EDP Sciences
Free Access
Issue
A&A
Volume 606, October 2017
Article Number A96
Number of page(s) 15
Section Extragalactic astronomy
DOI https://doi.org/10.1051/0004-6361/201730819
Published online 20 October 2017

© ESO, 2017

1. Introduction

thumbnail Fig. 1

Rest-frame optical spectra of two AGNs from our X-ray/SDSS sample. For each target, in the upper left corner of the large panel we show the MJD, the plate and the fibre numbers which uniquely identify the SDSS spectrum. The dashed vertical lines mark the location of the most prominent emission lines, besides some features crucial for the analysis presented in this paper and/or in Paper I. Fit results obtained in Paper I, from the multicomponent simultaneous fit in the Hβ-[O III] and Hα-[NII] regions are also shown: best-fit narrow component (NC) and broad line region component (BC) profiles are highlighted with green curves; outflow component (OC) and Fe II emission are shown with blue and magenta curves, respectively. Finally, black Gaussian profile in the low panel shows a second set of BC used to fit the complex BLR profiles (see Paper I for further details). The insets on the right of each panel show an enlargement of the sulphur lines. The spectra display faintness and blending problems affecting density-sensitive [S II]λλ6716, 6731 and temperature-sensitive [O III]λ4363 emission lines. In particular, the object in the first panel displays a well detected and modelled sulphur doublet, from which we can easily derive the density-sensitive flux ratio, but faint oxygen λ4363 line; the spectrum in the lower panel shows blended sulphur doublet features and oxygen line at λ4363. For both the AGNs it is therefore not possible to derive temperature-sensitive flux ratios.

Open with DEXTER

According to the most popular active galactic nuclei (AGN)-galaxy evolutionary models (e.g. Hopkins et al. 2008; Menci et al.2008; Sijacki et al. 2015), the growth of a super-massive black hole (SMBH) has a significant impact on host galaxy evolution. The accompanying released accretion energy, coupling with the interstellar medium (ISM), has been postulated to regulate both the star formation processes in the host and the accretion onto the SMBH (e.g. King & Pounds 2015; Zubovas & King 2014).

The presence of AGN-driven outflows is nowadays quite well established through high resolution observations of local and high-redshift galaxies, and it is now possible to study in detail the feedback phenomena, characterising the galaxy-wide extension and the morphology of the ejected material as well as the masses and the energetics related to outflows (e.g. Bischetti et al. 2017; Brusa et al. 2016; Cresci et al. 2015; Feruglio et al. 2015; Harrison et al. 2012, 2014; Husemann et al. 2016; Perna et al. 2015a,b; Rupke & Veilleux 2013; see also Fiore et al. 2017, for an updated and complete list). However, the physical processes responsible for the coupling between AGN winds and the ISM remain largely unknown. Furthermore, outflow effects on host galaxy evolution, as derived from outflow energetics (mass outflow rate, kinetic and momentum powers), remain mostly unknown.

In Perna et al. (2017a; Paper I hereinafter) we presented a sample of 563 X-ray selected AGN at z< 0.8, for which SDSS spectra are available. We combined ionised emission line and neutral absorption feature information as modelled through multicomponent simultaneous fitting (Brusa et al. 2015), non-parametric measurement (Zakamska & Greene 2014), and a penalised pixel fitting procedure (pPXF; Cappellari & Emsellem 2004; Cappellari 2017) analysis, to derive kinematic properties of both warm and cold gas components of the ISM. The modelling of optical spectra allowed us to derive the incidence of ionised (~ 40%) and atomic (< 1%) outflows covering a wide range of AGN bolometric luminosity, from 1042 to 1046 erg/s and to relate the presence of ionised outflows with different AGN power tracers. We also derived X-ray and bolometric luminosities and we discuss our results in the context of an evolutionary sequence allowing for two distinct stages of the feedback phase: an initial stage characterised by X-ray and optical obscured AGNs in which the atomic gas is still present in the ISM and the outflow processes involve all the gas components, and a later stage associated with unobscured AGNs, which line of sight has been cleaned and the cold components have been heated or exhausted.

In this second paper we focus on the physical conditions of ionised gas, studying in details stacked spectra and small and medium-size sub-samples of X-ray/SDSS sources characterised by the presence of well detected optical diagnostic lines. In particular, we focus on the estimate of the electron temperature (Te) and density (Ne) of the unperturbed and outflowing ionised gas in the narrow line region (NLR).

Plasma properties of NLR unperturbed gas are now well constrained to average Ne of the order of 250–400 cm-3 and Te of ~ 1.5 × 104 K (e.g. Vaona et al. 2012; Zhang et al. 2013). The physical conditions within the outflowing regions are instead mostly unknown, because of the faintness of the outflow wings of the emission lines involved in the diagnostics used to derive such information (see e.g. De Robertis & Osterbrock 1986; Rice et al. 2006; Vaona et al. 2012). Knowledge of these properties is crucial for the investigation of the mechanisms responsible of outflows: it can significantly reduce the uncertainties in the outflow energetics (up to a factor of ten), improving our understanding of the AGN outflow phenomenon and its impact on galaxy evolution.

The paper is organised as follows. We first give a brief description of the spectroscopic modelling results obtained in Paper I (Sect. 2). In Sect. 3 we review the current knowledge of the NLR plasma conditions and its role in deriving the outflow energetics. In Sects. 3.2 and 3.1 we derive temperature- and density-sensitive flux ratios from single targets and stacked spectra. Section 3.3 displays the median plasma properties. In Sect. 4 we investigate several diagnostics to infer information about outflowing gas ionisation mechanisms. Finally, we summarise our results and their implications in the Discussion section (Sect. 5).

2. Previous work – optical spectroscopic analysis

In Paper I we presented modelling results for a sample of 563 AGNs. Thanks to multicomponent simultaneous fit technique, for each emission line in the wavelength range between He Iλ4687 and [S II]λλ6716, 6731 doublet, we were able to separate NLR gas in virial motion, modelled with narrow Gaussian components [NC, which full width at half maximum (FWHM) has been constrained to be 550 km s-1], from perturbed outflowing material, modelled with outflow components (OC; FWHM> 550 km s-1; see, e.g. Figs. 1 and A.1; see also Paper I, Fig. 3). The outflow component detection has been tested considering chi-square minimisation and signal-to-noise (S/N, see Paper I). In this paper we take advantage of SDSS spectra modelling results to construct stacked spectra and select well defined sub-samples of sources with the intention of studying NLR plasma properties in presence of AGN-driven outflows.

3. The plasma properties in the outflowing gas

Electron density and electron temperature within AGN-driven outflow regions are largely unknown. In the current literature, these quantities are important sources of uncertainty in outflow kinematic estimates for the ionised phase. Carniani et al. (2015) showed how Ne and Te enter in the determination of the ejected mass Mout and, in consequence, of the mass outflow rate out, kinetic and momentum powers ( and , respectively; Vout is the outflow velocity). Here we report their Eq. (5) which they derived for the [O III]λ5007 line (but the same considerations apply when Balmer emissions are used instead of [O III]; see, e.g. Cresci et al. 2015; Liu et al. 2013): (1)where mp is the proton mass, is the “condensation factor”, L[OIII] is the [O III]λ5007 luminosity, 10[O/H] − [O/H] is the metallicity, j[OIII] the oxygen emissivity. The emissivity term shows a weak dependence on electron density over several orders of magnitude, but also a relevant dependence on the electron temperature: a difference of a factor of three easily emerges when we consider Te of 1 × instead of 2 × 104 K1. Moreover, the outflow mass shows an inverse proportionality to the electron density.

Outflow energetics have usually been derived in the past assuming given values for electron temperature and density. While a general consensus is found for a Te = 1 × 104 K (e.g. Harrison et al. 2014; Carniani et al. 2015; Bischetti et al. 2017; Nesvadba et al. 2006; but see Liu et al. 2013; Cresci et al. 2015), several values are used for the electron density, spanning one order of magnitude or more: for example, 1000 cm-3 has been assumed by Cano-Díaz et al. (2012), 500 cm-3 by Carniani et al. (2015), and 100 cm-3 by a large number of other authors (e.g. Brusa et al. 2015; Cresci et al. 2015; Harrison et al. 2014; Kakkad et al. 2016; Liu et al. 2013).

A few diagnostic ratios involving forbidden lines can be used to derive these properties in regions with densities 106 cm-3 (depending on the critical density of the involved forbidden transitions). In particular, [S II]λλ6716, 6731 and [O III] flux ratio (involving [OIII]λλ4959, 5007 and [OIII]λ4363) diagnostics, are potentially useful to measure Ne and Te because of their optical wavelengths2, through the equations: where R[OIII] = [f(λ5007) + f(λ4959)] /f(λ4363) and R[SII] = f(λ6716) /f(λ6731) (Osterbrock & Ferland 2006).

Unfortunately, the faintness of the involved emission lines (in particular, [O III]λ4363 and [S II] doublet) and the interdependence between Te and Ne (see Eq. (3)) make the measurement of these quantities challenging. As an example, we show in Fig. 1 (top panel) the spectrum of a source for which it is possible to model the sulphur lines well, separating OC and NC, but with faint [O III]λ4363. In this case, therefore, it is possible to derive R[SII] but not R[OIII]. The bottom panel of Fig. 1 shows instead how the vicinity of sulphur and oxygen lines with broad line region (BLR) Balmer emission Hα and Hγ in type 1 AGNs also complicates the situation. These are the reasons for which, for example, Vaona et al. (2012) derived reliable estimates of the electron temperature of the unperturbed NLR gas only for ~500 objects, starting from a parent sample of ~2500 SDSS AGNs.

The fact that the OC can be fainter than the unperturbed narrow components (see Paper I, Fig. C.1), makes it even more difficult to derive such diagnostic information for the outflowing ionised material. Only for a handful of previous studies was it possible to derive such physical properties, although with large uncertainties. These works are generally based on single luminous targets (e.g. Brusa et al. 2016; Perna et al. 2015a) or, in the best cases, on a small number of sources (e.g. Westmoquette et al. 2012). Harrison et al. (2012) used a stacked spectrum of z ~ 2.4 ultra-luminous galaxies to estimate the electron density. Genzel et al. (2014), assuming a pressure equilibrium between outflowing gas and the ionised gas in star-forming regions, proposed an electron density of ~80 cm-3 for the outflowing gas, as derived from star-forming ionised gas in the discs and centres of star-forming galaxies at z ~ 2 (see also Kaasinen et al. 2017). These few results point to large ranges of values for the electron densities, from 102 to >103 cm-3 (see, e.g. Rodríguez Zaurín et al. 2013). To the best of our knowledge, the electron temperature has been derived for only six targets (those presented by Villar-Martín et al. 2014; Nesvadba et al. 2008; and Brusa et al. 2016), with Te ≈ 1.5 × 104 K.

thumbnail Fig. 2

Median stacked spectrum around the doubly ionised oxygen emission lines obtained combining the spectra of faint/obscured sources with evidence of ionised outflows (see the text for more details). Dashed curves highlight 1σ uncertainties. Red curve represents the best-fit result we obtained fitting simultaneously the emission lines displayed. Green Gaussian represent NC and BC components, blue profiles mark OC.

Open with DEXTER

thumbnail Fig. 3

Magnification of the region of [O III]λ4363-Hγλ4342 lines for the 44 sources for which we could determine R[OIII] flux ratios. For each spectrum, the best-fitting components presented in Sect. 3.1 are superimposed: solid green curves represent the systemic narrow component (NC); blue curves the OC.

Open with DEXTER

3.1. Electron temperature diagnostics

To constrain at best the electron temperature in ionised gas in our X-ray/SDSS selected AGN sample, we use two different approaches. With the first approach, we constructed a stacked spectrum to derive temperature-sensitive diagnostic ratio ROIII for both unperturbed and outflow emitting gas. Then, we considered well selected individual objects and derived the same quantity to verify the reliability of the results found with the former approach.

Our X-ray/SDSS sample shows a large variety of AGN types, from typical type 1 AGNs with blue spectra and prominent BLR lines to reddened and/or faint AGNs (see Paper I, Fig. 3). Because of the not-so-large number of X-ray/SDSS sources, instead of stacking galaxies in bins of AGN luminosity and/or obscuration, we chose to stack continuum-subtracted spectra. We considered only those sources for which we were able to model the continuum over the entire wavelength range covering the separation between the doubly ionised oxygen lines using pPXF best-fit technique (Paper I). From these, we constructed the stacked spectrum combining the continuum-subtracted spectra of all those sources with evidences of outflow in the [O III] line (75 AGNs). These sources are associated with faint and/or obscured AGNs (1040<L[OIII]< 1042.5, with a median luminosity of 1041.2 erg/s); this allowed us to reduce the possible Hγλ4342 BLR emission in the vicinity of [O III]λ4363.

Figure 2 shows the median stacked spectrum and the best-fit results we obtain fitting simultaneously all the prominent features in the displayed wavelength range and following the strategy presented in Paper I. The presence of outflow components with FWHM ≈ 1000 km s-1 is noticeable in all displayed emission lines. From simultaneous multicomponent fit, we derived the flux ratios and (the errors are computed employing a bootstrap method; Peterson et al. 2004).

The stacked spectrum allowed us also to check for the possible presence of [Fe II] contamination of [O III]λ4363. In fact, Curti et al. (2017) found that faint iron lines at 4288 Å and 4360 Å concomitantly emerge in high-metallicity stack galaxy spectra, resulting in an overestimation of the oxygen line flux. The absence of [Fe II]λ4288 permitted us to reasonably exclude the possible contribution of iron contamination of oxygen 4363 Å line.

With the second approach, we carefully select a sample of sources with well detected and unblended [O III]λ4363 line from the total X-ray/SDSS sample. To allow the analysis of the outflow wings in such faint emission line we narrowed down the sample selecting only those targets with well detected (S/N> 10) [O III]λ4363. Furthermore, to mitigate blending problems, we discarded all those sources with broad Hγ BLR profiles after a visual inspection. This reduced the sample to eight sources. For each target, we fitted simultaneously the oxygen line at 4363 Å and the Balmer line at 4342 Å imposing the same systemics, widths and sets of Gaussian components as obtained from the simultaneous fit in the Hα and Hβ regions (Paper I). From such analysis, we derived ROIII flux ratios for both the NC and OC components.

For sources without evidence of outflows in [O III]λ5007, the spectral analysis does not require any decomposion between NC and OC emission; we could therefore relax the requirement on the S/N, imposing a S/N> 5, and derive temperature-sensitive flux ratios also for this sub-sample. In this way we obtained R[OIII](NC) ratios for an additional ten AGNs. The spectra around the region of [O III]λ4363 and the fit results for these eight plus ten sources are reported in Fig. 3. The derived ROIII distributions for NC (18 AGNs) and OC (8 AGNs) are shown in Fig. 4 (left panel).

We note that the ROIII(OC) distribution is located closely around the median position of that of ROIII(NC) (i.e. ROIII ~ 90). This could suggest that, on average, NC and OC share similar electron temperatures (Eq. (2)). These samples are however very small to point to any conclusion. We therefore tested this thesis using additional 26 targets with [O III]λ4363 detected with 5 <S/N< 10 and showing evidence of ionised outflows. If we assume the same temperature for both outflowing and systemic ionised gas, the amplitude fractions OC/NC should be the same in [O III]λ5007 and [O III]λ4363 (see Eq. (2)). We fitted the emission lines with this additional constrain. The fit results are shown in the last part of Fig. 3. All the temperature-sensitive flux ratios are reported in Table A.2.

We note that the profiles are generally well reproduced under the assumption that Te(NC) = Te(OC). Of course, the low quality of the spectra does not allow a strong result significance. In Fig. 4, left, we show the distribution obtained adding these 26 targets to the outflow sample. In the figure, we also show the results from the stack analysis; the overlap between 68% confidence intervals for NC and OC flux ratios also tends to support our assumption in deriving the distribution of flux ratio measurements from single targets. From the final distribution, we derive the median value R[OIII](NC) ⟩ = ⟨ R[OIII](OC) ⟩ = 55 ± 28, with the uncertainty defined by the 68% confidence interval.

All median flux ratios obtained so far are collected in Table 1.

thumbnail Fig. 4

left: [O III] ratio distributions. Grey shaded areas mark the R[OIII] distribution for the unperturbed ionised gas (i.e. NC). Blue shaded area denotes the outflow emission R[OIII] distribution of S/N> 10 sources. The blue dashed line shows the R[OIII] histogram of the outflow components of all S/N> 5 sources (see the text for detailed analysis description). In each panel, median and 68% confidence intervals obtained from the analysis of stacked spectra are also shown for comparison (black and blue symbols, representing line ratios associated with NC and OC, respectively). Right: [S II]λλ6716, 6731 ratio distributions. The grey solid line marks the distribution for the AGN sample without evidence of outflows from our line fitting routine. The black and blue shaded areas denote the distributions for NC and OC sulphur ratios obtained from the sub-sample of 28 AGNs.

Open with DEXTER

Table 1

Plasma properties.

3.2. Electron density diagnostics

As in the previous section, we used two different approaches to derive the electron density of AGN-ionised gas. We first constructed a stacked spectrum combining the spectra of all those sources with evidence of ionised [O III]λ5007 outflows and without blending problems between sulphur and Hα BLR emission. For each spectrum satisfying such conditions (90 AGNs), we subtracted the continuum emission and normalised the fluxes to the Hα peak. Figure 5 shows the median stacked spectrum obtained combining the normalised, continuum-subtracted spectra. In order to best model the sulphur profile and distinguish between systemic and perturbed emitting gas, we fitted simultaneously Hα, [N II] and [S II] lines (see Paper I for further details). We clearly reveal and separate OC, with FWHM ≈ 800 km s-1, from a narrower unperturbed component, with FWHM of 350 km s-1, in all emission lines. From the best-fit results (shown in Fig. 5), we derive and (as before, the errors were computed employing a bootstrap method).

The second approach employed the analysis of optical spectra of single objects. As a first step, we measured the NLR diagnostic ratio R[SII] for all AGNs with no severe and ambiguous blending with Hα BLR emission and without signatures of outflows revealed in simultaneous fits, resulting in a sample of 121 AGNs. In fact, when OC components are revealed, the doublet lines are usually severely blended and, in general, the fitting procedure gives ambiguous results (see, e.g. Perna et al. 2015a; Rodríguez Zaurín et al. 2013; Villar-Martín et al. 2014). From this sample we obtained the distribution shown in Fig. 4, right (grey histogram).

To study the electron density of the outflowing regions, we focussed the analysis on those AGNs with the simplest spectral profiles, meaning those with well defined [S II] wings, modelled with two kinematic components (NC + OC). and, as before, without strong blending with BLR emission. In Paper I we noticed that approximately 40% of X-ray/SDSS AGNs of our parent sample display signatures of ionised outflows. The above mentioned conditions are, however, satisfied by only 28 sources. This reflects the difficulties in the electron density measurements. The fitted spectra are shown in Fig. A.1. From this sample, we computed R[SII] ratios for both NC and OC. The R[SII](NC) distribution (Fig. 4, right) has a smaller spread when compared with that of the AGN sample without OC components, because of the particular selection. Instead, the R[SII](OC) distribution (blue area) covers a larger range of values and is peaked at lower ratios. From these distributions, we derived the median values and (the errors define 68% confidence intervals; see Table A.1 for the compilation of all flux ratio measurements).

From each [S II] intensity ratio, in principle, we could derive an estimate of the electron density from Eq. (3), taking into account the dependence on the electron temperature. This means that for each source, in order to derive at best Ne, we should be able to detect and analyse all the emission lines needed for electron density and temperature diagnostics. Unfortunately, the faintness of the temperature-sensitive emission line [O III]λ4363 does not allow spectral analysis for all but seven of our targets selected to study the [S II] emission (see, e.g. the spectrum in Fig. 1, panel a). Therefore, we chose to follow a statistical approach, and derive median electron densities of outflowing and systemic gas for given median electron temperatures.

thumbnail Fig. 5

Median stacked spectrum around the [S II] emission line doublet obtained combining the spectra of sources with evidence of ionised outflows and without blending problems with Hα BLR emission. Dashed curves highlight 1σ uncertainties. The red curve represents the best-fit result we obtained fitting simultaneously Hα, [N II] and [S II] doublets. The green Gaussians represent NC and BC components, blue profiles mark OC.

Open with DEXTER

3.3. Results: the plasma properties

We compute fiducial estimates of the electron temperatures of both NC and OC through Eq. (2), using the median doubly ionised oxygen flux ratios derived from stack analysis, Te(NC) = (1.3 ± 0.2) × 104 K and K. We also use the median value of the final R[OIII] distribution (Fig. 4, left, blue dashed histogram), to derive an additional estimate of the electron temperature in the outflow region, K. Because of the similarities in the distributions of NC and OC, and the fact that this temperature value is also consistent within ~ 1σ with the stack analysis measurements, we consider Te = 1.7 × 104 K as a median electron temperature for the entire NLR emitting gas (i.e. NC and OC emitting material).

We therefore adopt such median temperature to derive fiducial electron densities from Eq. (3), for both best-fit NC and OC results obtained from stack analysis: cm-3 and cm-3. In the same manner, we also derive fully consistent electron densities from the median values of sulphur ratio distributions: cm-3 and cm-33.

The targets we used to construct the stacked spectra have observed [O III]λ5007 luminosity range of 1040L[OIII] ≲ 1042.5, with a median luminosity of 1041.2 erg/s, similarly to those of the individual spectra reported in Tables A.1 and A.2. Because of their faint and/or obscured nature, we also report the median values of doubly ionised luminosity corrected for the extinction, erg/s (derived using Balmer decrement arguments and assuming Case B ratio of 3.1 and the SMC dust-reddening law; Perna et al. 2015a) and of the intrinsic X-ray luminosity, LX ~ 1042.5 erg/s. The limitation in AGN luminosity regime is due to the above mentioned conditions required to analyse [S II] and [O III]λ4363 which, for instance, preclude the inclusion of the majority of unobscured type 1 AGNs. Therefore, we note that the results presented here may be relevant only to characterise the outflow plasma properties of faint and moderately luminous AGNs (1040.5LX ≲ 1044, or ).

Narrow component plasma properties have been derived for larger samples of SDSS Seyferts by other authors in the past decade (e.g. Zhang et al. 2013; Vaona et al. 2012). Our estimates are totally consistent with the median values indicated by these authors (Ne ≈ 400 and Te ≈ 1.5 × 104 K).

Although with large uncertainties, the outflow plasma condition estimates presented in this work are, to the best of our knowledge, the first average estimate from stack analysis and medium-size samples (30 targets) of AGNs. Moreover, despite the fact that our median values for Ne(NC) and Ne(OC) are still comparable within the errors, our analysis suggests that the outflowing gas may actually be characterised by a large range of electron densities, possibly favouring the high-density regime. As a consequence, the most common assumption to derive outflow energetics (Ne = 100 cm-3) may overestimate such measurements even by a factor of ten.

thumbnail Fig. 6

Top panels: [O III]/Hβ versus [N II]/Hα, [S II]/Hα and [O I]/Hα BPT diagrams for the 28 sources used to derive the outflow electron density. Solid and dashed black lines represent the curve used to separate the purely SF galaxies, AGNs and LINER loci. Shock model grids are overplotted, with increasing velocities, from 100 to 1000 km s-1 (red to green lines), and magnetic field (blue to purple curves). Bottom panels: FWHM plotted against log([N II]/Hα), log([S II]/Hα), and [O I]/Hαfrom left to right. Both NC (open black) and OC (solid blue symbols) are shown. Positive correlations are found between the three ionisation state and the gas kinematics tracers.

Open with DEXTER

3.4. Possible bias in Te due to [O III]λ4363 selection

The fact that we compute the electron temperature for those sources with intense [O III]λ4363 could bias the results obtained with the second approach, by favouring targets with higher Te (Eq. (2)). We compute 1σ upper limits for all those sources without clear detection and, from the median value of their ROIII distribution, we derive an upper limit on the electron temperature of 3 × 104 K. Unfortunately, this value is not useful to our purpose. However, the fact that we have not observed any difference in the median Te between the S/N> 10 and 5 <S/N< 10 samples could suggest that the bias is, if any, negligible. The same behaviour can be seen for the large sample studied by Zhang et al. (2013), for which the same NLR unperturbed gas electron temperature has been found for both their two [O III]λ4363 luminosity subclasses of Seyferts. Indeed, the scarce relevance of bias may be also stressed by the consistent results we obtained from stack analysis, for which only faint and obscured AGNs have been taken into account.

4. Outflowing gas ionisation mechanisms

The unambiguous separation between NC and OC components in the [S II] emission lines for the medium size sample of the 28 AGNs from which we derived electron density estimates, enables the study of the diagnostic diagrams [O III]/Hβ versus [S II]/Hα (see, e.g. Kewley et al. 2006). Furthermore, we find well detected [O I]λ6300 emission lines in all sources (see Fig. A.1). Therefore, taking advantage from simultaneous analysis results, we fit this faint emission constraining the systemics and the FWHM of NC and OC components and derive the intensity ratio needed for a second diagnostic diagram, [O III]/Hβ vs. [O I]/Hα (Kewley et al. 2006).

Figure 6 shows three diagnostic diagrams, [O III]/Hβ versus [N II]/Hα (already presented in Paper I for the entire sample of X-ray/SDSS AGNs), [S II]/Hα and [O I]/Hα for both NC and OC. The lines drawn in the diagrams correspond to the optical classification scheme of Kewley et al. (2006, 2013): in the second and third diagram, the LINER locus is shown. The OC appears associated with the same level of ionisation of NC (i.e. same [O III]/Hβ ratios) and similar [N II]/Hα but larger [S II]/Hα and [O I]/Hα ratios. The second and third diagnostic diagrams clearly point to a LINER-like emission for the outflow components. Such line ratios are generally associated with ionisation by fast radiative shocks (e.g. Allen et al. 2008; but see also Belfiore et al. 2016). Shock model results have been made available for a large range of physical parameters: pre-shock density (to be distinguished by electron density we measured previously, and possibly associated with “post-shock” regions; see, e.g. Harrison et al. 2012), shock velocity, magnetic field, and abundances. We superimposed on the figure a grid of shock model with assumed solar abundance and a pre-shock density of 100 cm-3 (ITERA; Groves & Allen 2010). The grid shows different line ratios for various values of magnetic field and shock velocities (up to 1000 km s-1). The models, however, fail to reproduce the exact position of the OC line ratios in the BPT diagrams. We tested all available shock models from the ITERA library without any improvement.

It is possible that more extreme set of parameters are needed to reproduce the shocks in AGN-driven outflows: under particular assumptions, shock velocities and gas velocity dispersion can be similar, and therefore even larger than 1000 km s-1 (see McElroy et al. 2015, discussion). Radiation pressure-dominated photo-ionisation models easily reproduce the loci occupied by NC measurements (see the model grids in Westmoquette et al. 2012) but, as the shock models, fail to cover the highest [O I]/Hα and [S II]/Hα ratios. We notice however that these values might also be related to the presence of high metallicity regions within outflowing gas (see Villar-Martín et al. 2014), and that further observational efforts and theoretical investigation are required to discriminate between radiation pressur and shock models.

More strong evidence for shock excitation interpretation is given by the observed correlation between the gas kinematics and ionisation state (Dopita & Sutherland 1995). The line ratios produced in photo-ionised regions should be independent of the gas kinematics, while are expected to correlate with the kinematics of the shock-ionised material (see McElroy et al. 2015; Ho et al. 2014; Arribas et al. 2014). In the lower panels of Fig. 6 we show the FWHM against [N II]/Hα, [S II]/Hα and [O I]/Hα ratios for both NC and OC. We note that the FWHM should not suffer for strong degeneracy between the [S II] doublet and [N II]-Hα complex components, because derived from our simultaneous fits (Fig. A.1). We use Spearman rank correlation coefficients to determine the significance of the observed trends between line ratios and velocities. We find, on average, coefficients of 0.45 with probabilities of 0.01 for the correlation being observed by chance, for both NC and OC separately. The same correlation has been found by Arribas et al. (2014) for NC (but not for OC), confirming the complex kinematic conditions within the NLR.

5. Discussion and conclusions

In Perna et al. (2017), we analysed SDSS optical and X-ray spectra of a sample of 563 AGN at z< 0.8. In this paper, we have taken advantage of the optical spectra modelling presented in Paper I to derive the plasma properties for both narrow line and outflowing component emission. For the first time we derived estimates of electron temperature and density of the outflowing gas from stack analysis and medium size samples (30) of faint and moderately luminous AGNs.

Studying individual AGN spectra, we found indications suggesting that Te may be quite similar in both outflowing and unperturbed gas, with a median value K. The analysis of a stacked spectrum derived combining the spectra of faint and obscured X-ray detected AGNs allowed the determination of independent estimates for both NC and OC: Te(NC) = 1.3 × 104 K and Te(OC) ≈ 2.4 × 104 K. These values are still consistent (within 1σ) with the median value obtained from individual spectra analysis.

In Sect. 4 we presented BPT and ionisation state – velocity diagnostics diagrams for 28 AGNs selected from the X-ray/SDSS sample; such diagnostics indicate a shock excitation interpretation for the observed ionised outflows. If this is the case, the similar electron temperature of OC and NC could be consistent with theoretical arguments, which postulate that the (forward) shock accelerating the ISM gas is strongly cooled, so that the gas temperature rapidly returns to its pre-shock value (King 2014; see also its Fig. 2). However, if the bulk of emission is associated to shock excitation, we should observe much higher temperatures (Te> 5 × 105 K; Osterbrock & Ferland 2006). These temperatures are ruled out by our observations: in that case, we should observe much stronger [O III]λ4363 lines. Even considering possible dust extinction effects, which could affect the doubly ionised oxygen line ratios of a factor of 0.85 (assuming a SMC dust-reddening law; Prevot et al. 1984), our temperature estimates would be just ~ 10% higher.

The position of OC component flux ratios in the BPT may also be associated with different physical origins. For example, the high [S II]/Hα and [O I]/Hα OC ratios can be associated with hard ionisation radiation field, assuming the material is close to the AGN, or with high metallicities (e.g. Belfiore et al. 2016; Villar-Martín et al. 2014). Moreover, the kinematic separation between different emission line components does not ensure a common spatial distribution for all the emitting species in diagnostic diagrams. Hence, different emitting species could be characterised by distinct physical conditions. Spatially resolved information is required to better investigate the ionisation source of outflowing gas. The outflowing gas electron densities we derived analysing individual targets and a stacked spectrum displayed a wide range of values, with a distribution characterised by a median value of 1200 cm-3 and a 68% interval going from 700 to 3000 cm-3 (Table 1).

Xu et al. (2007) studied the NLR emission of ~100 SDSS type 1 AGNs. They observed a negative trend between the electron density and the blueshift of the [O III] wing. This result conflicts with our measurements. However, we note that all these results are currently limited to small number of sources: the trend suggested by Xu et al. (2007) is based on the measurement of low Ne in 6 out 54 targets with signature of outflows. Moreover, they derived electron density estimates without any separation between narrow and outflow components, making the comparison difficult.

On the other hand, other indications in literature suggest even higher Ne in the outflowing regions: Villar-Martín et al. (2015), studying high-ionisation lines such as [Fe X] and [Ne V], proposed electron densities up to 105 cm-3. However, such emission lines, although actually tracing outflowing gas (see, e.g. Lanzuisi et al. 2015), have high ionisation potential and critical densities (IP ≳ 100 eV and Nc ≳ 107 cm-3, to be compared to sulphur IP = 10.36 and Nc ≈ 2500 cm-3). They could be associated with more internal regions (see Rose et al. 2015a,b, in which such lines have been proposed to occupy a region between the BLR and the inner walls of the torus; see also Morse et al. 1998). When we consider kpc-scale outflows, typical electron density may be instead more similar to those of narrow emission lines in NLR rather than in the nuclear regions. The slightly higher density (≈ 1200 instead of ≈ 500 cm-3) may be explained by a possible compression due to the AGN wind on the ISM material.

In summary, all these considerations based on the results we obtained and on speculative arguments, suggest a more conservative approach in the estimate of the outflow energetics: the most typical assumption in deriving crucial outflow energetics (i.e. Ne = 100 cm-3) may, in fact, overestimate the outflow energetics by a factor of up to ten.


1

[O III] emissivity from PyNeb (Luridiana et al. 2015).

2

[OII]λ3727 doublet ratio allows a further density diagnostic. However, from the observational point of view, [O II] lines are so close in wavelength that only high spectral resolution observations permit the derivation of their flux ratio. This argument precludes the use of single ionised oxygen diagnostic for the sources presented in these works.

3

Similar results are obtained for the seven sources for which we were able to analyse both doubly ionised oxygen lines and sulphur lines (see Sect. 3.2): Ne(NC) ⟩ = 510 cm-3 and Ne(OC) ⟩ = 1100 cm-3.

Acknowledgments

M.P., G.L. and M.B. acknowledge support from the FP7 Career Integration Grant “eEASy” (“SMBH evolution through cosmic time: from current surveys to eROSITA-Euclid AGN Synergies”, CIG 321913). G.L. acknowledges financial support from ASI-INAF I/037/12/0. Support for this publication was provided by the Italian National Institute for Astrophysics (INAF) through PRIN-INAF-2014 (“Windy Black Holes combing galaxy evolution”). We thank the anonymous referee for his/her constructive comments to the paper. M.P. thanks A. Citro and S. Quai for useful discussion on stacked spectra analysis. Funding for the Sloan Digital Sky Survey (SDSS) has been provided by the Alfred P. Sloan Foundation, the Participating Institutions, the National Aeronautics and Space Administration, the National Science Foundation, the US Department of Energy, the Japanese Monbukagakusho, and the Max Planck Society. The SDSS Web site is http://www.sdss.org/. The SDSS is managed by the Astrophysical Research Consortium (ARC) for the Participating Institutions. The Participating Institutions are The University of Chicago, Fermilab, the Institute for Advanced Study, the Japan Participation Group, The Johns Hopkins University, Los Alamos National Laboratory, the Max-Planck-Institute for Astronomy (MPIA), the Max-Planck-Institute for Astrophysics (MPA), New Mexico State University, University of Pittsburgh, Princeton University, the United States Naval Observatory, and the University of Washington.

References

Appendix A: X-ray/SDSS sub-samples

In Fig. A.1 we report the multicomponent simultaneous fit results of the 28 sources for which it was possible to well distinguish between different kinematic components in the forbidden doublet of singly ionised sulphur (Sect. 3.2). The flux ratios R[SII] derived for the entire set of 28 AGNs are reported in Table A.1; the doubly ionised oxygen ratios obtained for the sub-samples of AGNs defined in Sect. 3.1 and used to construct the histograms in Fig. 4 are reported in Table A.2.

Table A.1

Plasma diagnostics sub-samples.

Table A.2

Plasma diagnostics sub-samples – [O III] flux ratios.

thumbnail Fig. A.1

Rest-frame optical spectra of 8 out 28 AGNs for which we derive density-sensitive flux ratios. For each object, we show the spectrum (black curve) and the best-fit models (red curves) obtained from multicomponent simultaneous fit in the regions around the Hβ and the Hα emission. The dashed vertical lines mark the location of Hβ, [O III] doublet, [O I], Hα, [N II] and [S II] doublets. Best-fit NC and BC profiles are shown with green curves; blue curves show OC emission.

Open with DEXTER

All Tables

Table 1

Plasma properties.

Table A.1

Plasma diagnostics sub-samples.

Table A.2

Plasma diagnostics sub-samples – [O III] flux ratios.

All Figures

thumbnail Fig. 1

Rest-frame optical spectra of two AGNs from our X-ray/SDSS sample. For each target, in the upper left corner of the large panel we show the MJD, the plate and the fibre numbers which uniquely identify the SDSS spectrum. The dashed vertical lines mark the location of the most prominent emission lines, besides some features crucial for the analysis presented in this paper and/or in Paper I. Fit results obtained in Paper I, from the multicomponent simultaneous fit in the Hβ-[O III] and Hα-[NII] regions are also shown: best-fit narrow component (NC) and broad line region component (BC) profiles are highlighted with green curves; outflow component (OC) and Fe II emission are shown with blue and magenta curves, respectively. Finally, black Gaussian profile in the low panel shows a second set of BC used to fit the complex BLR profiles (see Paper I for further details). The insets on the right of each panel show an enlargement of the sulphur lines. The spectra display faintness and blending problems affecting density-sensitive [S II]λλ6716, 6731 and temperature-sensitive [O III]λ4363 emission lines. In particular, the object in the first panel displays a well detected and modelled sulphur doublet, from which we can easily derive the density-sensitive flux ratio, but faint oxygen λ4363 line; the spectrum in the lower panel shows blended sulphur doublet features and oxygen line at λ4363. For both the AGNs it is therefore not possible to derive temperature-sensitive flux ratios.

Open with DEXTER
In the text
thumbnail Fig. 2

Median stacked spectrum around the doubly ionised oxygen emission lines obtained combining the spectra of faint/obscured sources with evidence of ionised outflows (see the text for more details). Dashed curves highlight 1σ uncertainties. Red curve represents the best-fit result we obtained fitting simultaneously the emission lines displayed. Green Gaussian represent NC and BC components, blue profiles mark OC.

Open with DEXTER
In the text
thumbnail Fig. 3

Magnification of the region of [O III]λ4363-Hγλ4342 lines for the 44 sources for which we could determine R[OIII] flux ratios. For each spectrum, the best-fitting components presented in Sect. 3.1 are superimposed: solid green curves represent the systemic narrow component (NC); blue curves the OC.

Open with DEXTER
In the text
thumbnail Fig. 4

left: [O III] ratio distributions. Grey shaded areas mark the R[OIII] distribution for the unperturbed ionised gas (i.e. NC). Blue shaded area denotes the outflow emission R[OIII] distribution of S/N> 10 sources. The blue dashed line shows the R[OIII] histogram of the outflow components of all S/N> 5 sources (see the text for detailed analysis description). In each panel, median and 68% confidence intervals obtained from the analysis of stacked spectra are also shown for comparison (black and blue symbols, representing line ratios associated with NC and OC, respectively). Right: [S II]λλ6716, 6731 ratio distributions. The grey solid line marks the distribution for the AGN sample without evidence of outflows from our line fitting routine. The black and blue shaded areas denote the distributions for NC and OC sulphur ratios obtained from the sub-sample of 28 AGNs.

Open with DEXTER
In the text
thumbnail Fig. 5

Median stacked spectrum around the [S II] emission line doublet obtained combining the spectra of sources with evidence of ionised outflows and without blending problems with Hα BLR emission. Dashed curves highlight 1σ uncertainties. The red curve represents the best-fit result we obtained fitting simultaneously Hα, [N II] and [S II] doublets. The green Gaussians represent NC and BC components, blue profiles mark OC.

Open with DEXTER
In the text
thumbnail Fig. 6

Top panels: [O III]/Hβ versus [N II]/Hα, [S II]/Hα and [O I]/Hα BPT diagrams for the 28 sources used to derive the outflow electron density. Solid and dashed black lines represent the curve used to separate the purely SF galaxies, AGNs and LINER loci. Shock model grids are overplotted, with increasing velocities, from 100 to 1000 km s-1 (red to green lines), and magnetic field (blue to purple curves). Bottom panels: FWHM plotted against log([N II]/Hα), log([S II]/Hα), and [O I]/Hαfrom left to right. Both NC (open black) and OC (solid blue symbols) are shown. Positive correlations are found between the three ionisation state and the gas kinematics tracers.

Open with DEXTER
In the text
thumbnail Fig. A.1

Rest-frame optical spectra of 8 out 28 AGNs for which we derive density-sensitive flux ratios. For each object, we show the spectrum (black curve) and the best-fit models (red curves) obtained from multicomponent simultaneous fit in the regions around the Hβ and the Hα emission. The dashed vertical lines mark the location of Hβ, [O III] doublet, [O I], Hα, [N II] and [S II] doublets. Best-fit NC and BC profiles are shown with green curves; blue curves show OC emission.

Open with DEXTER
In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.