Revealing the X-ray source in IRAS 13224-3809 through flux-dependent reverberation lags

0.28 MB·English

by  E. Kara

Mon.Not.R.Astron.Soc.000,1–6(2012) Printed10January2013 (MNLATEXstylefilev2.2)
Revealing the X-ray source in IRAS 13224-3809 through
flux-dependent reverberation lags
E. Kara1⋆, A. C. Fabian1, E. M. Cackett2, G. Miniutti3 and P. Uttley4
3 1Institute of Astronomy, Madingley Rd, Cambridge CB3 0HA
1 2Department of Physics and Astronomy, Wayne State University,Detroit, MI 48201, USA
0 3Centro de Astrobiologia (CSIC-INTA), Dep. de Astrofisica; LAEFF, PO Box 78, E-28691, Villanueva de la Can˜ada, Madrid, Spain
2 4Astronmical Institute ‘Anton Pannekoek’, Universityof Amsterdam, Postbus 94249, 1090 GE Amsterdam, the Netherlands

n
a
J Accepted 2012December 29.Received2012December13;inoriginalform2012October18

9

] ABSTRACT
E IRAS13224-3809wasobservedin2011for500kswiththeXMM-Newtonobservatory.
H We detect highly significant X-ray lags between soft (0.3 – 1 keV) and hard (1.2 – 5
. keV) energies. The hard band lags the soft at low frequencies (i.e. hard lag), while
h
the opposite (i.e. soft lag) is observed at high frequencies. In this paper, we study
p
the lag during flaring and quiescent periods. We find that the frequency and absolute
-
o amplitude of the soft lag is different during high-flux and low-flux periods. During
r the low flux intervals, the soft lag is detected at higher frequencies and with smaller
t
s amplitude.Assumingthatthesoftlagis associatedwiththelighttraveltime between
a
primary and reprocessed emission, this behaviour suggests that the X-ray source is
[
more compact during low-flux intervals, and irradiates smaller radii of the accretion

1 disc (likely because of lightbending effects). We continue with an investigationof the
v lagdependence onenergy,andfindthatisolatingthelow-fluxperiodsrevealsastrong
4 lag signature at the Fe Kα line energy, similar to results found using 1.3 Ms of data
2 on another well known Narrow-Line Seyfert I galaxy, 1H0707-495.
9
1 Key words: black hole physics – galaxies: active – X-rays: galaxies – galaxy: indi-
. vidual : IRAS 13224-3809.
1
0
3
1
v: 1 INTRODUCTION variability occurs on timescales of thelight crossing timeof
i thesource(orgreater),andthereforewecancalculateanup-
X TheX-rayemittingregioninaccretingblackholesystemsis
perlimitonthesizeoftheemittingregion,R<cδt(Terrell
notwellunderstood.Generally,itisagreedthattheangular
r 1967).FromobservationsoffastvariabilityinAGN,weknow
a momentum of accreting gas onto a black hole forms an op-
that the corona must be compact, likely within ∼ 100 r .
tically thick,geometrically thin accretion disc that radiates g
Recent observations from the X-ray variability seen from
as a series of blackbody components (Shakura& Sunyaev
microlensed quasars also show that the corona is compact
1973). For supermassive black holes, this thermal emis-
towithin 10 r orless (Chartas et al. 2009;Dai et al. 2010;
sion peaks in the UV. Some of the thermal disc photons g
Chartas et al. 2012).
are then inverseCompton upscattered to X-ray energies by
mildlyrelativisticelectronsinahotcloud,calledthecorona Variability studies of black hole binaries led to the
(Haardt & Maraschi 1991, 1993). While theX-ray emission discovery of the X-ray lag, where usually hard pho-
mechanismisrelativelywellunderstood,thegeometryofthe tons lag behind soft photons (e.g. Miyamoto et al. 1988;
corona where the X-rays are produced is an area of active Nowak & Vaughan1996;Nowak et al.1999).Thiseffectwas
research.Inthispaper,weprobethegeometryofthecorona alsolaterobservedinAGN(Papadakis, Nandra, & Kazanas
through an analysis of theX-ray reverberation lags. 2001; McHardy et al. 2004). The origin of the hard lag is
X-ray variability studies have been influential in start- still not well understood, though in the prevailing model,
ing to understand the size of the emitting region. AGN are the multiplicative propagation effects of fluctuations in the
known to be extremely variable in the X-ray band, com- accretionflowmodulateregionsemittingsoftphotonsbefore
monly doubling in amplitude on timescales of less than one thoseemitting hard photons(Kotov, Churazov, & Gilfanov
day for radio-quiet sources (McHardy 1988). Substantial 2001; Ar´evalo & Uttley 2006).
Recentdevelopmentsin X-rayvariabilityhaverevealed
a newtypeof lag, where soft photonslag behind hard pho-
⋆ E-mail:[email protected] tons. This lag, interpreted as a reverberation lag, offers a
(cid:13)c 2012RAS
2 Kara et al.
new perspective in which to study the X-ray emitting re-
gion in AGN. Understanding the phase lag between light
curves of different energy bands gives us an orthogonal ap- 400
proach to testingphysical models thatare often degenerate
in thetime-integrated energy spectrum.
Reverberation lags were first discovered in g (s) 200
Narrow-Line Seyfert 1 (NLS1) galaxy, 1H0707-495 a
L
(Fabian et al. 2009). Since then, lags have been
detected more than a dozen other Seyfert galax- 0
ies (Emmanoulopoulos, McHardy,& Papadakis 2011;
DeMarco et al. 2011, 2012; Zoghbi et al. 2012), including
most recently IRAS 13224-3809 (Fabian et al. 2012b). −200
10−4 10−3
Reverberation lags have also been observed on the order
Frequency (Hz)
of milliseconds or less in black hole binary GX 339-4
(Uttley et al. 2011).
Figure 1. Lag-frequency spectrum for the 500 ks observation.
The soft lags can be understood with the standard re-
Thelagiscalculatedbetweenthesoftenergyband(0.3-1.keV)
flectionscenario(Guilbert & Rees1988;Lightman & White
and the hard band (1.2 - 5. keV). The most negative lag (at
1988).Inthismodel,thebulkoftheX-raycontinuumispro- ν=4.1×10−4 Hz)is−92±31s.
duced in the corona. Some continuum photons fall towards
the surface of the accretion disc, are absorbed by photo-
electricabsorption,andreprocessedintoanX-rayreflection
chosen from a circular region of the same size, and were
spectrum comprised of a reflection continuumand emission
thesamedistancetothereadout nodeasthesourceregion.
lines (e.g. Ross & Fabian 2005). When reflection arises in
The background subtracted light curves were produced us-
the accretion disc, the whole reflection spectrum is blurred
ingthetool epiclccorr.The light curvesranged in length
due to Doppler and other relativistic effects close to the
from 83 ksto 124 kswith 10 s bins.
black hole. Soft lags, where thedirect continuumvariations
leadthoseinthesoft-bandreflection,aretheninterpretedas
theadditional light traveltimetakenbythereflected paths
between thecorona and theinner accretion disc. 3 RESULTS
IRAS 13224-3809 (z =0.066) is one of the most X-ray
3.1 Lag vs. Frequency
variable Seyfert 1 galaxies known (Boller, Brandt,& Fink
1996), and therefore is a useful source with which to probe Thelagiscomputedintheusualwayfollowingtheformulae
the environments of the innermost regions. It was first ob- described in Nowak et al. (1999). The frequency-dependent
served with XMM-Newton in 2002 for 64 ks (Boller et al. lag is calculated between light curves in the soft (0.3 –
2003; Gallo et al. 2004; Ponti et al. 2010), and most re- 1. keV) and hard (1.2 – 5 keV) energy bands, which are
cently, for 500 ks, which led to a significant detection of dominatedbythesoft excess(likelydisc-reflection) and the
the soft lag (Fabian et al. 2012b). In this letter, we follow primary powerlaw-like emission, respectively (Fabian et al.
up that recent work with an analysis of the soft lags of 2012b; Pontiet al. 2010). The Fourier transform of a light
IRAS13224-3809 in the low and high flux states. curve contains an amplitude and complex phase, such that
the Fourier transform of soft light curve s(t) is written
S = |S|eiφs. The cross product of the soft and hard band
Fourier transforms, H∗S = |H||S|ei(φs−φh), provides the
2 OBSERVATIONS AND DATA REDUCTION
phase difference between the two energy bands. This phase
The XMM-Newton satellite (Jansen et al. 2001) observed differenceisconvertedbackintoafrequency-dependenttime
IRAS 13224-3809 for 500 ks over four orbits from 2011 lag, where τ(f) = (φs−φh)/2πf. Given this definition for
July19to2011July 29(Obs.IDs0673580101, 0673580201, the lag, a negative amplitude lag means that the soft band
0673580301,0673580401).Forthisanalysisofthereverbera- lagsbehindthehardband,whilethepositiveamplitudelag
tionlags,wefocusonthehightime-resolutiondatafromthe shows the hard band lagging. We take note that the phase
EPIC-pncamera(Stru¨der et al.2001).Thefirstobservation lag is defined over the interval (−π,π), which causes phase
was taken in full window imaging model, and next three in wrapping at high frequencies (Nowak & Vaughan 1996).
large window imaging mode. All of the data were reduced Fig.1showsfrequency-dependenttimelagbetweenthe
in the same way, using the XMM-Newton Science Analysis softandhardbands,fromFabian et al.(2012b).Atlowfre-
System(SASv.11.0.0) andthenewest calibration files.The quencies,wefindalargepositivelagwherethehardfluxlags
details of the data reduction are explained in Fabian et al. behind the soft flux by hundreds of seconds. At higher fre-
(2012b). quencies,ν ∼[2−10]×10−4 Hz,thereisanegativelagwhere
Thedatawerecleanedforhighbackgroundflares,which the soft flux lags the hard by ∼100 s. Above ∼1.5×10−3
resultedinatotalexposuretimeof∼300ks.Thedatawere Hz,the signal becomes dominated by Poisson noise.
selected with the condition pattern 6 4. Pile-up effects
were not significant in any of theobservations.
3.2 Flux-resolved analysis
The source light curve was extracted from circular re-
gions of radius 35 arcsec, which were centered on the max- IRAS 13224-3809 is a highly variable source that exhibited
imum source emission. The background light curves were distinct flaring and quiescent periods during the500 ks ob-
(cid:13)c 2012RAS,MNRAS000,1–6
IRAS 13224-3809 flux-dependent reverberation lags 3
1
-1ct s) 00..68
d ( 0.4
Har 0.2
0
8
-1ct s) 6
Soft ( 24
0
0 5×104 105 0 5×104 105 5×104 0 5×104 105
Time (s)
Figure2.Hardband(1.2–5keV)andsoftband(0.3–1keV)lightcurvesforthe4orbitsin500sbins.Greenregionsdenotehighflux
segments andredregionsdenotelowfluxsegments.Thesehighandlowfluxsegments willbeusedforthelaganalysisshowninFig.3.
500 400
250
200
0
0
s) s)
g (−250 g (
a a
L L
−200
−500
−400
−750 Low Flux Low Flux
High Flux High Flux
−1000 −600
10−4 10−3 10−4 10−3
Temporal Frequency (Hz) Temporal Frequency (Hz)
Figure 3. Lag-frequency spectrum for the low (red) and high Figure4.Lag-frequencyspectrumofthe4thorbitalone,forthe
(green)fluxsegmentsshowninFig.2.Themostnegativelagfor low (red) and high (green) flux segments shown in Fig. 2. The
thehighfluxintervalsis−660±280satν=1.8×10−4 Hz,and most negative lag for the high flux interval is −350±130 s at
the most negative lag for the low flux is higher: −230±31 s at ν =2.2×10−4 Hz,andthe mostnegative lagforthe lowfluxis
ν=7×10−4 Hz.Again,note thatthemostnegative lagforthe higher:−118±42satν=6.5×10−4 Hz.
total observation(Fig.1)occursatν=4.1×10−4 Hz.
the total, indicating the source variability is occurring on
shorter timescales.
servation.Inthissection,weexaminethelagsfromlowand
Asacheck,wecomputeflux-dependentlagsusingorbit
high flux intervals. Fig. 2 shows light curves from the 4 or-
4alone,whichshowsdistinctlowandhighfluxsegments(as
bits.Thelow fluxsampleisshown inred,andthehigh flux
designatedingreenandredintherightmostpanelofFig.2).
is shown in green. We chose light curvesegments that were
The lag is computed for these two sets of continuous light
long in order to obtain information at low frequencies. The
power spectral density confirms that up to 1.5×10−3 Hz, curves, and shown in Fig. 4. We find that for continuous
segments (i.e. for one realisation of some underlying pro-
we are not in a Poisson noise dominated frequency regime
cess),thesameflux-dependentbehaviourofthelagisclear,
for both thelow and high fluxsamples.
justwithlowersignal-to-noise.Theanalysisthatfollowshas
been performed with low and high flux segments from the
total 500 ksobservation.
3.2.1 The lag-frequency spectrum
Fig. 3 shows the lag vs. frequency for the high and low
3.2.2 The lag-energy spectrum
flux segments. The lag is computed in the same way as in
Section 3.1, between the energy bands of 0.3 – 1 keV and Wealsoexaminehowtheflux-dependentlagevolveswithen-
1.2 – 5 keV. Focusing on the high flux points in green, we ergyatthefrequenciesofthenegativelag(i.e.[5.8−10.5]×
see that the soft lag occurs at lower frequencies than found 10−4 Hzfor thelow flux,and [1.4−2.8]×10−4 Hzfor high
fromthetotallightcurvesample,inFig.1.Thismeansthat fluxes). In this analysis, we measure the time lag of light
thevariability associated with thesoft lag occurs on longer curves in relatively narrow energy bins with respect to the
timescales duringthe flaring period. Also, the amplitude of light curve of a broader reference band from 0.3 – 0.8 keV
the lag is greater. Now looking at the red low flux points, (see Kara et al. 2012, for further details). We use the con-
we see that the soft lag occurs at higher frequencies than ventionthatapositivelagmeansthatthelightcurveinthat
(cid:13)c 2012RAS,MNRAS000,1–6
4 Kara et al.
10
-1V (a) (b) (c)
1
e
k
-1s
s 0.1
nt
u
o
d c 0.01
e
z
mali 10−3 high flux powerlaw low flux powerlaw
Nor high flux reflection low flux reflection
10−4
1 10 1 10 1 10
Energy (keV)
Figure 5. Energy spectra for the high and low flux samples in panels (a) and (b) respectively. For clarity, only the power-law and
reflectioncomponents areplotted. Panel(c)showsthepowerlawandreflectionforthelowandhighfluxspectraoverplotted.
250 forthehighandlowfluxsegments,respectively.Forclarity,
onlythepower-lawandtotalreflectioncomponentsareplot-
0 ted.Thehigh fluxspectrum isdominated bythepowerlaw,
while the low flux spectrum, shows a greater contribution
−250 from reflection.
s)
g ( Panel (c) of Fig. 5 shows the power-law and reflection
a
L
components of the low and high flux segments overplotted.
−750 We notice that the power-law flux increases by an order of
magnitude between the low and high flux spectra, but the
Low Flux
−1000 reflection component remains nearly unchanged, especially
High Flux
at theenergy of theFe K line.
1 10
InKara et al.(2012),weuseMonteCarlosimulationsto
Energy (keV)
showthatdilutionwillhaveaneffectonthemeasurementof
thelag.Thelagismeasuredbetweenasoftenergybandand
Figure6.Lag-energyspectraforthelow-flux(red)andhigh-flux
a hard band, where, to first order, we approximate the soft
(green) intervals, showing the energy dependence of the lag for
thefrequenciesofthesoftlag(ν=[5.8−10.5]×10−4 Hzforthe bandasreflectionandthehardbandaspowerlaw.However,
lowflux,andν=[1.4−2.8]×10−4 Hzforthehighflux). we understand that there is ‘contamination’ in each band
caused by the other varying component. As these compo-
nentshave different intrinsic time delays, the net measured
lag will be some weighted average of all the components.
energy bin lags behindthereference band,while a negative
In the ideal case, where there is no contamination, the in-
lag means that the light curve in that energy bin leads the
trinsic lag will be the measured lag. In the other extreme,
reference.
where there are equal parts of reflection and powerlaw in
Fig.6showsthehigh-frequencylag-energyspectrumfor
eachband,themeasuredlagwillbezero.Furthermore,con-
thelowandhighfluxsegments.Thelowfluxsegmentsshow
tamination bya non-varyingoran uncorrelated component
a much clearer signal than the high flux,likely due to poor
willnoteffectthemeasurementofthelag,asthelagismea-
statistics and lower coherence in the high flux.The general
sured between coherent signals.
trendsofthetwolag-energyspectraaresimilar,exceptthat
theamplitude of thelag between 1–4 keV is greater for the
It is clear from the spectra in Fig. 5 that we need to
high flux. The low-flux lag-energy spectrum shows a clear
accountfordilutioninthemeasurementofthelag.Herewe
peak at ∼ 6.5 keV, the energy of the Fe Kα line. There
quantifytheamountofdilution bymeasuring thereflection
does not appear to be a corresponding peak at high fluxes,
fractioninthehardandsoft bands.Forexample,inthelow
but as theerror bars are so large, it cannot beruled out.
fluxspectrum, thesoft band is composed of 80% reflection,
but also the hard band is composed of 45% reflection. The
netamountof reflectionbetween thetwobandsistherefore
3.2.3 The time-integrated energy spectrum
35%, meaning that the measured lag is only 35% of the
Wecomplementthelaganalysispresentedabovewithabrief intrinsic lag. Similarly for the high flux spectrum, the soft
investigation of the spectral properties in the low and high band is composed of 40% reflection, while the hard band
flux segments. We applied the best fit spectral model from iscontaminatedby10%reflection.Againthenetamountof
Fabian et al.(2012b)tothelowandhighfluxspectra,freez- reflectionis30%,meaningthatthemeasuredlaginthehigh
ing all physical parameters that are not likely to change fluxis only 30% of the intrinsic lag. Weconclude then that
betweenstate(i.e.inclination,galacticabsorption,Feabun- the effect of dilution is nearly the same for both the high
dance).Theresultsareshowninpanels(a)and(b)ofFig.5 and low flux lags, and therefore we can be confident that
(cid:13)c 2012RAS,MNRAS000,1–6
IRAS 13224-3809 flux-dependent reverberation lags 5
600 200 200
25
1H0707-495
IRAS 13224-3809
150 0 0
400
100 −25
g (s) g (s)−200 −50
La 50 La
−400 −75
0 0
1H0707-495 −100
−600 IRAS 13224-3809 (low flux state)
−50
−125
−200
10−4 10−3 0.01 1 10
Temporal Frequency (Hz) Energy (keV)
Figure 7. Thelag-frequency spectrum forthe total observation Figure 8. High-frequency lag-energy spectrum for the low flux
(sameasFig.1)inred,overplotted withthelag-frequencyspec- segments in IRAS 13224-3809 (red), overplotted with the high-
trumof1H0707-495fromKaraetal.(2012)inblue. frequencylag-energyspectrumof1H0707-495(ν=[0.98−2.98]×
10−3 Hz)inblue.
the intrinsic amplitude of the lag in the high flux sample is
indeed greater than thelag in thelow flux. fluxes,which couldimplythatthecoronaextends3.5 times
furtheraway in thehigh state than in thelow.
The energy spectra of the low and high flux segments
(Fig.5)agreewiththeinterpretationofanextendedcorona.
4 DISCUSSIONS
The low flux sample shows a stronger contribution from
(i) As shown in Fabian et al. (2012b), IRAS 13224-3809 reflection. This is expected from a more compact corona,
exhibitsfrequencydependentlagswherethehardbandlags where light bending effects close to the central black hole
the soft band at low frequencies below ∼ 2 × 10−4 Hz, causeagreaterfractionofcontinuumemissiontobedirected
whilethesoftbandlagsthehardbandathigherfrequencies towards the disc. Spectral modelling of 1H0707-495 during
(Fig. 1). a low flux state in 2011 also shows a more compact corona
(ii) In this work, we take a step further and study the (Fabian et al. 2012a).
flux-dependenceof the observed lags. We find that the soft Thespectraldifferencebetweenlowandhighfluxsam-
lag has a larger amplitude and occurs on longer timescales plesappearstobecausedbyachangeinpowerlawflux,and
duringhigh fluxstates than duringlow flux states (Fig. 3). not from a change in reflection. Long timescale variations
(iii) According to the lag interpretation presented by in continuum that are decoupled from reflection have also
Fabian et al. (2009), Zoghbi et al. (2010) and others, soft been observed in MCG-6-30-15 (Vaughan & Fabian 2004;
lags are reverberation lags due to the light travel time Miniutti et al. 2007), and explained within the framework
between the primary source and the inner accretion disc. of strong light bending. We note that while the reflec-
Therefore, the low flux state (smaller reverberation lags tion spectrum appears unchanged on these long timescales,
at shorter timescales) refers to a compact emitting source the frequency-resolved covariance spectrum (as shown for
that is closer to the central black hole, while the high flux 1H0707-495 in Kara et al. 2012) does show that reflection
state(bigger lag at longer timescales) refers toan extended varies with thecontinuum on short timescales.
emitting region further from the black hole. Moreover, we
point out that a more compact corona will inevitably ir-
radiate smaller disc radii because of light bending (e.g. 4.1 Comparison with 1H0707-495
Miniutti & Fabian2004),thusprovidingaqualitativelyself-
ThelaganalysisofIRAS13224-3809showsstrikingsimilari-
consistent explanation for both the higher frequency and
tiestotheresultsfoundfor1H0707-495using1.3Msofdata
smaller amplitude of thesoft lags duringlow flux states.
(Kara et al.2012).Fig.7showsthelag-frequencyspectrum
Thespatialextentofthecoronacanbeinterpretedasan ofIRAS13224-3809 overplottedwith1H0707-495 (notedif-
increase in theheight of thecorona abovethedisc, or some ferent y-axis scales). We find that IRAS 13224-3809 shows
radial expansion of the corona, or some combination of the same general positive-to-negative trend as seen in 1H0707-
two. An increase in the amplitude of the lag is easily un- 495,justwithalargeramplitudenegativelagthatoccursat
derstood for avertically extendedcorona asthelight travel lower frequencies. The negative lag in 1H0707-495 is mea-
betweenthecoronaandthediscincreases.Foraradiallyex- sured to be−31.9±4.2 s at 1.33×10−3 Hz, while theneg-
tendedsource,onecanimagine thatthecoronairradiates a ative lag in IRAS 13224-3809 is a factor of ∼ 3.33 times
largerportionofthedisc,causingalongeraveragelag.Some more(−92±31s)atafrequencythatis∼3.25timeslower
complexities may arise with this interpretation because the (4.1 × 10−4 Hz). Both the lag and the frequency change
emissivitydecreasesgreatlywithlargerdiscradius,however roughly by a factor of 3.3. Assuming a simple mass scaling
we cannot ruleout a radially extended corona from theob- relationship,thisimpliesthatIRAS13224-3809 is3.3 times
servations presented here. Both the lag amplitude and the moremassivethat1H0707-495. Accordingtothemassscal-
frequencychangebyafactorof∼3.5betweenlowandhigh ingtrendsinDe Marco et al.(2012),thelagsinIRAS13224-
(cid:13)c 2012RAS,MNRAS000,1–6
6 Kara et al.
3809areconsistentwithablackholemassof107M⊙.Inthis Boller T., Tanaka Y., Fabian A., Brandt W. N., Gallo L.,
work,however, weuseIRAS13224-3809 toshow that there Anabuki N., Haba Y., Vaughan S., 2003, MNRAS, 343,
issomeflux-dependencetothereverberationlag.Therefore, L89
we should take some caution in predicting the mass of the Chartas G., Kochanek C. S., Dai X., Poindexter S.,
central black hole from the amplitude and frequency of the Garmire G., 2009, ApJ, 693, 174
lag. While observations used to measure reverberation lags Chartas G., Kochanek C. S., Dai X., Moore D., Mosquera
arerelativelylong,itispossibletosamplesomeintrinsically A.M., BlackburneJ. A., 2012, ApJ,757, 137
high or low flux state. This would introduce systematics to DaiX.,KochanekC.S.,ChartasG.,Kozl owskiS.,Morgan
thesimple mass scaling relationship. C. W., Garmire G., Agol E., 2010, ApJ,709, 278
We compare the low-flux lag-energy spectrum of De Marco B., Ponti G., Uttley P., Cappi M., Dadina M.,
IRAS 13224-3809 with the lag-energy spectrum of 1H0707- Fabian A. C., Miniutti G., 2011, MNRAS,417, L98
495(Fig.8).Theshapesofthespectraforthesetwodifferent De Marco B., Ponti G., Cappi M., Dadina M., Uttley
sources are strikingly similar, but theamplitudeof thelags P., Cackett E. M., Fabian A. C., Miniutti G., 2012,
is greater for the low flux segments in IRAS 13224-3809. arXiv:1201.0196
The peak at ∼ 6 keV and the dip at 3 – 4 keV are clearly Emmanoulopoulos D., McHardy I. M., Papadakis I. E.,
present in both data sets. Being able to isolate the low- 2011, MNRAS,416, L94
fluxsegments of IRAS13224-3809 has allowed usto clearly Fabian A.C., et al., 2009, Nature, 459, 540
uncover the reverberation signal, as was only possible with Fabian A.C., et al., 2012, MNRAS,419, 116
1.3 Msofdatawith 1H0707-495. 1H0707-495 doesnot have Fabian A.C., et al., 2012, arXiv:1208.5898v1
such distinct low-flux segments as IRAS 13224-3809, and Gallo L. C., Boller T., Tanaka Y., Fabian A. C., Brandt
so a similar flux-resolved analysis was not possible for this W.N.,WelshW.F.,AnabukiN.,HabaY.,2004,MNRAS,
source. 347, 269
The high-frequency lag-energy spectra of IRAS 13224- Guilbert P. W., Rees M. J., 1988, MNRAS,233, 475
3809 and 1H0707-495 (Figs. 6 and 8) are consistent with a Haardt F., Maraschi L., 1991, ApJ, 380, L51
reflectionmodel,wherethenegativelagsareassociatedwith Haardt F., Maraschi L., 1993, ApJ, 413, 507
thelighttraveltimebetweenthecontinuum-emittingcorona Jansen F., et al., 2001, A&A,365, L1
and the accretion disc. The lag-energy spectrum shows the KaraE.,FabianA.C.,CackettE.M.,SteinerJ.F.,Uttley
1 – 4 keV band (powerlaw dominated) leads the 0.3 – 0.8 P., Wilkins D.R., Zoghbi A., 2012, MNRAS,264
keVreferenceband(wherereflectionisappreciable).Similar Kotov O., Churazov E., Gilfanov M., 2001, MNRAS, 327,
to 1H0707-495, we see a distinct peak at ∼ 6 keV, the rest 799
frameenergyoftheFeKline,andadipat3–4keV,which Lightman A.P., WhiteT. R., 1988, ApJ,335, 57
hasbeeninterpretedasasignaturefromtheredwingofthe McHardy I., 1988, MmSAI,59, 239
FeKline,originatingfromtheinnermostradius.Thisinter- McHardy I. M., Papadakis I. E., Uttley P., Page M. J.,
pretationisdiscussedfurtherinKara et al.(2012),wherewe Mason K. O., 2004, MNRAS,348, 783
model the high-frequency lag-energy spectrum of 1H0707- Miyamoto S., Kitamoto S., Mitsuda K., Dotani T., 1988,
495. Natur,336, 450
Miniutti G., Fabian A. C., 2004, MNRAS,349, 1435
Miniutti G., et al., 2007, PASJ,59, 315
4.2 Conclusions Nowak M. A., Vaughan B. A., 1996, MNRAS,280, 227
NowakM.A.,VaughanB.A.,WilmsJ.,DoveJ.B.,Begel-
Looking at the flux-resolved reverberation lags reveals new
man M. C., 1999, ApJ,510, 874
observational constraints that can help us distinguish be-
Papadakis I. E., Nandra K., Kazanas D., 2001, ApJ, 554,
tween the physical nature of the low and high flux states.
L133
We learn that flares are likely connected to the expansion
Ponti G., et al., 2010, MNRAS,406, 2591
of the corona out vertically and possibly radially, too, and
Ross R. R.,Fabian A. C., 2005, MNRAS,358, 211
that when the source is in quiescence, the corona remains
ShakuraN. I., SunyaevR.A., 1973, A&A,24, 337
compact and close to thecentral region.
Stru¨der L., et al., 2001, A&A,365, L18
Terrell J., 1967, ApJ, 147, 827
Uttley P., Wilkinson T., Cassatella P., Wilms J.,
PottschmidtK.,HankeM.,B¨ockM.,2011,MNRAS,414,
ACKNOWLEDGEMENTS
L60
This work is based on observations obtained with XMM- Vaughan S., Fabian A.C., 2004, MNRAS,348, 1415
Newton, an ESA science mission with instruments and ZoghbiA.,FabianA.C.,UttleyP.,MiniuttiG.,GalloL.C.,
contributions directly funded by ESA Member States and ReynoldsC.S.,MillerJ.M.,PontiG.,2010,MNRAS,401,
NASA. EK thanks the Gates Cambridge Scholarship. ACF 2419
thankstheRoyal Society. Zoghbi A., Fabian A. C., Reynolds C. S., Cackett E. M.,
2012, MNRAS,422, 129
REFERENCES
Ar´evalo P., Uttley P., 2006, MNRAS,367, 801
Boller T., Brandt W. N., Fink H., 1996, A&A,305, 53
(cid:13)c 2012RAS,MNRAS000,1–6