Mon.Not.R.Astron.Soc.000,1–??(2003) Printed2February2008 (MNLATEXstylefilev1.4) Exploring the complex X-ray spectrum of NGC 4051. K.A.Pounds1, J.N.Reeves2, A.R.King1 and K.L.Page,1 1 Department of Physics and Astronomy, University of Leicester,Leicester, LE1 7RH, UK 2 Laboratory for High Energy Astrophysics, NASA Goddard Space Flight Center, Greenbelt, MD 20771, USA Accepted ;Submitted 4 0 0 ABSTRACT 2 Archival XMM-Newton data on the nearby Seyfert galaxy NGC 4051, taken in rela- n tivelyhighandlowfluxstates,offerauniqueopportunitytoexplorethecomplexityof a itsX-rayspectrum.We findthe hardX-raybandtobe significantlyaffectedbyreflec- J tion from cold matter, which can also explain a non-varying,narrowFe K fluorescent 6 line.We interpretmajordifferencesbetweenthe highandlowfluxhardX-rayspectra 1 in terms of the varying ionisation(opacity) of a substantialcolumn of outflowing gas. An emission line spectrum in the low flux state indicates an extended region of pho- 2 toionised gas. A high velocity, highly ionised outflow seen in the high state spectrum v can replenish the gas in the extended emission region over ∼103 years, while having 7 sufficient kinetic energy to contribute significantly to the hard X-ray continuum. 5 2 Key words: galaxies:active – galaxies:Seyfert: general– galaxies:individual: NGC 0 4051 – X-ray: galaxies 1 3 0 / h p 1 INTRODUCTION NGC 4051 is a low redshift (z = 0.0023) narrow line - Seyfert 1 galaxy, which has been studied over much of o TheadditionalsensitivityofXMM-NewtonandChandrahas the history of X-ray astronomy. Its X-ray emission often r t emphasised the complexity in the X-ray spectra of AGN. varies rapidly and with a large amplitude (Lawrence et al. s While there is broad agreement that the X-ray emission 1985,1987),occasionallylapsingintoextendedperiodsofex- a : is driven by accretion onto a supermassive black hole, the treme low activity (Lamer et al. 2003). When bright, the v detailed emission mechanism(s) remain unclear. Significant broadbandX-rayspectrumofNGC4051appearstypicalof i X complexity - and diagnostic potential - is introduced by re- a Seyfert 1 galaxy, with a 2–10 keV continuum being well processing of the primary X-rays in surrounding matter. representedbyapowerlawofphotonindexΓ∼1.8–2,witha r a Scattering and fluorescence from dense matter in the pu- hardeningofthespectrumabove∼7keVbeingattributable tative accretion disc has been recognised as a major fac- to‘reflection’from‘cold’,densematter,whichmightalsobe tor in modifying the observed X-ray emission of bright the origin of a relatively weak Fe K emission line (Nandra Seyfert galaxies since its discovery 13 years ago (Nandra andPounds1994).However,NGC4051alsoexhibitsstrong et al. 1989, Pounds et al. 1990). Additional modification of spectral variability, apparently correlated with source flux. the observed X-ray spectra arises by absorption in passage Thenatureof thisspectral variability hasremained contro- throughionisedmatterinthelineofsighttothecontinuum versialsincetheGINGAdatawerealternativelyinterpreted X-ray source. The high resolution X-ray spectra obtained as a change in power law slope (Matsuoka et al. 1990) and with XMM-Newton and Chandra have shown the consider- by varying partial covering of the continuum source by op- ablecomplexityofthis‘warmabsorber’(egSakoetal.2001, tically thick matter (Kuniedaet al. 1992). Kaspiet al. 2002), including recentevidencefor high veloc- LaterROSATobservationsprovidedgoodevidencefora ity outflows (eg Chartas et al. 2002, Pounds et al. 2003a,b; flux-linked variable ionised absorber, and for a ‘soft excess’ Reevesetal.2003)whichconstituteasignificantcomponent below ∼1 keV (Pounds et al. 1994, McHardy et al. 1995, inthemassandenergybudgetsofthoseAGN.Inthispaper KomossaandFink1997).ExtendedASCAobservationsled wereport on thespectral analysis of twoXMM-Newton ob- Guainazzi et al.(1996) to report a strong and broad Fe K servations ofthebright,nearbySeyfert 1galaxy NGC4051 emission line (implying reflection from the inner accretion takenfrom theXMM-Newton dataarchive.Wefindfurther disc),andapositivecorrelation ofthehardpowerlawslope supportforthesuggestionmadeinanearlysurveyofXMM- withX-rayflux.A3-yearmonitoringcampaignofNGC4051 NewtonSeyfertspectra(PoundsandReeves2002),thatthe with RXTE, including a 150-day extended low interval in full effects of ionised absorption in AGN have often been 1998, produced clear evidence for the cold reflection com- underestimated. ponent (hard continuum and narrow 6.4 keV Fe K line) re- (cid:13)c 2003RAS 2 K.A.Pounds et al. mainingconstant,whileagainfindingtheresidualpowerlaw slope to steepen at higherX-ray fluxes(Lamer et al. 2003). More surprisingly, a relativistic broad Fe K line component wasfoundtobealwayspresent,evenduringtheperiodwhen theSeyfertnucleuswas‘switchedoff’(Guainazzietal.1998, Lameretal.2003).Oneotherimportantcontributiontothe extensiveX-rayliteratureonNGC4051camefromanearly Chandra observation which resolved two X-ray absorption line systems, with outflowing velocities of ∼2300 and ∼600 km s−1, superimposed on a continuum soft excess with sig- nificant curvature(Collinge et al.2001). Of particular inter- estinthecontextofthepresentanalysis,thehighervelocity outflow is seen in lines of the highest ionisation potential. The Chandra data also show an unresolved Fe K emission line at ∼6.41 keV(FWHM ≤2800 km s−1). In summary, no clear picture emerges from a review of the extensive data on the X-ray spectrum of NGC 4051, Figure 1. Background-subtracted EPIC pn data for the 2001 with thespectral variability being (mainly) dueto a strong May(black)and2002November(red)observationsofNGC4051 power law slope - fluxcorrelation, or tovariable absorption in(asubstantialcolumnof)ionisedmatter.Supportforthe former view has recently come from a careful study of the availableforspectralfittingfrom the2001 observation were soft-to-hard flux ratios in extended RXTE data (Taylor et 81.7 ksec (pn), 103.6 ksec (MOS2), 114.3 ksec (RGS1) and al. 2003), while the potential importance of absorption is 110.9 ksec(RGS2).Forthe2002 observation thefinalspec- underlined by previous spectral fits to NGC 4051 requiring tral data were of 46.6 ksec (pn), 101.9 ksec (MOS1and 2), columndensitiesoforder∼1023cm−2(egPoundsetal.1994, 51.6 ksec (RGS1) and 51.6 ksec (RGS2). Data were then McHardy et al. 1995). binned to a minimum of 20 counts per bin, to facilitate Given these uncertainties we decided to extract XMM- use of the χ2 minimalisation technique in spectral fitting. Newton archival data on NGC 4051 in order to explore its Spectral fitting was based on the Xspec package (Arnaud spectralcomplexities.Aftersubmissionofthepresentpaper, 1996). All spectral fits include absorption due to the NGC anindependentanalysisofthe2002NovemberEPICpndata 4051line-of-sightGalacticcolumnofNH =1.32×1020cm−2 by Uttley et al. (2003) was published on astro-ph, reaching (Elvisetal. 1989). Errorsarequotedatthe90% confidence different conclusions to those we find. We comment briefly level (∆χ2=2.7 for oneinteresting parameter). on these alternative descriptions of the spectral variability We analysed the broad-band X-ray spectrum of NGC of NGC 4051 in Section 9.4. 4051 integrated over the separate XMM-Newton observa- tions, noting the mean flux levels were markedly different, andperhapsrepresentativeofthe‘highstate’and‘lowstate’ 2 OBSERVATION AND DATA REDUCTION X-ray spectra of this Seyfert galaxy. [In fact the 2001 May X-ray flux is close to the historical mean for NGC 4051, NGC 4051 was observed by XMM-Newton on 2001 May but we will continue to refer to it as the ‘high state’ for 16/17 (orbit 263) for∼117ksec,andagain on 2002Novem- convenience]. To obtain a first impression of the spectral ber 22 (orbit 541) for ∼52 ksec. The latter observation was change we compare in figure 1 the background-subtracted timedtocoincidewithanextendedperiodoflowX-rayemis- spectra from the EPIC pn camera for orbits 263 and 541. sion from NGC 4051. These data are now public and have ThesamecomparisonfortheEPICMOS2data(notshown) been obtained from the XMM-Newton data archive. X-ray isessentially identical.From∼0.3–3 keVthespectralshape data are available in both observations from the EPIC pn is broadly unchanged, with the 2001 flux level being a fac- (Stru¨deretal.2001)andMOS2(Turneretal.2001)cameras, tor∼5higher.From∼3keVuptotheveryobviousemission andtheReflection GratingSpectrometer/RGS (denHerder lineat∼6.4keVthefluxratiodecreases,indicatingaflatter et al. 2001). The MOS1 camera was also in spectral mode continuumslope in thelow state spectrum over thisenergy in the 2002 observation. Both EPIC cameras were used in band.Onthissimplecomparisonthe∼6.4keVemissionline small window mode in the first observation, together with appears essentially unchanged in energy, width and photon the medium filter, successfully ensuring negligible pile-up. flux.Wewill deferamore detailed comparison of the‘high’ The large window mode, with medium filter, was used in and‘low’statedatauntilSection5,afterfirstmodellingthe thesecond,lowfluxstateobservation.TheX-raydatawere individualEPIC spectra. first screened with the latest XMM SAS v5.4 software and eventscorrespondingtopatterns0-4(singleanddoublepixel events) were selected for the pn data and patterns 0-12 for MOS1 and MOS2. A low energy cut of 300 eV was applied 3 HIGH STATE EPIC SPECTRUM toallX-raydataandknownhotorbadpixelswereremoved. 3.1 Power law continuum We extracted EPIC source counts within a circular region of 45′′ radius defined around the centroid position of NGC WebeganouranalysisoftheEPICdatafor2001Mayinthe 4051, with the background being taken from a similar re- conventionalwaybyfittingapowerlawoverthehardX-ray gion, offset from but close tothesource. Thenet exposures (3–10keV)band,therebyexcludingthemoreobviouseffects (cid:13)c 2003RAS,MNRAS000,1–?? X-ray spectrum of NGC 4051 3 energy, width and equivalent width free. The best-fit ob- served line energy was 7.15±0.05 keV (pn) and 7.05±0.05 keV (MOS) in the AGN rest-frame, with an rms width of 150±50 eV,andanEW of100±20 eV.Theaddition ofthis gaussian absorption line gave a further highly significant improvement to the overall fit, with χ2/dof = 1802/1730. Fitting the less compelling absorption feature at ∼8–9 keV with a second absorption line was not statistically signifi- cant. However, an absorption edge did improve the fit to χ2/dof=1767/1728, foranedgeenergyof8.0±0.1keVand optical depth 0.15±0.05. In summary, the 3-10 keV EPIC data from the high state2001MayobservationofNGC4051isdominatedbya power law continuum, with a photon index (after inclusion of cold reflection plus an emission and absorption line) of 1.90±0.02(pn)and1.84±0.02 (MOS).Thenarrowemission line at ∼6.4 keV is compatible with fluorescence from the Figure2.Ratioofdatatopowerlawfitsoverthe3–10keVband same cold reflecting matter, while - if identified with reso- forthepn(black)andMOS(red)spectrainthehighstate2001 nanceabsorption ofFeXXVIorFeXXV-the∼7.1 keVline MayobservationofNGC4051. implies a substantial outflow of highly ionised gas. We find norequirementforthepreviouslyreportedstrong,broadFe ofsoftX-rayemissionand/orlowenergyabsorption.Thisfit K emission line, the formal upper limit for a line of initial yielded a photon index of Γ∼1.85 (pn) and Γ∼1.78 (MOS), energy 6.4 keV being 70 eV. butthefitwas poorwith significant residuals. Inparticular thepresenceofanarrowemission linenear6.4keV,andin- creasingpositiveresidualsabove9keV(figure2),suggested 3.3 Soft Excess theadditionofacoldreflectioncomponenttorefinethecon- Extendingtheabove3–10 keVcontinuumspectral fitdown tinuumfit,whichwethenmodelledwithPEXRAVinXspec to 0.3 keV, for both pn and MOS data, shows very clearly (MagdziarzandZdziarski1995).Sincethereflectioncompo- (figure3)thestrongsoftexcessindicatedinearlierobserva- nent was not well constrained by the continuum fit, we left tions of NGC 4051. free only the reflection factor R (= Ω/2π, where Ω is the To quantify the soft excess we again fitted the com- solid angle subtended at the source), fixing the power law cut-offat200keVanddiscinclinationat20◦,withallabun- bined pn and MOS data, obtaining a reasonable overall dances solar. The outcome was an improved fit, with ∆χ2 fit with the addition of blackbody continua of kT ∼120 of 40 for R = 0.8±0.2. The power law index Γ increased by and 270 eV, together with absorption edges at ∼0.725 keV (τ∼0.24)and∼0.88keV(τ∼0.09).Basedonthisbroadband 0.1forbothpnandMOSfits.Inallsubsequentfitswethen fit we deduced soft X-ray flux levels for the 2001 May ob- set R = 0.8 (compatible with the strength of the 6.4 keV servation of NGC 4051 of 2.9×10−11 erg s−1 cm−2 (0.3–1 emission line). Based on this broad band fit we obtained a keV), with ∼61 percent in the blackbody components, and 2-10 keV flux for the 2001 May observation of NGC 4051 of 2.4×10−11 erg s−1 cm−2 corresponding to a 2-10 keV 1.1×10−11 erg s−1 cm−2 (1–2 keV). Combining these re- luminosity of 2.7×1041 erg s−1 (H0 =75 kms−1Mpc−1). sluumltsinwositithytohfeNhiGghCer40e5n1erignytfihtey‘iheilgdhs’asntaotveeroafll70×.31–01401keerVg s−1 (H0=75 kms−1Mpc−1). 3.2 Fe K emission and absorption The power law plus reflection continuum fit at 3–10 keV leaves several residual features in both pn and MOS data, 4 LOW STATE EPIC SPECTRUM thesignificanceofwhichareindicatedbythecombinedχ2of 2068 for 1740 degrees of freedom (dof). Visual examination Theaboveprocedurewasthenrepeatedinanassessmentof of figure 2 shows, in particular, a narrow emission line near the 2002 November EPIC data, when the X-ray flux from 6.4keVandevidenceofabsorptionnear∼7keVandbetween NGC 4051 was a factor ∼4.5 lower (figure 1). ∼8–9 keV. Fitting the hard X-ray continuum was now more un- To quantify these features we then added furtherspec- certain since the spectrum was more highly curved in the tral components to the model, beginning with a gaussian lowfluxstate(comparefigs4and2),makinganunderlying emission line with energy, width and equivalent width as power law component difficult to identify. To constrain the free parameters. This addition improved the 3–10 keV fit, fitting parameters we therefore made two important initial toχ2/dofof1860/1735, withalineenergy(intheAGNrest assumptions. The first, supported by the minimal change frame) of 6.38±0.01 keV (pn) and 6.42±0.03 keV (MOS), apparent in thenarrow Fe K line, was to carry forward the rms width ≤60 eV and line flux of 1.6±0.4 ×10−5 photon cold reflection (normalisation and R) parameters from the s−1cm−2(pn)and1.4±0.6×10−5photons−1cm−2(MOS), ‘high state’ spectral fit (in fact, as noted above, appropri- corresponding toan equivalent width (EW) of 60±15 eV. ately at a flux level close to the historical average for NGC Next,wefittedthemostobviousabsorptionfeaturenear 4051). The second assumption was that the power law con- 7 keV with a gaussian shaped absorption line, again with tinuum changed only in normalisation, but not in slope (as (cid:13)c 2003RAS,MNRAS000,1–?? 4 K.A.Pounds et al. Figure 3.Extrapolation to0.3keV ofthe 3–10 keV spectral fit Figure 4.Ratioof data to power law plus continuum reflection (detailedinsection3.2)showingthestrongsoftexcessinbothpn model fit over the 3–10 keV band for the pn (black) and MOS (black)andMOS(red)spectraduringthe2001Mayobservation (red)spectrainthelowstate2002NovemberobservationofNGC ofNGC4051. 4051. foundintheextendedXMM-NewtonobservationofMCG-6- 30-15,FabianandVaughan2003).Thisisincontrasttothe conclusions of Lamer et al. (2003) but - as we see later - is consistentwiththedifferencespectrum(figure8),whichfits quite well at 3–10 keV to a power law slope of Γ∼2, while also showing nosignificant residual reflection features. With these initial assumptions, the 3–10 keV fit to the lowstatespectrumyieldedthedata:modelratioshowninfig- ure4.Avisualcomparisonwithfigure2showsaverysimilar narrowemissionlineat∼6.4keV,butwithstrongcurvature to the underlying continuum, and significant differences in theabsorptionfeaturesabove7keV.Thesestrongresiduals resultedinaverypoorfitat3–10keV,withχ2 of1610/990. Wenotethespectralcurvatureinthe3–6keVbandisrem- iniscent of an extreme relativistic Fe K emission line; how- ever, since our high state spectrum showed no evidence for suchafeature,anditmightinanycasebeunexpectedwhen Figure 5. Partial covering model spectrum fitted over the 3–10 thehard X-ray illumination of theinnermost accretion disc keVbandforthe2002NovemberobservationofNGC4051.Also ispresumablyweak,weconsideredinsteadamodelinwhich shown are the separate components in the fit: the unabsorbed a fraction of the power law continuum is obscured by an powerlaw(green),absorbedpowerlaw(red)andGaussianemis- ionised absorber. Weinitially modelled thispossibility with sion line (blue). See Section 4.1 for details. For clarity only the ABSORI in Xspec, finding both the 3–6 keV spectral cur- pndataareshown. vatureandtheabsorptionedgeat∼7.6keVwerewellfitted with∼60percentofthepowerlawcoveredbyionisedmatter ofionisationparameterξ(=L/nr2)∼25andcolumndensity ThislendssupporttoourinitialassumptionthatbothEPIC NH ∼1.2×1023 cm−2. spectra include a ‘constant’ reflection component, illumi- The main residual feature was then the narrow Fe K nated by the long-term average hard X-ray emission from emission line. NGC 4051. With the addition of this narrow emission line the overall 3–10 keV fit obtained with the partial covering model was thengood (χ2/dof = 1037/1037). Figure 5 illus- 4.1 The narrow Fe K emission line trates the unfolded spectrum and spectral components of thisfit. A gaussian line fit to the emission line at ∼6.4 keV in the low state EPIC data was again unresolved, with a mean energy (in the AGN rest frame) of 6.41±0.01 keV (pn) and 4.2 Soft Excess 6.39±0.02keV(MOS),andlinefluxesof1.9±0.3×10−5pho- ton s−1 cm−2 (pn) and 2.0±0.4 ×10−5 photon s−1 cm−2 Extrapolation of the above partial covering 3–10 keV spec- (MOS), corresponding to an EW against the unabsorbed tralfitdownto0.3keVshowsasubstantialsoftX-rayexcess power law component of 500±75 eV. The important point remains (figure 6), with a similar relative strength to the isthat,withinthemeasurementerrors,themeasuredfluxes power law component seen in the high state data. We note of the ∼6.4 keV line are the same for the two observations. thatthe‘softexcess’,ierelativetothepowerlawcomponent, (cid:13)c 2003RAS,MNRAS000,1–?? X-ray spectrum of NGC 4051 5 Figure 6. Partial covering model fits over the 3–10 keV band Figure7.Extrapolationto0.3keVofthe3–10keVpartialcov- extendedto0.3keV,forthepn(black)andMOS(red)datafrom ering fit of fig 5 showing the strong soft excess modelled by a thelowstate2002November observationofNGC4051. blackbody component (blue), and a broad absorption trough at ∼0.76keV.Forclarityonlythepndataareshown. would have been extremely strong (data: model ratio∼8) had we taken the simple power law fit (Γ∼1.4) to the low subtractedlowstatedatafromtheequivalenthighstatedata state3–10keVdata.Extendingthepartialcoveringmodelto (corrected for exposure). To improve the statistical signifi- 0.3keV,withtheaddition-asinthehighstate-ofablack- cance of the higher energy points the data were re-grouped bodycomponentofkT∼125eV(thehottercomponentwas foraminimumof 200counts.Theresultingdifferencespec- notrequired),gavean initially poorfit(χ2 of2348 for1265 trumiscomparedinfigure8withapowerlawfittedat3–10 dof for the pn data), with a broad deficit in observed flux keV.Severalpointsareofinterest.First,thepowerlawindex at ∼0.7-0.8 keV being a major contributor (figure 6). The of the difference spectrum, Γ∼2.04 (pn) and Γ∼1.97 (M2), additionofagaussianabsorptionlinetothepartialcovering is consistent with the assumed ‘constant’ value in the indi- modelgavealargeimprovementtothebroad-bandfit(toχ2 vidual spectral fits. Second, the narrow Fe K emission line of1498for1262dof),foralinecentredat0.756±0.003 keV, and high energy data upturn are not seen, supporting our with rms width 50±15 eV and EW∼40 eV. We show this initialassumptionofa‘constant’coldreflectioncomponent. complexspectralfitinfigure7,andcommentthatthemodel The narrow feature observed at ∼7 keV corresponds to the dependencyofunfoldedspectraisrelativelyunimportantin absorptionlineseen(only)inthehighstatespectrum,while illustrating such strong, broad band spectral features. Sig- we shall see in Section 6 that the deficit near 0.55 keV in nificantly,thebroad-band spectral fit remains substantially the MOS data (which has substantially better energy reso- inferior to the similar fit to the high state data. Examina- lution there than thepn) is probably explained by a strong tionofthespectral residualsshowsthisisduetoadditional and ‘constant flux’emission lineof OVII.Finally, thesmall fine structure in the soft band of the low state spectrum, peak near 8 keV can be attributed to the absorption edge structurethatisalsoevidentinfigures6and7.Weexamine shifting to lower energy as the photoionised gas recombines theRGSdatainSection6toexplorethenature(absorption in the reduced continuumirradiation. or emission) of this structure. Whilethearithmeticdifferenceof twospectraprovides ThededucedsoftX-rayfluxlevelsforthe2002Novem- asensitivecheckforthevariabilityofadditivespectralcom- berobservationofNGC4051were6.3×10−12 ergs−1 cm−2 ponents, a test of the variability of multiplicative compo- (0.3–1keV),with∼53percentintheblackbodycomponent, nents is provided by the ratio of the respective data sets. and1.8×10−12ergs−1cm−2(1–2keV).Combiningthesere- Figure9reproducestheratioofthehighandlowstatedata sultswitha2-10keVfluxof5.8×10−12 ergs−1cm−2yielded (pnonly)afterre-groupingtoaminimumof500countsper an overall 0.3–10 keV luminosity of NGC 4051 in the ‘low’ bin.From∼0.3–3keVthefluxratioaverages∼5,asseenin state of 1.5×1041 erg s−1 (H0=75 kms−1Mpc−1). figure 1, falling to higher energies as the mean slope of the low state spectrum hardens. The large positive feature at ∼0.7–0.8 keV is of particular interest, indicating a variable multiplicativecomponent,almostcertainlycorrespondingto 5 COMPARISON OF THE HIGH AND LOW enhancedabsorptioninthelowstatespectrum.Infactthat STATE EPIC DATA feature can be clearly seen in the low state EPIC data in The above spectral fitting included two important assump- figures 6 and 7. We suggest the broad excess at ∼1–2 keV tions, that the cold reflection was unchanged between the can be similarly explained by greater absorption affecting high and low flux states, and the variable power law com- thelowstatespectrum,lendingsupporttoouroverallinter- ponentwasofconstantspectralindex.Wenowcomparethe pretation of the spectral change. Finally, we note that the EPIC data for the two observations to further explore the narrow dip in the ratio plot at ∼6.4 keV is consistent with nature of the spectral change. Figure 8 illustrates the dif- the Fe K emission line having unchanged flux, but corre- ference spectrum obtained by subtracting the background- spondingly higher EW in thelow state spectrum. (cid:13)c 2003RAS,MNRAS000,1–?? 6 K.A.Pounds et al. Figure 8. High minus low state difference spectral data (pn- Figure 10. Fluxed RGS spectrum from the XMM-Newton ob- black, M2-red) compared with a simple power law, as described servationofNGC4051in2001May. inSection5. Figure 11. Fluxed RGS spectrum from the XMM-Newton ob- Figure9.Ratioofhighstatetolowstatespectraldata(pnonly), servationofNGC4051in2002November. asdescribedinSection5. identified with resonance absorption in He- and H-like ions 6 SPECTRAL LINES IN THE RGS DATA ofC,N,OandNe.Incontrast,thecombinedRGSdatafor the low state data from 2002 November showed a mainly Both EPIC spectra show a strong soft excess, with the low emission line spectrum, more characteristic of a Seyfert 2 state (2002) spectrum also having more evidence of fine galaxy (eg Kinkhabwala et al.2002). Significantly, the NVI, structure. To study the soft X-ray spectra in more detail OVII and NeIX forbidden lines are seen in both high and we then examined the simultaneous XMM-Newton grating low state RGS spectra at similar fluxlevels. Taking note of data for both observations of NGC 4051. Figures 10 and that fact we then analysed the low state (2002) data first, 11 reproduce the fluxed spectra, binned at 35m˚A, to show and subsequently modelled the RGS high-minus-low differ- bothbroadandnarrowfeatures.Thecontinuumfluxlevelis encespectrum,togetatruermeasureoftheabsorptionline higherinthe2001data(consistentwiththelevelsseeninthe strengths in thehigh state (2001) spectrum. EPIC data), with a more pronounced curvature longwards of∼15˚A.Numeroussharpdatadropshintatthepresenceof manynarrowabsorptionlines.Incontrast,the2002Novem- 6.1 An emission line spectrum in the low state ber RGS spectrum exhibits a lower and flatter continuum data flux,and a predominance of narrow emission lines. Webeganananalysisofeachobservationbysimultane- To quantify the emission lines in the 2002 spectrum we ously fitting the RGS-1 and RGS-2 data with a power law addedgaussianlinestothepowerlawplusblackbodycontin- and black body continuum (from the corresponding EPIC uumfit in Xspec,with wavelength andfluxas free parame- 0.3–10 keVfits)andexaminingthedata:model residualsby ters.Ineachcasethelinewidthwasunresolved,indicatinga eye. For the 2001 May observation the strongest features FWHM≤300kms−1.Detailsofthe8strongestlinesthereby were indeed narrow absorption lines, most being readily identifiedarelistedinTable1.Thestatistical qualityofthe (cid:13)c 2003RAS,MNRAS000,1–?? X-ray spectrum of NGC 4051 7 fit was greatly improved by the addition of the listed lines, with a reduction in χ2 of 251 for 16 fewer dof. When ad- justedfortheknownredshiftofNGC4051alltheidentified lines are consistent with the laboratory wavelengths indi- cating that theemitting gas has amean outflow (orinflow) velocity of ≤200 km s−1. Figure12illustratestheOVIItriplet,showingthedom- inant forbidden line and strong intercombination line emis- sion,butnoresidualresonancelineemission(at21.6˚A).The lineratios,consistentwiththosefoundintheearlierChandra observation (Collinge et al. 2001), give a clear signature of aphotoionisedplasma,withanelectrondensity≤1010cm−3 (PorquetandDubau2000).Asimilarly dominantforbidden line in the NVI triplet yields a density limit a factor ∼10 lower. We note theabsence of the OVIIresonance emission line may be due to infilling by a residual absorption line of similar strength. Figure 12. Emission lines dominate the 2002 November RGS AfterremovaloftheemissionlineslistedinTable1,sev- data.TheOVIItripletisillustratedwithonlytheforbiddenand eral additional emission features (see figure 11) remained. intercombination lines clearly visible. The gaussian line fits in- Although narrow and barely resolved, the wavelength of cludeonlytheRGSresolutionshowingtheemissionlinesarein- these features allows them to be unambiguously identified trinsicallynarrow.SeeSection6.1fordetails. with the radiative recombination continua (RRC) from the same He- and H-like ions of C, N, O and (probably) Ne. Table 2 lists the properties of these RRC as determined by sation parameter derived from our XSTAR fit to the RGS fittinginXspecwiththeREDGEmodel.WhiletheRRCof absorption spectrum (Section 7). The scale of the soft X- CV, CVI, NVI and OVII are well determined, we fixed the ray emitting gas is apparently much greater than theBLR, otherthreshold energies at theirlaboratory valuesto quan- for which Shemmer et al. 2003 find a value of 3.0±1.5 light tify the measured equivalent widths. What is clear is that days(∼3−10×1015cm).Infactithasoverallproperties,of the RRC are very narrow, a combined fit yielding a mean density, temperature and velocity consistent with the NLR temperature for the emitting gas of kT ∼3 eV (T ∼4×104 in NGC 4051. K).Wenotethislowtemperatureliesinaregionofthermal The above emission lines and RRC provide an accept- stability for such a photoionised gas (Krolik et al. 1981). ablefittotheRGSdataforthe2002Novemberobservation Furthermore, the low temperature indicates collisional ion- of NGC 4051. However a coarse binning of the data:model isation and excitation will be negligible, and radiative re- residuals (figure13) shows abroad deficit of fluxremaining combination should bethe dominant emission process. at ∼15−17˚A. It seems likely that this feature is the same AdditionalconstraintsontheemittinggasinNGC4051 as that seen in the broad band fits to the EPIC data for can be derived by noting that the 2002 November XMM- 2002 November (Section 4) and tentatively identified with Newtonobservationtookplacesome20daysafterthesource an unresolved transition array (UTA) from Fe M-shell ions entered an extended low flux state. Furthermore, the emis- (Behar et al. 2001). When fitted with a gaussian absorp- sion line strength of the OVII forbidden line is essentially tion line we find an rms width of σ= ∼30 eV and EW of the same as when NGC 4051 was much brighter in 2001 25 eV against the low state continuum, consistent with the May. This implies that the emission spectrum arises from absorption trough required in thepartial coveringfitto the ionisedmatterwhichiswidelydispersedand/orofsuchlow low state EPIC data (section 4.2). density that the recombination time is >∼2×106s. At a gas temperatureof∼4×104K,therecombinationtimeforOVII isoforder150(n9)−1s,wheren9isthenumberdensityofthe ionisedmatterinunitsof109cm−3(ShullandVanSteenberg 6.2 Absorption lines in the high state difference spectrum. 1982).Thepersistentlowstateemission wouldthereforein- dicate a plasma density ≤105cm−3. The observed wavelengths of the main emission lines in the Assuming a solar abundance of oxygen, with 30 per- 2002 spectrum and their equivalent absorption lines in the cent in OVII, 50 percent of recombinations from OVIII di- 2001 spectrum are the same within the resolution of our rect to the ground state, and a recombination rate at kT gaussian line fitting. (At higher resolution the absorption ∼3 eV of 10−11 cm3 s−1 (Verner and Ferland 1996), we lines appear to have a mean outflow velocity of ∼500 km deduce an emission measure for the forbidden line flux of s−1,whiletheemission linesareclosetothesystemicveloc- order 2×1063cm−3. That corresponds to a radial extent of ityofNGC4051.)Furthermore,fromouranalysisinSection >∼3×1017cm for a uniform spherical distribution of pho- 6.1itseemsclearthattheemissionlinespectrumrepresents toionised gas at the above density of ≤105cm−3. Coinci- an underlying component that responds to some long-term dentally,thealternativeexplanationforaconstantemission average flux level of the ionising continuum of NGC 4051. line flux,via an extended light traveltime, also requires an We therefore first subtracted the 2002 RGS spectrum from emittingregionscalesizeof>∼1017cm.Wenote,furthermore, the 2001 spectrum with the aim of obtaining a truer mea- thatthesevaluesofparticledensityandradialdistancefrom sureof theabsorption line strengths in thehigh state data. the ionising continuum source are consistent with the ioni- Quantifying the main absorption lines by adding gaussian (cid:13)c 2003RAS,MNRAS000,1–?? 8 K.A.Pounds et al. Table 1. Principal emission lines identified in the 2002 November RGS spectrum of NGC 4051. Wavelengths are in Angstroms and adjustedtothesourcerestframeandlinefluxesareinunitsof10−5 photons cm−2 s−1. Line λsource λlab flux EW(eV) CVILyα 33.75±0.03 33.74 4±1 2.5±0.6 NVI1s-2p(f) 29.55±0.04 29.53 3.5±0.5 3.2±0.5 NVIILyα 24.79±0.03 24.78 3±0.8 3.5±1 OVII1s-2p(f) 22.14±0.02 22.10 11±1.5 14±2 OVII1s-2p(i) 21.79±0.03 21.80 5±1 6±2 OVIIILyα 18.99±0.03 18.97 8±1.3 10±1.5 FeXVII2p-3s 17.06±0.03 17.2 2±0.6 5±2 NeIX1s-2p(f) 13.75±0.05 13.70 3±0.6 7±2 Table 2. Radiative recombination continua identified in the 2002 November RGS spectrum of NGC 4051. All wavelengths are in Angstroms and linefluxes are inunits of 10−5 photons cm−2 s−1. The threshold wavelengths fixed (f) at the laboratory values areas indicated. RRC λsource λlab flux EW(eV) NeX 9.13(f) 9.11 0.3±0.2 6±4 NeIX 10.39(f) 10.37 0.75±0.3 13±5 OVIII 14.19(f) 14.15 1.7±0.5 9±3 OVII 16.85 16.81 3.2±0.8 11±3 NVI 22.50 22.45 1.7±0.8 2±1 CVI 25.41 25.35 5.7±1 7±1 CV 31.71 31.64 2±1.5 3±2 replacedthegaussianabsorptionlinesintheabovefitswith a model comprising a grid of photoionised absorbers based ontheXSTARcode(Kallmanetal.1996).Wemodelledthe RGSdifferencespectrum as thebest measure of low ionisa- tion matter; however, in the case of the EPIC data, where weexpecttheparticledensitytobehigher(andrecombina- tion timeshorter) than in theextendedemission region, we modelledthedirecthighstatedata.TheXSTARmodelab- sorberscoverawiderangeofcolumndensityandionisation parameter, with outflow (or inflow) velocities as a variable parameter.AllabundantelementsfromCtoFeareincluded withtherelativeabundancesasavariableinputparameter. To limit processing timewe assumed a fixedwidth for each absorption line of 1000 km s−1 FWHM. We first attempted a fit to the RGS data over the 8.5–35˚Aband, where the power law plus blackbody contin- uum gave χ2 of 5388 for 4197 dof. The addition of a two- Figure13.Ratioofthe2002NovemberRGSdatatotheemission component ionised absorber significantly improved this fit, lineandRRC model described inSection 6.1. A broaddeficit of fluxat∼15−17˚Amaybeattributedtoanunresolvedtransition to χ2 of 4829/4183, with an ionisation parameter logξ of array(UTA)inweaklyionisedFe. 2.7±0.1 and column density of NH∼6×1021cm−2, and an ionisation parameter logξ of 1.4±0.1 and column density of NH∼2×1021cm−2. The relative abundances of C,N,O,Ne, lines to the corresponding continuum fit then produced the and Fe, tied for both components, were determined to be line list in Table 3. 0.5, 0.6,0.35,0.3 and 1.0. The apparent redshift from the fit was 5.2±1.4×10−4, indicating an outflow velocity of ∼600 km s−1 at the redshift of NGC 4051. We note this velocity is probably a lower limit for those lines where a significant 7 AN IONISED ABSORBER MODEL FIT TO emissioncomponenthasbeensubtracted.Whilethepresent THE 2001 MAY XMM-Newton DATA. analysis is only intended toobtain a rough characterisation Tobetterquantifythehighlyionised matterresponsible for of thelow ionisation outflow, several points stand out. theobservedabsorptionfeaturesinthe‘highfluxstate’spec- Asignificantlybetterfitwiththe2-componentabsorber trumofNGC4051andcheckforphysicalconsistencyofthe suggests a range of ionisation parameter exists in the out- candidatelineidentifications, includingthehigh energy ab- flow, a point emphasised more strongly when the Fe-K ab- sorptionfeaturesseenabove7keVintheEPICdata,wenext sorption line is added (see below). As noted earlier the (cid:13)c 2003RAS,MNRAS000,1–?? X-ray spectrum of NGC 4051 9 Table3.Principalabsorptionlinesidentifiedinthe2001MayRGS‘difference’spectrumofNGC4051.TheNeIXresonancelineappears tobebroadandisprobablyblendedwithFeXIX.Wavelengths areinAngstromsandadjustedtothesourcerestframe Line λsource λlab EW(mA) CVILyα 33.67±0.02 33.74 170±30 CVILyβ 28.42±0.02 28.47 60±15 CVILyγ 26.95±0.03 26.99 20±10 NVI1s-2p 28.73±0.02 28.79 80±15 NVIILyα 24.73±0.03 24.79 100±25 OVIIILyα 18.93±0.01 18.97 100±15 OVIIILyβ 15.98±0.02 16.01 25±10 OVII1s-2p 21.57±0.03 21.60 60±15 OVII1s-3p 18.62±0.03 18.63 40±112 OVII1s-4p 17.76±0.04 17.77 15±8 NeIX1s-2p 13.43±0.03 13.45 40±15 Repeating the XSTAR fit for the only likely alterna- tive identification of the ∼7.1 keV line, with the resonance (1s-2p) absorption of Fe XXV, required a lower ionisation parameter for the third component, of logξ=3.3±0.1 to- gether with a column density of NH∼1023cm−2. The ap- parent redshift from the fit was then -0.058, indicating an outflow velocity of ∼16500 km s−1 at the redshift of NGC 4051. Again assuming a cone angle of the highly ionised outflowofπ sr,thecorrespondingoutflowmassrateisthen ∼5×10−2 M⊙yr−1,withakineticenergyof2×1042ergs−1. 8 AN IONISED ABSORBER MODEL FIT TO THE 2002 NOVEMBER EPIC DATA. Implicit in our analysis so far is the constant nature of the Figure 14.Partofthe2001MayRGS‘difference’spectrumfit- extended, low ionisation gas. In contrast, our partial cov- tedwiththephotoionisedabsorbermodeldescribedinSection7. ering fit to the low state EPIC data requires a substantial Themainabsorptionlinesshownare,fromtheright,OVII1s-2p columndensityin alower ionisation state(thanin thehigh (21.60˚A), OVIIILyα(18.97˚A)andOVII1s-3p(18.63˚A) state EPIC fit), covering ∼60 per cent of the hard X-ray continuum source. It seems a reasonable assumption that thisnewabsorbingcomponentisformedfromthepreviously strongest observed lines are all from H- and He-like ions of highlyionisedoutflowastheionisation parameterfallswith thelightermetals.Themodelisseentofitthetheresonance the reduced hard X-ray flux. To quantify this change, and linesofOVIIandOVIIIquitewell(figure14),perhapssince obtainanalternativefittotheoveralllowstatespectrumof thesestrongestlinesdrivethefit.However,thehigherseries NGC4051, we then replaced theABSORImodel of section linesofOVIIarenoticeablystrongerthaninthemodel,sug- 5 with XSTAR. gesting theresonance line is saturated in the core. TheresultsupportstheABSORIfitting.Figure15illus- The ∼7.1 keV absorption line, most conservatively at- tratestheXSTARfittothelowstateEPICdata,theparam- tributedtoblue-shiftedFeXXVILyαresonanceabsorption, eters being, a power law of Γ= 1.93±0.03, ∼57 per cent of required a third, more highly ionised component in the ab- whichiscoveredbyacolumnofNH∼3.6×1023cm−2andin- sorbinggas,withionisationparameterlogξ∼3.8andcolumn termediateionisationparameterlogξ=1.4±0.1.Theremain- densityofNH∼2×1023cm−2.Theapparentredshiftfromthe ing power law component, a black body of kT ∼125 eV, fitwas-0.02,indicatinganoutflowvelocityof∼6500kms−1 and a narrow Fe K emission line are covered by an ionised attheredshiftofNGC4051.Theionisationparameter,and absorber similar to that fitted to the RGS high state data, hencecolumndensity,arenotwellconstrainedbythissingle withlogξ∼2.8andNH∼8×1021cm−2,togetherwithacold linefit.However, themain uncertaintyin thecolumn prob- columnof∼2×1020 cm−2,slightlygreaterthantheGalactic ably lies on the upside, since the line is apparently broader column. It is interesting to note that a similar high column thanintheXSTARmodelfit,whilethefittedionisationpa- densityisrequiredforEPICfitsinbothhighandlowstates, rameter is close to that for a peak abundance of FeXXVI. with the dramatic spectral change being attributed to part Assumingaconeangleforthehighlyionisedoutflowofπsr, of that matter recombining in response to the lower hard thecorresponding outflow mass rate is ∼7×10−3 M⊙yr−1, X-rayflux.Furthermore, when left as a free parameter, the withanassociatedkineticenergyof∼1041ergs−1,compara- power law slope in the partial covering fit has a preferred bletothehigh (mean)statehardX-rayluminosity ofNGC valueconsistentwiththatinthehighstate.Inotherwords, 4051. thesecondimportantinitialassumptionwemadeinsection (cid:13)c 2003RAS,MNRAS000,1–?? 10 K.A.Pounds et al. orsuper-Eddington)accretionrate,theevidenceforcolumn densitiesofhighlyionised gas inexcessofNH∼1023cm−2 is becoming more common for Seyfert 1 galaxies (eg Bianchi etal.2003).Suchcolumnsthenpotentiallyofferanalterna- tiveexplanation totheextremebroadFeKline, viapartial covering of the power law continuum. In thepresent analy- sis of theXMM-Newton observations of NGC 4051 we have explored the partial covering alternative, noting that the observation ofFeXXVI(orFeXXV)absorption in thehigh state requires a similar column density (of highly ionised gas), which would have recombined as a result of the re- ducedionisingfluxpersistingfor∼20dayspriortothe2002 Novemberobservation.Weconcludethatvariableopacityin thisoutflow, respondingto thereduced ionising fluxduring theextendedlowstateofNGC4051,providesanaturalex- planationofthedramaticchangeobservedinthebroadband X-rayspectrumofNGC4051.Wesuggestthattheeffectsof Figure 15. The XSTAR 0.3–10 keV partial covering fit to the absorption by line-of-sight ionised gas may have been gen- lowstate2002 Novemberobservation ofNGC4051, showingthe erally underestimated in the analysis and interpretation of strong soft excess and a broad absorption trough at ∼0.76 keV. AGN spectra, and note that a similar explanation was pro- Alsoshownaretheseparatecomponentsinthefit:theunabsorbed posed by Costantini et al. (2001) in reporting a large scale power law (red), absorbed power law (pink), Gaussian emission line (blue) and blackbody (green). For clarity only the pn data spectralchangein theSeyfertgalaxy NGC3516 observedin areshown. two BeppoSAX observations 4 monthsapart. The important detection of a high velocity outflow in the high state spectrum of NGC 4051 (given independent 5 (in addition to constant cold reflection) is also supported support by the recent report of an outflow at 4500 km s−1 in thisanalysis. from a second Chandra observation; van der Meer et al. 2003) raises the additional question of what fraction of the hard X-ray emission may arise, not from the disc/corona, 9 DISCUSSION but from shocks in this flow? We showed in Section 7 that if the inner flow has a wide cone angle, the associated ki- 9.1 Hard X-ray emission and re-processing in netic energy is comparable with the hard X-ray luminosity outflowing gas. inNGC4051.Ifthetriggerforamassiveoutflowis-assug- Giventhegeneral acceptancethatAGNarepoweredbyac- gested by King and Pounds (2003a) - accretion at or above cretion onto a supermassive black hole it seems reasonable theEddingtonlimit,thenmightthisapplyforNGC4051?A that the usually-dominant optical-XUV flux arises as ther- recentreverberationmeasurement(Shemmeretal.2003)has mal radiation from the accretion disc. However, the origin indicated the black hole mass in NGC 4051 to lie between ofthehardX-raypowerlawcomponent(andsoft X-rayex- 2−10×105 M⊙,anunusuallylowvalue(foranAGN),but cess) remains less clear, with up-scattering of disc photons onesupportedbyafurtherrecentanalysisoftheX-rayvari- in ahigh temperature‘corona’ beingapopular mechanism. abilty(McHardyetal.2003).Suchalowmasssuggeststhat The strengthening view that viscosity in the disc is largely inthe‘typical’brightstateofNGC4051, asweobservedin of magnetic origin offers an appealing way of transferring 2001 May when the total X-ray luminosity was ∼ 1042 erg accretion energy to the coronal electrons by re-connection s−1, the bolometric luminosity of NGC 4051 might indeed in buoyant magnetic flux. Reprocessing of hard X-rays in havereached, or exceeded,the Eddington limit. the disc may then explain a major part of the ‘continuum reflection’andfluorescentFeKemission often seeninAGN spectra. 9.2 Extended photoionised gas in NGC 4051. Themostdirectevidencegiveninsupportofthispicture hasusuallybeentherapid,highamplitudeX-rayvariability The low central continuum flux during the 2002 November (implying a small emission region) and the broad, skewed observation of NGC 4051 allowed the emission spectrum profile of the Fe K line, indicating an origin in reflection from an extended photoionised gas to be observed in the from the innermost accretion disc where strong relativis- RGSdata,arare opportunitytoobservethiscomponentin tic effects are expected. Recently some doubts have been a Seyfert 1 galaxy (see also Turner et al. 2003 for a similar raisedonthewideapplicability ofthismodel.Inparticular, observation of NGC 3516). The detection of several RRC improved X-ray spectra from Chandra and XMM-Newton show the temperature of the gas to be ≤5×104K, a region have failed to confirm the relativistic Fe K emission line of thermal stability, while ouranalysis of thestrong forbid- in a majority of AGN. Furthermore, new evidence of mas- denlineemissionofOVIIprovidesanestimateofthe(mini- sive outflows of highly ionised matter in a number of AGN mum)extentas3×1017cm.Acorrespondingminimummass has drawn attention to the need to take due account of re- for this extended gas envelope is then ∼10 M⊙. Assuming processing in overlying (as well as disc) matter. While in a mean outflow velocity of 100 km s−1, the flow time (to the previous cases where high velocity outflows have been reach3×1017cm)is∼103 years.Itisinterestingtonotethat confirmed they appear to be linked to a high (Eddington themassoftheextendedlowionisation region would bere- (cid:13)c 2003RAS,MNRAS000,1–??