Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z\sim1.5 lensed quasars
G. Tozzi, G. Cresci, A. Marasco, E. Nardini, A. Marconi, F. Mannucci, G. Chartas, F. Rizzo, A. Amiri, M. Brusa, A. Comastri, M. Dadina, G. Lanzuisi, V. Mainieri, M. Mingozzi, M. Perna, G. Venturi, C. Vignali
AAstronomy & Astrophysics manuscript no. high-z_ufo © ESO 2021March 3, 2021
Connecting X-ray nuclear winds with galaxy-scale ionised outflowsin two z ∼ . lensed quasars G. Tozzi , , G. Cresci , A. Marasco , E. Nardini , , A. Marconi , , F. Mannucci , G. Chartas , F. Rizzo , A. Amiri , ,M. Brusa , , A. Comastri , M. Dadina , G. Lanzuisi , V. Mainieri , M. Mingozzi , M. Perna , , G. Venturi , , andC. Vignali , Dipartimento di Fisica e Astronomia, Università di Firenze, Via G. Sansone 1, 50019, Sesto Fiorentino (Firenze), Italye-mail: [email protected] INAF - Osservatorio Astrofisico di Arcetri, Largo E. Fermi 5, 50127, Firenze, Italy Department of Physics and Astronomy, College of Charleston, Charleston, SC, 29424, USA Max-Planck Institute for Astrophysics, Karl-Schwarzschild Str. 1, D-85748, Garching, Germany Dipartimento di Fisica e Astronomia dell’Università degli Studi di Bologna, via P. Gobetti 93 /
2, 40129 Bologna, Italy INAF - Osservatorio di Astrofisica e Scienza dello Spazio, via P. Gobetti 93 /
3, 40129, Bologna, Italy European Southern Observatory, Karl-Schwarzschild-Str. 2, D-85748 Garching, Germany Space Telescope Science Institute, 3700 San Martin Drive, Baltimore, MD 21218, USA Centro de Astrobiología, (CAB, CSIC-INTA), Departamento de Astrofísica, Cra. de Ajalvir Km. 4, 28850 – Torrejón de Ardoz,Madrid, Spain Instituto de Astrofísica, Facultad de Física, Pontificia Universidad Católica de Chile, Casilla 306, Santiago 22, ChileMarch 3, 2021
ABSTRACT
Aims.
Outflows driven by active galactic nuclei (AGN) are expected to have a significant impact on host galaxy evolution, but thematter of how they are accelerated and propagated on galaxy-wide scales is still under debate. This work addresses these questions bystudying the link between X-ray, nuclear ultra-fast outflows (UFOs), and extended ionised outflows, for the first time, in two quasarsclose to the peak of AGN activity ( z ∼ ff ective. Methods.
Our selected targets, HS 0810 + + z ∼ . λ Results.
We detected spatially resolved ionised outflows in both galaxies, extended more than 8 kpc and moving up to v > / s.We derived mass outflow rates of ∼
12 M (cid:12) yr − and ∼ (cid:12) yr − for HS 0810 + + Conclusions.
Compared with the co-hosted UFO energetics, the ionised outflow energetics in HS 0810 + + ff ers by about two orders of magnitudefrom theoretical predictions, requiring either a massive molecular outflow or a high variability of the AGN activity to account for sucha discrepancy. By additionally considering our results together with those from the small sample of well-studied objects (all local butone) having both UFO and extended (ionised, atomic, or molecular) outflow detections, we found that in 10 out of 12 galaxies, thelarge-scale outflow energetics is consistent with the theoretical predictions of either a momentum- or an energy-driven scenario ofwind propagation. This suggests that such models explain the acceleration mechanism of AGN-driven winds on large scales relativelywell. Key words. galaxies: evolution – galaxies: active – quasars: emission lines – ISM: jets and outflows – techniques: imaging spec-troscopy
1. Introduction
Feedback mechanisms from active galactic nuclei (AGN) arewidely considered to play a key role in galaxy formation andevolution (Silk & Rees 1998; King 2010b; Fabian 2012; King& Pounds 2015). AGN feedback is indeed included in all theo-retical, semianalytic, and numerical studies of galaxy formationand evolution (e.g. Granato et al. 2004; Di Matteo et al. 2005;Ciotti et al. 2010; Guo et al. 2016; Nelson et al. 2019) as it al-lows us to reconcile theoretical predictions with observed galaxyproperties. Such AGN activity is considered the main factor re-sponsible for the quenching of star formation in more massivegalaxies (so-called ‘negative feedback’). The energy output of asupermassive black hole (SMBH) accreting close to the Edding- ton limit is large enough to drive massive, wide-angle outflowson large scales (e.g. Zubovas & King 2012, 2014; Costa et al.2014; King & Pounds 2015; Pontzen et al. 2017) that are capa-ble of either sweeping the gas out of the host galaxy, thus reduc-ing the host galaxy gas reservoir for star formation, or heatingthe intergalactic medium of the host galaxy through the injec-tion of thermal energy, thus preventing the gas from cooling andcollapsing to form stars (‘ejective’ versus ‘preventive’ feedback;e.g. see Woo et al. 2017; Cresci & Maiolino 2018). Both pro-cesses are expected to halt the accretion onto the central blackhole and, consequently, to give rise to the SMBH mass valuesobserved to correlate with the physical properties of the hostgalaxy bulge (i.e. its mass, velocity dispersion and luminosity;e.g. Ferrarese & Merritt 2000; Kormendy & Ho 2013). Addi-
Article number, page 1 of 19 a r X i v : . [ a s t r o - ph . GA ] M a r & A proofs: manuscript no. high-z_ufo tional observational evidence supporting the mutual influence ofthe central SMBH and the host galaxy comes from the observedsimilarity between the star formation (SF) and BH accretion his-tories across cosmic time. In fact, both activity histories are seento peak at z ∼ z ∼ − z ∼ − ff ectsin action since this is the time when AGN feedback is expectedto be more e ff ective.Thanks to advanced integral-field spectroscopic (IFS) fa-cilities, AGN-driven outflows have been exhaustively observedfrom optical to IR and mm bands, in both local (e.g. Feruglioet al. 2010, 2015; Cicone et al. 2012, 2014) and high-redshiftgalaxies (e.g. Cano-Díaz et al. 2012; Maiolino et al. 2012; Car-niani et al. 2015; Cresci et al. 2015). It is worth mentioning thatwhile theoretical predictions usually refer to the whole outflow-ing gas, observations usually trace the emission produced by asingle gas phase of the outflow. Therefore, in order to properlycompare model predictions with observational results, it is fun-damental to obtain a complete, multi-phase description of theoutflow (e.g. Cicone et al. 2018; Harrison et al. 2018).Even though the existence of AGN-driven outflows has beenwidely confirmed thanks to observations, there are a number ofrelevant open questions that remain unanswered, considering,for instance, how the energy released by the accreting BH is cou-pled with the interstellar medium (ISM), thus driving large-scaleoutflows, and how e ffi cient the coupling is between the nuclearand galaxy-scale outflows.Theoretical models (e.g. King 2003, 2005; King & Pounds2015) predict fast ( v ∼ . c ), highly ionised winds, acceleratedon sub-pc scales by the AGN radiative force as the origin of thestrong galaxy-scale feedback. As the nuclear wind impacts theISM of the host galaxy, it produces an inner reverse shock thatslows down the wind, along with an outer forward shock accel-erating the galactic ISM. Depending on the e ffi ciency of coolingprocesses (typically radiative) in removing energy from the hotshocked inner gas, there are two main wind driving-modes (e.g.King 2010a; Costa et al. 2014; King & Pounds 2015). If the cool-ing occurs on a timescale that is shorter than the wind flow time,most of the inner wind kinetic energy is lost (usually via inverseCompton scattering) and, therefore, the wind momentum is theonly conserved physical quantity (‘momentum-driven’ regime).Vice versa, if the cooling is negligible, the postshock gas retainsall the mechanical energy and expands adiabatically (‘energy-driven’ regime), sweeping up a significant amount of the hostgalaxy gas. According to a widely accepted picture, the observedscaling relations are the result of the e ff ect of AGN feedback act-ing on the host galaxy through two distinct, subsequent phases(e.g. Zubovas & King 2012; King & Pounds 2015): an initialmomentum-driven regime lasting so long as the BH mass hasnot yet reached the M BH − σ relation and the outflow is confinedwithin ∼ ff ord et al.2013, 2015; Nardini et al. 2015) via the presence of stronglyblueshifted absorption lines of highly-ionised metals (typically iron, e.g. Fe XXV and Fe XXVI). UFOs are found in at least 40%of the local sources (Tombesi et al. 2010, 2011, 2012, 2013),with typical mass outflow rates of ∼ . − (cid:12) yr − and ki-netic powers of log ˙ E K ∼ −
45 erg s − (Tombesi et al. 2012).Moving to high redshifts ( z > z > λλ n ∼ cm − ), sub-pc scale of the AGN broad line region (BLR). In thepresence of outflows, the [O III] line profile is highly asymmet-ric with a broad, blueshifted wing corresponding to high speedsalong the line of sight ( v (cid:38) − (see e.g. Carniani et al.2015; Cresci et al. 2015; Perna et al. 2015; Brusa et al. 2016;Marasco et al. 2020).This work is aimed at studying the connection between nu-clear, X-ray UFOs, and the ionised phase of large-scale outflows,for the first time, in two QSOs close to the peak of AGN activ-ity ( z ∼ λ ff erent scales.In doing so, we aim to highlight (and take advantage of) the cru-cial role of gravitational lensing as powerful tool for overcomingthe current observational limits and going on to investigate thephysical properties of distant QSOs.This paper is organised as follows. In Sect. 2, we present theselected targets and describe the observation and data reductionprocedure. In Sect. 3, we show our data analysis and spectralfitting. The inferred results are then presented in Sect. 4. In Sect.5, we discuss the wind acceleration mechanism in our two QSOsand compare our results with findings from the literature. Finally,in Sect. 6, we outline our conclusions. We adopt a Λ CDM flatcosmology with Ω m , = . Ω Λ , = .
73 and H =
70 km s − Mpc − throughout this work.
2. Description of the observed QSOs
Our sample consists of two z ∼ . + + Article number, page 2 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . the program 0102.B-0377(A) (PI: G. Cresci). These objects werespecifically selected as they are known to host UFOs (Chartaset al. 2016; Chartas et al. 2021) and to be at redshifts ( z ∼ . z ∼ . + + λ ∼ . − . µ m). Con-sequently, this selection in redshifts corresponds to study ob-jects at epochs close to the peak of AGN activity ( z ∼ z >
1) QSOs with UFOdetection, of which seven are found in the literature (amongthem, HS 0810 + + L −
10 keV > ergs − , except for PID352 L −
10 keV ∼ erg s − ) - QSOs athigh redshift; this is either because they are intrinsically lumi-nous (of the total 14-QSO sample, only HS 1700 + + + + + + ; Hasinger et al. 2002; Chartas et al. 2003,2007, 2009, 2016; Dadina et al. 2018). Thus, these objects de-liver high quality X-ray spectra which clearly exhibit UFO ab-sorption features, in spite of their high redshift ( z > z > + z ∼ .
9; Hasinger et al.2002) has been the only z > + + z > + + HS 0810+2554
HS 0810 + (cid:46)
500 km s − ) QSO at z ∼ .
5, which was discoveredby Reimers et al. (2002). It consists of four lensed images ina typical fold lens configuration with the two southern, bright-est images in a merging pair configuration (A + B), as shown inthe
HST image in Fig. 1 ( left panel). The lens galaxy (labelledwith G) is detected in the
HST image, and its redshift is es-timated to be z l ∼ .
89 from the separation and the redshiftdistribution of existing lenses (Mosquera & Kochanek 2011).Quadruply lensed QSOs occur in strong gravitational lensingregimes (e.g. Narayan & Bartelmann 1996), when the compactand bright UV accretion disk and X-ray corona emission regionsoverlap the lens caustics. This leads to high magnification fac-tors, whose values strongly depend on the image and lens posi-tions. As a consequence, a small change in the input parametersto the lens models (image and lens positions) can lead to a sig-nificant change in the image magnifications. For HS 0810 + µ in di ff erent spectral bandsare found in the literature, in particular for the X-ray ( µ ∼ µ ∼ µ ∼
25; Jackson et al. 2015). Here, we list only the sources with published results on UFO de-tection, but also the remaining unpublished objects are known to begravitationally lensed (Chartas et al. 2021).
HS 0810+2554 SDSS J1353+1138
Fig. 1.
Lensed images of HS 0810 + left ) and SDSS J1353 + right ). Left: HST
ACS F555W image of HS 0810 + + B isblended together. At the centre, we can see the emission from the fore-ground lens galaxy.
Right: V and H -band images of SDSS J1353 + upper panels) and corresponding imagesafter the subtraction of A and B components ( lower panels), clearlyshowing the lens galaxy (component G). Image from Inada et al. (2006). HS 0810 + Chandra and
XMM-Newton observations (Char-tas et al. 2016) provided definitive proofs for the presence of ahighly ionised, relativistic wind in the source nuclear region. Thestrongly blueshifted absorptions of highly-ionised metals (i.e. FeXXV, Si XIV) indicate that the outflow velocity components arewithin the range of 0 . − . c . The VLT / UVES spectrum of HS0810 + v CIV = ,
400 km s − (Chartas et al. 2014, 2016). Even though is classified as radio-quiet object, VLA observations at 8.4 GHz (Jackson et al. 2015)indicate that HS 0810 + + = →
2) emission, suggesting the presence of a mas-sive molecular outflow on kpc-scales. With our characterisa-tion of the ionised outflow in HS 0810 + SDSS J1353+1138
Unlike HS 0810 + + z ∼ .
6. Inada et al. (2006) dur-ing the University of Hawai’i 88-inch Telescope (UH88) follow-up observations of SDSS J1353 + V , R , I, and H -band images of the source. The two lensed images are well- Article number, page 3 of 19 & A proofs: manuscript no. high-z_ufo
Target name α (J2000) δ (J2000) z a scaleHS 0810 + h m s . + ◦ (cid:48) (cid:48)(cid:48) . ± .
002 8.67 kpc / (cid:48)(cid:48) SDSS J1353 + h m s . + ◦ (cid:48) (cid:48)(cid:48) . . ± .
002 8.69 kpc / (cid:48)(cid:48) Table 1:
Properties of our two-QSOs sample. a Redshifts are measured from the [O III] systemic component in integrated spectra extracted fromthe nuclear region of both sources (Sects. 3.1 and 3.2). distinguishable (see the right panel in Fig. 1), with an angularseparation of ∆ ∼ . (cid:48)(cid:48) (Inada et al. 2006).More recently, on 2016 January 13, SDSS J1353 + XMM-Newton . The analysis of the X-ray spec-trum (Chartas et al. 2021) revealed a significant absorption at ∼ . ∼ . c UFO.
SINFONI observations of HS 0810 + + ff erent nights inFebruary and March 2019, respectively, in the near-IR J -band( λ ∼ . − . µ m) and with a spectral resolution R = ,using the 0 . (cid:48)(cid:48) × . (cid:48)(cid:48) pixel scale which provides a totalfield of view (FOV) of 8 (cid:48)(cid:48) × (cid:48)(cid:48) , which is essential for map-ping the gas dynamics on galaxy scales. The airmasses aredi ff erent for each target, spanning the ranges of ∼ . − . ∼ . − . + + + + ff erent positions of theFOV about 4 . (cid:48)(cid:48) apart, to perform the sky subtraction througha nodding technique. A dedicated star observation to measurethe point-spread-function (PSF) was not available in eithercase but the estimated angular resolution is ∼ (cid:48)(cid:48) ( ∼ (cid:48)(cid:48) ) forHS 0810 + + ff ectsconsisting in a significant change of the AGN continuum emis-sion across the FOV of both sources. This is a consequence ofthe di ff erential atmospheric dispersion at di ff erent wavelengths,which makes it so the measured spectra not ‘straight’ as ex-pected. In practical terms, as the wavelength increases, an in-creasingly larger fraction of the emission gets deposited intoadjacent pseudo-slits, producing a coherent spatial shift of thetarget as a function of λ . Additionally, possible flexures in theinstrument may contribute to producing the observed shift. Inorder to limit the impact of these optical distortions, we spa-tially aligned the emission centroid channel-by-channel in eachsingle-exposure, sky-subtracted cube by adopting the followingprocedure.For each cube, we first determined, in every spectral chan-nel, the average position of the emission centroid on the FOV The SINFONI adaptive optics module (AO-mode) was not availableat the time of the observations, since SINFONI had been moved fromUT4 to UT3 for the last few months of its research activity. through a 2D-Gaussian fit. Then we calculated the shift of thecentroid mean position with respect to the centroid position inthe first spectral channel, assumed as the reference channel. Inboth spatial directions on the FOV we found an increasing trendof the shift at increasing wavelengths, which we modelled witha two-degree polynomial to neglect the presence of some spikesthat are due to noisier channels. The spatial shift, totally ob-served from the bluest to the reddest spectral channel, spannedthe range of ∼ . − × × .
25 Å channel width and cov-ers the spectral range ∼ . − . µ m, corresponding to about4400 − As both objects are gravitationally lensed QSOs, lens modelsare required in order to infer the intrinsic (i.e. unlensed) physi-cal properties of the outflow, such as the intrinsic radius and un-lensed flux, which are key ingredients for calculating the outflowenergetics.Both QSOs are lensed by a foreground elliptical galaxy anddetailed lens models for these two objects can be found in the lit-erature. In particular, for HS 0810 + ff erent spectral bands (e.g. from VLA-radio datain Jackson et al. 2015, from ALMA-mm data in Chartas et al.2020). In this work, we adopted, for HS 0810 + HST -WFC3 IR observations: images and lens galaxy positions havebeen measured from direct F140W wide imaging (central wave-length ∼ − z l ∼ .
89 for the lens galaxy (Mosquera & Kochanek2011), Nierenberg et al. (2020) modelled the deflector mass dis-tributions with a singular isothermal ellipsoid (SIE), plus an ex-ternal shear to account for tidal perturbations from nearby ob-jects.Detailed lens models for SDSS J1353 + i -band with the Magellan Instant Camera(MagIC) on the Clay 6.5m Telescope and in the K -band withthe Subaru Telescope adaptive optics system, respectively. In-ada et al. (2006) modelled the lens mass distribution using ei-ther a SIE model, or a singular isothermal sphere (SIS) modelplus a shear component ( γ ), and estimated the lens redshift to be z l ∼ . µ are 3.81 and Article number, page 4 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . + γ and SIE model, respectively. Also assuming z l ∼ .
3, Rusu et al. (2016) found slightly lower values for totalmagnification: µ ∼ .
47 (SIS + γ ), µ ∼ .
42 (SIE) and µ ∼ . + γ ). We used all these magnification values from the liter-ature to determine the unlensed flux carried by the ionised out-flow in SDSS J1353 + + µ out ∼
2) starting from our data. Such values will be discussedfurther in Sect. 4.2 and Appendix A.
3. Data analysis
For the spectral analysis of SINFONI data, we adopted the fit-ting code used in Marasco et al. (2020) to analyse MUSE data oftwo local QSOs. Here, we implemented the code to also handlethe SINFONI data, introducing adjustments and new function-alities depending on the specific necessities of our data. In thefollowing, we illustrate the basics of our spectral-fitting method,highlighting the required changes for the analysis of our SIN-FONI data (see Marasco et al. 2020 for the detailed descriptionof the fitting code).The entire fitting procedure aims at performing the kinemat-ical analysis of the di ff use ionised gas, with a primary focus onthe [O III] λ Phase I.
We built a template model for the bright BLR emis-sion using an integrated, high signal-to-noise ratio (S / N) spec-trum.
Phase II.
The BLR template built in phase I was used to mapspaxel-by-spaxel the contribution of the BLR to the emissionacross the entire FOV. The resulting BLR model cube was thensubtracted from the data cube.
Phase III.
Finally, we performed, spaxel-by-spaxel, a finermodelling of the faint emission lines produced by the di ff useionised gas, originating on galactic scales.Hereafter, we refer to the di ff use gas emission lines as ‘nar-row’ in order to distinguish them from the typical ‘broad’ emis-sion lines (FWHM > − ; e.g. Osterbrock 1981) origi-nating from within the dense and highly turbulent BLRs.A single noise value has been associated to each channelin our SINFONI data cubes, computed as the root mean square(rms) of the fluxes extracted spaxel-by-spaxel in a region with nosignificant emission from the target. The details of the spectralanalysis of SINFONI data of both QSOs are provided below. I. Modelling the BLR emission
The fitting code starts with modelling the bright BLR emissionin a spectrum extracted from the nuclear region, while also fit-ting the other spectral components. The spectral components tobe fitted are: the AGN continuum, BLR emission lines, and nar-row emission lines from the di ff use gas. In principle, we shouldalso have the stellar continuum emission, but in our data theAGN continuum is entirely dominant. The BLR model is built bythe code as the sum of two independent components: the broadBalmer hydrogen emission lines (H β in HS 08010 + β and H γ in SDSS J1353 + λ ∼ − ff useemission instead consists of the [O III] emission doublet and thenarrow components of the Balmer hydrogen lines. For HS 0810 + / N spectrum froman aperture of 0.3 (cid:48)(cid:48) radius, centred on the observed blended emis-sion of A + B images (see Fig. 1), and fitted all the previouslymentioned components simultaneously. We modelled the AGNcontinuum through a 1st-degree polynomial. The Fe II emissionlines were modelled using the semi-analytic templates of Ko-vaˇcevi´c et al. (2010), while the BLR component of H β was fittedby two broad Gaussian components. The narrow emission lineswere fitted through two Gaussian components. Given the com-plexity of the BLR-H β line profile in HS 0810 + + + II. Mapping the unresolved BLR emission across the FOV
As the BLR emission goes unresolved in our data, we expectthat its observed spatial variation follows the PSF of our obser-vations. Therefore, we allowed the BLR template obtained inphase I to change only in amplitude across the FOV and we pro-ceeded to fit the whole data cubes with the software pPXF (Cap-pellari 2017). For the modelling of the narrow emission lines, weused multiple Gaussian components and adopted a statistical ap-proach (a Kolgomorov-Smirnov test) to select, spaxel-by-spaxel,the minimal, optimal number of Gaussian components to aptlyreproduce the emission line profiles (see Marasco et al. 2020 fordetails). For both HS 0810 + + III. Modelling the narrow emission lines
In phase III, we focused on the finer modelling of the narrowemission lines that remained after the subtraction of the BLRand AGN continuum emission components. The only signifi-cant residual emission in our data is the [O III] emission dou-blet, while the narrow components of Balmer hydrogen lines arevery weak and marginally resolved. Therefore, we modelled theresidual narrow emission lines through multiple Gaussian com-ponents, adopting some reasonable constraints: the two emissionlines of the [O III] doublet were fitted by imposing the samecentral velocity and velocity dispersion, with the intensity ra-tio I (5007) / I (4959) fixed at 3 according to the theoretical ex-pectations of the atomic theory; whereas, for the weak narrowcomponents of Balmer hydrogen lines, we assumed the same [OIII] λ Article number, page 5 of 19 & A proofs: manuscript no. high-z_ufo H β [OIII] [OIII] HS 0810+2554 [OIII] [OIII]H β H γ SDSS J1353+1138
Fig. 2.
Representative scheme of our fitting procedure.
Top panels:
SINFONI J -band spectra of HS 0810 + left ) and SDSS J1353 + right ), zoomed in the spectral region of [O III] and Balmer hydrogen emission lines. Both spectra were extracted using an aperture of ∼ (cid:48)(cid:48) -radius, centred on the peak of the AGN continuum emission (located on image A in SDSS J1353 + ff usegas (light blue) represented as sum of multiple Gaussian components. Middle panels: J -band spectra extracted from the subtracted cubes, with thesame aperture used in the top panels. Subtracted data (black lines) are compared to the best-fit models (red lines) resulting from the finer emissionlines (EL) modelling implemented in phase III. The green lines highlight the outflow component alone.
Bottom panels:
Residuals obtained bysubtracting the full EL model from the subtracted spectra. statistical approach to determine spaxel-by-spaxel the optimalnumber of components required. In both QSOs, most of the lineprofiles are well reproduced by two Gaussian components: anarrow, bright component close to the systemic velocity, plus abroad blueshifted component to reproduce the [O III] blue wingobserved in most of the FOV, which we identify with approach-ing outflow emission. In HS 0810 + + ffi cient to repro-duce the most complex line profiles, as we do not detect the [OIII] red wing anywhere.In order to study the physical and dynamical properties ofthe outflow emission, which is the main focus of this work, wehad to properly identify the [O III] emission due to the high-speed outflowing gas by disentangling its contribution to the [OIII] emission line due to the gas bulk motion within the host galaxy. To do so, we adopted the same selection criterium usedby Marasco et al. (2020). For each Gaussian component used toreproduce the [O III] line profile in a given spaxel, we focused onthe fraction of the total line flux contained in the line wings witha velocity shift | v − v peak | larger than a certain threshold width w th , where v peak is the peak velocity of the line in each spaxel.If the fraction of total flux in the line wings was higher than agiven threshold τ , the Gaussian component has been classifiedas a possible ‘outflow’ component, to be confirmed by the fol-lowing kinematical analysis (Sect. 4.1); otherwise, it has beenclassified as a ‘narrow’ component, due to systemic gas motionsin the host galaxy. We verified the decomposition in several rep-resentative spaxels to select the optimal threshold values. In HS0810 + τ = . w th =
300 km s − : the Gaus-sian component reproducing the narrowest, brightest emissionnear the systemic velocity has been typically classified as nar-row; while any additional Gaussian component used to modeleither the blue or the red wing in [O III] profile has been identi- Article number, page 6 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . fied as outflow component. In SDSS J1353 + w th =
250 km s − ).Figure 2 summarises our strategy. The top panels show J -band spectra of HS 0810 + + ∼ (cid:48)(cid:48) -radius, centred on the peak of the observed emission (locatedon image A in SDSS J1353 + ff use gas model is plotted assum of multiple components. We notice that in HS 0810 + ∼ − ff ect the overall fitting pro-cedure, as the observed Fe II emission is reproduced well bythe templates at all the other wavelengths. The middle panelsshow the spectra extracted from the subtracted cubes using thesame aperture as above, along with the results from the finermulti-Gaussian fit of the di ff use gas emission lines (phase III).In both QSO-subtracted spectra, the [O III] line profile exhibitsa prominent, asymmetric blue wing that is already visible in thefull spectra shown above. This strongly suggests the presenceof high-speed outflowing material moving towards the observer,which we discuss in greater detail in Sect. 4.1. As noted in Sect. 3.1, for SDSS J1353 + β and [O III] lines, while comparing the nuclearspectrum of the brighter image A (spectrum A) with that of thefainter image B (spectrum B), shown in the upper and lower pan-els of Fig. 3, respectively. In particular, while in the former it ispossible to easily identify the [O III] emission lines, we did notdetect any counterpart in the latter. Moreover, the H β line pro-file in spectrum B is broader, with an evident brighter blue wing.Both e ff ects are likely due to an overall increase in the Fe II emis-sion in image B, as the H β line width is not expected to intrinsi-cally vary between di ff erent lensed images. The anti-correlationbetween Fe II and [O III] emissions in AGN spectra reflects awell-known e ff ect known as Eigenvector-1 (Boroson & Green1992) and represents one of the most frequent di ff erences amongAGN properties. Even though it has been the subject of manystudies, a clear physical understanding of its origin is still lack-ing. Boroson & Green (1992) suggest that high column densitiesin the BLR enhance Fe II, while reducing the ionising radiationable to reach the NLR. In a spectral analysis of AGN principalcomponents in SDSS, Ludwig et al. (2009) argue instead that thecovering factor of the NLR is the likely cause of the range in [OIII] strength, while Ferland et al. (2009) suggest that the highercolumn densities required for the infall in more luminous AGNswould additionally account for the observed correlation of Fe IIstrength with L / L Edd .In spite of its still unclear origin, there are two main possibleexplanations for the observed significant variation in the Fe IIemission between the two lensed images. The first one is basedon the typical short time scales (i.e. days, weeks; e.g. Kaspi et al.2000) on which the BLR is seen to vary. Because of the di ff erentpath followed by the light from the background QSO, the twolensed images are produced with a time delay of about 16 days(Inada et al. 2006). This temporal shift is comparable with thetypical BLR variation timescale, therefore, it could be su ffi cient H γ H β [OIII] [OIII] H γ H β [OIII] [OIII] Fig. 3.
Best-fit models of nuclear spectra of SDSS J1353 + (cid:48)(cid:48) radius, centred on image A (upperpanel) and on image B (lower panel). The various spectral components,the total model and data are represented with di ff erent colours (see theplot legend). In spectrum A, a broken power law distribution is perfectlysuited to reproduce the BLR-H β profile, while in spectrum B, an addi-tional broad Gaussian component was required to adequately reproducethe broad peak in the BLR-H β line profile, which is entirely dominantover the barely detected H β narrow component (solid light-blue line).The two spectra clearly di ff er from each other mostly for the lack of [OIII] detection and the presence of a prominent blue wing in the H β lineprofile in spectrum B. to have a significant change in the BLR emission explaining thee ff ect we observed. Given the short time scale probed here, thiscould carry interesting implications for the accretion variationsin the AGN and in the consequent response of the BLR gas. Al-ternatively, the observed variation could be the consequence ofmicrolensing e ff ects (e.g. see Nierenberg et al. 2020) producedby either single stars or low-mass dark matter halos interven-ing along the line of sight. Microlensing e ff ects typically a ff ectonly the emission originated on small scales, while the emissionfrom the NLR is insensitive. Of the two possibilities, the lat-ter seems to be less likely, as we do not observe any significantcounterpart variation in the strength of the BLR-H β component,in addition to that observed in the Fe II strength. However, a re- Article number, page 7 of 19 & A proofs: manuscript no. high-z_ufo markable simultaneous variation in both H β and Fe II strengthis what we would expect in the case when the two emissionsare strictly co-spatial, whereas we know that the BLR is strati-fied and that microlensing magnifies the emission from the mostcompact regions more strongly. Therefore, we do not excludethe microlensing hypothesis. A detailed analysis of the di ff erentBLR spectra from the two images is beyond the scope of thiswork and will be presented in a forthcoming paper. Therefore, inthis work, we focus on how we accounted for this e ff ect duringthe spectral analysis.Consequently, in SDSS J1353 + (cid:48)(cid:48) radius aperture centred on the emissionpeak of each lensed image. We proceeded to fit them separately.In both spectra we used a first-degree polynomial to model theAGN continuum that is still dominant over the stellar continuum.Unlike the BLR modelling of HS 0810 + ffi cient to reproduce the broader and complexprofile of the broad Balmer emission lines, especially the H β linein spectrum B. In fact, even though both H β prominent wingsare likely due to the Fe II emission, as discussed above, the FeII templates employed by the code were not able to reproducesuch observed emission. Therefore, we modelled both wings aspart of the broad H β line profile without any focus on their physi-cal interpretation, as we were simply interested in identifying theoverall BLR spectrum in order to remove it. For the modelling ofthe broad Balmer emission lines observed in SDSS J1353 + β and H γ ) , we used a broken power law distribution con-volved with a Gaussian profile (Nagao et al. 2006): F ( λ ) = F × (cid:16) λλ (cid:17) + α , for λ < λ F × (cid:16) λλ (cid:17) − β , for λ > λ , (1)where the free parameters of the fit, for each line, are the centralwavelength λ , the two power-law indices α and β , the normali-sation F , and the width σ of the Gaussian kernel. In spectrum A(upper panel), the H β and H γ were modelled separately throughthe line profile described in Eq. (1). Spectrum B (lower panel) re-quired instead an additional broad Gaussian component to suit-ably reproduce both the more extended red wing and the broadpeak ( σ ∼
800 km s − ) in the H β profile. Moreover, we had toconstrain the H γ line profile through that of H β . Di ff erent Fe IItemplates have been selected in the BLR best-models for the twonuclear spectra. To model the narrow emission line profiles, weused three Gaussian components in spectrum A, while a singleGaussian component in spectrum B, since we did not actuallyobserve any narrow component.Then we proceeded to fit the whole data cube following theprocedure described in Sect. 3.1, with the main di ff erence thatin each spatial pixel, pPXF considered a linear combination ofthe two BLR models, weighing their relative contribution andproviding their most suitable combination as the BLR model inthat specific spaxel. In general, in those spaxels close to one ofthe lensed images, we basically got the BLR model directly ob-tained in the modelling of the respective nuclear spectrum; whilea combination of the two BLR-models in those spaxels ended uproughly between the two images, as we indeed expected. Before analysing the outflow kinematics across the FOV, wetested whether the emission of the detected ionised outflows was spatially resolved. This is crucial for the calculation of the out-flow energetics. As our observed data have missed a dedicatedPSF star, we compared the spatial extent of the [O III] outflowemission with that of the BLR emission (both obtained from theprevious spectral modelling). In fact, given that the latter is unre-solved in our data, it is suitable for reproducing the instrumentalresponse. We fitted a 2D-Gaussian profile to the BLR flux map,obtained by integrating in wavelength our BLR model, in orderto estimate the angular resolution of our seeing-limited observa-tions. The resulting best-fit Gaussian profiles are not circularlysymmetric, especially in the case of HS 0810 + ff ects and to the blending of A andB images, rather than to a possible intrinsic asymmetry of thePSF. Therefore, for both sources we assumed the angular size ofthe minor axis of the best-fit Gaussian profile as representativefor the true PSF extent, as this is heading in the direction wherethe lens stretching e ff ects are expected to be minimised. For HS0810 + + θ res ) of 0.7 (cid:48)(cid:48) and 0.8 (cid:48)(cid:48) , respectively. These NIR valuesare slightly smaller than the optical seeing measurements ob-tained with the di ff erential image motion monitor (DIMM) dur-ing the observations, namely, 0 . (cid:48)(cid:48) and 1 . (cid:48)(cid:48) , respectively (Oyaet al. 2016).To test whether the detected ionised outflows are spatially re-solved, we adopted the following procedure. First, we created theflux maps for the [O III] outflow and BLR components. Then wecalculated spaxel-by-spaxel the ratio of the [O III] outflow ([OIII] out ) flux over the BLR flux, and produced the maps reportingthe flux-ratio values across the FOV.The ratio maps obtained for both QSOs are shown in the leftpanels of Fig. 4. The increasing trend in [O III] out -to-BLR ratioswith the distance from the emission peak indicates that the [OIII] outflows are spatially resolved in both QSOs. Moreover, theratio map of HS 0810 + + B. Unlike HS 0810 + + ff ected by significant lens stretching e ff ects. As a consequence,the [O III] out -to-BLR ratio maps shows an isotropic pattern of in-creasing ratio values moving outwards from the centre of imageA (we recall that we detected the [O III] emission only from thisimage, as previously discussed in Sect. 3.2).What we discuss above, based on the use of 2D-ratio maps,can be better appreciated using spatial profiles. We determinedthe spatial profiles of the [O III] outflow and BLR emissions, aswell as of their ratio values and studied their variability withincreasing distance from the peak of the overall emission. InHS 0810 + θ ∼ ◦ ) with a width of five spaxels, alongwhich we calculated the emission spatial profiles. On the con-trary, given the isotropic pattern of the whole emission in SDSSJ1353 + (cid:48)(cid:48) -value, that is, to the value at the peak position ofthe overall emission; in the case of the [O III] outflow and BLRprofiles, the 0 (cid:48)(cid:48) -value corresponds also to their own peak value.This is helpful in further confirming our previous conclusion thatthe [O III] outflows are spatially resolved in both galaxies sincethe [O III] outflow profiles are broader than the respective BLR Article number, page 8 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . S D SS J + H S + Fig. 4.
Spatially resolved ionised outflows in HS 0810 + + Left panels:
Maps of the ratio between the[O III] outflow flux and the BLR flux (both from best-fit models). Coloured pixels refer to a S / N (cid:38) / N (cid:38)
3) on the full [O III] emission line (i.e.narrow + outflow components) for HS 0810 + + + ’, while the dotted white lines indicate the contour levels of the BLR emission at 75%, 50%, and 25% of its peak; for SDSSJ1353 + Right panels:
Normalised intensity profiles along the pseudo-slit (black dotted-dashed linesin the ratio map) and in circular annuli of increasing radius for HS 0810 + + + profiles. A unique exception occurs in SDSS J1353 + (cid:48)(cid:48) from the centre where we can ob-serve a clear bump in the BLR emission profile: this is due tothe flux contribution from image B. Therefore, this does not af-fect our previous conclusion. As a further test, in addition fromthat obtained from our fitting procedure, we determined the [OIII] outflow profile by collapsing the spectral channels in thesubtracted-data cube, including the blue wing of the [O III] lineprofile (4976 − − + + , we focused on the ratio values of the [OIII] outflow flux over the BLR flux. These are plotted in loga-rithmic scale (on the right-hand side of the plots), after havingbeen rescaled to 1 in the central pixel. In this way, we can easilyidentify values higher than 1 as regions producing a significant[O III] outflow emission and, hence, we can determine the spa-tial extent of resolved ionised outflows. The associated errorbarswere computed by propagating the uncertainties on the [O III]and BLR fluxes in the spatial pixels involved. These were com- To be still corrected for lens e ff ects. puted by propagating the error (mostly due to the noise) associ-ated to the spectral channels, over which we integrated to get thetotal flux contained in that spatial pixel. We took the maximumdistance, including solely ratio values not consistent with 1, asboth radius within which to spatially integrate the flux of the [OIII] outflow component, and outflow observed radius ( R obs ) tobe still corrected for the SINFONI-PSF and lensing stretchinge ff ects. To correct for the PSF-smearing, we applied the correc-tion R PSF = (cid:113) R − ( θ res / , where R PSF is the PSF-correctedradius, R obs is the radius observed in the image plane and θ res is our seeing estimate (0.7 (cid:48)(cid:48) and 0.8 (cid:48)(cid:48) for HS 0810 + + + R obs = . (cid:48)(cid:48) , find-ing a total observed flux of (3 . ± . × − erg s − cm − and R PSF ∼ . (cid:48)(cid:48) . In SDSS J1353 + R obs = . (cid:48)(cid:48) and assessed a total observed flux of (8 . ± . × − erg s − cm − and R PSF ∼ . (cid:48)(cid:48) for the [O III] outflow. In Sect. 4.2,we account for the lens e ff ects and estimate the intrinsic extent(by correcting for stretching e ff ects) and unlensed flux (by cor- The estimates for R PSF are here provided with no uncertainty. Weevaluate the error on the outflow intrinsic radius in Sect. 4.2, after cor-recting for the lensing e ff ects. Article number, page 9 of 19 & A proofs: manuscript no. high-z_ufo recting for magnification e ff ects) of the ionised outflows in HS0810 + +
4. Results
The main purpose of this work is to map the kinematics of the[O III] emission, with a primary focus on the outflow compo-nent. Figure 5 shows a global overview of the distribution and thekinematics of the ionised gas resulting from the modelling of the[O III] emission line. The moment-0 (intensity field), moment-1 (velocity field), and moment-2 (dispersion field) maps for thenarrow and the outflow components are shown separately in or-der to better trace their distinct spatial and velocity distributions.All maps have been produced reporting only spatial pixels witha S / N equal or higher than 2 for HS 0810 + + | v | (cid:38)
500 km s − and σ (cid:38)
600 km s − , respec-tively; see moment-1 moment-2 maps in Fig. 5 relative to theoutflow component). Such velocity dispersions are well abovethe values measured in typical star-forming systems at theseredshifts ( σ ∼
100 km s − , e.g. Cresci et al. 2009; Law et al.2009) and, along with the overall blue-shifted motion, they pro-vide clear evidence for large-scale outflows in these galaxies.Moreover, while in HS 0810 + + | v | (cid:46)
50 km s − , σ (cid:46)
200 km s − for HS 0810 + | v | (cid:46)
100 km s − , σ (cid:46)
300 km s − forSDSS J1353 + + σ (cid:46)
300 km s − ), as the line profileis modelled with a unique Gaussian component, given that the[O III] emission line is fainter and the S / N is lower. This couldindicate that the [O III] outflow component is still present, butcannot be isolated from the [O III] narrow component becauseof its faintness and the worse S / N.Similarly to Marasco et al. (2020) and given the complexityof the [O III] line profile across the FOV, we preferred adoptingthe following definitions of velocity and width for the outflowcharacterisation (e.g. see also Harrison et al. 2014; Zakamska &Greene 2014; Cresci et al. 2015; Carniani et al. 2015; Brusa et al.2016), rather than the moment-1 and moment-2 values. The lat-ter are indeed more a ff ected by geometrical projection and dustabsorption e ff ects. In each spatial pixel, we determined the 10thand 90th velocity percentiles ( v and v ) of the overall emissionline profile (i.e. narrow + outflow components if present), as rep-resentative velocities of the approaching and receding outflowcomponents, respectively. The null velocity value correspondsto the systemic velocity peak of the narrow component in thecentral spectrum. From v and v , we computed the line width W defined as v − v . The W width is approximately equal to the full width at half maximum (FWHM) for a Gaussian profile.Maps of v , v and W are shown in Fig. 6.The maps of v show highly blueshfited velocities in mostof the field of HS 0810 + + − − ; in the latter, the outflow region is pref-erentially elongated in the NE-SW direction with highest veloc-ity values (up to − − ) at the SW end of the stronglyblueshifted region. In HS 0810 + v map, where the outflow is seen receding fromus at velocities up to about 1730 km s − along the line of sight.Looking at the W maps, we observe extremely large values(1100 km s − (cid:46) v (cid:46) − ) in the outflow regions, whichis consistent with other z ∼ v map for each QSO, wenote that the shape of the [O III]-outflow moment-1 map reflectsthe bluest region in the v map, suggesting that any additionalGaussian component added to model the wings in the [O III]profile has been correctly classified as outflow component (com-pare also [O III]-outflow moment-2 maps and the respective W maps).We rule out the possibility of alternative scenarios, such asgalactic inflows or a galaxy merger event. In fact, in the few re-ported cases of their detection, galactic inflows have been ob-served mostly in absorption and with quite small bulk veloci-ties ( ∼
200 km s − ) and velocity dispersions (e.g. Bouché et al.2013). Moreover, for the inflowing gas theoretical modellingpredicts a small covering factor (e.g. Steidel et al. 2010), mak-ing its direct observation rare especially at high redshift (e.g.Cresci et al. 2010). Finally, we exclude also the galaxy-mergerscenario since the deep optical images of both HS 0810 + + ff ects. While these are expected not to significantly a ff ectthe observed gas kinematics (hence the outflow velocity), theystrongly alter the observed gas spatial distribution and the ob-served surface brightness: fluxes are magnified and spatial di-mensions are stretched. Therefore, the obtained maps could notbe used to infer directly the outflow intrinsic radius and its totalflux, which are key ingredients, along with the outflow velocity,in the computation of the outflow energetics. We first need toquantify the lensing e ff ects and then we can derive the unlensedphysical properties of the outflow. This aspect is discussed inSect. 4.2 (see also Appendix A). As we discuss in Sect. 4.1, we performed both the spectral anal-ysis and the kinematical study in the lens plane. For this reason,we had to correct for the lens e ff ects to determine the actual ex-tent and flux of the outflow.There are several adaptive-mesh fitting codes which, givena mass distribution for the lens and a surface brightness profilefor the background source, fit the resulting forward lensed im-age to the observed data and use a statistics test (e.g. the min-imum χ method) to establish the best-fit models for both thelens and the source. These algorithms usually require the knowl-edge of the instrumental PSF to allow a correct comparison with Article number, page 10 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . [OIII]-full line [OIII]-narrow [OIII]-outflow HS 0810+2554 m o m e n t m o m e n t m o m e n t [OIII]-full line [OIII]-narrow [OIII]-outflow SDSS J1353+1138
Fig. 5.
Moment-0 (intensity field), moment-1 (velocity field), and moment-2 (dispersion field) maps of the [O III] line emission in HS 0810 + left ) and in SDSS J1353 + right ). The maps for the total, narrow and outflow components are shown separately, reporting only spatial pixelswith a S / N equal or higher than 2 for HS 0810 + + + ’ indicates the emission centroid in eachQSO, while the dotted lines represent the contour levels of the BLR emission at 75%, 50%, and 25% of its peak. v10% v90% W80% H S + S D SS J + Fig. 6. v , v , and W maps of the [O III] emission line in HS0810 + + / N as in the moment maps of Fig. 5, thatis, with a S / N equal or higher than 2 and for HS 0810 + + the observed data. The output of these fitting-codes is a 2D or3D reconstruction (depending on the code used) of the unlensed source. In order to achieve an accurate reconstruction, it is re-quired that the lensed images are all detected and spatially re-solved , as their position depends on the first derivative of thegravitational potential of the lens, while their flux on the secondderivative (e.g. Jackson et al. 2015; Nierenberg et al. 2020). Inother words, the knowledge of the position of the multiple lensedimages and of their flux provides strong constraints on the lensand background source models.Unfortunately, we could not use such fitting codes to fullyreconstruct the unlensed outflow in the source plane for HS0810 + + + µ out = . ± . R out = (8 . ± .
7) kpc, with z = . ± .
002 as theredshift measured from the nuclear spectrum extracted for theBLR-modelling (described in Sect. 3.1) and, here, adopted toconvert the outflow angular size into kpc units. Our z measure-ment is consistent with the centroids of the ALMA CO(J = → = →
1) emission lines of HS 0810 + µ out the observed [O III]outflow flux (determined in Sect. 3.3), we found the unlensedoutflow flux to be F out = (1 . ± . × − (2 . /µ out ) erg s − cm − . In addition or alternatively to single lensed images, fitting codes han-dle also lensed arcs. Article number, page 11 of 19 & A proofs: manuscript no. high-z_ufo
Our µ out ∼ + ff ers (up totwo orders of magnitude) from the values from the literature,presented in Sect. 2.1. This follows from the fact that the latterare estimates of the magnification of the emission from the morecompact (UV disk and X-ray corona) region, while our estimateis relative to a large-scale ( ∼ + + i and K -band, respectively. Our assumptions rely on the fact that inSDSS J1353 + ff ects are supposed tobe smaller than in HS 0810 + + ff erential magnification.On the basis of the first argument, we neglected the stretchinglens e ff ect and approximated the unlensed outflow angular sizeto R PSF = . (cid:48)(cid:48) ± . (cid:48)(cid:48) (determined in Sect. 3.3). Consider-ing instead the lens magnification, we expect total magnificationfactors of a few units that is weakly dependent on the geomet-rical details of the flux distribution for background emissionswith comparable spatial extent. Consequently, we used the av-erage between the i -band (i.e. µ = .
81 and µ = .
75; Inadaet al. 2006) and K -band total magnification factors (i.e. µ = . µ = .
42 and µ = .
53; Rusu et al. 2016) as a proxy for the totaloutflow magnification µ out , under the assumption of comparableunlensed physical sizes. Given the unknown real unlensed fluxdistribution of the J -band outflow, we conservatively assumedan uncertainty of 10% on our adopted µ out value, thus obtain-ing µ out = . ± .
4. Correcting, finally, for the lens magnifi-cation and converting to kpc-units, we found the unlensed fluxfor the outflow to be F out = (2 . ± . × − (3 . /µ out ) ergs − cm − and its intrinsic radius to be R out = (9 . ± .
1) kpc,using z = . ± .
002 as measured from the nuclear spectraextracted during the BLR-modelling (described in Sect. 3.2).In Table 2, we summarise the main outflow properties for HS0810 + + v , v and W maps (described in Sect. 4.1), respec-tively, referred to as v max10 , v max90 and W max80 . Then we report ourlens-corrected estimates of R out and F out (inferred as discussedabove), the latter corrected for the outflow magnification factors, µ out , shown in the last column. Most of these physical quantitiesare also employed in the computation of the outflow energeticsin Sect. 4.3. We derived the physical properties of the large-scale ionised out-flows in HS 0810 + + λ L [OIII] = (cid:90) V (cid:15) [OIII] f dV , (2)where V is the volume occupied by the ionised outflow, f is thefilling factor of the [O III] emitting clouds in the outflow, and (cid:15) [OIII] is the [O III] line emissivity which, at the temperaturestypical of the NLR ( ∼ K), is weakly dependent on the tem-perature ( ∝ T . ) and can be written as: (cid:15) [OIII] = . × − E [OIII] n O + n e erg s − cm − , (3)with E [OIII] as the energy of the [O III] photons, n O + and n e ,the volume densities of the O + ions, and of the electrons, re-spectively. Then assuming that most of the oxygen in the ionisedoutflow is in the form of O + , it follows that: (cid:15) [OIII] ≈ × − E [OIII] n [O / H] erg s − cm − , (4)where [O / H] gives the oxygen abundance in solar units. Themass of the outflowing ionised gas can be derived from the fol-lowing expression: M out = (cid:90) V . m H n e f dV , (5)where m H is the mass of the hydrogen atom and the factor of1.27 follows from including the mass contribution of helium. Bycombining Eqs. (2) and (5), we get: M out = . × (cid:18) L [OIII] erg s − (cid:19) (cid:18) (cid:104) n e (cid:105) cm − (cid:19) − C [O / H] M (cid:12) , (6)where (cid:104) n e (cid:105) is the electron density averaged over the ionised out-flow volume (i.e. (cid:104) n e (cid:105) = (cid:82) V n e f dV / (cid:82) V f dV ) and C = (cid:104) n e (cid:105) / (cid:104) n (cid:105) is the so-called ‘condensation factor’. Under the simplifying hy-pothesis that all ionising gas clouds have the same density, weget C = Ω and a radial extent, R out , and that it consists of a collection of ionised gas clouds, uni-formly distributed within the cone and outflowing with a speed v out . The mass outflow rate is given by:˙ M out = (cid:104) ρ (cid:105) v out Ω R , (7)where (cid:104) ρ (cid:105) is the average mass density computed over the totalvolume V occupied by the conical outflow . By substituting (cid:104) ρ (cid:105) in Eq. (7) with its definition in terms of ˙ M out (using Eq. 6) and V , we obtain: ˙ M out = (cid:18) L [OIII] erg s − (cid:19)(cid:18) (cid:104) n e (cid:105) cm − (cid:19) − (cid:18) v out km s − (cid:19)(cid:18) R out kpc (cid:19) − (cid:12) yr − [O / H] (8)where we have assumed C =
1. The mass outflow rate thus in-ferred does not depend on either the opening angle Ω of the out-flow cone or the filling factor f of the emitting clouds. We note that unless f =
1, in general (cid:104) ρ (cid:105) (cid:44) . m H (cid:104) n e (cid:105) , with (cid:104) n e (cid:105) averaged over the volume occupied by the emitting clouds and not overthe whole conical volume.Article number, page 12 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . QSO v max10 v max90 W max80 R out F out µ out km s − km s − km s − kpc erg s − cm − HS 0810 + − ±
70 1730 ±
60 3360 ±
110 8 . ± . . ± . × − . ± . + − ±
80 2270 ±
130 3850 ±
90 9 . ± . . ± . × − . ± . Table 2:
Directly measured properties of the [O III] outflows in HS 0810 + + v , v and W maps( v max10 , v max90 and W max80 , respectively), intrinsic outflow radius ( R out ), unlensed [O III] outflow flux ( F out ) corrected for the outflow magnification factor( µ out ), reported in the last column. Finally, we calculate the kinetic energy ( E kin ), kinetic lumi-nosity ( L kin ) and momentum rate ( ˙ p out ) of the outflow by meansof the following expressions: E kin = . × (cid:18) M out M (cid:12) (cid:19) (cid:18) v out km s − (cid:19) erg , (9) L kin = . × (cid:18) ˙ M out M (cid:12) yr − (cid:19) (cid:18) v out km s − (cid:19) erg s − , (10)˙ p out = . × (cid:18) ˙ M out M (cid:12) yr − (cid:19) (cid:18) v out km s − (cid:19) dyne . (11)Equations (6)-(11) require the knowledge of di ff erent physicalproperties of the outflow, some of which we were able to derive,while others had to be assumed. These are: the oxygen abun-dance, which we fixed to the solar abundance, and the electrondensity, which we assumed to be n e ∼ − , in agreementwith the values measured in similar studies at high redshift (seee.g. Perna et al. 2017; Förster Schreiber et al. 2019). The sec-ond assumption, in particular, a ff ects the derived outflow ener-getics (Davies et al. 2020; Kakkad et al. 2020) but it is necessary,nonetheless, since we cannot measure n e directly from our data.We now focus on the physical quantities we were able tocalculate. In Sect. 4.2, we provided the values of the intrinsicradius R out and flux F out of the ionised outflows, traced by the[O III] line emission. From F out we calculated the intrinsic [OIII] line luminosity used in the mass outflow expression (Eq. 6).We found ( µ out -corrected) [O III] luminosity values of L [OIII] = (2 . ± . × erg s − and L [OIII] = (4 . ± . × erg s − for HS 0810 + + v and v , shown in Sect.4.1. The spectral analysis and the study of the gas kinematicsin HS 0810 + + v . Moreover, in HS 0810 + v out = max ( | v max10 − v sys | , | v max90 − v sys | ) , (12)where v max10 and v max90 are the maximum value, respectively, ob-served in the v and v maps (described in Sect. 4.1), and v sys is the bulk (or systemic) velocity of the galaxy, inferred from thenuclear spectrum used for the BLR-fitting (in Sect. 3.1) and setto the value of 0 km s − , as previously described in Sect. 4.1. This definition is required by the unknown geometry and orien-tation of the outflow with respect to the line of sight: since weignore the true angle of the outflow with respect to the line ofsight, and given that the bulk of the outflow unlikely points to-wards the observer, we assume that the best representation of theoutflow speed is provided by the velocity ‘tail’ of the line pro-file, that is, v and v in Eq. (12). These values are thought tobe more suited to represent v out than the mean (or median) veloc-ity of the line, which strongly depends on projection e ff ects anddust absorption (e.g. Cresci et al. 2015).The same argument holds also for the determination of theoutflow velocity in SDSS J1353 + ff erent orientation with respect to the ob-server and higher dust absorption, we did not observe any asym-metric red wing in the [O III] profile produced by the recedingpart of the outflow, as stressed in Sect. 4.1. Therefore, for SDSSJ1353 + v out = v max10 as out-flow velocity for SDSS J1353 + R out , L [OIII] and v out , and a typical uncertainty of 50%on n e (e.g. Perna et al. 2017; Förster Schreiber et al. 2019).The physical properties of the ionised outflows detected in HS0810 + + ffi ciency and of themomentum-boost. The former is defined as the ratio betweenthe outflow kinetic luminosity, L kin , (defined in Eq. 10) and theAGN bolometric luminosity, L Bol , (corrected for the lens magni-fication), while the latter is defined as the ratio between the mo-mentum rate of the outflow ( ˙ p out ) and the momentum initiallyprovided by the AGN-radiation pressure (i.e. L Bol / c ), which isapproximately identified also with the momentum rate of theX-ray UFO. The values of L Bol adopted in this work are for(2 . ± . × erg s − and (39 ± × erg s − HS 0810 + + ∼ (cid:12) yr − and ∼
12 M (cid:12) yr − ,and kinetic e ffi ciencies of ∼ × − and ∼ × − forSDSS J1353 + + + L Bol ∼ × erg s − of the ˙ M out − L Bol and L kin − L Bol scaling relations (Carniani et al. 2015; Fiore et al.2017) for the ionised outflow component. On the contrary, theinferred ˙ M out value for SDSS J1353 + ∼ L Bol ∼ × erg s − . Our findings are howeverrelative only to the ionised phase of large-scale outflows, while asignificant amount of outflowing gas may be in neutral molecu-lar and / or atomic phase. Given the typical bolometric luminosi-ties of our galaxies ( L Bol ∼ − erg s − ), outflow massrates predicted for the molecular component may indeed exceedour measurements for the ionised gas by a factor ∼ Article number, page 13 of 19 & A proofs: manuscript no. high-z_ufo et al. (2020) have recently claimed the tentative detection of amassive ( M out , mol ∼ × M (cid:12) ) CO-molecular outflow in HS0810 + ∼
400 M (cid:12) yr − and 1040 km s − , respectively. Compared to molecular outflows,which have been observed in z >
5. Discussion
The main purpose of this work is to shed light on the accelerationmechanism of ionised outflows on large scales. In this regard, wecompared the energetics of the galaxy-scale ionised outflows tothe UFOs present at nuclear scales, in order to test whether theyare causally connected (i.e. whether they are subsequent phasesof the same AGN-accretion burst). In the case of a causal con-nection, we could go on to investigate the nature of the acceler-ation mechanism distinguishing between momentum-driven andenergy-driven winds.We show the energetics measurements of the large scaleionised outflows in HS 0810 + + + + Chandra observation acquiredin 2016 using the updated version of the photoionisation code
XSTAR (Bautista & Kallman 2001).As done in Marasco et al. (2020), we followed Nardini &Zubovas (2018) in order to re-compute the UFOs energeticsin a consistent way based on the same assumption we madein the calculation of the large-scale outflow energetics. We as-sumed that the UFO is launched from the escape radius r esc ≡ GM BH /v of the BH and we derived the mass outflow ratefor the nuclear wind as:˙ M UFO (cid:39) . (cid:18) Ω π (cid:19)(cid:18) N H cm − (cid:19)(cid:18) M BH M (cid:12) yr − (cid:19)(cid:18) v UFO c (cid:19) − M (cid:12) yr − , (13)where Ω is the solid angle subtended by the UFO, M BH theblack hole mass, N H the hydrogen gas column density, and v UFO the wind speed. We took the values of v UFO and N H inferredby Chartas et al. (2021). For HS 0810 + ff er from those previously pub-lished (i.e. v UFO , = . + . − . c , N H , = . + . − . × cm − and v UFO , = . + . − . c , N H , = . + . − . × cm − for the twodistinct UFO components; Chartas et al. 2016), but are still inagreement. The adopted M BH values are virial estimates based e n e r gy - d r i v e n m o m e n t u m - d r i v e n ■ ● ■ ● Fig. 7.
Momentum-boost versus wind velocity diagram for both UFOs(represented as circles) and galactic outflow components (representedas squares) of HS 0810 + + + + molecular large-scale outflow is shown,identified by the star symbol. on H β for HS 0810 + + M BH for SDSS J1353 + f = Ω π = . + . − . , on the basis that about 40%of local AGNs have been observed to host UFOs (Tombesi et al.2010; Go ff ord et al. 2013). With all these ingredients, we calcu-lated the mass rate ˙ M UFO (with Eq. 13), the momentum rate ˙ p UFO (i.e. ˙ p UFO = ˙ M UFO v UFO ) and the momentum-boost of the nuclearwinds defined as ˙ p UFO / ( L Bol / c ). As uncertainty on all quantities,the minimum-maximum range of possible values was consid-ered, considering the values of v UFO , N H and M BH as inferred inChartas et al. (2021). For L Bol we took the average between the µ -corrected values obtained in Chartas et al. (2021) with two dif-ferent methods: the 2-10 keV bolometric correction (Lusso et al.2012) and the estimate from the continuum luminosity at 1450Å (Assef et al. 2011; Runnoe et al. 2012).Table 4 summarises the physical properties of the UFOsin HS 0810 + + + M UFO and ˙ p UFO / ( L Bol / c ) refer to thewhole hosted UFO, as sum of the two detected UFO compo-nents at di ff erent speeds (for which we separately report N H and v UFO in Table 4), originally discovered in Chartas et al. (2016)and confirmed in Chartas et al. (2021).Figure 7 shows the momentum-boost as a function of the out-flow speed for HS 0810 + + Article number, page 14 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . QSO v out M out ˙ M out E kin L kin / L Bol ˙ p out km s − M (cid:12) M (cid:12) yr − erg 10 − L Bol / c HS 0810 + ±
70 15 ± ± ± ±
500 1 . ± . + ±
80 2 . ± . . ± . . ± . ± . ± . Table 3:
Derived properties of the ionised outflow in HS 0810 + + ffi ciencyand momentum-boost. The values of L Bol are shown in Table 4.
QSO log( M BH ) a L Bol b N H v UFO f ˙ M UFO ˙ p UFO M (cid:12) erg s − cm − c M (cid:12) yr − L Bol / c HS 0810 + . + . − . . ± . . + . − . . + . − . . ± . . + . − . . + . − . . + . − . . + . − . SDSS J1353 + . + . − . ± . + . − . . + . − . . ± . + − . + . − . Table 4:
Physical properties of the UFOs in HS 0810 + + M BH values are virial estimates based on H β in HS 0810 + + M BH in SDSS J1353 + L Bol are corrected for the lens magnification and computed as average of the two independent estimates of the AGN bolometric luminosity,obtained through the 2-10 keV bolometric correction (Lusso et al. 2012) and from the continuum luminosity at 1450 Å (Assef et al. 2011; Runnoeet al. 2012). we observe that while the [O III] outflow of HS 0810 + + ∼
100 than the predictions for a momentum-drivenpropagation, therefore, it is far from any tentative connectionwith the UFOs on nuclear scale. We reiterate that our measure-ment is a lower limit, as we did not detect [O III] emission fromthe second lensed image and, moreover, we approximated theoutflow intrinsic radius to the one observed (just PSF-corrected).Nevertheless, even accounting for all these issues and approxi-mations we made, such a discrepancy between observations andtheoretical predictions could hardly be explained, as it amountsto about two orders of magnitude. Even associating half of the [OIII] flux observed in image A to image B, on the basis of the fluxratio measurement of the two H -band images (Inada et al. 2006),we would obtain a momentum-boost larger than the previous oneby only a factor ∼ .
5. Similarly, it is unlikely that we are under-estimating the stretching e ff ects so much as to overestimate R out by a factor of ∼ ff erent origin. The more plau-sible hypotheses are: 1) either the likely presence of a massivemolecular outflow in this galaxy that our work is not account-ing for; or 2) the possibility that the observed UFO is causedby an extraordinary burst episode (see e.g. Zubovas & Nardini2020) in the BH accretion activity of SDSS J1353 + Γ ∼ .
2, which is typical of Narrow-line Seyfert1 galaxies (Leighly 1999; Vaughan et al. 1999) and a typicalsignature of highly accreting systems (Huang et al. 2019). TheEddington ratio λ Edd (defined as L Bol / L Edd ) estimated for SDSSJ1353 + λ Edd = . ± .
02 (Chartas et al. 2021), whichis larger by a factor ∼ + λ Edd = . ± .
03; Chartas et al. 2021). Such results couldsupport the recent post-burst scenario. Certainly, each hypothe-sis does not necessarily exclude the other and the real situationcan be a combination of the two (e ff ects). The small values of momentum-boost ( ∼ . −
2) and ofkinetic e ffi ciency ( ∼ − × − ), inferred for the ionisedoutflows in SDSS J1353 + + ff ect: whilethe sources accreting at high rates (close to the Eddington limit)are actually the most promising candidates for hosting an ac-tive UFO (e.g. Nardini et al. 2019), they usually present a verybright Fe II emission and a faint, outshined [O III] emission. Asa consequence, the [O III] may be not ideal to trace the ionisedphase of the outflow in AGNs accreting at high rates since thebulk of the ionised gas could be in the form of di ff erent chemicalspecies.It is also possible we underestimated the uncertainty of the[O III] outflow in HS 0810 + p out , tot / ( L Bol / c ) ∼ v out , tot ∼ − ). Giventhe two orders of magnitude of di ff erence between ionised andmolecular outflow masses, the mass-weighted velocity is essen-tially the molecular outflow velocity ( v out , tot = − ;Chartas et al. 2020).In Fig. 7, we report the CO + [O III] point with its uncer-tainty. Once the contribution of the molecular component is in-cluded, the energetics of the overall large-scale outflow in HS0810 + Article number, page 15 of 19 & A proofs: manuscript no. high-z_ufo deeper observations are required to confirm the CO-outflow de-tection and to constrain its energetics.
Our study has revealed that the energetics of the galaxy-scaleionised outflow in HS 0810 + + + z ∼ . + + M BH , N H , and v UFO . All measurements for molecular outflows have been re-scaled to the same luminosity-to-mass conversion factor α CO = . − pc ) − M (cid:12) , typical of QSOs, starburst and submil-limeter galaxies (Downes & Solomon 1998; Bolatto et al. 2013;Carilli & Walter 2013). The main AGN, X-ray-wind, and large-scale outflow properties of the known QSO-sample are listed inTable B.1 in Marasco et al. (2020).Figure 8 represents the updated version of Fig. 9 in Marascoet al. (2020), including the measurements relative to HS0810 + + p out and ˙ p UFO , respectively), ordered according to the increasing L Bol of the host AGN. This is an alternative way (with regard toFig. 7) to compare observational results with theoretical pre-dictions for each QSO and to distinguish between the two mainregimes. The energetics of the large-scale outflow in 10 out of 12QSOs results to be consistent (or nearly consistent) with eithera momentum-driven or an energy-driven regime: in seven (six)and three (four) objects, respectively, if we exclude (include)the contribution of the tentatively detected molecular outflowin HS 0810 + + + p out / ˙ p UFO in SDSSJ1353 + + p out / ˙ p UFO inferred for IRAS 17020 + + λ Edd ∼ .
7; Longinotti et al. 2015). Finally, we donot observe any remarkable trend in ˙ p out / ˙ p UFO values with L Bol ,nor any evident dependence of wind acceleration mechanism ongalaxy redshift upon separately inspecting the results obtainedfor the high-redshift (our two objects plus APM 08279 +
6. Conclusions
Galaxy-wide outflows powered by AGN activity are thought toplay a fundamental role in shaping the evolution of galaxies, asthey allow us to reconcile theoretical models to observations.However, even though observations have widely confirmed theirpresence in both local and high-redshift galaxies, a clear un-derstanding of the mechanism which accelerates these power-ful, galaxy-scale winds is still lacking. To test the predictions ofthe current theoretical models, we need to compare, in a givenobject, the energetics of the sub-pc wind with that of the galaxy-wide outflow. The optimal sources in attempting the make a con-nection between di ff erent scales are the powerful QSOs near thepeak of AGN activity ( z ∼ ff ective.Given such a perspective, this work focuses on two z ∼ . + + + z ∼ . + + λλ ∼ v ∼ − in the image lens plane.2. After correcting for the gravitational lensing e ff ects, wefound that the ionised outflow in HS 0810 + + Article number, page 16 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . Fig. 8.
Ratio between the galaxy-scale and sub-pc scale outflow momentum rates for di ff erent QSOs hosting UFOs. Measurements for individualobjects are shown in blue with the respective errorbars, using di ff erent markers according to the gas phase of the observed large scale outflow. Thegalaxy points are ordered by increasing L Bol . The horizontal dashed line shows the prediction for a momentum-driven wind ( ˙ p out / ˙ p UFO = α and [O III] emission, respectively. For HS 0810 + to be unrelated to the nuclear scale energetics, likely requir-ing either the presence of a massive molecular outflow or ahigh variability among the QSO activity.3. By comparing our inferred results with those of the smallsample of known QSOs from the literature, each hosting bothsub-pc scale UFOs and neutral or ionised winds on galaxyscales, we found that the momentum- and energy-drivenframeworks describe all the observed targets very well, withthe exception of SDSS J1353 + + + ff erent scales observed in SDSSJ1353 + Acknowledgments.
We thank the anonymous referee for com-ments and suggestions, which have improved the paper. We acknowl-edge support from the Italian Ministry for University and Research(MUR) for the BLACKOUT project funded through grant PRIN2017PH3WAT.003. MP is supported by the Programa Atracción deTalento de la Comunidad de Madrid via grant 2018-T2 / TIC-11715.MP acknowledges support from the Spanish Ministerio de Economíay Competitividad through the grant ESP2017-83197-P, and PID2019-106280GB-I00. Some data shown in this work were obtained from theMikulski Archive for Space Telescopes (MAST).
References
Aird, J., Coil, A. L., Georgakakis, A., et al. 2015, MNRAS, 451, 1892Aird, J., Nandra, K., Laird, E. S., et al. 2010, MNRAS, 401, 2531Assef, R. J., Denney, K. D., Kochanek, C. S., et al. 2011, ApJ, 742, 93Bae, H.-J. & Woo, J.-H. 2018, ApJ, 853, 185Bautista, M. A. & Kallman, T. R. 2001, ApJS, 134, 139Bayliss, M. B., Sharon, K., Acharyya, A., et al. 2017, ApJ, 845, L14Bolatto, A. D., Wolfire, M., & Leroy, A. K. 2013, ARA&A, 51, 207Boroson, T. A. & Green, R. F. 1992, ApJS, 80, 109Bouché, N., Murphy, M. T., Kacprzak, G. G., et al. 2013, Science, 341, 50Brusa, M., Cresci, G., Daddi, E., et al. 2018, A&A, 612, A29 Brusa, M., Perna, M., Cresci, G., et al. 2016, A&A, 588, A58Cano-Díaz, M., Maiolino, R., Marconi, A., et al. 2012, A&A, 537, L8Cappellari, M. 2017, MNRAS, 466, 798Carilli, C. L. & Walter, F. 2013, ARA&A, 51, 105Carniani, S., Marconi, A., Maiolino, R., et al. 2015, A&A, 580, A102Carniani, S., Marconi, A., Maiolino, R., et al. 2016, A&A, 591, A28Carniani, S., Marconi, A., Maiolino, R., et al. 2017, A&A, 605, A105Cazzoli, S., Arribas, S., Maiolino, R., & Colina, L. 2016, A&A, 590, A125Chartas, G., Brandt, W. N., & Gallagher, S. C. 2003, ApJ, 595, 85Chartas, G., Brandt, W. N., Gallagher, S. C., & Garmire, G. P. 2002, ApJ, 579,169Chartas, G., Cappi, M., Hamann, F., et al. 2016, ApJ, 824, 53Chartas, G., Cappi, M., Vignali, C., et al. 2021, ApJ (submitted)Chartas, G., Charlton, J., Eracleous, M., et al. 2009, New A Rev., 53, 128Chartas, G., Davidson, E., Brusa, M., et al. 2020, MNRAS, 496, 598Chartas, G., Eracleous, M., Dai, X., Agol, E., & Gallagher, S. 2007, ApJ, 661,678Chartas, G., Hamann, F., Eracleous, M., et al. 2014, ApJ, 783, 57Cicone, C., Brusa, M., Ramos Almeida, C., et al. 2018, Nature Astronomy, 2,176Cicone, C., Feruglio, C., Maiolino, R., et al. 2012, A&A, 543, A99Cicone, C., Maiolino, R., Sturm, E., et al. 2014, A&A, 562, A21Ciotti, L., Ostriker, J. P., & Proga, D. 2010, ApJ, 717, 708Coatman, L., Hewett, P. C., Banerji, M., et al. 2017, MNRAS, 465, 2120Concas, A., Popesso, P., Brusa, M., et al. 2017, A&A, 606, A36Costa, T., Sijacki, D., & Haehnelt, M. G. 2014, MNRAS, 444, 2355Cresci, G., Hicks, E. K. S., Genzel, R., et al. 2009, ApJ, 697, 115Cresci, G., Mainieri, V., Brusa, M., et al. 2015, ApJ, 799, 82Cresci, G. & Maiolino, R. 2018, Nature Astronomy, 2, 179Cresci, G., Mannucci, F., Maiolino, R., et al. 2010, Nature, 467, 811Dadina, M., Vignali, C., Cappi, M., et al. 2018, A&A, 610, L13Davies, R., Baron, D., Shimizu, T., et al. 2020, MNRAS, 498, 4150Di Matteo, T., Springel, V., & Hernquist, L. 2005, Nature, 433, 604Downes, D. & Solomon, P. M. 1998, ApJ, 507, 615Eisenhauer, F., Abuter, R., Bickert, K., et al. 2003, in Society of Photo-OpticalInstrumentation Engineers (SPIE) Conference Series, Vol. 4841, InstrumentDesign and Performance for Optical / Infrared Ground-based Telescopes, ed.M. Iye & A. F. M. Moorwood, 1548–1561Faber, S. M. & Jackson, R. E. 1976, ApJ, 204, 668Fabian, A. C. 2012, ARA&A, 50, 455Ferland, G. J., Hu, C., Wang, J.-M., et al. 2009, ApJ, 707, L82Ferrarese, L. & Merritt, D. 2000, ApJ, 539, L9Feruglio, C., Ferrara, A., Bischetti, M., et al. 2017, A&A, 608, A30Feruglio, C., Fiore, F., Carniani, S., et al. 2015, A&A, 583, A99Feruglio, C., Maiolino, R., Piconcelli, E., et al. 2010, A&A, 518, L155Fiore, F., Feruglio, C., Shankar, F., et al. 2017, A&A, 601, A143Fluetsch, A., Maiolino, R., Carniani, S., et al. 2020, arXiv e-prints,arXiv:2006.13232Fluetsch, A., Maiolino, R., Carniani, S., et al. 2019, MNRAS, 483, 4586Förster Schreiber, N. M., Übler, H., Davies, R. L., et al. 2019, ApJ, 875, 21Fruchter, A. S. & Hook, R. N. 2019, Drizzle: A Method for the Linear Recon-struction of Undersampled Images (PASP), 114–144Go ff ord, J., Reeves, J. N., McLaughlin, D. E., et al. 2015, MNRAS, 451, 4169Go ff ord, J., Reeves, J. N., Tombesi, F., et al. 2013, MNRAS, 430, 60Gonzaga, S., Hack, W., Fruchter, A., & Mack, J. 2012, The DrizzlePac Handbook(Baltimore, STScI) Article number, page 17 of 19 & A proofs: manuscript no. high-z_ufo
Granato, G. L., De Zotti, G., Silva, L., Bressan, A., & Danese, L. 2004, ApJ,600, 580Guo, Q., Gonzalez-Perez, V., Guo, Q., et al. 2016, MNRAS, 461, 3457Harrison, C. M., Alexander, D. M., Mullaney, J. R., & Swinbank, A. M. 2014,MNRAS, 441, 3306Harrison, C. M., Costa, T., Tadhunter, C. N., et al. 2018, Nature Astronomy, 2,198Hasinger, G., Schartel, N., & Komossa, S. 2002, ApJ, 573, L77Huang, Y.-K., Hu, C., Zhao, Y.-L., et al. 2019, ApJ, 876, 102Inada, N., Oguri, M., Morokuma, T., et al. 2006, ApJ, 653, L97Jackson, N., Tagore, A. S., Roberts, C., et al. 2015, MNRAS, 454, 287Kakkad, D., Mainieri, V., Vietri, G., et al. 2020, A&A, 642, A147Kaspi, S., Smith, P. S., Netzer, H., et al. 2000, ApJ, 533, 631King, A. 2003, ApJ, 596, L27King, A. 2005, ApJ, 635, L121King, A. & Pounds, K. 2015, ARA&A, 53, 115King, A. R. 2010a, MNRAS, 408, L95King, A. R. 2010b, MNRAS, 402, 1516Kormendy, J. & Ho, L. C. 2013, ARA&A, 51, 511Kovaˇcevi´c, J., Popovi´c, L. ˇC., & Dimitrijevi´c, M. S. 2010, ApJS, 189, 15Lanzuisi, G., Giustini, M., Cappi, M., et al. 2012, A&A, 544, A2Law, D. R., Steidel, C. C., Erb, D. K., et al. 2009, ApJ, 697, 2057Leighly, K. M. 1999, ApJS, 125, 317Longinotti, A. L., Krongold, Y., Guainazzi, M., et al. 2015, ApJ, 813, L39Ludwig, R. R., Wills, B., Greene, J. E., & Robinson, E. L. 2009, ApJ, 706, 995Lusso, E., Comastri, A., Simmons, B. D., et al. 2012, MNRAS, 425, 623Madau, P. & Dickinson, M. 2014, ARA&A, 52, 415Madau, P., Ferguson, H. C., Dickinson, M. E., et al. 1996, MNRAS, 283, 1388Maiolino, R., Gallerani, S., Neri, R., et al. 2012, MNRAS, 425, L66Marasco, A., Cresci, G., Nardini, E., et al. 2020, A&A, 644, A15Marconi, A., Risaliti, G., Gilli, R., et al. 2004, MNRAS, 351, 169Mosquera, A. M. & Kochanek, C. S. 2011, ApJ, 738, 96Nagao, T., Marconi, A., & Maiolino, R. 2006, A&A, 447, 157Narayan, R. & Bartelmann, M. 1996, arXiv e-prints, astroNardini, E., Lusso, E., & Bisogni, S. 2019, MNRAS, 482, L134Nardini, E., Reeves, J. N., Go ff ord, J., et al. 2015, Science, 347, 860Nardini, E. & Zubovas, K. 2018, MNRAS, 478, 2274Nelson, D., Pillepich, A., Springel, V., et al. 2019, MNRAS, 490, 3234Nierenberg, A. M., Gilman, D., Treu, T., et al. 2020, MNRAS, 492, 5314Osterbrock, D. E. 1981, ApJ, 249, 462Oya, S., Terada, H., Hayano, Y., et al. 2016, Experimental Astronomy, 42, 85Peng, C. Y., Impey, C. D., Rix, H.-W., et al. 2006, ApJ, 649, 616Perna, M., Brusa, M., Cresci, G., et al. 2015, A&A, 574, A82Perna, M., Lanzuisi, G., Brusa, M., Cresci, G., & Mignoli, M. 2017, A&A, 606,A96Pontzen, A., Tremmel, M., Roth, N., et al. 2017, MNRAS, 465, 547Reimers, D., Hagen, H. J., Baade, R., Lopez, S., & Tytler, D. 2002, A&A, 382,L26Rizzo, F., Vegetti, S., Fraternali, F., & Di Teodoro, E. 2018, MNRAS, 481, 5606Ross, N. R., Assef, R. J., Kochanek, C. S., Falco, E., & Poindexter, S. D. 2009,ApJ, 702, 472Runnoe, J. C., Brotherton, M. S., & Shang, Z. 2012, MNRAS, 422, 478Rupke, D. S., Veilleux, S., & Sanders, D. B. 2005, ApJS, 160, 115Rusu, C. E., Oguri, M., Minowa, Y., et al. 2016, MNRAS, 458, 2Silk, J. & Rees, M. J. 1998, A&A, 331, L1Spingola, C., McKean, J. P., Vegetti, S., et al. 2020, MNRAS, 495, 2387Stacey, H. R., McKean, J. P., Powell, D. M., et al. 2021, MNRAS, 500, 3667Steidel, C. C., Erb, D. K., Shapley, A. E., et al. 2010, ApJ, 717, 289Sulentic, J. W., Zwitter, T., Marziani, P., & Dultzin-Hacyan, D. 2000, ApJ, 536,L5Tombesi, F., Cappi, M., Reeves, J. N., & Braito, V. 2012, MNRAS, 422, L1Tombesi, F., Cappi, M., Reeves, J. N., et al. 2013, MNRAS, 430, 1102Tombesi, F., Cappi, M., Reeves, J. N., et al. 2011, ApJ, 742, 44Tombesi, F., Cappi, M., Reeves, J. N., et al. 2010, A&A, 521, A57Vaughan, S., Reeves, J., Warwick, R., & Edelson, R. 1999, MNRAS, 309, 113Vegetti, S. & Koopmans, L. V. E. 2009, MNRAS, 392, 945Vignali, C., Iwasawa, K., Comastri, A., et al. 2015, A&A, 583, A141Woo, J.-H., Son, D., & Bae, H.-J. 2017, ApJ, 839, 120Zakamska, N. L. & Greene, J. E. 2014, MNRAS, 442, 784Zubovas, K. & King, A. 2012, ApJ, 745, L34Zubovas, K. & King, A. R. 2014, MNRAS, 439, 400Zubovas, K. & Nardini, E. 2020, MNRAS, 498, 3633 Article number, page 18 of 19. Tozzi et al.: Connecting X-ray nuclear winds with galaxy-scale ionised outflows in two z ∼ . Appendix A: Approximated reconstruction of theunlensed outflow in HS 0810+2554
As mentioned in Sect. 4.2, for HS 0810 + ff er-ing in radius and aperture angle, while the position angle of thecone was fixed to θ ∼ ◦ , namely, the direction in which weobserved the major [O III] outflow emission in the lens plane(see Sect. 3.3). In order to reduce further the number of free pa-rameters characterising the outflow geometry, we fixed the originof all simulated cones to the position of the emission centroid ofthe unlensed outflow, obtained through a first tentative full 2D-reconstruction with the lens fitting-code by Rizzo et al. (2018)and a lens model fixed to that found in Nierenberg et al. (2020).In fact, even though it could not be used to establish the intrinsicextent of the [O III] outflow because of the presence of residuale ff ects from the PSF-deconvolution, it still provided reliable con-straints of the centre position of the background outflow emis-sion. Each cone was modelled as a uniform distribution of ho-mogeneous point-like sources, and then each point-source wasindividually lensed forward (i.e. singularly mapped and magni-fied in the lens plane) through the reconstruction-algorithm ofRizzo et al. (2018), keeping the lens model fixed to Nierenberget al. (2020). As a result, we obtained the whole forward lensedimage of each starting background cone, di ff ering in radius andaperture angle.In order to establish the intrinsic size of the outflow, we se-lected those cones with a radius compatible with the detected[O III] outflow emission by visually comparing the extent of theforward lensed emission produced by a given radius backgroundcone, with the maximum distance at which we observe the [O III]outflow, once corrected for the SINFONI-PSF, that is, R max = . . In Fig. A.1, we show an example of our forward lensingmethod: starting, for instance, from a background homogeneouscone with radius of 8 pixels and aperture of 60 ◦ (left) and itsforward lensed image (right). The solid red circumference has aradius equal to R max = . R max . Taking the averageof these more plausible radii and the maximum deviation fromthe mean as error, and converting into kpc-units, we estimatedthe intrinsic radius of the outflow to be R out = (8 . ± .
7) kpc,with z = . ± .
002 as the redshift we measured from thenuclear spectrum extracted during the BLR-fitting (described inSect. 3.1).For the estimate of the magnification factor with its error, weconsidered multiple aperture angles: 30 ◦ , 45 ◦ , 60 ◦ , and 90 ◦ , foreach defined-radius cone, and calculated the magnification fac- For simplicity, as it is the estimate of the outflow radius based ona qualitative comparison, we report the projected distances in the de-scription of the procedure in units of SINFONI pixels, thus recallingthat R max ∼ . ∼ y [ p i x e l ] x [pixel] x [pixel] Fig. A.1.
Example of forward lensing (Rizzo et al. 2018) of a back-ground homogeneous cone with radius of 8 pixels and aperture of 60 ◦ (left) into the respective lensed image (right). In both images, the or-ange stars, red and black ‘ + ’ indicate, respectively, the position of thecone origin, of the emission peak observed by SINFONI, and of the lenscentre. The dashed and solid red circumferences in the right panel havea radius equal to the intrinsic size of the cone in the source plane (i.e. 8pixel in the example shown here) and to the maximum extent reached bythe [O III] emission observed by SINFONI, i.e. R max = . + ’). Thedashed and dotted black lines identify the position angle ( ∼ ◦ ) andthe aperture (here 60 ◦ ) of the intrinsic cone. We note that the conicalsource does not intercept the centre of the lens, thus subject to a lowmagnification. tor as the ratio between the total flux in the lens plane and the to-tal flux in the source plane. All simulated conical configurationsprovided low magnification factors that are weakly dependent onthe assumed geometry of the cone (they di ff er by a factor ∼ . − . µ out = . ± .
2, where the maximumdeviation from the mean has been taken as error. We used thisvalue to correct the observed [O III] outflow flux (determined inSect. 3.3), thus obtaining F out = (1 . ± . × − (2 . /µ out ) ergs − cm − . List of Objects ‘HS 0810 + +1138’ on page 1