due to both low and high ionization gas in MR 2251-178.

Absorption lines due to more highly ionized gas are also significantly detected in the HETG spectrum. He- and H-like lines of O, Ne, Mg, and Si are all detected (with the exception of the O vii 1s → 2p line due to the lack of S/N below 0.6 keV in the MEG spectrum). In some cases, higher order 1s → np lines are detected, especially in the case of Ne ix where the series of resonance lines up to 1s → 6p are seen. Higher ionization L-shell lines of iron are also present, e.g., from Fe xix–xxii. The spectra over the S and Fe K band are also shown in Figure 4, although note that neither strong emission nor absorption features appear to be present in these parts of the spectrum. In the S band, weak absorption may be present at the expected energies of the He- and Li-like lines of S, although they are below the formal detection threshold. The details of the iron K band spectrum will be discussed further in Section 4.4.

It is also apparent from Table 2 that most of the measured rest frame energies of the absorption lines are close to the known atomic energies. This suggests that the outflow velocity of the soft X-ray absorbing gas is relatively small. We discuss below some of the velocity profiles of the strongest detected H- and He-like lines.

The RGS provides an energy coverage of 0.3–2.0 keV with high throughput and therefore provides a high quality view of the soft X-ray warm absorber, with a lower energy bandpass than the Chandra HETG. The initial analysis of the absorption line spectrum suggests that multiple absorption components may be required in order to model the wide range of ionization states of the gas, e.g., covering, for instance, Fe vii–xxii, Ne v–x, or Mg vi–xii.

Indeed, enlarged portions of the RGS spectrum of MR 2251-178 are shown in Figures 5 and 6. Note that these are plotted in the observed frame and not in the rest frame. The warm absorber is clearly complex, comprising a wealth of atomic features. Notably, inner-shell (Li-like and below) and higher-order (i.e., the 1s → np transitions, where n ⩾ 3) absorption lines are detected throughout the spectrum, due to C, N, O, Ne, and Mg. Figure 5 shows that the higher-order line series of C vi is particularly prominent, while N vi, N vii, O vii, and O viii also have higher-order line series, with each ion reaching at least the 1s → 4p transition.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Complementing the array of absorption lines is also some interesting interplay between emission and absorption components; e.g., see the O vii line at 517–539 eV (23–24 Å) in the observed frame in Figure 5. The O vii (1s2p → 1s²) emission line complex is superimposed by three narrow absorption lines corresponding to inner-shell absorption due to O v (line 11, Figure 5) and the two lines that make up the O vi (1s²2s → 1s2s2p) doublet (lines 13 and 14; Figure 5). Again, similar structures are present at other energies, with N vii, O viii, and Ne ix all showing emission superposed by absorption. The nature of the emission line spectrum will be discussed further in Section 5.

From panels (a) and (b) of Figure 6, both neon and magnesium also show evidence for inner-shell absorption from at least their Be-like ionization states (Behar & Netzer 2002). Indeed, the inner-shell lines for Mg in particular occur throughout the ∼1.2–1.3 keV energy range, as per the HETG. This appears to be the origin of the absorption trough visible in Figure 1 and first noted in the lower resolution Suzaku spectrum of MR 2251-178 published by Gofford et al. (2011). The complete list of atomic lines identified in the RGS data—including details such as the responsible ion, the electron transition, and the centroid energies in the source rest frame—is given in Table 3.

We constructed velocity profiles of the strongest H- and He-like absorption lines identified in the above HETG and RGS spectra. In each case, the profiles were constructed by taking the ratio of the data to the best-fit parameterization of the continuum model described above and transposing them into velocity space around the known lab frame energy (wavelength) of each line. For the H-like ions, the C vi, N vii, O viii, Ne x, Mg xii, and Si xiv profiles have been produced, with the profiles plotted in Figure 7. Note that the C vi line corresponds to the 1s − 3p absorption line (as the 1s − 2p line at the redshift of MR 2251-178 is close to the edge of the RGS bandpass), while the other profiles correspond to the 1s − 2p lines. Similarly, profiles were also constructed for the He-like resonance lines of N vi, O vii, Ne ix, Mg xi, and Si xiii and are shown in Figure 8 (note that only the first four profiles are actually plotted here). In the case of the He-like ions, the 1s − 3p lines of O vii and Ne ix are used instead of the 1s − 2p lines, due to contamination with other lines present in the spectrum. Overall, the profiles from the C, N, and O lines were derived from the RGS data in the soft X-ray part of the spectrum (taking the mean of RGS 1 and RGS 2 where both were available), while the Ne, Mg, and Si profiles were derived from the HEG data at higher energies. Note that negative velocities indicate blueshifts throughout this paper.⁸ The profiles are as measured from the data, without correcting for the spectral resolution of the instrument.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The subsequent lines were fit with Gaussian profiles and the results of the fits are shown in Table 4, which gives both the overall velocity shift (v_out) of the line profile (as determined from the centroid of the Gaussian profile) as well as the observed 1σ velocity width of the profile (σ_obs). First, it can be seen both from the profiles themselves and from the fits that the outflow velocities of the lines tend to decrease in magnitude with increasing ionization state, e.g., from C through Si. For instance, for the H-like ions, the C vi and N vii profiles have a velocity shift of v_out ∼ −450 km s⁻¹, with the Mg xii profile having a formal upper limit on the outflow velocity of only v_out < 40 km s⁻¹, while the velocity centroids for O viii and Ne x are somewhat intermediate in value. We note that a similar possible trend was found in emission in the Seyfert 2 galaxy NGC 1068 (Kinkhabwala et al. 2002), whereby the higher energy (excitation) lines had somewhat lower velocities.

Table 4. Gaussian Fits to Velocity Profiles of H- and He-like Absorption Lines

Line	Instrument	σobsa	σintb	voutc	Δχ2 d
H-like:
C vi Ly-β (a)	RGS	320 ± 80	<270	−444 ± 73	43.0
C vi Ly-β (b)e	RGS	...	...	−1840 ± 100e	16.0
N vii Ly-α (a)	RGS	480 ± 95	$360^{+120}_{-140}$	−450 ± 94	57.5
N vii Ly-α (b)e	RGS	...	...	−2020 ± 120e	8.2
O viii Ly-α	RGS	297 ± 65	<200	−353 ± 62	47.0
Ne x Ly-α	HEG	415 ± 135	$395^{+170}_{-145}$	−227 ± 123	33.1
Mg xii Ly-α	HEG	<120	...	<40	31.2
Si xiv Ly-α	HEG	<125	...	<340	8.7
He-like:
N vi He-α (a)	RGS	320 ± 90	<280	−428 ± 88	26.2
N vi He-α (b)e	RGS	...	...	−1990 ± 120e	12.1
O vii He-β	RGS	600 ± 180	$460^{+220}_{-280}$	−900 ± 180	22.3
Ne ix He-β	HEG	440 ± 190	$420^{+180}_{-200}$	<234	14.1
Mg xi He-α	HEG	310 ± 175	<455	<170	7.0
Si xiii He-α	HEG	<120	...	<163	8.8

Notes. ^aObserved 1σ width of the absorption line in km s⁻¹. ^bIntrinsic 1σ width of the absorption line in km s⁻¹ after correcting for instrumental spectral resolution. ^cVelocity shift of the absorption line in km s⁻¹. Negative values denote a blueshift. Upper limits on outflow velocities are expressed as absolute values for clarity. ^dImprovement in fit statistic after modeling the Gaussian absorption profile. ^ePossible higher velocity component of the absorption line.

Download table as:  ASCIITypeset image

The velocity profiles and fits also indicate that a second higher velocity component may be present in the lower energy lines of C vi, N vi, and N vii, with an outflow velocity of v_out ∼ −2000 km s⁻¹. Such a component is not present in the higher energy lines. The outflow velocities of the possible higher velocity components are also given in Table 4, noting that the line width of this component was assumed to be the same as for the respective lower outflow velocity lines. Thus, while we note the possible presence of a higher velocity component in some of the lines, we do not discuss this further here, as the improvement in the fit statistic upon adding this second velocity component (see Table 4) was generally less than the more robust low velocity component that is always present.

The observed velocity widths of the Gaussian profiles (σ_obs) are also given in Table 4. These are not corrected for instrumental resolution, however, for comparison, the σ widths of the RGS (RGS 1+2 combined) varies between σ = 300 and 380 km s⁻¹ for C vi to O viii. For the HEG, the corresponding values are between σ = 120 and 230 km s⁻¹ for Ne x to Si xiv. The intrinsic line widths corrected for instrumental resolution (σ_int) are also given in Table 4. Thus, some of the line profiles appear resolved, with typical widths of σ_int = 300–400 km s⁻¹, while the higher energy lines (e.g., Mg xii and Si xiii) appear to be unresolved, similar to the potential trend in outflow velocity discussed above.

We first considered the RGS spectrum. The initial analysis of the absorption line spectrum from the HETG and RGS observations in Section 3 suggests that multiple absorption components may be required in order to model the wide range of ionization states of the gas, e.g., covering, for instance, Fe vii–xxii, Ne v–x, or Mg vi–xii.

In order to model the absorption spectrum, we successively added individual components of absorbing gas, fully covering the line of sight to the source, until the fit statistic was no longer improved at the 99.9% confidence level. Three components of fully covering gas are formally required in the RGS model, which are listed as components 1–3 in Table 5. The lowest ionization absorber (component 1) was modeled by the low ionization xstar grid as described above and components 2–3 were modeled by the higher ionization grid. We note that the continuum itself was assumed to be a power law with a variable photon index, absorbed by the Galactic column, while we no longer retain either the ad-hoc absorption edge or the simple neutral partial coverer in the models. However, we do allow for at least one additional component of partially ionized absorbing gas (as modeled by an xstar grid) to partially cover the X-ray source, in addition to the three fully covering components of gas described above, which appears to be required statistically to achieve a good fit. Soft X-ray emission lines are also added to the model as Gaussians when statistically required by the data at >99% and will be discussed in detail later. Thus, the phenomenological form of the spectral model fitted to the RGS data is

$$\begin{eqnarray} F(E) &=& {\rm tbabs} \times {\rm comp1} \times {\rm comp2} \times {\rm comp3} \nonumber\\ && \times [{\rm pow}_{\rm uncov} + {\rm Gauss} + ({\rm pc_{1}} \times {\rm pow}_{\rm cov})], \end{eqnarray} \tag{ 1 }$$

where comp 1–3 represent the three fully covering warm absorber components, Gauss represents the Gaussian emission lines, and tbabs represents the Galactic absorption. The partial covering absorber is represented by pc₁, which covers a fraction f_cov of the line of sight to the X-ray source, while 1 − f_cov is subsequently unattenuated by the partial covering component. Thus, the fraction of the continuum that is absorbed is simply given by the respective ratio of the power-law normalizations, i.e.: f = N_cov/(N_cov + N_uncov). The spectral parameters of the RGS fit are listed in Table 5.

Table 5. Warm Absorber Parameters from the RGS and HETG Spectra

Component	Parameter	RGS 2011	HETG 2011	HETG 2002
Power law	Γ	2.32 ± 0.08	$2.13^{+0.11}_{-0.10}$	=2011i
(uncovered)	funcova	$0.39^{+0.03}_{-0.02}$	0.23 ± 0.03	0.18 ± 0.02
Warm Absorber	NHb	2.12 ± 0.07	$2.10^{+0.19}_{-0.23}$	$1.43^{+0.34}_{-0.37}$
(Component 1)	log (ξ/erg cm s−1)c	1.27 ± 0.02	1.15 ± 0.05	0.91 ± 0.16
	voutd	−480 ± 40	−315 ± 40	−290 ± 150
ΔC or Δχ2 e	...	2065	176.2	...
Warm Absorber	NHb	1.50 ± 0.20	$1.5^{+0.3}_{-0.5}$	$1.2^{+0.9}_{-0.7}$
(Component 2)	log (ξ/erg cm s−1)c	$2.04^{+0.04}_{-0.07}$	$2.14^{+0.10}_{-0.11}$	$2.03^{+0.23}_{-0.13}$
	voutd	−470 ± 60	$-260^{+30}_{-60}$	$-150^{+130}_{-140}$
ΔC or Δχ2 e	...	244.6	148.1	...
Warm Absorber	NHb	3.6 ± 1.3	$1.7^{+0.7}_{-0.6}$	$2.3^{+2.5}_{-1.4}$
(Component 3)	log (ξ/erg cm s−1)c	$2.80^{+0.05}_{-0.07}$	$2.88^{+0.11}_{-0.14}$	$2.9^{+0.4}_{-0.3}$
	voutd	<130	<70	$-380^{+200}_{-220}$
ΔC or Δχ2 e	...	18.5	33.7	...
Partial Coverer	NHb	60.0f	$55^{+2}_{-3}$	=2011i
(pc1)	log (ξ/erg cm s−1)c	1.0f	$1.04^{+0.08}_{-0.11}$	=2011i
	fcov1g	0.61 ± 0.05	0.40 ± 0.10	0.39 ± 0.07
ΔC or Δχ2 e	...	213.1	193.1	...
Partial Coverer	NHb	...	$690^{+90}_{-100}$	=2011i
(pc 2)	log (ξ/erg cm s−1)c	...	1.04f	=2011i
	fcov2g	...	0.37 ± 0.10	0.43 ± 0.11
ΔC or Δχ2 e	...	...	31.6	...
Total Fluxh	F0.5 − 2.0	1.80 ± 0.01	1.33 ± 0.01	0.75 ± 0.01
	F2.0 − 10.0	...	3.8 ± 0.1	2.5 ± 0.1

Notes. ^aUncovered fraction of the power-law component. ^bHydrogen column density, in units × 10²¹ cm⁻². ^cLog ionization parameter. ^dOutflow velocity in units of km s⁻¹. Negative values indicate outflow. ^eImprovement in either the C-statistic (HETG) or χ² (RGS) upon the addition of the component to the model. ^fIndicates parameter is fixed. ^gCovering fraction of the partial covering component. ^h0.5–2.0 keV or 2–10 keV band flux, in units × 10⁻¹¹ erg cm⁻² s⁻¹. ⁱParameter is tied to the 2011 HETG value.

Download table as:  ASCIITypeset image

Overall, the three warm absorber components that are required to model the RGS spectrum cover the range in column density from N_H = 1.5–3.6 × 10²¹ cm⁻² and ionization parameter from log (ξ/erg cm s⁻¹) = 1.27–2.80. Consistent outflow velocities are found for the low and medium ionization components 1 and 2, with v_out = −480 ± 40 km s⁻¹ and v_out = −470 ± 60 km s⁻¹, respectively. However, the highest ionization component 3 does not require an outflow velocity (formally consistent with zero) and only a limit can be placed with v_out < 130 km s⁻¹. We note that the lack of any outflow velocity of component 3 also appears consistent with the velocity profile analysis in Section 3.4, where the velocities of the higher excitation lines appear to be lower. The column density (6 × 10²² cm⁻²) and ionization (log (ξ/erg cm s⁻¹) = 1) of the partial covering component are not well constrained in the RGS fit, mainly because the limited higher energy bandpass of the RGS makes it difficult to constrain multiple continuum components, while the partial coverer itself does not impart discrete detectable lines upon the soft X-ray spectrum (but it does impart continuum curvature). Thus, its column and ionization have been fixed in the model, while we note that these values are consistent with those obtained with the HETG in Section 4.2.

Nonetheless, the partial coverer is certainly required in the model; the fit statistic is increased by Δχ² = 192.4 upon removing the partial coverer from the model and refitting; its exclusion leads to systematic broad residuals in the data/model ratio suggesting that the continuum is inadequately modeled. The covering fraction of the partial coverer is f_cov = 0.61  ±  0.05. Overall, the fit statistic for the best-fit warm absorber model is χ²/dof = 2991.7/2562, while the continuum photon index upon modeling all three required components of warm absorption is steeper, with Γ = 2.32 ± 0.08. The warm absorber model reproduces well the absorption lines observed in the RGS spectrum, as shown by the solid line in Figures 5 and 6. We also note that in addition to the warm absorption, an additional neutral component of absorption is required in the rest frame of MR 2251-178. However, its column density is quite small, N_H = (2.8 ± 0.3) × 10²⁰ cm⁻², and it may plausibly be associated with absorption in the quasar host galaxy rather than the AGN.

The relatively low turbulence velocity (σ = 100 km s⁻¹) of the warm absorber components aides in the modeling of the higher order lines, as some of the 1s → 2p lines may lie on the saturated part of the curve of growth. This means the some of the higher order lines can be of comparable strength as the 1s → 2p lines, while some of the line series are detected up to 1s → 6p. Indeed, the warm absorber model matches well the profiles of the higher order lines, as can be seen in Figures 5 and 6.

To correctly account for the intensity of the low ionization lines, the absorbing grid requires a much softer (steeper) input continuum than the other higher ionization absorption components (which have Γ_input = 2.0), in order not to over-ionize the gas and reduce their depth in the model. The necessary power-law continuum required by the xstar grid in order to model the low ionization lines is Γ_input = 2.5. This is much softer than what has typically been found for MR 2251-178 assuming a fully-covering absorption model, which is of the order of Γ = 1.6–1.7 (Pan et al. 1990; Mineo & Stewart 1993; Kaspi et al. 2004; Gibson et al. 2005). However, the underlying soft X-ray photon index recovered in the RGS spectrum (Γ = 2.32 ± 0.08), after the required absorbing layers of gas are accounted for, is in reasonable agreement with the photon index required to reproduce the soft X-ray lines. This lends weight to the notion that MR 2251-178 may, indeed, have an intrinsically soft continuum that is partially covered by a complex and stratified absorber. We discuss this further in Section 4.2.

The above best-fit model was then applied to the 2011 HETG spectrum, allowing the continuum and warm absorber parameters to vary between the datasets. A second partial covering component of higher column density of ∼7 × 10²³ cm⁻² was added to the model, as the direct application of the RGS model gave a slight excess at higher energies in the HETG spectrum. Otherwise, the model construction applied to the HETG data is identical to the RGS that applied to data.

The absorber fit parameters applied to the 2011 HETG spectrum are also listed in Table 5. The parameters of the three warm absorber components are rather similar to those obtained from the RGS data, with most of the values consistent within the errors between the observations. Similar to the RGS, the warm absorber column densities cover the narrow range N_H = 1.5–2.1 × 10²¹ cm⁻², while the ionization spans a range from log (ξ/erg cm s⁻¹) = 1.15–2.9. There is evidence for a small change in the ionization of the warm absorber of the low ionization component 1, increasing from log (ξ/erg cm s⁻¹) = 1.15 ± 0.05 to log (ξ/erg cm s⁻¹) = 1.27 ± 0.02 between the HETG and the RGS, following the same direction as the 0.4–2.0 keV continuum flux that also increased from the HETG to the RGS; we discuss this further in Section 4.3 below. The column density of component 1 is consistent between observations, with N_H = 2 × 10²¹ cm⁻², although the outflow velocity is slightly smaller,⁹ with v_out = −315 ± 40 km s⁻¹. The ionization and columns of components 2 and 3 are consistent within the errors, while as per the RGS, the highest ionization component 3 does not require any outflow, as noted above.

Figure 9 shows the relative contributions of each of the three warm absorber components against a power-law continuum. The lowest ionization component 1 (top panel) contributes the lower ionization ions, i.e., O v–vii, Ne v–viii, Mg vi–ix, and Si viii–xi, as well as M-shell iron, as expected. The higher ionization components produce most of the He and H-like ions, as well as the higher ionization (L-shell) iron ions (see the lower panels).

Zoom In Zoom Out Reset image size

4.2.1. The Nature of the Photoionizing Continuum

The HETG has a wider bandpass and higher resolution than the RGS, which allows some additional tests to be applied to the inner-shell lines in particular. Figure 10 shows a comparison between the fit to the warm absorber when the low ionization component (component 1) of xstar absorption has a Γ_input = 2.5 input photoionizing continuum (blue line) or Γ_input = 2.0 (red line). For the case of the harder Γ = 2 input continuum, the model is clearly unable to account for the depth of the inner-shell (Li-like and below) charge states of Ne or Mg, whereas the Γ_input = 2.5 absorber is able to model the low ionization absorption lines. This suggests that a softer input continuum is strongly required to model the absorption. The absorption grid with the steeper continuum also provides a better fit to the Fe M-shell UTA and also the silicon inner-shell lines. These differences are reflected in the fit statistic, which for the Γ = 2 grid is C = 2665.9 for 2335 dof. On the other hand, for the Γ = 2.5 grid, the fit statistic is C = 2542.4 for the same number of dof, corresponding to a difference of ΔC = 123.5.

Zoom In Zoom Out Reset image size

Overall, the photon index of the continuum recovered after modeling all the layers of absorption is Γ = 2.13 ± 0.10. Thus, the index is somewhat flatter than in the RGS (Γ = 2.32), but this may reflect the fact that the RGS is more sensitive at soft X-ray energies than the HETG, especially if the intrinsic continuum has subtle curvature, becoming slightly steeper toward lower energies. Note that Figure 2 also shows the level of the intrinsic continuum (the dashed blue line) after correcting for all the absorbing layers of gas. Thus, the observed continuum without modeling the absorption (which would otherwise appear to have a very hard photon index of Γ = 1.3) does not necessarily represent the intrinsic emission, where Γ ≳ 2, more typical of radio-quiet quasars (e.g., Reeves & Turner 2000; Porquet et al. 2004; Scott et al. 2011).

The partial covering components also appear to be required by the data. The moderate column partial covering component (called pc1; Table 5) appears well constrained, with N_H = 5.5 ± 0.3 × 10²² cm⁻² and $\log (\xi /\rm {erg\,cm\,s}^{-1})= 1.04^{+0.08}_{-0.11}$, while its covering fraction is f_cov = 0.4 ± 0.1. The highest column component (pc2; Table 5) is less well constrained, but the fit is still worse by ΔC = 31.6 if this component is removed from the model and the continuum refitted. The removal of the pc2 absorber results in the fitted photon index hardening from Γ = 2.13 ± 0.10 to Γ = 1.77 ± 0.05. Furthermore, if the more moderate column partial coverer (pc1) is also removed, then the fit is considerably worse (ΔC = 213.1) and the photon index then becomes an unphysical Γ = 1.49 ± 0.03.

Such a hard continuum slope also poses a problem for the modeling of the warm absorber components, as the low ionization (inner shell) absorption requires a soft input photoionizing continuum of Γ ∼ 2.5 as above, which cannot be recovered in the model without applying the partial covering absorption. The other possibility is that the intrinsic continuum shape and high energy spectral energy distribution are unusual, consisting of a rather hard power-law component Γ ∼ 1.5 (much harder than usually observed in radio-quiet quasars), then softening to an index of Γ ≳ 2.5 at soft X-ray energies. The broadband continuum modeling will be explored in more detail in a forthcoming paper (E. Nardini et al. 2013, in preparation), where the XMM-Newton EPIC and Optical Monitor data will be considered, as well as archival Suzaku and Swift/BAT observations, thereby covering the optical/UV through the hard X-ray bandpass.

We note that although a softer continuum does provide a better fit to the inner-shell lines and some improvement to the Fe UTA, the model fits to these inner-shell features is dependent on the calculation of the ionization balance for these elements. For example, in their analysis of the 900 ks HETG spectrum of NGC 3783, Netzer et al. (2003) noted that their best warm absorber model did not accurately reproduce the Fe UTA due to the predicted iron being too highly ionized. Netzer et al. (2003) suggested that the problem was the lack of accurate low-temperature (Δn = 0) dielectronic recombination (DR) rates for the M-shell sequence of iron (Fe ix–Fe xvi). Following this, Netzer (2004) and Kraemer et al. (2004) incorporated estimated Δn = 0 DR rates into the codes ION (Netzer 1996) and Cloudy (Ferland et al. 1998), respectively, and demonstrated that such rates would shift the overall ionization balance of M-shell iron downward, hence solving the problem described by Netzer et al. (2003).

More recently, DR rates have been computed (Badnell 2006) for the M-shell states of iron, which are included within xstar. These are an order of magnitude greater than the radiative recombination rates for these ions and several times greater than the estimated DR rates from Netzer (2004) and Kraemer et al. (2004). Furthermore, these rates have been confirmed in storage-ring experiments (Schmidt et al. 2006). However, while for the same physical parameters as those used in Netzer et al. (2003), Cloudy models using the new DR rates predict similar C, N, and O column densities, the predicted Fe ionization is now too low to fit the UTA. Although it may be possible to recover the fit by changing model parameters (e.g., the continuum slope), these results may also indicate that some process that mitigates the effects of the new DR rates is not being accurately treated. One possibility is (multi-electron) autoionization following inner-shell ionization (D. Savin 2013, private communication). In any event, given such sensitivity to the accuracy and availability of atomic data, the exact parameterization of the low ionization absorber could differ, with the ionization perhaps somewhat lower than currently inferred by xstar.

The best-fit absorption model to the 2011 HETG spectrum was also applied to the earlier 2002 HETG observation. The signal to noise of the 2002 observation is substantially lower, due to the overall lower flux level (and count rate) and shorter exposure of this observation (see Table 1), which means that most of the individual absorption lines were not detected (see Gibson et al. 2005 for a description of this dataset). However, the same spectral model can still be applied to the 2002 data, allowing the continuum and warm absorber parameters to vary between the observations. For ease of comparison the photon index of the 2002 observation was tied to that of the 2011 observation, i.e., Γ = 2.13. The column and ionization of the partial covering components were also fixed to the 2011 values, as otherwise they are less well determined, although the covering fractions were allowed to vary. The warm absorber parameters (column density, ionization, outflow velocity) were allowed to vary between the observations.

The absorber parameters of the 2002 observation are shown in Table 5. Again, the absorption values are largely consistent between the 2002 and 2011 HETG observations, as well as with the 2011 RGS observations, suggesting that the absorber components appear stable over time. The main parameter that does appear to change is the ionization of the low ionization component 1 absorber. Indeed, if the 2011 RGS observation is also considered, the ionization of component 1 appears to increase from $\xi =8.1^{+3.5}_{-2.5}$ erg cm s⁻¹ (2002 September/HETG) to ξ = 14.1 ± 1.6 erg cm s⁻¹ (2011 September/HETG) to ξ = 18.6 ± 0.8 erg cm s⁻¹ (2011 November/RGS). Indeed, the changes in ξ appear to increase in direct proportion with the observed 0.5–2.0 keV band flux, varying from 0.75 ± 0.01 × 10⁻¹¹ erg cm⁻² s⁻¹ (2002 September/HETG) to 1.33 ± 0.01 × 10⁻¹¹ erg cm⁻² s⁻¹ (2011 September/HETG) to 1.80 ± 0.01 × 10⁻¹¹ erg cm⁻² s⁻¹ (2011 November/RGS). Thus, from the lowest ionization to the highest, ξ increases by a factor × 2.3, while the soft X-ray flux increases by the same factor. This would appear to suggest that the low ionization absorber is in photoionization equilibrium with the continuum. In contrast, there appears to be no change in the higher ionization components 2 and 3, within the errors. Note that this behavior is also consistent with a 2002 December (80 ks) Chandra Low Energy Transmission Grating (LETG) observation (not analyzed here), which showed about a 35% lower flux than the 2002 HETG observation, but observed the low ionization absorber to have an even lower ionization (log (ξ/erg cm s⁻¹) = 0.63 ± 0.06; Ramírez et al. 2008).

We also illustrate the apparent change in ionization further in Figure 11, which plots the change in the xstar model varying the ionization of warm absorber component 1 against the 2011 RGS data in the Fe M-shell UTA band. The top panel of Figure 11 plots the best fit model obtained, with an ionization parameter of log (ξ/erg cm s⁻¹) = 1.27 for component 1, as reported in Table 5. Then, the ionization parameter was lowered (and fixed) to log (ξ/erg cm s⁻¹) = 1.15, equal to the value found for component 1 in the 2011 HETG spectrum. This results in a worse fit, as seen in panel (b) of Figure 11; indeed, even allowing the other warm absorber and continuum parameters in the fit to vary resulted in a worse fit by Δχ² = 24.6. Similarly, if the ionization parameter is lowered still further, to log (ξ/erg cm s⁻¹) = 0.91 as obtained from the 2002 Chandra HETG data, the fit is substantially worse by Δχ² = 125.4, compared with the best fit case shown in panel (a). Indeed, this can be seen in panel (c) of Figure 11, whereby the drop in the Fe M-shell UTA region observed at 17.5–18.5 Å is too shallow compared with the data, while the spectrum is then too absorbed redward of this feature. Thus, overall, the Fe M-shell UTA region appears to be quite sensitive to the ionization state of the spectrum.

Zoom In Zoom Out Reset image size

The other possible change in the spectra is in the partial covering absorption. Considering all three grating observations, the uncovered fraction (or 1 − f) of the power law (in other words, the fraction that is not obscured by the partial covering absorption) appears to increase as the flux increases from the 2002 to the 2011 observations, from (1 − f) = 0.18 ± 0.02 to (1 − f) = 0.39 ± 0.03. This may suggest that the AGN is more obscured when it is in a lower flux state, which has been claimed in several Seyferts to date, e.g., NGC 3516 (Turner et al. 2005, 2008), PG 1211+143 (Bachev et al. 2009; Pounds & Reeves 2009), H 0557-385 (Longinotti et al. 2009), and NGC 4051 (Terashima et al. 2009; Lobban et al. 2011), and, indeed, variable X-ray absorption was first suggested from soft X-ray band variations in MR 2251-178 itself (Halpern 1984). This variability behavior will be investigated further in a subsequent paper (D. Porquet et al. 2013, in preparation), considering a broadband X-ray analysis of all the contemporary and archival observations of MR 2251-178.

Previous studies of MR 2251-178 using a 2009 Suzaku observation (Gofford et al. 2011) and the 2002 HETG observation (Gibson et al. 2005) suggested the presence of a highly ionized and possibly strongly outflowing absorption component in the iron K band. Such absorption could be similar to the very highly ionized outflows (or "ultra fast outflows") detected in about 40% of local type I AGNs with XMM-Newton (Tombesi et al. 2011) and Suzaku (Gofford et al. 2013). Thus, we have analyzed the higher energy 2011 Chandra HETG observation above 2 keV, using the HEG spectrum, to assess whether such a component is present in the new data. The 2002 HETG spectrum was also re-analyzed for comparison, while the results are also compared with the Suzaku analysis in Gofford et al. (2011).

Figure 12 shows the data/model residuals of the 2011 HEG spectrum from the best-fit absorption model discussed above, plotted over the Fe K band in the quasar rest frame, further binning the spectrum to 20 counts per bin to increase the signal to noise. First, we consider the iron K-band emission. The lack of any strong iron Kα emission is quite apparent in the residuals. Indeed, the limit on the equivalent width (EW) of a narrow 6.4 keV line is 11 ± 6 eV and is only very marginally required at ∼95% confidence in the fit, with ΔC = 6.3. The limit on the width of the line is σ < 28 eV or σ < 1300 km s⁻¹. No other iron K emission component is required in the spectrum, either narrow or broad. The weakness of the iron Kα line in MR 2251-178 has also been noted previously (Gofford et al. 2011 and references therein) and is much weaker that the typical narrow iron line EW ∼50–100 eV observed in most Seyfert 1s (e.g., Nandra et al. 1997; Patrick et al. 2012; Tatum et al. 2013). The weakness of the iron K line may be accounted for by the X-ray Baldwin effect, whereby the EW of the iron Kα line appears to decrease with AGN X-ray luminosity (e.g., Iwasawa & Taniguchi 1993; Nandra et al. 1997; Reeves & Turner 2000; Page et al. 2004; Bianchi et al. 2007; Shu et al. 2010). The 2–10 keV X-ray luminosity of MR 2251-178 in this observation is 3.7 × 10⁴⁴ erg s⁻¹ (5.8 × 10⁴⁴ erg s⁻¹ when corrected for absorption), higher than most local Seyfert 1 s.

Zoom In Zoom Out Reset image size

There does appear to be a broad but shallow absorption trough in the 2011 data at 7.3 keV. Fitting the trough with a Gaussian absorption profile gives a rest frame centroid energy of E = 7.34 ± 0.08 keV with an EW = −58 ± 24 eV and the fit statistic improves by ΔC = 15.2. Note that this appears to be consistent with the high energy absorption line that was previously claimed in the 2002 HETG observation by Gibson et al. (2005); there, the line centroid was at E = 7.26  ±  0.04 keV. Furthermore, Gofford et al. (2011) claimed an absorption trough in the Suzaku observation at an energy of $E=7.57^{+0.19}_{-0.12}$ keV, which is only marginally inconsistent at a 90% confidence level with the line energy measured in the 2011 Chandra data, while the EW = $-26^{+18}_{-12}$ eV is consistent. In Figure 12, the 2002 HEG spectrum has been overlaid on the 2011 data, with the normalization of the 2002 spectrum allowed to vary to account for the overall lower flux level in the 2002 observation; it appears that the trough in the 2002 data has a profile in both energy and depth consistent with the 2011 data. With the 2002 and 2011 data fitted together with a single Gaussian profile, consistent parameters were obtained: a line energy of E = 7.32  ±  0.06 keV and an EW = −60  ±  18 eV. The fit statistic was improved by ΔC = 26.2 with respect to a model without the absorption line. The profile appears to be resolved compared with the HETG resolution, with a width of $\sigma =120^{+50}_{-40}$ eV or $\sigma =4900^{+2100}_{-1600}$ km s⁻¹. Note that if the absorption line is associated with the Fe xxvi (H-like) 1s → 2p doublet at 6.97 keV, then the velocity shift implied is −15000 ± 2600 km s⁻¹. We also note that no significant iron Kα emission was required from refitting the 2002 HEG spectrum, although the upper limit on its EW is less well determined (<40 eV) and is consistent with the 2011 measurement.

We attempted to model the Fe K-band absorption with a highly ionized xstar grid. Unlike for the warm absorber components, a high turbulence velocity grid was used, with σ = 5000 km s⁻¹, consistent with the observed line width and an illuminating hard X-ray continuum with Γ = 2. The ionization parameter is not so well constrained, with $\log (\xi /\rm {erg\,cm\,s}^{-1})=4.8^{+1.0}_{-0.8}$, but suggests that either H-like or He-like iron contributes to the absorption. The column density was found to be largely degenerate with the ionization parameter (i.e., as the ionization increases the column density increases to compensate) and only a lower limit can be obtained of N_H > 1.5 × 10²³ cm⁻². The outflow velocity derived was consistent with the line analysis, with v_out = −15600 ± 2400 km s⁻¹, and is consistent with the Gibson et al. (2005) value of v_out = −12700 ± 2400 km s⁻¹. However, we also note that at this velocity, the absorption is only marginally excluded at the 90% confidence level from being associated with a local z = 0 absorber.

We also tested whether the iron K-shell region could instead be fit with a photoelectric edge, from neutral or mildly ionized iron, without any velocity shift, as was implied by the highly ionized absorption model. Indeed, fitting the Chandra data with a simple edge model results in a equally good fit statistically, with a best-fit edge energy of E = 7.15  ±  0.05 keV and an optical depth τ = 0.15  ±  0.05. Such an edge component could plausibly result from a partial covering absorber with a column density typically exceeding N_H > 10²³ cm⁻² as has been discussed; this may also be required from fitting the broader band HETG spectrum. Thus, it is not possible to distinguish here between the high velocity absorber and possible partial covering cases in MR 2251-178; higher resolution data in the Fe K bandpass, such as from the calorimeter to be flown on Astro-H, would be required to differentiate between these cases.

Thus, the detection of the Fe K-band absorption trough appears to be confirmed from the two Chandra observations, with the parameters consistent in both and at the same rest frame energy, although its exact origin remains uncertain. Gofford et al. (2011) also claimed further blueshifted absorption features at lower energies from the Suzaku data; in particular, absorption lines at E = 2.52 ± 0.02 keV and E = 2.79 ± 0.03 keV in the quasar rest frame, which were identified as blueshifted S xv and S xvi 1s → 2p, respectively. A 1.3 keV absorption trough was present in the Suzaku data near 1.3 keV and tentatively identified as blueshifted iron L-shell transitions. In the latter case, the much higher resolution HETG and RGS spectra resolve the 1.3 keV absorption into a series of lower ionization lines of inner shell Mg from Mg vi–ix, with only a modest outflow velocity of ∼ − 400 km s⁻¹. However, the absorption line at 2.52 keV appears to be only marginally detected in the 2011 Chandra spectrum at ∼99% confidence (ΔC = 9.3) at E = 2.521 ± 0.002 keV in the quasar rest frame (or 4.92 Å) with EW = −2.0 ± 1.2 eV; these parameters are entirely consistent with those measured by Suzaku. An absorption line is not detected at 2.79 keV, however, the limit of EW < 4 eV from Chandra is consistent with the Suzaku measurement of −5 ± 2 eV. Thus, the presence of this possible higher velocity component appears uncertain based on the current data and such a component does not appear to be present in the line profiles of C to Si.

The exposure times of both the HETG and RGS observations allow us to perform an unprecedented high signal-to-noise and high resolution spectroscopic study of the properties of both the primary continuum and the ionized absorption and emission features in the quasar MR 2251-178. We summarize our main observational results below.

In the soft X-ray range, numerous absorption features are clearly detected:

• 1.  
  a deep absorption trough between 0.7 and 0.8 keV most likely identified as a UTA, due to 2p → 3d transitions from low ionization M-shell iron, i.e., Fe vii–x;
• 2.  
  a multitude of inner K-shell lines of O, Ne, Mg, and Si, due to charge states corresponding to Li-, Be-, B-, C-, N-, and O-like ions, etc.
• 3.  
  several higher ionization L-shell (2p → 3d) lines of iron (i.e., Fe xvii–xxiv).
• 4.  
  resonance (1s → 2p) lines from He- and H-like ions of C, N, O, Ne, Mg, and Si and, in some cases, higher order 1s → np lines up to n = 6.

In most cases, the (strongest) absorption line profiles are narrow or not resolved, with velocity widths typically σ ≲ 300 km s⁻¹. Similarly, the outflow velocities inferred from the measured rest frame energies of the absorption lines are small or consistent with zero, of the order of v_out ≲ 400 km s⁻¹.

The spectral fit using photoionized xstar model grids shows that three fully-covering warm absorber (WA) components are required in order to model the wide range of ionization states of the gas, with N_H = 1.5–3.6 × 10²¹ cm⁻² and log (ξ/erg cm s⁻¹) = 1.27–2.80. The small outflow velocities found for the low and medium ionization components 1 and 2 are consistent with each other, with v_out = −480 ± 40 km s⁻¹ and v_out = −460 ± 60 km s⁻¹, respectively, while the highest ionization component 3 does not require an outflow velocity with v_out < 130 km s⁻¹. Notably, the necessary power-law continuum required by the xstar grid in order to model the low ionization lines (component 1) is Γ_input = 2.5, which is softer than that required for the higher ionized lines (i.e., Γ_input = 2.0). Moreover, one additional component of partially (covering factor ∼61%) ionized absorbing gas with N_H ∼ 6 × 10²² cm⁻² and log (ξ/erg cm s⁻¹) ∼ 1 is required to achieve a good fit. Interestingly, after the required absorbing layers of gas are accounted for, the soft X-ray photon index found (Γ = 2.32 ± 0.08) is in good agreement with what is required to reproduce the soft X-ray inner-shell absorption lines (i.e., Γ ∼ 2.5). Therefore, MR 2251-178 may have an intrinsically soft continuum, at least below 2 keV, which is partially covered by a complex and stratified absorber.

For the 2011 HETG spectrum, the parameters of the three fully-covering WA are rather similar to those obtained from the RGS spectra with N_H = 1.5–2.1 × 10²¹ cm⁻² and log (ξ/erg cm s⁻¹) = 1.15–2.9, but a second partial covering component (covering factor ∼40%) of higher column density ∼7 × 10²³ cm⁻² seems to be required based on the spectral curvature above 2 keV. As for the RGS spectrum, a softer input continuum is strongly required to model the low ionization warm absorber component 1. However, there is evidence for a small but significant change in its ionization parameter that appears to be correlated with the soft X-ray flux. Applying this model to the 2002 HETG spectrum, we confirm that the change of the ionization parameter is in direct proportion with the soft X-ray flux, suggesting that this component is in photoionization equilibrium with the continuum. The other possible change in the spectra is in the partial covering absorption. Considering all three grating observations, the uncovered fraction of the power law appears to increase as the flux increases from the 2002 to the 2011 observations, from 0.18 ± 0.02 to 0.39 ± 0.03, suggesting that this AGN is more obscured in lower flux states.

The soft X-ray spectra also display several emission lines from a photoionized emitter from He- and H-like ions of C, N, O, and Ne. Notably, a strong and broad emission line near 0.56 keV is clearly detected in the RGS 1 spectrum at the expected energy of the O vii triplet and is well represented by a blend of the forbidden (dominant), intercombination, and resonance emission lines with a common velocity of ∼7300 km s⁻¹ (FWHM). Three narrow absorption lines corresponding to inner-shell absorption due to O v and the two lines that make up the O vi (1s²2s → 1s2s2p) doublet are superimposed on the broad O vii triplet profile. Similar structures are present at other energies, with N vii, O viii, and Ne ix all showing absorption superimposed on the emission.

In the hard X-ray energy band of the HETG spectrum, there is a lack of any strong iron Kα emission, with EW = 11 ± 6 eV. This could be accounted for by the X-ray Baldwin effect, since MR2251-178 has a much higher 2–10 keV luminosity than most local Seyfert 1s. However, we found the presence of a significant absorption feature at 7.3 keV consistent with what was previously reported from the 2002 HETG observation by Gibson et al. (2005), but only marginally inconsistent at the 90% confidence level with the line energy measured in the 2009 Suzaku observation by Gofford et al. (2011). This Fe K-band absorption is well modeled by a highly ionized xstar grid with a high turbulence velocity of 5000 km s⁻¹ and an outflow velocity of ∼−15,600 km s⁻¹. However, an alternative origin from a low ionization partial covering absorber, without requiring any velocity shift cannot be excluded. The much higher spectral resolution of both the HETG and RGS data allows us to resolve the 1.3 keV absorption feature—first observed in the lower resolution 2009 Suzaku X-ray Imaging Spectrometer (XIS) spectrum (Gofford et al. 2011) and tentatively identified as blueshifted iron L-shell transitions—into a series of lower ionization lines of inner shell Mg from Mg vi–ix, with only a modest outflow velocity of ∼ − 400 km s⁻¹.

6.2.1. Constraints from the O vii Line Triplet

6.2.2. Constraints from the Variability of the Warm Absorber Components

The component 1 warm absorber appears to respond to the overall increase in the continuum between the 2002 and 2011 observations, but also in the ∼40 day timescale between the 2011 XMM-Newton and Chandra observations (Section 4.2); it therefore appears to be in photoionization equilibrium. We can accordingly attempt to place a lower limit on the density of this absorber via the recombination timescale. For this, we use the recombination timescale formula from Bottorff et al. (2000) that accounts simultaneously for the cascade into the population of X_i ions from the population of X_{i + 1} ions and the cascade out of the population of X_i ions into the population of X_{i − 1} ions:

$$\begin{equation} t(X_{i})=\frac{1}{\alpha (X_{i})n_{e} \left[\frac{f(X_{i+1})}{f(X_{i})}- \frac{\alpha (X_{i-1})}{\alpha (X_{i})} \right]}, \end{equation} \tag{ 2 }$$

where f(X_i) is the ionic fraction of the X_i ion, α(X_i, T_e) is the recombination coefficient of the X_i ion at the electronic temperature T_e, and n_e is the electron density (∼1.2 n_H for cosmic abundance). We apply this formula to O vii. At log ξ = 1.27, the ratio O vii/O viii is 6.0 and T_e is 4 × 10⁴ K. Using the recombination coefficient from Nahar & Pradhan (2003) and a recombination time of t ≲ 40 days between observations, we find a lower limit for the hydrogen density of 3.8 × 10⁴ cm⁻³. Hence, this implies an upper limit for the radial distance (R_var) of 5.3 × 10¹⁹ cm (i.e., ≲17 pc) or a few parsecs. Moreover, as discussed below in Section 6.2.3, the minimum radius for component 1 is about 2.8 × 10¹⁹ cm (i.e., ≳ 9 pc). Therefore, the location of component 1 is well constrained between 9 pc and 17 pc. For comparison, the expected distances of the torus and the NLR are about 7 pc and 140 pc, respectively, using the following formulae of Krolik & Kriss (2001) and Mor et al. (2009):

$$\begin{equation} R_{\rm torus}\sim L_{\rm ion,44}^{1/2} ({\rm pc}) \end{equation} \tag{ 3 }$$

$$\begin{equation} R_{\rm NLR} = 295 \times L_{46}^{0.47\pm 0.13} ({\rm pc}). \end{equation} \tag{ 4 }$$

For MR 2251-178, the ionizing (1–1000 Rydberg) luminosity was taken as L_{ion, 44} = 20 (in units of 10⁴⁴ erg s⁻¹) and L₄₆ = 0.43¹⁰ is assumed as the bolometric luminosity (Dunn et al. 2008), in units of 10⁴⁶ erg s⁻¹.

Therefore, component 1 appears to be located at distances consistent with the parsec scale torus and/or the scale of the inner NLR radius. We note that R_var is much greater than the BLR distance that is only about 75 light-days (using the recent R_BLR–λL_λ(5100 Å) relationship from Bentz et al. 2013 and the average 5100 Å  flux from Lira et al. 2011), i.e., 0.06 pc. The lack of response from 2002–2011 of the higher ionization (components 2 and 3) absorbers may place this gas at greater distances. However, a more intense monitoring campaign (over weeks to months) would be needed to place a firmer constraint on the density and therefore the radial location of the absorbers.

6.2.3. Warm Absorber Properties: Radii, Outflows Rates, and Energetics

We estimate the lower and upper limits of the distance, mass outflow rate, and kinetic power of the WAs following the assumptions and definitions outlined in Tombesi et al. (2013) for the fully covering warm absorbers (components 1, 2, and 3) and for the highly ionized absorber discussed in Section 4.4. An upper limit on the radial location of an absorber can be derived from the definition of the ionization parameter and the requirement that the thickness of the absorber does not exceed its distance to the supermassive black hole, i.e., N_H ≃ n_HΔR < n_HR:

$$\begin{equation} r_{\mathrm{max}} \equiv L_{\mathrm{ion}}/\xi N_\mathrm{H}. \end{equation} \tag{ 5 }$$

Note that the material cannot be farther away than this given the observed ionization and column density. An estimate of the minimum distance can be derived from the radius at which the observed velocity corresponds to the escape velocity:

$$\begin{equation} r_{\mathrm{min}} \equiv 2 G M_{\mathrm{BH}}/ v_{\mathrm{out}}^{2}. \end{equation} \tag{ 6 }$$

Here, the black hole mass estimate for MR 2251-178 is taken as 2.4 × 10⁸M_☉ (Dunn et al. 2008).

For the calculation of the mass outflow rate, we use the expression derived by Krongold et al. (2007) that is appropriate for a biconical wind-like geometry and that does not rely on the estimate of the covering and filling factors (see Tombesi et al. 2013 for details):

$$\begin{equation} \dot{M}_{\mathrm{out}} \equiv f(\delta, \phi) \pi \frac{n_\mathrm{H}}{n_\mathrm{e}} m_\mathrm{p} N_\mathrm{H} v_\mathrm{out} r, \end{equation} \tag{ 7 }$$

where f(δ, ϕ) is a function that depends on the angle between the line of sight to the central source and the accretion disk plane, δ, and the angle formed by the wind with the accretion disk, ϕ (see Figure 12 of Krongold et al. 2007). As in Krongold et al. (2007) and Tombesi et al. (2013), we assume f(δ, ϕ) ≃ 1.5, corresponding roughly to a vertical disk wind (ϕ ≃ π/2) and an average line of sight angle of δ ≃ 30° for a type I AGN, while n_H/n_e is about 1/1.2 for solar elemental abundances, so:

$$\begin{equation} \dot{M}_{\mathrm{out}} \simeq 6.6\times 10^{-24} N_\mathrm{H} v_\mathrm{out} r [{\rm g\ s}^{-1}]. \end{equation} \tag{ 8 }$$

To determine the $\dot{M}_{\mathrm{out}}$ interval range, we use the values of r_max and r_min inferred from Equations (5) and (6), except for component 1, for which with use the value found above due to the recombination timescale of O vii, i.e., r_var (see values reported in Table 7) for r_max.

Table 7. Properties of the Fully Covered Warm Absorber Components ("Components 1, 2, and 3") and the Highly Ionized Component ("Component High") Discussed in Section 4.4

Parameters	Component 1	Component 2	Component 3	Component High
NH (×1021 cm−2)	2.12 ± 0.07	1.50 ± 0.20	3.6 ± 1.3	>150
log (ξ/erg cm s−1)	1.27 ± 0.02	$2.04^{+0.04}_{-0.07}$	$2.80^{+0.05}_{-0.07}$	$4.8^{+1.0}_{-0.8}$
voutc (km s−1)	−480 ± 40	−470 ± 60	<130	−15600 ± 2400
rmin (cm)/(pc)	2.8 × 1019/9.0	2.9 × 1019/9.4	3.8 × 1020/122	2.6 × 1016/0.008
rmaxa (cm)/(pc)	5.3 × 1019/17.2	1.2 × 1022/3940	8.8 × 1020/285	2.1 × 1017/0.068
$\dot{M}_{\mathrm{out}}$ (×1025 g s−1)	[1.9–3.6]	[1.4–560]	[11.8–27.3]	>4.0
$\dot{M}_{\mathrm{out}}$ ($\dot{M}$ yr−1)	[0.3–0.6]	[0.2–89]	[1.9–4.3]	>0.6
$\dot{E}_\mathrm{K}$/Lbolb (%)	[5.1 × 10−4–9.8 × 10−4]	[3.5 × 10−4–0.14]	[2.4 × 10−4–5.4 × 10−4]	>1.1
$\dot{P}_\mathrm{out}$/$\dot{P}_\mathrm{rad}$ (%)	[0.63–1.2]	[0.45–184]	[1.1–2.5]	>46

Notes. The values of N_H, log ξ, and v_out are those inferred from the 2011 RGS observations (see Table 5) for components 1, 2, and 3, and from the HETG observations for the highly ionized absorber. See the text for full definitions of the parameters. ^ar_max is inferred from Equation (5), except for component 1, for which r_max corresponds to r_var inferred from the recombination timescale; see Section 6.2.2. ^bL_bol = 4.3 × 10⁴⁵ erg s⁻¹ (Dunn et al. 2008) and L_Edd = 3.0 × 10⁴⁶ erg s⁻¹. ^cVelocity shift of the absorber in km s⁻¹. Negative values denote a blueshift. Upper limits on the outflow velocity are expressed as absolute values for clarity.

Download table as:  ASCIITypeset image

Neglecting additional acceleration of the outflow, i.e., assuming that it has reached a constant terminal velocity, the kinetic (or mechanical) power can consequently be derived as

$$\begin{equation} \dot{E}_\mathrm{K} \equiv \frac{1}{2} \dot{M}_{\mathrm{out}} v_\mathrm{out}^2. \end{equation} \tag{ 9 }$$

We also calculated the outflow momentum rate as $\dot{P}_\mathrm{out} \equiv \dot{M}_\mathrm{out} v_\mathrm{out}$, and subsequently compared it with the momentum flux of the radiation field, $\dot{P}_\mathrm{rad} \equiv L_\mathrm{bol}/c$. All values are reported in Table 7.

The inner and outer radii of component 1 are the best determined, between 9 and 17 pc, with the upper bound being set by the 40 day timescale response of the absorber to the continuum. The higher ionization component 3 is constrained between ∼120 and 290 pc; this is consistent with being placed outside component 1, noting that no response of this absorber was detected to continuum variations, consistent with a lower density. Component 2 is the least well determined and is consistent with the radial estimates for components 1 and 3 (Table 7). Thus, the locations of components 1 and 3 are consistent with the torus and the NLR, respectively, as estimated above. The possible highly ionized (iron K band) absorber (Table 7, component high), with an outflow velocity of ∼15, 000 km s⁻¹, would appear to be located much closer to the black hole (≲ 0.01 pc) with a location perhaps consistent with an accretion disk wind (Tombesi et al. 2013).

The kinetic power of the three warm absorbers (components 1–3) appears to be ≲ 0.01% of the bolometric luminosity, while for the highly ionized absorber, we found a minimum value of 1% of the bolometric luminosity, hence its mechanical power can potentially affect the host galaxy via feedback (Hopkins & Elvis 2010). Nonetheless, the mass outflow rates of components 1–3, as well as the highly ionized absorber, are rather similar, the lower limits on $\dot{M}_{\rm out}$ vary between 0.2 and 1.9 M_☉ yr⁻¹ for components 1–3, while the highly ionized absorber shows $\dot{M}_{\rm out}\gtrsim 0.6 \ M_{\odot }$ yr⁻¹. In comparison, for a bolometric luminosity of 4.3 × 10⁴⁵ erg s⁻¹ and assuming an accretion efficiency of η = 0.06, the expected mass accretion rate of MR 2251-178 is $\dot{M}_{\rm acc}\sim 1.3 \ M_{\odot }$ yr⁻¹; thus, the combined mass outflow rate from MR 2251-178 is likely to be at least equal to (or somewhat above) the accretion rate onto the black hole. Finally, the outward momentum rate of the putative highly ionized absorber is estimated to be at least ∼50% of L_bol/c, which suggests efficient (τ ∼ 1) scattering between photons and electrons in a Thomson scattering driven outflow, as may be expected in a highly ionized accretion disk wind (King & Pounds 2003).

UV absorption has also been found previously in the spectrum of MR 2251-178. Using Hubble Space Telescope/Faint Object Spectrograph (HST/FOS) data obtained in 1996, Monier et al. (2001) found absorption lines due to Lyα, N v, and C iv with a systematic blueshift of ∼300 km s⁻¹ with a total hydrogen column density of about 5 × 10²¹ cm⁻². From the comparison between HST data taken with FOS in 1996 and Space Telescope Imaging Spectrograph data obtained in 2000, the C iv absorption in particular showed variability—both in terms of its velocity and column density—over a period of roughly four years. This relatively short timescale variability showed that this UV absorption is truly intrinsic and constrained the absorption clouds to within r ≲ 2.4 kpc of the continuum source (Ganguly et al. 2001), consistent with the estimate of Monier et al. (2001). Kaspi et al. (2004) reported for the first time the entire FUSE spectrum of MR 2251-178 and detected at least four blueshifted absorption systems of C iii, H i, and O vi; one at −580 km s⁻¹ and at least three other blended components with centroid velocities at about −150, −300, and −430 km s⁻¹.

We note that the velocity profiles obtained here from the X-ray data, e.g., from C vi and O viii, appear to be consistent with these UV profiles, with the X-ray absorption line profiles having typical velocity shifts of the order ∼ − 300 to 400 km s⁻¹, as shown in Figure 7. The only exception may be the highest ionization lines, such as Mg xii and Si xiv, which do not require any net blueshift, but this very highly ionized gas may be more apparent in the X-ray spectrum than in the UV. The total depth of the far-ultraviolet (FUV) and UV absorption lines appeared larger than the underlying continuum, which indicates that the broad UV emission lines are absorbed by the UV absorber and therefore the UV absorber lies outside the BLR. This is similar to what is found in the X-ray spectrum presented here, whereby the broad soft X-ray lines (BLR) are absorbed by the narrow lines from the X-ray warm absorber.

The UV absorber properties discussed above are similar to those found here for the fully-covered soft X-ray warm absorber components (namely components 1, 2, and 3); indeed, both the column densities and the outflow velocities appear consistent. Moreover, the relatively tight constraint for the location of component 1 shows that it lies outside the BLR region too, with the components 2 and 3 consistent with being further out due to their lack of variability. In conclusion, the UV and soft X-ray warm absorption components cover a similar range of column densities and appear to be kinematically consistent with each other in terms of their outflow velocities, although the X-ray absorption likely originates from more highly ionized gas. A more detailed comparison with the 2011 Chandra and XMM-Newton observations and a contemporaneous 2011 HST/Cosmic Origins Spectrograph (COS) observation will be deferred until future work. We note, however, that from a preliminary analysis of the HST/COS spectrum, as well as optical spectroscopy (M. Crenshaw 2013, private communication), the FWHM widths of the C iv and Hβ emission lines appear in the range 3200–3600 km s⁻¹. This is similar to, if somewhat smaller than, the widths of the X-ray emission lines such as O vii, with a FWHM ∼7300 km s⁻¹. This could suggest that the broad X-ray lines originate from the innermost part of the BLR, which would likely be more highly ionized.

The Seyfert 1 galaxy Mrk 509 (z = 0.03450)—which has a similar black hole mass (1–3 × 10⁸ M_☉) and a bolometric luminosity of only a factor of about two to three smaller than that of MR 2251-178 (Raimundo et al. 2012)—has also been recently monitored (in 2009) from UV to hard X-rays (HST/COS, XMM-Newton, Chandra, Swift, and INTEGRAL) to constrain the location of the outflow components (Kaastra et al. 2011b). Thus, a comparison between MR 2251-178 and Mrk 509 may be informative, given their similar properties at the higher luminosity end of the Seyfert population, while both AGNs have long XMM-Newton or Chandra exposures. The deep (600 ks) XMM-Newton/RGS spectrum of Mrk 509 revealed the presence of a multitude of blueshifted absorption lines from three slow velocity absorber components (∼ − 13 km s⁻¹, ∼ − 320 km s⁻¹, and −770 km s⁻¹), with two strong and discrete ionization parameter peaks in the log (ξ/erg cm s⁻¹) = 1–3 range at about log (ξ/erg cm s⁻¹) = 2.0 and log (ξ/erg cm s⁻¹) = 2.8 (Detmers et al. 2011). The ionization parameters of the UV components with similar outflow velocities are much lower than those found in X-rays, which could indicate that the UV and X-ray absorbers are cospatial but have different densities, as also inferred from the LETG spectrum (Ebrero et al. 2011). The presence of a possible fast outflow with v_out ≃ −14, 000 km s⁻¹ was claimed using the summed spectrum of previous XMM-Newton observations (Ponti et al. 2009) and was only marginally detected in the LETG 2009 observation and the XMM-Newton/pn spectrum (Ponti et al. 2013).

The location of the outflowing components in Mrk 509 are claimed to be consistent with a torus wind or NLR origin (Kaastra et al. 2012). While the kinetic luminosity of the outflow is small in Mrk 509, the mass carried away is larger than the likely 0.5 M_☉ yr⁻¹ accreting onto the black hole. These properties appear to be similar to those presented here for MR 2251-178. Observationally, the X-ray column densities, outflow velocities, and ionization parameters cover a very similar range, while in terms of radial location, the warm absorbers of both AGNs appear commensurate with a parsec scale wind, consistent with the outermost torus or inner NLR.