Mon.Not.R.Astron.Soc.000,1–??(2015) Printed21January2016 (MNLATEXstylefilev1.4) Detection of a second high velocity component in the highly ionized wind from PG 1211+143 Ken Pounds 1, Andrew Lobban 1, James Reeves 2 and Simon Vaughan 1 1 Department of Physics and Astronomy, University of Leicester,Leicester, LE1 7RH, UK 6 2 School of Physical Sciences, Keele University,Keele, ST5 5BG, UK 1 0 2 n Accepted ;Submitted a J 0 ABSTRACT 2 An extended XMM-Newton observation of the luminous narrow line Seyfert galaxy PG1211+143 in 2014 has revealed a more complex highly ionized, high velocity out- ] A flow.ThedetectionofpreviouslyunresolvedspectralstructureinFeKabsorptionfinds asecondoutflowvelocitycomponentofthehighlyionizedwind,withanoutflowveloc- G ityofv∼0.066±0.003c,inadditiontoastillhighervelocityoutflowofv∼0.129±0.002c h. consistent with that first seen in 2001. We note that chaotic accretion, consisting of p many progradeand retrogradeevents, offers an intriguing explanationof the dual ve- - locity wind. Inthat contextthe persisting outflow velocitiescould relateto physically o distinct orientations of the inner accretion flow, with prograde accretion yielding a r t higher launch velocity than retrograde accretion in a ratio close to that observed. s a Key words: galaxies:active – galaxies:Seyfert: quasars:general– galaxies:individ- [ ual: PG1211+143– X-ray: galaxies 1 v 7 6 3 1 INTRODUCTION factor, with a persistent wind having sufficient mechanical 5 energy to disrupt the bulge gas in the host galaxy (Pounds 0 X-ray spectra from an XMM-Newton observation of the lu- 2014a). An indication how much of that energy could be . minous Seyfert galaxy PG1211+143 in 2001 provided the lost before reaching a still-active star forming region came 1 first detection in a non-BAL AGN of strongly blue-shifted withevidenceofaUFOshockingagainsttheISMorslower- 0 absorption lines of highly ionized gas, corresponding to a moving ejecta in the low mass Seyfert galaxy NGC4051 6 1 sub-relativistic outflow velocity of ∼0.09c (Pounds et al. (Poundsand Vaughan 2011, Poundsand King2013). : 2003). That velocity was based primarily on the identifica- Determination of the correct wind velocity, v, is crit- v tionofastrongabsorptionlineat∼7keVwiththeresonance ically important in estimating the mass and energy rates i X Lyman-α transition in Fe XXVI. The subsequent inclusion of AGN winds, with the latter being dependent on v3 in r of additional absorption lines, of Ne, Mg, Si and S, led to a radial outflow. However, significant uncertainty remains a there-identification of the∼7 keV absorption line with the when the deduced velocity depends on the identification of resonance 1s-2p transition in Fe XXV, and a revised out- asingleblue-shiftedabsorptionline,typicallyofFeK,asin flow velocity of 0.14±0.01c (Pounds and Page 2006). Fur- many early UFO detections (eg Tombesi et al. 2010). Spec- ther observations of PG1211+143 over several years with tralmodellingoverawiderenergybandreducesambiguities XMM-Newton,ChandraandSuzakufoundthehighvelocity in line identification, as first demonstrated in Pounds and outflow to be persistent but of variable strength (Reeveset Page(2006),whilealsoprovidingadditionaloutflowparam- al. 2008). Evidence that the mean outflow in PG1211+143 eters, including the mean ionization parameter and column wasbothmassiveandenergetic-withpotentialimportance density.Suchbroad-bandspectralmodellingwasusedinthe for galaxy feedback - was obtained from the detection of P second paper on the XMM-Newton archival data (Tombesi Cygni and other broad emission features by combining the et al. (2011) and byGofford (2013) in theirsimilar analysis 2001, 2004 and 2007 XMM-Newton EPIC spectra (Pounds of Suzaku data. and Reeves 2009). In thepresent paper we make a further addition, in si- Examination of archival data from XMM-Newton and multaneously modelling the outflow of PG1211+143 with Suzaku subsequently showed ultra-fast, highly-ionized out- both photoionized absorption and emission grids. To quan- flows (UFOs) to be relatively common in nearby, luminous tifyabsorptionandemission structureweemployphotoion- AGN(Tombesi2010,2011;Gofford2013).Thefrequencyof ized grids of pre-computed spectra based on the xstar code these detections appeared toconfirm a substantial covering (Kallman et al. 1996), with the publicly available grid 25 (cid:13)c 2015RAS 2 Ken Pounds et al. (turbulence velocity 200 km s−1) being found to provide pn 2014 4 a satisfactory match to the highly-ionized absorption, and 1. velocity-broadenedemissionadequatelymodelledbygrid22 (turbulencevelocity 3000 km s−1). Initially employing that procedure in re-modelling the 2 2001 outflow we re-affirm a column density N of ∼1023 1. H cm−2, ionization parameter logξ∼2.9 erg cm s−1 and out- flowvelocityv∼0.14±0.01c.Bycomparisonwiththetypical o properties ofUFOs(Tombesi etal. 2011) andwith thenew rati 1 observationsreportedhere,itappearstheunusuallylowout- flow ionization - and correspondingly high opacity - in the 2001observationwasaprimaryreasonforitbeingamongst 8 themost highly visible UFOsyet seen. 0. 2 5 10 observed energy (keV) 2 A NEW XMM-Newton OBSERVATION OF pn 2001 PG1211+143 IN 2014 InordertofurtherexploretheionizedwindinPG1211+143 1.5 anextendedXMM-Newtonobservationwascarriedoutdur- ing 7 spacecraft orbits over the period 2014 June 2 to 2014 July9.Thepresentpaperdescribestheanalysisof highen- ergy spectral data, with subsequent papers reporting the o detection of a low ionization outflow in the soft x-ray spec- ati 1 r trum, and evidence for short-term variability. We use data from the European Photon Imaging Cameras (EPIC) pn CCD(Struederetal.2001),operatedinlargewindowmode. On-targetexposuresforindividualorbitsweretypically ∼100 ks, apart from the fifth observation (rev2664) of ∼55 0.5 ks,givingatotalon-targetexposureof∼630ks.Fulldetails oftheXMM-Newtonobservingloganddataextractionpro- 2 5 10 cedures, and of accompanying Swift observations, are given observed energy (keV) in Lobban et al.(2015), reporting the results of a detailed Figure 1. Comparison of the stacked pn data from the 2014 timing analysis. observationofPG1211+143withthatfromthe2001observation, Raw data were processed using version 14.0 of the both plotted as aratiotoadouble power-lawcontinuum. While ⋆ XMM-Newton Scientific Analysis Software package, fol- themostprominentabsorptionlinenear7keVisweakerthanin lowing standard procedures. Source events were extracted 2001,themuchlonger2014exposureallowstheoverallspectrum from 20arcsec circular regions centred on the AGN, while to be better defined. In particular, the Fe K emission profile is background events were taken from much larger regions resolvedintohighandlowionizationcomponents andadditional away from the target source and from other nearby back- absorptionstructureisindicatedathigherenergies groundsources.ForourEPIC-pnanalysis,weutilizedsingle and doublegood pixelevents(PATTERN ≤ 4). Theexclusion 3 CONFIRMATION OF A HIGH VELOCITY of short intervals of background flaring resulted in a final WIND IN PG1211+143 ∼570ks of high quality pn data. The total residual back- groundcountratewas<1percentofthesourcerateforall Whileexaminationofpnspectrafromthe7individualorbits threeEPIC cameras in each observation (Figure 2). in2014showedstrongvariabilityinthesoftx-rayband,the WeassumeanAGNredshiftofz=0.0809(Marzianiet harder continuum changes relatively little above ∼2 keV. al. 1996). Spectralmodelling isbased ontheXSPECpackage To obtain a baseline hard x-ray spectrum of PG1211+143 (Arnaud 1996) and includes absorption due to the line-of- with maximum statistical quality we therefore summed low sight Galactic column of N = 2.85×1020cm−2 (Murphy background pn data from all 7 orbits for an initial spectral H etal. 1996). 90%confidenceintervalson modelparameters analysis. The resulting low background source exposure is arebasedon∆χ2=2.706.Publishedestimatesfortheblack a factor ∼10 greater than for any of the previous XMM- holemassinPG1211+143rangefrom3×107 M⊙ (Kaspiet Newton observations in 2001, 2004 and 2007. al.2000)to1.5×108 M⊙ (Bentzetal.2009),withthelower Figure 1 compares the stacked pn spectrum from 2014 value making the mean luminosity close to the Eddington withthatfrom 2001,bothdatasetsbeingplottedasaratio limit. All derived velocities are corrected for therelativistic toadoublepower-lawcontinuumofphotonindexΓ∼1.6and Dopplereffect. Γ∼3tomatchavariable‘soft excess’atthelowest energies. Visual comparison of the two plots confirms that emission and absorption features are significantly better defined in the longer 2014 exposure. The emission near 6 keV is now ⋆ http://xmm.esac.esa.int/sas/ resolved into low and high ionization components, and the (cid:13)c 2015RAS,MNRAS000,1–?? Detection of a second high velocity 3 prominent absorption line at ∼7 keV is again seen in the pn 2014 2014 data,albeit beinglessdeepthanin2001.Otherdiffer- encesinthe2014spectrumincludeweakerabsorptioninthe lowermassions,butaclearindicationofadditionalabsorp- 0−4 1 tion structure in the Fe K band between 6-10 keV. Those differences suggest the fast wind in 2014 was more highly m−2 c ieollninizgedthtehabnroinad20b0a1n,dasspuegcgtersutmionfrwomen2o0w14cownfiitrhmtbhye msaomde- eV−1 0−5 k 1 photoionized grids used to parameterise the2001 outflow. s−1 Figure 3 (top panel) shows a section of the stacked pn nts spectrum covering the Fe K region. The data have been ou groupedwithaminimumof25counts(forχ2compatibility) c 0−6 1 andwithamaximumof3datapointsperpncameraresolu- tion(FWHM)tooptimisethevisibilityofspectralstructure. The emission near 6 keV is resolved into components with rest energies close to the neutral Fe-K fluorescent emission 5 6 7 8 9 10 observed energy (keV) line and the 1s-2p resonance emission lines of He-like and H-like Fe, respectively. The absorption line at ∼7 keV is a Figure 2. Comparison of the source (black) and background factor ∼3 weaker than the corresponding line in 2001, with spectra(red)forthestacked2014pndata. Thesourcespectrum an equivalent width of 37±5 eV.In contrast, thedeep 2014 is plotted without background subtraction and is near identical exposure reveals additional structure at ∼6–10 keV, with to that including background subtraction, confirming the Fe K absorption features are not due to fluorescent Cu and Zn lines apparent absorption lines at ∼6.9, ∼7.3, ∼7.5, ∼7.8, ∼8.2 † fromthedetector structure and ∼8.6 keV. Gaussian fittingofthosespectralfeaturesisdetailedin Section 4.1, and while the significance of individual lines is limited, we emphasise that findinga common velocity for a line series can be highly significant, again underlining the velocityof1500±1200kms−1 intheAGNrest-frame.Criti- importanceof takingproperaccount ofmultipleabsorption cally,although reducedin depth,theabsorption features at linesinthestudyofx-rayoutflows.Analysisandinterpreta- ∼7–9 keV in Fig.3 (top) remained clearly visible. tionoftheFeKabsorptionlinestructure,providingfurther Havingmodelledtheunderlyingcontinuum,aphotoion- detail on the highly ionized wind in PG1211+143, is the ized absorber was then added to match that absorption main purpose of this paper. structure, obtaining a significant improvement to the spec- Beforetestingforionizedabsorptionitisespecially im- tral fit (χ2ν of 1470/1361), for a column density NH of portant to correctly model the underlying continuum for 7.2±5.0 × 1022 cm−2, ionization parameter logξ=3.3±0.1 such high quality data. In particular, scattering from opti- erg cm s−1 and outflow velocity of 0.119±0.003c. The ad- callythickmatter(orso-called ’reflection’foundtobecom- dition of a photoionized emission spectrum, with tied ion- mon in AGN spectra (Nandra and Pounds 1994)) may im- ization parameter, reproduced the ionized emission lines at pose a step change near the Fe K absorption edge (∼6.5 ∼6.2 and ∼6.45 keV, and further improved the fit (χ2 of ν keV in observer space), and an indication of such an ef- 1445/1359), with the column density of the absorber de- fect can be seen in Figure 3 (top panel). To quantify that creasing toNH of 5.8±3.2×1022 cm−2.Thetied ionization ’reflection’ we added a Xillver component (Garcia et al. parameter (logξ=3.42±0.05 erg cm s−1) and absorber ve- 2013) to the power law continuum model in XSPEC, si- locity were unchanged in this second fit, while untying the multaneously matching the fluorescent Fe K emission line ionization parameters of the emission and absorption spec- and continuum reflection from optically thick ionized mat- tra made little difference to χ2. The ionized emission spec- ter. When previously modelled with a Gaussian, the Fe K trum was found to have a blueshift (relative to the AGN), line energy of ∼5.96 keV (∼6.44 keVat theAGN redshift), corresponding to an outflowvelocity of 4200±1200 km s−1, and EW of 73 eV yielded a 2–10 keV spectral fit statistic which we take to represent a mean value averaged over an of(χ2 of1513/1364). Re-fittingtheGaussian withXillver extendedionized outflow. ν providedacorrespondingadjustmenttothecontinuumfrom Figure3(lowerpanel)illustratesthekey4-9keVsection reflection, with a further improvement in χ2 of 1495/1364. of the single velocity outflow model, with absorption lines ν The reflection component had an ionization parameter of seeninthedataat∼7and∼7.3keVidentifiedwiththe1s-2p logξ∼2.1, with inclination fixed at 45◦ and a small outflow resonancelinesofFeXXVandXXVI,andabsorptionlinesat ∼8.2and ∼8.6keVwith He-β and ablendofLyman-β and He-γ of the same ions. Finding a match to four absorption features with the same outflow velocity adds confidence in † The absorption lines at ∼8.2 and ∼8.6 keV lie close to fluo- therobustness of the spectral fit. rescent x-rayemission lines of Cu and Zn arisingfrom energetic The photoionized emission spectrum in Figure 3 corre- particleimpactsonthepncameraelectronicsboard(Struederet spondingly matchesthetwo high energy componentsin the al. 2001). Fortunately, the low particle background throughout ratioplot,beingidentifiedwiththeHe-andH-likeresonance most of the 2014 XMM-Newtonobservations (Figure 2) ensured such background features have a negligible effect on the source lines, in a ratio set by the linked ionization parameter. En- spectrum,anoutcomeconfirmedbyobtainingaverysimilarratio couragingly,theHe-β emissionlinecanalsobeseenat∼7.7 plottothatinFigure3whennobackgrounddataissubtracted. keVin both data and model, with a similar blueshift. (cid:13)c 2015RAS,MNRAS000,1–?? 4 Ken Pounds et al. pn 2014 Table1.Finalparametersofthehighlyionizedoutflowobtained 4 froma2–10 keV spectral fit tothe 2014 pndata, withtwo pho- 1. toionized absorbers, defined by ionization parameter ξ (erg cm s−1),columndensityNH (cm−2)andoutflow velocity(v/c), to- gether with a photoionized emission spectrum modelled by an 2 ionization parameter and outflow velocity. Extracted or added 1. luminosities(ergs−1)overthefittedspectralband2–10keVand theimprovementinχ2arealsogivenforeachphotoionizedmodel o ati component r 1 comp logξ NH(1023) v/c Labs/em ∆χ2 abs 4.0±0.2 3.7±2.9 0.129±0.002 5×1041 15/3 abs 3.4±0.1 2.0±1.0 0.066±0.003 1.8×1042 27/3 8 0. emi 3.5±0.1 1(f) 0.011±0.003 7×1041 34/3 4 5 6 7 8 9 observed energy (keV) blending with He-α from the high velocity wind to provide 0−3 a bettermatch to thebroad absorption feature observed at 1 ∼6.9keV.Aconsequenceofthatlineblendistoincreasethe × V)−1 2 relative strength of the higher velocity Lyman-α line and m s ke−2−1 1.5×10−3 hcvoeellnuoccmeitny-dsisuenbasslisttoayniotnifcartllehyaast-edflbooswltihgchottmhlyepioinonnetinhztea.tdiTouhnaelphvaierglaohmceiretytoeufirttfla,ontwdo ons c 0−3 0.129A±n0.i0m0p2ocr.tantconsequenceofsimultaneouslymodelling ot 1 h theabsorptionand(re-)emissionspectraisdemonstratedby P keV (2 ×10−4 tvhioeuHsley-αpaarbtlsyorhpitdiodnenlinbye itnhetheemlioswsieornvleinloeciatty∼ou6t.5flokwe,Vp.rIen- 5 turn,thepartialabsorptionoftheFeLyman-αemissionline explains the relative weakness of that emission component 0 in the data ratio plot. Comparing Figures 3 and 4, and the 4 5 6 7 8 9 observed energy (keV) fit residuals, confirms that a significant contribution to the improved dual absorber fit lies in better matching the ob- Figure 3.(top) Spectral structure inthe stacked 2014 pnspec- served opacity at ∼7 keV, while the lower velocity He-like trumfromFig1resolvesFeKemissioncomponents correspond- ingtotheFeKfluorescencelineandresonanceemissionfromFe absorption line models the sharp drop in the data at ∼6.6 XXV and Fe XXVI ions. (lower) The addition of photoionized keVasitcutsintothebluewingoftheH-likeemissionline. emission and absorption spectra match the resonance emission From the dual velocity outflow spectral fit we find a lines observed at ∼6.2 keV and ∼6.45 keV, with photoionized mean 2-10 keV source luminosity of ∼6×1043 erg s−1. The absorptionimprintedonthehardpower lawcontinuum (red). A total extracted (absorbed) and added (emission) luminosi- single outflow velocity identifies absorption lines at ∼7.0, ∼7.3, ties are ∼2.2×1042 erg s−1 and ∼7×1041 erg s−1, respec- ∼8.2and∼8.6keVwiththeαandβ lines,againofFeXXVand tively, with the lower velocity, lower ionization absorber XXVI.ThemodelincludescontinuumreflectionandlinkedFeK contributing ∼80% of the 2-10 keV opacity. Table 1 sum- fluorescencelineemission(black)andiscompletedwithasofter, marisesthemainparametersofthephotoionizedabsorption unabsorbed power law (green) required by inter-orbit difference and emission spectra describing the highly ionized outflow spectra(seetext) in PG1211+143 in 2014. To summarise, in XSPEC terms the final model is: TBabs(pl + (pl )(mtable)(mtable) + Xillver + at- 3.1 A second velocity component in the highly soft hard able), where pl is the dominant continuum component ionized wind of PG1211+143. hard (Γ∼1.6),subjecttoionizedabsorptionfromthetwooutflow Examination of Figure 3 shows that while the absorption components. Xillver represents the reflection continuum lines seen at ∼7, ∼7.3, ∼8.2 and ∼8.6 keV are matched by and associated Fe K fluorescence line, as described in the resonance and higher order transitions in Fe XXV and Fe text. pl is a soft continuum component (Γ∼2.9) found soft XXVIinthephotoionizedoutflow,thecomparisonofmodel from inter-orbit difference spectra where the subtraction of and data is incomplete. To seek a further improvement in low flux spectra from high flux spectra show the residual the fit a second absorber was therefore added, with ioniza- spectrumiswelldescribedbyasimplepowerlaw.Whilethis tionparameter, columndensityandvelocityagain free.Re- variable,softcontinuumcomponenthaslittleimpactonthe fitting the 2-10 keV spectrum showed a further substantial present analysis, it is included for consistency with thesoft improvement to the fit (χ2 of 1415/1356), with the second x-ray analysis of XMM-Newton grating spectra (Pounds et ν absorption component having a significantly lower outflow al. 2015). velocity of 0.066±0.003c. Figure 4 (lower panel) compares the stacked 2014 pn Figure 4 (top) illustrates the dual velocity absorber datawith thedualvelocity outflowmodel(plottedimmedi- model,withtheLyman-αlineofthelowervelocityflownow ately above), while Figure 5 shows the spectral model plot (cid:13)c 2015RAS,MNRAS000,1–?? Detection of a second high velocity 5 pn 2014 pn 2014 10−3 m−2 10−4 × c m s keV)−2−1−1 1.5×102−3 s sskeV−1 −1 −1 1100−−65 otons c 10−3 count 10−7 h P keV (2 ×10−4 o 1.2 5 ati 1 r 0.8 0 4 5 6 7 8 9 2 4 6 8 10 observed energy (keV) observed energy (keV) 0−4 Figure 5. (top) Comparisonof the dual velocity outflow model 1 andpndataover thefull2-10keVband, showingtheindividual spectralcomponents,withthepowerlawandreflectioncontinua (red,greenandblack),thefluorescentFeK(alsoblack)andion- cm−2 0−5 izedemissionlines(blue) V−1 5×1 e k s−1 4 SUPPORTING EVIDENCE FOR DUAL s nt OUTFLOW VELOCITIES u o c 0−5 4.1 Gaussian line fitting to the observed spectral 1 × structure 2 Gaussian fitting to the emission and absorption line struc- ture seen in the stacked pn data provides a model- 4 5 6 7 8 9 observed energy (keV) independent check on the robustness of the above spectral analysis. The 4–10 keV ratio plot shown in Figure 6, ob- Figure 4. Comparison of stacked pn data and the final outflow tained by removing the photoionized absorption and emis- model fit including a second highly ionized absorption compo- nent. (top) The principal emission and absorption lines in the sionspectrafromthebestfitmodelofSection3,wasscanned two-component absorber model now provide a good match to sequentiallywith positive andnegativeGaussians, havinga spectralfeaturesinthedata. Colourcodingoftheemissionlines minimum(1σ)linewidthof50eVcomparabletothepnde- includesblackforthefluorescentFeKline,andblueforthehigh- tectorresolution. Allfeaturesgiving ∆χ2 >∼6wererecorded ionization emission. The hard power law and unabsorbed power and thesweep was then repeated with two broader absorp- law components are shown in red and green respectively. Sepa- tion features at ∼7 keV and ∼8.2 keV re-fitted as narrow rate absorption line sequences of Lyman-α, He-α, Lyman-β and linepairs,making9absorptionlinesinall.Theemissionand He-βforoutflowvelocitiesof∼0.066cand∼0.129caremarkedon absorptionblendat∼6.5–6.6keVwasresolvedbyfixingthe theplot.(lower)Whilelackingthevisualclarityofasimilarplot emission line width at 100 eV. The resulting fit is shown in obtainedwithhigherresolutionspectradata,individualemission Figure 6, with themeasured line energies, equivalent width and absorption features remain visible when folded through the CCDresponsefunction. and proposed identifications listed in Table 2. Of the 9 narrow absorption lines in Table 2, those at ∼6.62, ∼6.87, ∼7.82 and ∼8.17 keV have observed line en- ergiesconsistentwithanoutflowvelocityintherange0.065- 0.069cwhenidentifiedwiththeαandβ linesofFeXXVand FeXXVI. Taken as a line set, the weighted mean velocity is0.067±0.001c andthecombined significanceishigh (∆χ2 overthefull2–10 keVenergyrangetobetterillustrate con- = 51/8), providing strong model-independent support for tinuum and line emission components. the lower velocity absorber in Table 1. Four (of 5) remain- FromthelowerpanelofFigure5wenotetheabsenceof ing Gaussians, at ∼7.04, ∼7.31, ∼8.34 and ∼8.70 keV, are a hard excess in the data-to-model ratio confirms that con- consistentwiththehigheroutflowvelocityfoundinspectral tinuumreflectionisadequatelymodelledbyXillver,atleast modelling, with individual values - based on identification up to 10 keV in our spectral fit, and does not indicate the with the same set of resonance absorption lines - ranging strong reflection reported in Zoghbi et al.(2015). Although from 0.125-0.134c, and a mean value v∼0.128±0.002c. For discussedfurtheratthispointwenotethestrongestresidual this higher velocity line-set the statistical improvement of in the lower part of Figure 5 is excess emission near 9 keV, Gaussian linefittingisalso highly significant (∆χ2 =57/8). which - if confirmed - could represent blue-shifted FeXXVI Absorptionline5doesnotfitthedualoutflowvelocitypat- RRC,perhaps indicative of a cooling flow. tern,butis interestingin that thepossible association with (cid:13)c 2015RAS,MNRAS000,1–?? 6 Ken Pounds et al. pn 2014 locity outflowinthat earlier observation,from are-analysis of the soft x-ray spectrum (Pounds 2014b) and in an ear- 4 1. lier partial-covering spectral fit which required an absorber moving at ∼0.07c to explain continuum curvature (Pounds and Reeves2009). 2 The properties of powerful AGN winds are reviewed 1. in King and Pounds (2015), where a highly ionized wind o is envisaged being launched at the local escape velocity by rati continuum photons from a SMBH accreting at modest Ed- 1 dington ratios m˙ = M˙/M˙Edd ∼ 1, finding - for accretion from a disc - the excess accreting matter is expelled in a quasi-sphericalwind,with alaunchvelocityv≃ ηc∼0.1c. m˙ Observationalsupportforthatpictureisprovidedbyrecent 8 0. archival searches (Tombesi et al. 2010, 2011; Gofford et al. 2013) which find a substantial fraction of luminous AGN 6 7 8 9 having a highly ionized wind with a velocity in the range observed energy (keV) ∼0.03–0.3c Figure 6. Gaussian line fitting to spectral structure in Thenewobservationsindicateamorecomplexpicture, the stacked pn data from the XMM-Newton observation of with two distinct outflow velocities, co-existing in observa- PG1211+143in2014.Inthisplotdatabinninghasbeenrelaxed tions 13 years apart. Chaotic accretion (King and Pringle withthe removalof thelimitof3data points perresolutionele- 2006), consisting of many prograde and retrograde events, menttoprovidesmootherGaussianprofiles.Ninepossibleabsorp- offersanintriguingexplanationofsuchadualvelocitywind, tionlinesaredetected,numberedfromlefttorightasabs1–abs thepersistenceoftwodistinctoutflowvelocitiesperhapsre- 9inTable2,wherethemeasuredenergy,proposedidentification lating to physically distinct orientations of the inner accre- andoutflow velocityarelisted tionflow,bothclosetoEddington,andwithdifferingvalues oftheaccretionefficiencyηandhenceofvelocity.Thatpos- the FeXXV resonsnce line would match a third outflow ve- sibilityisdiscussedinmoredetailbyKingandNixon(2016) locityofv∼0.19cwhichgaveamarginalimprovementtothe Higher resolution hard x-ray spectra from the forth- dual velocity spectral modelling described in Section 3.2. coming Astro-H observatory should show how common are such complex highly ionized AGN winds as reported here for PG1211+143. 5 DISCUSSION The extended observation of PG1211+143 in 2014 has pro- vided high quality hard x-ray spectra revealing previously 6 CONCLUSION unseen spectral structure of a UFO in the ∼6-10 keV en- AnextendedXMM-Newtonobservationoftheluminousnar- ergy band. Spectral modelling has identified the observed row line Seyfert galaxy PG1211+143 in 2014 has revealed absorptionstructurewithresonanceandhigherorderlinesof previously unseen spectral structure in Fe K absorption, highlyionizedFe,consistentwithtwodistinctoutflowveloc- findinga second high velocity component of thehighly ion- ities. Thefaster wind component hasavelocity of v∼0.13c, ized wind. In identifying that additional complexity within similar to that first seen in 2001, but with a higher column the limits of CCD energy resolution, the 2014 observation densityand ionization parameter. Thesecond outflowcom- benefited critically from the high statistical significance of ponent,notresolvedintheEPICdatain2001,hasasimilar the EPIC data resulting from the unusually long observa- column density buta lower velocity v∼0.066c. tion. The outcome promises further revelations of the dy- Keyfactorsinidentifyingthelowervelocityflowin2014 namicalstructureofAGNwindsinaneweraofhighresolu- were the high quality of the stacked 2014 spectral data, al- tion hard x-ray spectroscopy,heralded with thenear-future lowing spectral features to be identified near the limit of launch of Astro-H. CCD energy resolution, and the simultaneous modelling of highly ionized emission, which allowed - in particular - the lowervelocity FeXXVHe-αabsorption linetobedetected. Thephotoionizedemissionspectrumalsosuccessfullyrepro- ACKNOWLEDGEMENTS duces the higher energy features observed in the stacked 2014 pn data, identified with resonance emission from He- XMM-Newton is a space science mission developed and op- and H-likeFe ions in a similar ratio to that seen in absorp- eratedbytheEuropeanSpaceAgency.Weacknowledgethe tion.Comparisonofabsorbedand(re-)emissionluminosities excellent work of ESA staff in Madrid in planningand con- indicates a substantial covering factor of the highly ionized ducting the XMM-Newton observations. The UK Science wind. and Technology Facilities Council funded the postdoctoral An important outcome of the present analysis is in researchassistantshipofAL.Weacknowledgethecontinued finding that the ultra-fast outflow discovered in an XMM- cooperation with theLeicester theorygroup led byAndrew Newton observation in 2001 is again detected in 2014. Al- King. The present text has benefited significantly in clarity though not resolved in the Fe K spectrum in 2001, there is from several constructive suggestions from the anonymous independent evidence for the co-existence of the lower ve- referee. (cid:13)c 2015RAS,MNRAS000,1–?? Detection of a second high velocity 7 Table 2. Sequential Gaussian fits to the positive and negative features in the pn 2014 data shown in Figure 6. Absorption lines (abs 1-9)haveafixedwidthcomparabletothepndetector resolution,withthefitted lineenergy, proposedidentification andcorresponding outflow velocity listed in each case. Improvement in ∆χ2 is after re-fitting the data at 4–10 keV following each added line. All line identifications arewithtransitionsinFeXXVorFeXXVI.Theequivalent widthofeachlinehasbeenestimatedseparatelyinXSPEC component obsenergy(keV) EW(eV) restenergy(keV) lineid energy(keV) v/c ∆χ2 FeK-α 5.98±0.01 80±7 6.46±0.01 FeK-α 6.40 0.009±0.002 139/3 Fe25 6.26±0.02 19±7 6.77±0.02 He-α 6.70 0.010±0.003 8/2 Fe26 6.49±0.07 73±15 7.02±0.07 Ly-α 6.96 0.01±0.01 34/2 abs1 6.62±0.02 -31±11 7.16±0.02 He-α 6.70 0.066±0.003 12/2 abs2 6.87±0.01 -22±6 7.43±0.01 Ly-α 6.96 0.065±0.002 16/2 abs3 7.04±0.02 -11±7 7.61±0.02 He-α 6.70 0.127±0.003 7/2 abs4 7.31±0.02 -22±8 7.90±0.02 Ly-α 6.96 0.126±0.003 25/2 abs5 7.49±0.03 -9±7 8.10±0.03 He-α 6.70 0.188±0.004 7/2 abs5 7.49±0.03 -9±7 8.10±0.03 Ly-α 6.96 0.151±0.004 7/2 abs6 7.82±0.03 -16±9 8.45±0.03 He-β 7.88 0.069±0.003 10/2 abs7 8.17±0.02 -20±8 8.83±0.02 Ly-β 8.25 0.068±0.002 13/2 abs8 8.34±0.04 -16±9 9.02±0.04 He-β 7.88 0.134±0.005 6/2 abs9 8.70±0.02 -41±11 9.40±0.02 Ly-β 8.25 0.129±0.003 19/2 abs9 8.70±0.02 -41±11 9.40±0.02 He-γ 8.29 0.125±0.003 19/2 REFERENCES ArnaudK.A. 1996,ASPConf.Series,101,17 Bentz M.C.,Peterson B.M.,Pogge R.W.,Vestergaard M. 2009, ApJL,694,166 GarciaJ.etal.2013,ApJ,768,146 Gofford J., Reeves J.N., Tombesi T., Braito V., Turner T.J., MillerL.,CappiM. 2013,MNRAS,430,60 Kallman T., Liedahl D., Osterheld A., Goldstein W., Kahn S. 1996,ApJ,465,994 Kaspi S. Smith P.S., Netzer H., Maoz D., Jannuzi B.T., Giveon U.etal.2000, ApJ,533,631 KingA.R.andPringleJ. 2006,MNRAS,373,90 KingA.R.andPoundsK.A. 2015, ARA&A,53,115 KingA.R.andNixonC. 2016, inpreparation LobbanA.,VaughanS.A.,PoundsK.A.,ReevesJ.N. 2016,MN- RAS,accepted (2015arXiv151202587L) MarzianiP.,SulenticJ.W.,Dultzin-HacyanD.,ClavaniM.,Moles M. 1996, ApJS,104,37 MurphyE.M.,LockmanF.J.,LaorA.,ElvisM.1996,ApJS,105, 369 NandraK.andPoundsK.A, 1994,MNRAS,268,405 Pounds K.A.,Reeves J.N.,King A.R.,Page K.L.,O’BrienP.T., TurnerM.J.L. 2003, MNRAS,345,705 PoundsK.A.andPageK.L. 2006, MNRAS,360,1123 PoundsK.A.andReeves J.N. 2009, MNRAS,397,249 PoundsK.A.andVaughanS. 2011,MNRAS,415,2379 PoundsK.A.andKingA.R. 2013, MNRAS,433,1369 PoundsK.A. 2014a, SpaceScienceReviews,183,339 PoundsK.A. 2014b, MNRAS,437,3221 Pounds K.A, Lobban A., Reeves J.N., Costa M., Vaughan S. 2015,MNRAS,submitted Reeves J.N.etal.2008,MNRAS,385,L108 StruederL.etal.2001,A&A,365,L18 Tombesi F., Cappi M., Reeves J.N., Palumbo G.C., Yaqoob T., BraitoV.,DadinaM. 2010,ApJ,742,44 Tombesi F., Cappi M., Reeves J.N., Palumbo G.C., Braito V., DadinaM. 2011, A&A,521,A57 ZoghbiA.etal.2015,ApJL,799,L24 (cid:13)c 2015RAS,MNRAS000,1–??