Received2006Sep27;accepted2006Dec24 PreprinttypesetusingLATEXstyleemulateapjv.11/12/01 XMM-NEWTON AND CHANDRA SPECTROSCOPYOF THE VARIABLE HIGH-ENERGY ABSORPTION OF PG 1115+080: REFINED OUTFLOW CONSTRAINTS G. Chartas,1 W. N. Brandt,1 S. C. Gallagher,2 and D. Proga3 Received 2006 Sep 27; accepted 2006 Dec 24 ABSTRACT We present results from multi-epoch spectral analysis of XMM-Newton and Chandra observations of the mini broad absorption line (BAL) quasar PG 1115+080. This is one of the few X-ray detected mini-BAL quasars to date that is bright enough in the X-ray band, mostly due to large gravitational- lensing magnifications, to allow in-depth spectral analysis. The present XMM-Newton observations of 7 PG1115+080haveprovidedthe highestsignal-to-noiseX-rayspectraofamini-BAL quasarobtainedto 0 date. BymodelingthespectraofPG1115+080wehaveobtainedconstraintsonthe columndensityand 0 ionization state of its outflowing absorbing gas. A comparison between these constraints over several 2 epochs indicates significant variability in the properties of the outflowing absorbers in PG 1115+080. n The depths of the high-energy broad absorption features in PG 1115+080 show a significant decrease a betweenthefirsttwoobservationepochsseparatedbyarest-frametimescaleof∼1year. Thisvariability J supports the intrinsic nature of these absorbers. Assuming the interpretation that the high-energy 4 absorption features arise from highly ionized Fe XXV we constrain the fraction of the total bolometric energy released by quasars PG 1115+080 and APM 08279+5255 into the IGM in the form of kinetic 1 energy to be ǫ = 0.64+0.52 (68% confidence), and ǫ = 0.09+0.07, respectively. According to recent k −0.40 k −0.05 v theoreticalstudies this rangeofefficiencies is largeenoughto influence significantlythe formationofthe 4 host galaxy and to regulate the growth of the central black hole. 0 1 Subject headings: galaxies: active — quasars: absorption lines — quasars: individual (PG 1115+080) 1 —quasars: individual (APM08279+5225)—X-rays: galaxies— gravitationallensing 0 7 1. introduction luminosity is thermally coupled via these outflows to the 0 surrounding gas. They find that the growth of the black- In recent years there has been mounting evidence from / hole mass is self-regulated and eventually saturates at a h both theoretical and observational studies for the impor- p tance of quasar outflows in regulating the growth of su- finalvaluethatdependsontheinitialamountofgasavail- - ableforaccretion. Theseauthorsalsofoundthatfeedback permassive black holes, controlling the formation of host o fromstarsandquasarsintheirsimulationsofgalaxymerg- r galaxies, and enriching the Intergalactic Medium (IGM). t Models ofstructure formationin a Λ-CDM cosmologyare ers can heat andexpel gas fromthe centers of the merged s galaxies. This loss of gas during the final stages of the a not consistent with observations unless feedback, either : from star formation or active galactic nuclei, is included merger can halt nuclear starburst activity and cause the v merger galaxy to become a gas-poor elliptical. i in the simulations. Recently, the potential importance of X quasar outflows has been explicitly demonstrated in theo- With the advent of XMM-Newton and Chandra it has become possible to infer the kinematic and ionization r retical models of structure formation and galaxy mergers a properties of highly ionized X-ray absorbers in Seyfert 1 that incorporate the effects of quasaroutflows [e.g., Scan- galaxies(e.g., Kaspiet al. 2002;Kaastraet al. 2002;Net- napieco & Oh 2004 (SO04); Granato et al. 2004 (G04); zer et al. 2003). Approximately 60% of Seyfert 1 galaxies Springel, Di Matteo, & Hernquist 2005 (SDH05); Hop- show outflowing X-ray and UV absorption by ionized gas kins et al. 2005, 2006]. A basic assumption in the mod- with velocities up to ≈ 2,500 km s−1 (e.g., Crenshaw et els of SO04 is that all quasars host outflows. SO04 find al. 1999; Kriss 2002). The absorbing outflow properties that by choosing the fraction of the total bolometric en- of more luminous quasars appear to differ in several ways ergy released over a quasar’s lifetime into the Interstellar from those of Seyfert 1s. The fraction of occurrence of Medium (ISM) and IGM in the form of kinetic energy X-ray and UV absorption has been reported to be lower to be ǫ =0.05, they can successfully model the observed k in quasars (e.g., Ganguly et al. 2001;George et al. 2002); evolution of the B-band quasar luminosity function be- however,theclaimsforaloweroccurrenceofX-rayabsorp- tweenredshifts of0.25and6.25. Ina recentstudy SDH05 tion have been challenged by Porquet et al. (2004) who simulated the growth of black holes in gas-rich galaxies find that about half of their sample of 21 low-redshift PG with and without the presence of accretionfeedback. The quasars harbor ionized absorbers. UV spectroscopic ob- feedback in their model is thought to occur through en- servationsindicate that about20% ofquasarsshow broad ergetic quasar outflows that interact with the gas of the absorption lines blueward of their resonant UV emission host galaxy. It is assumed that about 5% of the radiated lines(e.g.,Hewett&Foltz2003). Outflowvelocitiesofthe 1 Department of Astronomy & Astrophysics, Pennsylvania State University, University Park, PA 16802, [email protected], [email protected] 2 Department of Physics & Astronomy, University of California – Los Angeles, Mail Code 154705, 475 Portola Plaza, Los Angeles, CA 90095-1547, [email protected] 3 DepartmentofPhysics,UniversityofNevada, 4505SouthMarylandParkway,LasVegas,NV89154, [email protected] 1 2 CHARTAS ET AL. UV absorbers in BAL quasars have been found to be as Spectrometer (ACIS; Garmire et al. 2003) on board the high as ≈ 60,000 km s−1. Chandra X-ray Observatory (hereafter Chandra) on 2000 Recent X-ray observations of the mini-BAL quasar June 2 and 2000 Nov 3 for 26.8 ks and 10.0 ks, respec- PG 1115+080 and BAL quasar APM 08279+5255 have tively. The spectral analysis of the Chandra observations suggested the presence of relativistic outflows of highly of PG 1115+080 has been presented in Gallagher et al. ionized (ionization parameter of the order of logξ = 3.5) (2002)andChartasetal. (2003). Becauseofrecentsignif- absorbing material detected in the iron region (> 6 keV) icant improvements in the calibration of the instruments with velocities of up to ∼0.4c (Chartas et al. 2002,2003). on board Chandra and XMM-Newton since the publica- The inferred hydrogen column densities ranging between tion of thePG 1115+080 results, we have re-analyzed all 1022−23cm−2 andrelativisticvelocitiesoftheseoutflowing observations. Updates on the calibration of Chandra and X-rayabsorbersimplymass-outflowratesthatarecompa- XMM-Newton are reported on the Chandra X-ray Cen- rable to the estimated accretion rates. The presence of ter (CXC) and XMM-Newton Science Operations Centre massive, highly ionized, and high-velocity outflows from (SOC)WorldWide Web(WWW) sites, respectively.4 We quasarsindicatesthatthesewindsmaybeprovidingsignif- analyzed the XMM-Newton data for PG 1115+080 with icant feedback to the surrounding gas. Additional obser- the standard analysis software SAS version 6.1 provided vational evidence to support the presence of quasar feed- by the XMM-Newton SOC. The Chandra observations of back came with the detection of high-velocity blueshifted PG 1115+080 were analyzed using the standard software absorption-line features in the X-ray spectra of several CIAO 3.2providedby the CXC.A logofthe observations quasars and Narrow-LineSeyfert 1 galaxies (Reeves et al. that includes observation dates, observed count rates, to- 2003; Pounds et al. 2003a, 2003b). We note, however, tal exposure times, and observational identification num- that someof these claims havebeendisputed inrecentre- bers is presented in Table 1. We note that the count analysesofthedata(e.g.,Kaspi&Behar2006;McKernan rate and total number of counts (∼31,430 counts from et al. 2004, 2005). the three XMM-Newton observations) for PG 1115+080 TherelativelysmallvelocityspreadobservedfortheUV are the highest of any mini-BAL quasar X-ray spectrum absorptioninPG1115+080,comparedtothetypicalrange observed to date. of 5,000 – 25,000 km s−1 observed in BAL quasars, sug- For the reduction of the XMM-Newton observations gests that PG 1115+080be classified as a mini-BALQSO we filtered the PN (Stru¨der et al. 2001) and MOS (Turnshek 1988; Barlow, Hamann, & Sargent 1997). We (Turner et al. 2001) data by selecting events correspond- willconsiderPG1115+080asamini-BALthroughoutthis ing to instrument PATTERNS in the 0–4 (single and dou- paper withthe followingmodelinmind. Accordingto the ble pixel events) and 0–12 ranges, respectively. Sev- ”unification” model for BAL quasars most quasars have eral moderate-amplitude background flares were present outflowing winds, however, BAL and mini-BAL quasars during the XMM-Newton observations of UM 425 and correspondto quasarswith relatively largeinclination an- PG 1115+080. The PN and MOS data were filtered to gles. Inparticular,inBALquasarsitiscommonlythought exclude times when the full-field count rates exceeded that our line of sight intersects a large portion of the out- 20 cnts s−1 and 4 cnts s−1, respectively. The extracted flowing absorber whereas in mini-BAL quasars our line of spectra from the PN and MOS were grouped to obtain a sightintersectsashorterportion. Accordingtothismodel minimumof100and40counts,respectively,ineachenergy mini-BAL quasars would be expected to have smaller ab- bin, allowing use of χ2 statistics. Background spectra for sorbing columns, and smaller velocity gradients than tra- the PN and MOS detectors were extracted from source- ditional BAL quasars. free regions near UM 425 and PG 1115+080. The PN In this work we present recent results from monitoring and MOS spectra were then fitted simultaneously with a X-ray observationsof the mini-BAL quasarPG1115+080 variety of models employing XSPEC version 11.3 (Arnaud (z = 1.72). The goalof these observationswasto monitor 1996). TheenergyrangesusedforfittingthePNandMOS the timevariabilityoftheabsorptionfeaturesandthereby data were 0.2–10 keV and 0.4–9 keV, respectively. constrain the kinematic, ionization, and absorption prop- For the reduction of the Chandra observations we used ertiesofthequasaroutflowsinthisX-ray-brightmini-BAL standardCXCthreadstoscreenthedataforstatus,grade, quasar. Suchconstraintswillallowustoestimatethecon- andtime intervalsofacceptable aspectsolutionandback- tribution of quasar winds to the enrichment of the ISM ground levels. The pointings of the observatory placed and IGM and obtain a better understanding of the con- PG 1115+080 on the back-illuminated S3 chip of ACIS. nection between black-hole and bulge growth in the host To improve the spatial resolution we removed a ± 0′.′25 galaxy. randomization applied to the event positions in the CXC ThroughoutthispaperweadoptaΛ-dominatedcosmol- processing and employed a sub-pixel resolution technique ogy with H = 70 km s−1 Mpc−1, Ω = 0.7, and Ω = developed by Tsunemi et al. (2001). 0 Λ M 0.3. In both the XMM-Newton and Chandra analyses we tested the sensitivity of our results to the selected back- 2. observations and data analysis groundand source-extractionregions by varying the loca- tions of the background regions and varying the sizes of PG1115+080wasobservedwithXMM-Newton (Jansen the source-extractionregions. We did not find any signifi- et al. 2001) on 2001 Nov 25, 2004 June 10, and 2004 cantchangeinthebackground-subtractedspectra. Forall June 26, for 62.9 ks, 81.2 ks and 86.3 ks, respectively. models of PG 1115+080 we included Galactic absorption It was also observed with the Advanced CCD imaging 4 The WWW sites listing the updates are located at http://asc.harvard.edu/ciao/releasenotes/history.html and http://xmm.vilspa.esa.es/external/xmm sw cal/calib/rel notes/index.shtml,respectively. Refined Outflow Constraints of PG 1115+080 3 due to neutral gas with a column density of N =3.5 × amorecomplexsituation. Thetemperatureofthe ionized H 1020 cm−2(Starketal. 1992). All quotederrorsareatthe absorberisnotcalculatedself-consistentlybyproperther- 90% confidence level unless mentioned otherwise with all mal balance in the absori model but considered as an parameters conservatively taken to be of interest except input parameter. We found a significant improvement in absolute normalization. fitqualityatthe>99.9%confidencelevelaccordingtothe F-test (see model 3 of Table 2). The best-fit rest-frame 2.1. Spectral Analysis of the XMM-Newton Observations energiesandwidths ofthe absorptionfeaturesformodel3 of PG 1115+080. for epochs 1, 2 and 3 are listed in Table 2. We note that thebest-fitenergiesofthesefeaturesarequitestablewhen WefirstsimultaneouslyfittedthePN,MOS1andMOS2 trying different models for the low-energy spectral com- spectra of PG 1115+080at each of the three epochs with plexity. As an independent check of the accuracy of the a model consisting of a power law with neutral intrinsic fits that used the absori model we also repeated several absorption at z = 1.72 (model 1 of Table 2). These fits ofthese fits using the warmabsorbermodel XSTAR.XSTAR support the presence of an intrinsic absorber with col- umn densities of N = 2.2+0.5 × 1021 cm−2, 4.6+0.5 × calculates the physical conditions and emission spectra of H −0.5 −0.5 photoionized gases. In the current analysis we use a re- 1021 cm−2, and 2.9+−00..77 × 1021 cm−2, for the three XMM- centimplementation of the XSTARmodel that canbe used Newton observations of PG 1115+080. The fits are not within XSPEC. acceptable in a statistical sense (see the reduced χ2 val- We foundthatthebest-fitparametersofthe low-energy ues in Table 2). The fit residuals show significant absorp- ionizedabsorberderivedusing the XSTARmodel werecon- tion atobserved-frameenergiesof2.5–5keV.Toillustrate sistentwiththe valuesfoundusingtheabsorimodel. We the presence of these features and absorption features be- performedfits tothespectraofPG1115+080atthe three low 0.6 keV, we fit the spectra from observed-frame 2– epochs using both absori and XSTAR to compare results 2.5 keV and 5–10 keV with a power-law model (modified and performed the estimates of the confidence contours by Galactic absorption) and extrapolated this model to of the best-fit parameters using the more time-efficient the energy ranges not fit (see Figure 1). For clarity we absori model. only showthe higherS/NratioPNdatainFigure 1;how- We investigatedwhetherthe observedfluxvariabilityof ever, all fits were performed simultaneously using the PN the high-energy absorption features was consistent with a andMOS1+2dataunlessmentionedotherwise. Thelower decrease in the strength of two absorption lines used to panelsinFigure1showthe∆χresidualsbetweenthebest- model the absorption features. The energies and widths fit power-law model and the PN data. We note that the of the absorption lines were held fixed between epochs. apparent variability in the amplitude of the absorption Specifically, we performed a simultaneous fit to all epochs features is not due to variations in the best-fit value of Γ incorporating model 2 of Table 2 with the difference that for the spectral fits performed within the observed-frame the energies and widths of the absorption lines were con- ranges of 2–2.5 keV and 5–10 keV. The best-fit values of strainedtobethesameandonlythenormalizationsofthe the photon indices for epochs 1, 2 and 3 were 1.91+0.10, −0.10 absorptionlines wereallowedto vary. Thesefits wereper- 1.88+−00..1100,and1.84+−00..1111. Forthepurposeofcomparingthe formed in the 1–10 keV band of the higher S/N PN data. absorptionresidualsbetweenepochsthephotonindicesfor Weobtainedaχ2of137.1for137dof,andthebest-fitrest- epochs 1, 2, and 3 were set to Γ = 1.9. We proceed in fit- frame energies of the lines were 7.5 ± 0.3 keV and 10.0 ± ting a variety of models to the data guided by the shape 0.4 keV. Allowing the energies and widths of the lines to and location of these identified absorption residuals. As vary resulted in a similar χ2 of 132.0 for 129 dof. We our first refinement we considered an absorbed power-law note that for the fits where the energies and widths were model with Gaussianabsorptionlines nearthe absorption constrainedto be the same between epochs we found that features appearing between observed-frame energies of 2 the best-fit values of the optical depths for the absorption and 5 keV(see model2 of Table 2). Hereafter,we refer to linesat7.5keVwereconsistentwithzeroforepochs2and theseabsorptionfeaturesasabs1andabs2. Thefitstothe 3. This was expected since an inspection of Figure 1 indi- spectra ofPG1115+080forepochs 1and2 containedtwo cates that there are no residuals for epochs 2 and 3 near absorption lines, and the fits for epoch 3 contained one the rest-frameenergyof7.5 keV(left verticaldashedline) absorption line. The inclusion of these absorption lines . We conclude that the strengths of the absorption lines in model 2 resulted in significant improvements of the fits for epochs 2 and 3 are too weak to infer any variability of comparedto thepreviousonesinmodel1atthe>99.9%, their energies or widths. 99.2% and 93.6% confidence levels (according to the F- We next replaced the ionized absorber with a partial- test) for epochs 1, 2 and 3, respectively. covering model and repeated the fits for each epoch. We The remaining most significant contributions to the find that fits that include partial-covering(see model 4 of large values of χ2 for model 2 arise from the residuals Table 2) provide χ2 values and quality-of-fit parameters below 0.6 keV. To model these residuals we replaced the comparable to those that included an ionized absorber. neutral absorber in our spectral model with an ionized Modelsthatincludedabsorptionedgesforthehigh-energy intrinsic absorber (see model 3 of Table 2). These low- absorptionfeaturesofPG1115+080forepoch 1werecon- energy residuals are commonly detected in moderate S/N sidered and rejected in the current analysis, that includes spectra of BAL quasars and are thought to arise from ab- the new calibration, as previously found in Chartas et al. sorptionbymultipleionsofO,Ne,Na,Mg,andFeand/or (2003). The basic problem with such edge models is that, partial covering. In particular,we used the absorimodel for plausible iron abundances, they predict too much ab- contained in XSPEC (Done et al. 1992). We note that the sorption at low energies. The inclusion of a broad (width absorimodelisjustafirstapproximationtowhatislikely 4 CHARTAS ET AL. allowed to vary in fits) Fe Kα emission line in all models withthephotoionizationcodeXSTARassumingaslab-like in Table2 forepoch1resultedinlargerreducedχ2 values stationary absorber with N = 9 × 1022 cm−2, logξ =2, H compared to those with no line included; therefore, there and logξ = 3.5. The assumed column density of 9 × is no evidence for statistically significant Fe emission. We 1022 cm−2 liesbetweentheestimatedvaluesofthecolumn alsotestedthe sensitivityofthe best-fitparametersto the densities of the high ionzation absorbers of PG 1115+080 energy ranges used for fitting the PN data by varying the listed in Table 4. At ionization levels of logξ ∼ 3.5 and low-energy boundary of the fits from 0.2 keV to 0.4 keV. rest-frame wavelengths in the range 1.5–2 ˚A (see panel a We did not find any significant change in the parameters of Figure 3) the Fe lines are quite isolated and are there- of the high-energy absorption lines. The exclusion of the fore apparent even with low-resolution spectroscopy. At 0.2–0.4keVPNdata,however,didresultinslightlypoorer rest-frame wavelengths in the range of 6.2–10 ˚A and at constraints on the ionization parameter of the gas caus- ionization parameters of logξ ∼ 2 the spectra contain a ing the low-energy absorption. A detailed description of larger density of strong absorption lines that are difficult the currentcalibrationstatusoftheXMM-Newton instru- to resolve and identify (see panel b of Figure 3). ments can be found on the XMM-Newton SOC WWW The derived column densities for the low-ionization ab- site. A recent energy-dependent re-working of the PN re- sorber of PG 1115+080(see Table 2) for the three epochs sponse has resulted in a significant improvement in the described in this paper are lower than those found in X- calibration down to 0.2 keV and brings the PN and MOS ray observations of typical BAL quasars. For example, detectors into better agreement. the column densities of BAL quasars as inferred from X- In Figure 2a we show the 68%and 90% confidence con- ray observations lie in the range of 0.4–250 × 1023 cm−2 tours(basedonmodel3ofTable2)forthephotonindices (e.g., Gallagher et al. 2002 and Punsly 2006 present versus column densities for the three epochs. The col- BAL-quasar column densities derived from X-ray obser- umn densities for the low-energy absorbers in models 3 vations). We emphasizethatmostcurrentcolumn-density and 4 are consistent with no variation. A possible weak estimatesderivedfromX-rayobservationsarepoorlycon- variation of the photon index between epochs 1 and 3 is strained due to the limited S/N of most existing X-ray detected. 68% and 90% confidence contours for the ion- spectra of BAL and mini-BAL quasars. The presence of ization parameter of the low-energy absorber versus col- ionized or partially covering absorption cannot be accu- umn density are presented in Figure 2b. The absorber rately inferred from the analysis of individual X-ray spec- in PG 1115+080 appears to have been more ionized dur- tra of most BAL quasars and if present would in gen- ing epoch 1. In particular, during epoch 1 the ionized eral lead to an underestimate of the column densities. absorber is characterized by an ionization parameter of On the other hand, the large lensing magnifications of ξ = L/nr2 = 145+84 erg cm s−1 and a hydrogen column APM 08279+5255and PG 1115+080have providedmod- −52 density ofN =1.33+0.58 ×1022 cm−2,andduringepoch erate S/N X-ray spectra for these objects resulting in im- H −0.35 2byξ =56+30 ergcms−1 andahydrogencolumndensity provedconstraintscomparedtounlensedquasars. Wenote −23 that the column density of PG 1115+080appears to have ofN =1.41+0.36 × 1022 cm−2, where L is the integrated H −0.27 varied significantly since an earlier observation. Specifi- 5 eV–300 keV incident luminosity, n is the electron num- cally,duringaROSATPSPCobservationofPG1115+080 ber density of the absorber,and r is the distance between on21 November 1991the 0.2–2 keVX-ray flux of this ob- the absorber and ionizing source. ject was found to be at a minimum level (low-state) of Figure 2 suggests that the ionization parameter de- about 0.4 × 10−13 erg s−1 cm−2 (see Figure 6 of Char- creased in the order of epochs 1, 3, and 2. This trend tas 2000) and the intrinsic column was constrained to be in ionization parameter is evident in the PN data shown 1.2 ± 1.1 × 1023 cm−2. This column density is consis- in Figure 1. For epoch 1, as shownin Figure 1a,the spec- tentwith the typicalvaluesobservedin BAL quasars. We trum at low energies ∼ 0.2–0.4keV appears less absorbed estimated the optical-to-X-ray spectral slope, quantified relative to epochs 2 and 3, as expected for a more highly as α = log(f /f )/log(ν /ν ) (Tanan- ox 2 keV ˚ 2 keV ˚ ionized absorber where the opacity from low-z metals is 2500A 2500A baum, et al. 1979), where f and f are the flux smaller. Additional confirmation for the trend in ioniza- 2 keV ˚ 2500A tion parameter is provided in §3.1. densities at 2 keV and 2500 ˚A in the quasar rest-frame, Our spectral analysis of PG 1115+080implies the exis- respectively. For the cases of unabsorbed(absorbed) flux tence of two distinct absorbers with very different ioniza- densities at 2 keV for the epoch 1 XMM-Newton observa- tion parameters. In particular, if we interpret the appar- tion of PG 1115+080we find α = -1.63(-1.67). We note ox ent detection of absorption lines at rest-frame energies of that from the absorbed flux density at 2 keV for the 1991 7.27+0.37 keV and 9.79+0.96 keV as being due to Fe, the ROSAT PSPC observation (low-state) of PG 1115+080 −0.10 −1.05 most conservative assignment (giving the lowest outflow- we find α = -1.97. Due to the limited quality of the ox ing velocity) is to highly ionized Fe XXV which requires ROSAT PSPC spectrum useful constraints on the unab- logξ ∼3.5. Incontrast,thesignificantabsorptionfeatures sorbed value of α could not be obtained for the 1991 ox belowrest-frameenergiesof∼1.6keVarebestfitwithan observation. Forcomparisonthevalue ofα expectedfor ox ionization parameter of the order of logξ = 2 (see Table a typical AGN with a 2500 ˚A luminosity density similar 2). A justification of the interpretation of the absorption to that of the epoch 1 observationof PG 1115+080based lines is given in §3 and presented in more detail in Char- on the recent empirical relation of Steffen et al. 2006 (see tas etal. (2002,2003). To illustrate better the absorption equation2ofSteffenetal.) isα =-1.58. Thermsofα ox ox lines that can arise at these two distinct ionization levels at log(L ) = 30.8 is about 0.2 (see Table 5 of Steffen ˚ 2500A we show in Figure 3 the expected absorption spectrum of et al.). We conclude that PG 1115+080 is slightly X-ray PG 1115+080in the 1−10 ˚A rest-frame rangeas derived Refined Outflow Constraints of PG 1115+080 5 weakduringthethreeobservationepochspresentedinthis PG1115+080andAPM08279+5255resultedinimprove- paperbutstillwithinthescatterofα detectedintypical ments in the fits comparedto models that did not include ox AGN. absorptionlinesatthegreaterthan99.9%confidencelevel. In the present work we also confirm this significant detec- 3. discussion tion by re-analyzing the same data with updated instru- mental calibration files. OuranalysesoftheChandra andXMM-Newton spectra d) We made a plausibility argument based on the en- ofPG1115+080andAPM08279+5255presentedhereand ergies of the identified lines, and the known energies of in the discovery papers of Chartas et al. (2002, 2003) in- absorption lines from all the abundant elements, that the dicate that the high-energy broad absorption features de- detected absorption lines are associated with highly ion- tectedintheseobjectsaresignificantatthe>99.9%confi- ized Fe Kα absorption. Based on this identification we dencelevel. Ourclaimeddetectionsofrelativisticwindsin estimatedthe velocities of the outflowing X-rayabsorbing the X-rayspectraofPG1115+080andAPM08279+5255 material. were based on the following steps: Severalof the key derivations that follow in this section a) We chose a standard approach to fitting the X-ray assumeourinterpretationthatthehigh-energyabsorption spectra of PG 1115+080 and APM 08279+5255. In par- is due to lines arisingfrom highly ionized Fe XXV and/or ticular, we first started with a simple model and added FeXXVIandthattheoutflowvelocitiesoftheseabsorbers components of increasing complexity motivated both by range between 0.1c and 0.4c. Here we provide a sum- the residuals of the spectral fits and our understanding of mary of our justification of this interpretation that was the physics and structure of mini-BAL and BAL quasars. presented in more detail in Chartas et al. (2002, 2003). The first model consisted of a simple power-law modified Our analyses of the Chandra spectrum of by intrinsic absorption. This choice was inspired by nu- APM 08279+5255 and the XMM-Newton spectrum of merousempiricalandtheoreticalstudiesthatindicatethat PG 1115+080 showed strong evidence for the presence of the mainX-raycontinuumemissioncomponentofquasars absorptionlinesatrest-frameenergiesof8.05+0.10keVand −0.08 (e.g., rest-frame 2–30 keV) is produced by inverse Comp- 9.79+0.20 keV for APM 08279+5255 and rest-frame ener- ton scattering of soft photons from the accretion disk by −0.19 giesof 7.27+0.37 keVand 9.79+0.96 keVfor PG1115+080. hot electrons in a corona (e.g., Haardt & Maraschi 1991; −0.10 −1.05 For APM 08279+5255 we assumed that the two lines Reeves & Turner 2000). An inspection of the residuals are produced by resonant absorption due to Fe xxv to fits with simple absorbed power-law models indicated 1s–2p (Fe XXVI Lyα), and inferred that the X-ray ab- thepresenceofsignificantcomplexabsorptionresidualsat sorbers are outflowing with velocities of ∼ 0.20c(0.15c) rest-frame energies above ∼6.4 keV and rest-frame ener- and ∼ 0.4c(0.36c), respectively. For PG 1115+080 we gies below ∼1.5 keV. assumed that the high energy absorption is due to two b) To account for these fit residuals we considered a lines produced by resonant absorption due to Fe xxv variety of spectral models. Several of these models were 1s–2p (Fe XXVI Lyα), and inferred that the X-ray ab- rejected in the discovery papers and this work. The list sorbersareoutflowingwithvelocitiesof∼ 0.1c(0.05c)and of rejected models includes the following: Broken power- ∼ 0.4c(0.36c), respectively. The wind geometry that we lawmodelswithandwithoutasoftblackbodycomponent, assume is based on the unified BAL model (e.g., Wey- models that included one and two absorptionedges to ac- mann et al. 1991; Murray et al. 1995; Proga 2000; Elvis countforthehigh-energyabsorption,modelsthatincluded 2000). In this model the AGN outflow originates in the neutral absorption at the source to account for the low- accretion disk from a narrow range of radii. This wind energy absorption,andabsorptionby possible intervening rises initially almost perpendicular to the accretion disk materialtoaccountforthehigh-energyabsorption. Single and becomes more radial and equatorial at larger radii anddoublepartial-coveringmodelsthattrytofitthehigh- to form a bi-cone. The unified BAL model proposes that energy features of PG1115+080with ironedges requirea most of the observed range of absorption line-widths can curvatureofthelow-energycontinuumthatisinconsistent beexplainedwithorientationandwithavelocitygradient with our data. Additional data would be required to rule in the outflowing stream. The hydrodynamical models of out even more complex partial-coveringmodels. Proga et al. (2000) indicate that the density of the wind c) Motivated by the structure of the low and high- peaks near an inclination angle of about 70 degrees and energy residuals, the fact that PG 1115+080 and mostofthe outflowisconfinedwithin±10degreesofthis APM 08279+5255 exhibit broad absorption lines in the angle. We notethatProga&Kallman(2004)showedthat UV (e.g., the properties of the UV broad absorption lines the wind inclination angle can be smaller and the wind in PG 1115+080 and APM 08279+5255 have been pre- doesnotneedtobeequatorialifitismostlydrivenbyUV sented in Michalitsianos et al. (1996) and Irwin et al. radiation from the accretion disk. (1998), respectively), and previous claims of detections For the velocity calculations we considered the special- of low-energy ionized absorption in the X-ray spectra of relativisticvelocitycorrectionandassumedthatthe angle BAL quasars, we fitted the spectra with a model that θ between the wind velocity and our line of sight is 20◦. consisted of a power-law modified by ionized absorption This angle is not constrained with the present data; how- at the source and two iron lines at the source to ac- ever, hydrodynamical simulations indicate that the BAL count for the high-energy absorption. This model pro- winddivergenceanglemayrangebetween10◦–30◦depend- ducedacceptablefitstotheX-rayspectraofPG1115+080 ing on the location of the inner radius of the disk. The and APM 08279+5255 and resulted in a significant im- velocity estimates are not sensitive to this range of an- provement over all other models considered. We note gles. For example, the inferred outflow velocity (assum- that models that included high-energy absorption lines in 6 CHARTAS ET AL. ing θ ∼ 20◦) for the 9.79 keV absorption line detected in all epochs. The best-fit 2–2.5 keV and 5–10 keV spectral PG1115+080wouldvaryfrom−6%to+13%forθranging slopes for all three epochs have essentially the same value between 10◦–30◦. ofΓ=1.9forepochs1,2,and3respectively,andtheS/N Of all the abundant elements, iron absorption lines of the three observations are similar with effective expo- would be the closest in energy to the observed features. sure times and source count rates listed in Table 1. We As shown in Figure 3 the strongest lines with rest-frame thereforedonotexpectvariationsofthebest-fitmodelsor energies indicated in parenthesis near the observed ones S/Nofthe spectratocontributestronglytothe variations are: Fe XXV 1s–2p (6.7 keV), Fe XXV 1s–3p (7.88 keV), in the high-energy residuals. We find total ∆χ residuals Fe XXVI Lyα (6.97 keV), Fe XXVI 1s–3p (8.25 keV), S of ∆χ = −40 ± 3, ∆χ = −26 ± 4, and ∆χ = −20 E1 E2 E3 XV 1s–2p (2.46 keV), S XVI Lyα (2.62 keV) Si XIV Lyα ± 3 for epochs 1, 2, and 3, respectively. (2.005 keV), Ar XVIII Lyα (3.321 keV), and Ca XX Lyα Asecondapproachtoquantifyingvariabilityofthehigh- (4.104keV)(basedontheenergiesofpermittedresonance energy absorption features relies on taking ratios of the lines of Verner et al. 1996). In this sense, our interpreta- spectra. This approach is independent of model assump- tion that the absorption lines are associated with highly tions andtakes into accountthe uncertaintiesin eachbin. ionized Fe K absorption is the most conservative one pos- Forthepurposeofthisanalysisthegroupingofthespectra sible (e.g., absorption lines from relativistic sulfur or oxy- was made identical for the three epochs. In Figure 4 we gen would require much larger blueshifts). We also inves- showtheratiosofthespectraofepoch1toepoch2(R ), 12 tigated whether intervening absorbers in the lens galax- epoch 1 to epoch 3 (R ), and epoch 2 to epoch 3 (R ). 13 23 ies or in possible damped Lyman alpha systems along the We find the mean of the ratioswithin the high-energyab- line of sight could explain the high-energy absorption fea- sorbed range 2.5–5.0 keV to be < R > = 0.83 ± 0.05, 12 tures. We concluded that the observed high-energy ab- <R > = 0.73 ± 0.05, and <R > = 0.88 ± 0.05. 13 23 sorption features cannot be produced by absorption in AsathirdmethodofestimatingthevariabilityoftheX- intervening systems and the most likely origin is intrin- ray broad absorption features we compared the strengths sic absorption by highly ionized iron. The intrinsic origin of the best-fit Gaussian lines for the three epochs. Specif- of the absorption has been confirmed with the detection ically, we found the total absorbed photon fluxes in the of significant flux variability of the absorption features in Gaussian absorption lines for epochs 1, 2, and 3 were APM 08279+5255 (see Figure 7 in Chartas et al. 2003) (2.4 ± 1.1) × 10−5 photons cm−2 s−1, (0.49 ± 0.20) × and now with the detection of significant variability in 10−6 photons cm−2 s−1, and (0.34 ± 0.16) × 10−6 pho- PG1115+080(seeFigure4)asdiscussedinmoredetailin tons cm−2 s−1, respectively. The decrease in the totalab- §3.1. From the present data we can infer significant vari- sorbed photon fluxes in the absorption lines is consistent ability of the normalization of the absorption lines, how- with the results of the other two methods. ever,higherquality spectrawill be requiredto distinguish All approachesimply significantvariabilityofthe X-ray if there is any energy variability of the absorption lines. broadabsorptionfeaturesinPG1115+080betweenepochs Assuming this interpretation for the relativistic X-ray 1 and 2 separated by 0.92 yr (rest-frame) and marginal absorbing material, we present in the following sections variabilityofthe X-raybroadabsorptionfeaturesbetween estimates of the mass-outflow rate and efficiency of the epochs 2 and 3 separatedby 5.9 d (rest-frame). The 5.9 d outflow in PG 1115+080 and APM 08279+5255. These variability, if real, is consistent with a relatively small estimates are important for addressing the fundamental launching radius of the X-ray absorber. Specifically, as- issue of whether quasar outflows are capable of influenc- suming a radiatively driven wind 5 and using equation 1 ing their host galaxies and black-hole growth. Since the of Chartas et al. (2003) we estimate that a launching ra- present observations of PG 1115+080 do not resolve the dius of about 7 R is needed for the absorber to reach a s lensed images we also estimate the influence of this effect terminal velocity of about 0.4 c. At a launching radius on our main conclusions. of 7 R the time to reach 90% of the terminal velocity is s about9 d. This time-scale is consistentwith the marginal 3.1. Variability of the outflow in PG 1115+080 detection of variability of the Fe absorption lines between OurspectralanalysisoftheXMM-Newton observations epochs 2 and 3. of PG 1115+080presented in §2.1 indicates possible vari- A model-independent confirmation of the trend of the ability of the X-ray broad absorption features between ionizationparameter presented in §2.1 can also be seen in different epochs. To quantify the significance of these Figure4. Inparticular,Figure4showssignificantpositive variations we determined the sum of the residuals of ∆χ residuals of the ratios R and R and negative residuals 12 13 between the best-fit 2–2.5 keV and 5–10 keV continuum of the ratio R within the 0.2–0.4 keV band. This indi- 23 model and data in the energy range 2.5–5.0keV based on cates that the 0.2–0.4 keV flux decreased in the order of thespectralfitspresentedinFigure1. Asdescribedin§2.2 epochs 1, 3, and 2. This trend in soft X-rayflux is consis- we fit the spectra within the energy ranges of 2–2.5 keV tent with the observedtrends of the 0.2–0.4keVresiduals and 5–10 keV with Galactic absorption and a power-law shown in Figure 1 and the observed trend in ionization model and extrapolated this model to the energy ranges parameter derived from our spectral analysis and shown not fit (see §2.1). For this summed-∆χ method to be in Figure 2. We note that the ratio of the EPIC PN spec- statistically valid, the best-fit 2–2.5 keV and 5–10 keV tra is not sensitive to the precise value of the calibrated models and the S/N of the spectra need to be similar for effective area (the effective area cancels out in the ratio) 5 WefindthebolometricluminosityofPG1115+080tobeLBol =3.3×1046 ergs−1. ToestimateLBol weusedtheempiricalrelationLBol = fBC(3000˚A)ν Lν(3000˚A)µ−1, where fBC(3000˚A) = 5.3 is the luminosity-dependent bolometric correction obtained from equation (21) of Marconi et al. (2004) and µ is the flux magnification assumed to be ∼ 25 based on lensing models of this system (e.g., Impey et al. 1998). ThevalueofLν(3000˚A)=1.57×1032 ergs−1 Hz−1 forPG1115+080wastakenfromNeugebauer etal. (1987). Refined Outflow Constraints of PG 1115+080 7 as long as the response of the PN detector remained the ofaboutξ > 1000ergcms−1 absorptionfromhighly ion- ∼ same between observations. The PN detector response is izedlinessuchasSiXIV,SXVIandFeXXVareexpected known to be stable with time accordingto the calibration tobeproduced. Asthe ionizationparameterincreasesthe status of the XMM-Newton instruments (see footnote 4). absorber becomes completely ionized and line-driving be- SignificantvariabilityofPG1115+080hasalsobeenre- comes no longer important. This interpretation predicts ported in the optical, UV and X-ray bands. Weymann rapid changes (compared to those observed in UV broad et al. (1980) report several absorption lines as being vis- absorption lines) in the observed energies and equivalent ible in the May 1980 observations of this quasar. Young widths of the X-ray BALs. We note that magnetically et al. (1982) detect only weak absorption features blue- drivenoutflowsisalsoapossibility,asthismightbeneeded wardsofthe CIVline intheir December 1980spectrumof toreachsuchhighvelocitiesifthematerialbecomeshighly PG1115+080anddo notdetectthe lines seeninthe May ionized quickly. 1980 observations. Strong and rapid variability of broad The average continuum flux of PG 1115+080 has not absorption lines blueward of OVI in PG 1115+080 were varied by more than ∼ 20% between epochs as indicated reported by Michalitsianos et al. (1996). Chartas et al. by the observed PN count rates listed in Table 1. The (2000) reported a decrease by a factor of about 13 of the fractional change of the continuum flux can also be seen 0.2–2keVfluxbetween1979December5and1991Novem- in the ratio plots shown in Figure 4. There is thus no ber 21 and an increase by a factor of about 5 between the correlation between the depth of the high-energy broad 1991 November 21 and 1994 May 27 observations of PG absorption features and continuum flux level. 1115+080. Several recent theoretical studies of high-velocity out- 3.2. Efficiency of the quasar outflow in PG 1115+080. flows from quasars (e.g., Everett & Ballantyne 2004; Sim One ofthe keyandhighly uncertainparametersusedin et al. 2005) assume that the high-ionization absorption recent theoretical models that describe black-hole growth lines of S XIV, S XVI and Fe XXV observed in several andstructureformationistheefficiencyofthequasarout- quasars (i.e., PG 1211+143 and PG 0844+349 reported flow. The efficiency is defined as the fraction of the total by Pounds at al. 2003a,b) originate from gas that has bolometricenergyreleasedoveraquasar’slifetimeintothe reached its terminal velocity. This terminal velocity is ISM and IGM in the form of kinetic energy injection and then approximated with the escape velocity from the re- can be expressed as: gion from which the wind is launched, resulting in the approximation R ∼ R (c/v )2, where v is the observed outflowlvaeulnochcity ansd Rsoibssthe Schwarzosbcshild ra- ǫ = 1M˙ivw2ind,i =2πf R(R/∆R )N m vw3ind,i (1) dius. Forv intherange0.1–0.4ctheexpectedlaunching k,i 2 L c,i i i i H,i p L obs Bol Bol radiiforPG1115+080willlieintherange100 – 6Rs. For where M˙ is the mass-outflow rate of component i, v an estimate of the black-hole mass we used the empirical i wind,i is the outflow velocity of the X-ray absorber of compo- relation (equation 7) of Vestergaard & Peterson (2006). nent i, f is the global covering fraction of the absorber Specifically, we find the mass of the black hole and the c,i of component i, ∆R is the thickness of the absorber at Schwarzschild radius to be Mbh = 1.1 × 109 M⊙ and Rs radius R of componient i, N is the hydrogen column = 2GM /c2 = 3.1 × 1014 cm, respectively, based on the i H,i bh density of component i, and L is the bolometric pho- FWHM of CIV of 4700 km s−1 taken from the Hale spec- Bol ton luminosity of the quasar. We focus on estimating ǫ trumofPG1115+080(Youngetal. 1982),andthe1350˚A k for the outflow in PG 1115+080observedduring epoch 1. luminosity density of L = 9.48 × 1031 µ−1 erg s−1 Hz−1 ν The velocities of the different outflowing X-ray absorbers taken from Neugebauer et al. (1987), where µ is the flux are listed in Table 3. We assume a conservatively wide magnificationassumedtobe∼25basedonlensingmodels range for the covering factor of f =0.1–0.3 based on the of this system (e.g., Impey et al. 1998). c observed fraction of BAL quasars (e.g., Hewett & Foltz Oneoftheassumptionsintheserecenttheoreticalstud- 2003) and a fraction R/∆R ranging from 1–10 based on ies is that the absorbers are observed near their termi- current theoretical models of quasar outflows (e.g., Proga nal velocities. As we proposed in our previous study of & Kallman 2000). As we argued in § 3.1 we are likely APM 08279+5255 another possibility is that we are ob- observing the absorbers as they are acceleratednear their serving the absorber as it is being accelerated near the launching radii. We have assumed absorber radii ranging launching radius. We speculate that initially the absorb- from 3–15 R based on the maximum observed X-ray ve- ingmaterialsuppliedbytheaccretiondiskhasarelatively s locity components. We used a Monte Carlo approach to low ionization parameter because of the large gas density estimatetheerrorsofM˙ andǫ . Thevaluesoff ,R/∆R, at the base of the wind. A low ionization parameter re- i k c andR wereassumedtohaveuniformdistributionswithin sultsinalargevalueoftheforcemultiplierthatcanleadto i their error limits. By multiplying these three distribu- significant acceleration through scattering in atomic reso- tions and with the appropriate constants from equation 1 nancelines(line-driving;theforcemultiplierrepresentsthe weobtainedthe distributionsofM˙ andǫ . We finally de- ratio by which the line opacity of the absorbing material i k increases the radiation force relative to that produced by termined the means of the distributions of M˙i and ǫk and Thomson scattering alone). Because of the intense radia- estimated the 68% confidence ranges. tion field of the central source the absorbing gas becomes We note that the last stable orbit can be even smaller increasingly ionized within a time that is much shorter in a Kerr black hole and the disk can extend down to than the time it takes for the absorber to reach its termi- the event horizon. This may also be possible even for nalvelocity. Whentheionizationparameterreachesvalues a Schwarzschild black hole, as argued recently by Agol & Krolik (2000). However, at radii smaller than 3 R s 8 CHARTAS ET AL. the gravitational redshift would begin to become impor- tobeabletoproduceasignificantchangeinthecombined tantwithE /E ∼p(1−2M /r )andonewould spectrum of PG 1115+080. obs emis bh emis havetoinvokeevenlargeroutflowvelocitiestoexplainthe detected blue-shifted Fe lines. 3.4. Comparison with APM 08279+5255 Thetotalhydrogencolumndensitiesofthe FeXXVKα BAL quasars are relatively weak in X-rays making the componentsabs1andabs2areconsistentwiththeonesde- detectionofbroadabsorptionfeaturesdifficultwithXMM- terminedinChartasetal. (2003). Themass-outflowrates Newton and Chandra. We have previously presented evi- and efficiencies of the outflowing components are listed in denceforarelativisticoutflowinthegravitationallylensed Table 4. z=3.91 BAL quasar APM 08279+5255. Following an ap- There are several systematic errors in estimating the proach similar to the one described in §3.1 and using radii and column densities of the absorbers that are diffi- the derived outflow properties presented in Chartas et culttoconstrainandhavenotbeenincludedinthepresent al. (2003), we find that the mass-outflow rates and ef- calculations of the outflow efficiencies. However, both ficiencies of the outflows associated with the two Fe ab- tthheastethseyystleemadattiocuunndceerretsatiinmtiaetsesaroeftohneeosuidtfleodwineffithcieensceinesse. sorbers in APM 08279+5255 are M˙abs1 = 1.7+−11..40 M⊙ Specifically, we have allowed the radii of the absorbers to yr−1, ǫk,abs1 = 10+−86 ×10−3, M˙abs2 = 3.3+−22..91 M⊙ yr−1, range between the last stable orbit of the black hole to 15 and ǫk,abs2 = 8+−75 × 10−2. The total efficiency of the R basedon the maximum inferredoutflow velocities. We outflow will be the sum of all components of all species s therefore expect any errorsin the estimates of the radiiof over time. Because of the low S/N of the available spec- the absorbers to lead to larger values of outflow efficien- traofAPM08279+5255additionalcomponentscannotbe cies. Possible saturation of the absorption lines will lead constrained, and we should consider the estimated total to anunderestimate of the column densities and therefore mass-outflowratesandefficienciesaslowerlimits. Thees- to an underestimate of the outflow efficiencies as well. timatedoutflowrateisanorderofmagnitudesmallerthan theaccretionrateinAPM08279+5255whichweestimate 3.3. Contamination from unresolved lensed images to be 1.8 × 10−3(L44/η)M⊙ yr−1 ∼ 40 M⊙ yr−1, where we assumed a typical accretion efficiency of η = 0.1. The lensed images of PG 1115+080 cannot be resolved with XMM-Newton, and therefore the spectra used in our 4. conclusions analyses contain contributions from all images. We esti- matehowthiseffectinfluencesourpreviousresultsbyfirst High-energy X-ray absorption is detected in three determiningX-rayfluxratiosbasedonrecentChandra ob- epochsofmonitoringobservationsofthemini-BALquasar servations of PG 1115+080that do resolve the lensed im- PG 1115+080. We interpret this absorption as due to ages. To estimate the X-ray flux ratios of PG 1115+080 linesarisingfromhighlyblueshiftedFeXXV1s–2pand/or wemodeledtheChandra imagesofA1,A2,BandC,with Fe XXVI Lyα. We find that the depths of the X-ray point-spreadfunctions(PSFs)generatedbythesimulation broad absorption features decreased significantly between tool MARX (Wise et al 1997). The X-ray event locations epochs 1 and 2 separated by 0.92 yr (proper-time) and were binned with a bin-size of 0′.′0246 to sample the PSF detect a marginal decrease between epochs 2 and 3 sepa- sufficiently (an ACIS pixel subtends 0′.′491). The simu- rated by 5.9 days (proper-time). The case for a relativis- latedPSFswerefittedtotheChandra databyminimizing tic flow on the basis of either the epoch 2 or 3 observa- theC-statisticformedbetweentheobservedandsimulated tions of PG 1115+080 alone is less compelling. We note images of PG 1115+080. The relative positions of the that models that included high-energy absorption lines images were fixed to the observed NICMOS values taken in PG 1115+080 and APM 08279+5255 resulted in im- fromtheCfA-ArizonaSpaceTelescopeLensSurvey(CAS- provements in the fits compared to models that did not TLES). In Figure 5 we show the Lucy-Richardson decon- include absorption lines at the greater than 99.9% confi- volvedimageoftheChandra observationofPG1115+080. dence level. We had previously reported rapid variability WefindthattherelativeX-rayfluxratiosinthe0.2–8keV over timescales of 1.8 weeks (proper-time) of X-ray BALs bandare[A1/C] =3.86±0.42,[A2/C] =1.85±0.22, in the quasar APM 08279+5255. Variability of similar full full and [B/C] = 0.93 ± 0.14. For comparison,the HST H- magnitudeandoversimilartime-scaleshasneverbeenob- full band flux ratios are [A1/C] = 4.06 ± 0.17, [A2/C] = served in UV BALs. H H 2.56 ± 0.12, and, [B/C] = 0.65 ± 0.04. The H-band Assuming the interpretation that the high-energy ab- H magnitudes were taken from the CASTLES WWW site.6 sorption features are due to lines arising from highly ion- TheX-rayfluxratiosofPG1115+080indicatethatabout ized Fe XXV we used the measured outflow velocities, 75%of the X-rayflux originatesfrom the close-separation column densities, and estimated launching radii to con- imagesA1andA2. Sincetheobservedtimedelaybetween strain the mass-outflow rates and outflow efficiencies for imagesA1andA2is ∆t =0.149±0.006dtheremay PG 1115+080 and APM 08279+5255. We find the out- be variability of the X-Ar1aAy2BALs within this time-scale flow efficiencies to be ǫk =0.64+−00..5420(68% confidence), and only if the absorbers are launched near the last stable or- ǫ = 0.09+0.07(68% confidence), respectively. These esti- k −0.05 bit of the black hole. The “long” time delays between mates include only contributions from observed compo- images werepresentedinSchechter etal. (1997)andhave nents and therefore should be considered as lower lim- image C leading images A1+A2 by 9.4 ± 3.4 d and image its. Our derived estimates of the efficiency of the out- C leading image B by 23.7 ± 3.4 d. Since images B and flows in mini-BAL quasar PG 1115+080and BAL quasar C are considerably weaker than A1+A2 they are unlikely APM08279+5255,whencomparedto values predictedby 6 TheCASTLESWWWsiteislocatedathttp://cfa-www.harvard.edu/glensdata/. Refined Outflow Constraints of PG 1115+080 9 recentmodelsofstructureformation(SO04;G04;SDH05), imply that these winds will have a significant impact on shaping the evolution of their host galaxies and in regu- lating the growth of the central black hole. We note that our reported values of the mass-outflow rates and outflow efficiencies represent instantaneous quantities. Because of the detected variability of the out- flows in PG 1115+080and APM 08279+5255it is appro- priate to compare the outflow efficiency incorporated in theoreticalmodelswiththeaverageofthisobservedquan- tity overa period thatcoversthe rangeof variability. Ad- ditional monitoring observations are needed to constrain better the variability of the outflow properties. The fraction of BAL quasars with highly blueshifted Fe absorption lines is not well constrained from avail- able X-ray observations because of the relatively poor- to-moderate S/N X-ray spectra of BAL quasars ob- served to date. We have shown that the two X-ray brightest mini-BAL and BAL quasars, APM 08279+5255 and PG 1115+080, show significant highly ionized and blueshifted Fe lines. The third brightest BAL quasar known to date CSO 755 shows a hint of an Fe absorption feature inthe rest-frameenergyrange7.1–9.3keV(Shem- mer et al. 2005). Higher quality observationsarerequired to confirmthis result. TheX-rayspectraofthe remaining observed BAL quasars are of lower S/N and do not pro- videinterestingconstraintsonthepresenceofhigh-energy Fe absorption features. Todate,themostinterestingconstraintsonquasarout- flows are derived from X-ray and UV spectroscopic anal- yses. Current observations of BAL quasars imply that the X-ray and UV bands may be sampling different parts of the absorbing outflows. A large fraction of the high velocity BAL outflow is apparently visible only in the X- ray band. Unfortunately, the limited number and qual- ity of available X-ray spectra leave large uncertainties in the mass-outflowrate. Absorptionstudies ofquasarssuch as PG 1115+080 and APM 08279+5255 with future high resolution, high throughput X-ray spectroscopic missions such as Constellation-X hold great promise for pinning down this crucial quantity. We thank Michael Eracleous for fruitful discussions re- lated to AGN outflows. We thank Ohad Shemmer for pointing us to the luminosity-dependent bolometric cor- rection. We greatly appreciate the useful comments made by the refereethatledtoimprovedestimates ofthe black- hole mass and bolometric luminosity of PG 1115+080. GC acknowledges financial support from NASA grants NNG04GK83G and NAG5-13543. WNB acknowledges fi- nancial support from NASA LTSA grant NAG5-13035. Support for SCG was provided by NASA through the Spitzer Fellowship Program, under award 1256317. DP acknowledges support provided by NASA through grants HST-AR-10305.05-A and HST-AR-10680.05-A from the Space Telescope Science Institute, which is operated by theAssociationofUniversitiesforResearchinAstronomy, Inc., under NASA contract NAS5-26555. 10 CHARTAS ET AL. REFERENCES Agol,E.,&Krolik,J.H.2000,ApJ,528,161 Kaastra,J.S.,Steenbrugge, K.C.,Raassen,A.J.J.,vander Meer, R.L.J.,Brinkman,A.C.,Liedahl,D.A.,Behar,E.,&deRosa, Aldcroft,T.L.,&Green,P.J.2003,ApJ,592,710 A.2002, A&A,386,427 Arav,N.,Li,Z.,&Begelman,M.C.1994,ApJ,432,62 Kaspi,S.,Brandt,W.N.,George, I.M.,etal.,2002,ApJ,574,643 Arav,N.,Barlow,T.A.,Laor,A.,Sargent,W.L.W.,&Blandford, Kaspi,S.,&Behar,E.2006,ApJ,636,674 R.D.1998,MNRAS,297,990 Kraemer, S. B., Crenshaw, D. M., Hutchings, J. B., et al., 2001, Arnaud,K.A.1996,ASPConf.Ser.101:AstronomicalDataAnalysis ApJ,551,671 SoftwareandSystemsV,5,17 Kriss, G. A. 2002, in ASP Conf. Ser 255, Mass Outflow in Active Barlow, T. A., Hamann, F., & Sargent, W. L. W. 1997, ASP Galacatic Nuclei: Perspectives, ed. D. M. Crenshaw, S. B. Conf.Ser.128:MassEjectionfromActiveGalacticNuclei,128,13 Kraemer,&I.M.George(SanFrancisco;ASP),69 Braito,V.,DellaCeca,R.,Piconcelli,E.,etal.,2004,A&A,420,79 Krolik,J.H.&Voit,G.M.1998, ApJ,497,L5 Chartas,G.2000,ApJ,531,81 McKernan,B.,Yaqoob,T.,&Reynolds,C.S.2004,ApJ,617,232 Chartas, G., Brandt, W. N., Gallagher, S. C., & Garmire, G. P. McKernan, B., Yaqoob, T., & Reynolds, C. S. 2005, MNRAS, 361, 2002,ApJ,579,169 1337 Chartas,G.,Brandt,W.N.,&Gallagher,S.C.2003, ApJ,595,85 Maloney,P.R.,&Reynolds,C.S.2000,ApJ,545,L23 Crenshaw, D. M., Kraemer, S. B., Boggess, A., Maran, S. P., Mathur,S.&Williams,R.J. 2002,ApJ,589,L1 Mushotzky, R.F.,&Wu,C.1999,ApJ,516,750 Michalitsianos,A.G.,Oliversen,R.J.,&Nichols,J.1996,ApJ,461, Crenshaw, D. M., Kraemer, S. B., Gabel, J. R., et al., 2003, ApJ, 593 594,116 Murray,N.,Chiang,J.,Grossman,S.A.,&Voit, G.M.1995, ApJ, Done,C.,Mulchaey,J.S.,Mushotzky,R.F.,&Arnaud,K.A.1992, 451,498 ApJ,395,275 Netzer,H.,Kaspi,S.,Behar,E.,etal.,2003, ApJ,599,933 Elvis, M., Wilkes, B. J., McDowell, J. C., Green, R. F., Bechtold, J., Willner, S. P., Oey, M. S., Polomski, E., & Cutri, R., 1994, Neugebauer, G., Green, R. F., Matthews, K., Schmidt, M., Soifer, ApJS,95,1 B.T.,&Bennett, J.1987,ApJS,63,615 Elvis,M.2000,ApJ,545,63 Porquet, D., Reeves, J. N., O’Brien, P., & Brinkmann, W. 2004, A&A,422,85 Everett, J.E.,&Ballantyne, D.R.2004,ApJ,615,L13 Pounds, K. A., King, A. R., Page, K. L., & O’Brien, P. T. 2003a, Everett, J.E.2005, ApJ,631,689 MNRAS,346,1025 Gallagher, S. C., Brandt, W. N., Chartas, G., & Garmire, G. P. Pounds, K. A., Reeves, J. N., King, A. R., Page, K. L., O’Brien, 2002,ApJ,567,37 P.T.,&Turner,M.J.L.2003,MNRAS,345,705 Garmire,G.P.,Bautz,M.W.,Ford,P.G.,Nousek,J.A.,&Ricker, Proga,D.,Stone,J.M.,&Kallman,T.R.2000, ApJ,543,686 G.R.2003,Proc.SPIE,4851,28 Proga,D.,&Kallman,T.R.2004,ApJ,616,688 Goodrich,R.W.1997,ApJ,474,606 Punsly,B.2006,ApJ,647,886 Granato, G. L., De Zotti, G., Silva, L., Bressan, A., & Danese, L. 2004,ApJ,600,580 Reeves,J.N.,O’Brien,P.T.,&Ward,M.J.2003, ApJ,593,L65 Green, P.J.,Aldcroft,T.L.,Mathur,S.,Wilkes,B.J.,&Elvis,M. Reeves,J.N.,&Turner,M.J.L.2000,MNRAS,316,234 2001,ApJ,558,109 Scannapieco, E.,&Oh,S.P.2004,ApJ,608,62 Green,R.F.,Pier,J.R.,Schmidt,M.,Estabrook,F.B.,Lane,A.L., &Wahlquist, H.D.1980,ApJ,239,483 Schechter,P.L.,Bailyn,C.D.,Barr,R.,etal.,1997,ApJ,475,L85 Hamann,F.,Barlow,T.A.,Beaver,E.A.,Burbidge,E.M.,Cohen, Shemmer, O., Brandt, W. N., Gallagher, S. C., Vignali, C., Boller, R.D.,Junkkarinen,V.,&Lyons,R.1995,ApJ,443,606 Th.,Chartas,G.,&Comastri,A., 2005, AJ,130,2522 Hasinger,G.,Schartel,N.,&Komossa,S.2002,ApJ,573,L77(H02) Sim,S.A.2005,MNRAS,356,531 Hewett, P.C.,&Foltz,C.B.2003, AJ,125,1784 Spitzer,L.1978,PhysicalProcessesintheInterstellarMedium(New York:Wiley) Hopkins, P. F., Hernquist, L., Cox, T. J., Di Matteo, T., Martini, P.,Robertson,B.,&Springel,V.2005,ApJ,630,705 Springel,V.,DiMatteo, T.,&Hernquist,L.2005,ApJ,620,L79 Hopkins,P.F.,Hernquist,L.,Cox,T.J.,DiMatteo,T.,Robertson, Srianand,R.,&Petitjean, P.2000,A&A,357,414 B.,&Springel,V.2006, ApJS,163,1 Stark,A.A.,Gammie,C.F.,Wilson,R.W.,Bally,J.,Linke,R.A., Impey, C. D., Falco, E. E., Kochanek, C. S., Leha´r, J., McLeod, Heiles,C.,&Hurwitz,M.1992,ApJS,79,77 B.A.,Rix,H.-W.,Peng,C.Y.,&Keeton,C.R.1998,ApJ,509, 551 Steffen, A. T., Strateva, I., Brandt, W. N., Alexander, D. M., Koekemoer, A. M., Lehmer, B. D., Schneider, D. P., & Vignali, Irwin,M.J.,Ibata, R.A.,Lewis,G.F.,&Totten, E.J.1998, ApJ, C.2006,AJ,131,2826 505,529 Stru¨der,L.,Briel,U.,Dennerl,K.,etal., 2001,A&A,365,L18 Jansen, F.,Lumb,D.,Altieri,B.,etal.2001,A&A,365,L1